East Asia and Western Pacific METEOROLOGY AND CLIMATE

UNIVERSITY OF HONG KOHG
LlBEAEY

East Asia and Western Pacific
METEOROLOGY
AND CLIMATE

International Conference on
East Asia and Western Pacific
METEOROLOGY
AND CLIMATE
Hong Kong 6-8 July 1989
Editors:
P Sham
Royal Observatory Hong Kong
Hong Kong
CP Chang
Naval Postgraduate Schol
USA
VLX World Scientific
iSff^ Singapore • New Jersey • London • Hong Kong

Published by
'*"*"" ,1: ,-.;•„.. ,.»... octjsia
^ __ * - - H—
7 - >
World Scientific Publishing Co. Pte. Ltd.
P 0 Box 128, Faner Road, Singapore 9128
USA office: 687 Hartwell Street, Teaneck, NJ 07666
UK office: 73 Lynton Mead, Totteridge, London N20 SDH
EAST ASIA AND WESTERN PACIFIC METEOROLOGY
AND CLIMATE
Copyright © 1990 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced
in any form or by any means, electronic or mechanical, including photocopying,
recording or any information storage and retrieval system now
known or to be invented, without written permission from the Publisher.
ISBN 981-02-0027-7.
Printed in Singapore by JBW Printers & Binders Pte. Ltd.

FOREWORD
The International Conference on East Asia and Western Pacific Meteorology
and Climate, held on 6-8 July 1989 in Hong Kong and jointly hosted by the Royal
Observatory Hong Kong and the Centre of Asian Studies, University of Hong Kong,
provided a rare opportunity for scientists in the region, especially those prominent
Chinese-speaking meteorologists and oceanographers in China, Hong Kong,
Singapore, Taiwan and U.S.A., to meet one another, to share their knowledge and to
exchange their research experiences.
More than 70 invited scientists participated in the Conference and 68 papers
were presented.
The Conference was considered a historic event and a great success which was
largely due to the contributions of many individuals and organizations. The support
given by the United States University Corporation for Atmospheric Research and the
World Scientific Publishing Company, Singapore is especially acknowledged.
The proceedings contain a collection of refereed papers printed in English in
the hope that it will achieve a wide readership, although many of these papers were
originally written and delivered to the Conference in the Chinese language. I should
like to record my heartfelt appreciation of the excellent efforts devoted by the referees
themselves in this regard. It is hoped that through the proceedings many scientists in
the world will be able to share the knowledge of meteorology and climate of the East
Asia and Western Pacific region and the knowledge will be applied for the benefit of
the world at large.
P. Sham
Director of the Royal Observatory
Hong Kong

INTERNATIONAL CONFERENCE ON
EAST ASIA AND WESTERN PACIFIC
METEOROLOGY AND CLIMATE
6-8 July 1989
Hong Kong

VHI
LIST OF PARTICIPANTS IN THE PHOTOGRAPH
1 ANTHES, Richard A.
2 CHAN, Hong Ping
3 CHAN, Johnny C.L.
4 CHAN, Man Yun
5 CHAN, Yuk Kwan
6 CHANG, Chih-Pei
7 CHANG, Kar Man
8 CHANG, Long -Nan
9 CHANG, Simon Wei -Jen
10 CHAO, Jiping
11 CHEN, Qiushi
12 CHEN, George Tai-Jen
13 CHENG, Tze Shan
14 CHIU, Hung Yu
15 CHOU, Jifan
16 CHUANG, Wen-Ssn
17 DING, Yihui
18 GINN, Edwin W.L.
19 GRAY, David M.
20 GUO, Xiaorong
University Corporation for Atmospheric
Research, U.S.A.
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Naval Postgraduate School, Monterey,
U.S.A.
Royal Observatory, Hong Kong
National Central University, Taipei,
China
Naval Research Laboratory, Washington
D.C., U.S.A.
National Research Centre for Marine
Environmental Forecast, State
Oceanographic Administration, Beijing,
China
Peking University, Beijing, China
National Taiwan University, Taipei, China
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Lanzhou University, Lanzhou, China
National Taiwan University, Taipei, China
Academy of Meteorological Science, State
Meteorological Administration, Beijing,
China
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
National Meteorological Centre, State
Meteorological Administration, Beijing,
China

21 HONG, Siu-Shung National Central University, Taipei,
China
22 HSU, Sheng-I
23 HUANG, Joseph Chi-Kan
24 HUANG, Shisong
25 HWU, Shung Cho
26 JI, Liren
27 JIANG Huo-Ming
28 KAU, Wen-Shung
29 KONG, Chi Wing
30 KOO, Elaine
31 KOT, S.C.
32 KRISHNAMURTI, T.N.
33 KYLE, William J.
34 LAM, Chiu Ying
35 LAM, H.K.
36 LAM, Hilda
37 LAU, Alexis Kai-Hon
38 LAU, Ngar-Cheung
39 LAU, Robert
40 LAU, Sharon, S.Y.
41 LAU, William K.-M.
42 LEE Cheng-Shang
43 LEE, Boon Ying
44 LEUNG, Wing Mo
Chinese University, Hong Kong
National Oceanic and Atmospheric
Administration, Rockville, U.S.A.
Nanjing University, Nanjing, China
Royal Observatory, Hong Kong
Academia Sinica, Beijing, China
National Central University, Taipei,
China
National Taiwan University, Taipei, China
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
University of Hong Kong, Hong Kong
Florida State University, Tallahassee,
U.S.A.
University of Hong Kong, Hong Kong
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Princeton University, Princeton, U.S.A.
Princeton University, Princeton, U.S.A.
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong
Goddard Laboratory for Atmospheres,
National Aeronautics and Space
Administration, U.S.A.
National Taiwan University, Taipei, China
Royal Observatory, Hong Kong
Royal Observatory, Hong Kong

45 LIM, Hock National University of Singapore,
Republic of Singapore
46 LIN, Ho National Taiwan University, Taipei, China
47 LIN, Pay-Liam National Central University, Taipei,
China
48 LIN, Yeong-Jer St. Louis University, St. Louis, U.S.A.
49. LIOU, Chi-Sann Naval Environmental Prediction Research
Facility, Monterey, U.S.A.
50 LIU, Cho-Teng National Taiwan University, Taipei, China
51 LIU, Chung-Ming National Taiwan University, Taipei, China
52 LIU, David Shiao-Kung Rand Corporation, Santa Monica, U.S.A.
53 LIU, Gin-Rong National Central University, Taipei,
China
54 LIU, Shida Peking University, Beijing, China
55 LO, K. Kenneth National Taiwan University, Taipei, China
56 LU, Daren Academia Sinica, Beijing, China
57 LUO, Huibang Zhongshan University, Guangzhou, China
58 MAX, Mankin University of Illinois, Urbana, U.S.A.
59 PAN, Yihang National Research Centre for Marine
Environmental Forecast, State
Oceanographic Administration, Beijing,
China
60 PENG, Guangyi Chinese Meteorological Society, Beijing,
China
61 PENG, Melinda S. Naval Postgraduate School, Monterey,
U.S.A.
62 POON Hoi To " Royal Observatory, Hong Kong
63 PU, Shuzhen First Institute of Oceanography, State
Oceanographic Administration, Qingdao,
China
64 RAKTABUTR, Theeranun Meteorological Department, Bangkok,
Thailand
65 SAWYER-CROUCH, Karyn University Corporation for Atmospheric
Research U.S.A.

XI
66 SHAM, Pak
67 SHIEH, Shinn-Liang
68 SOONG, Su-Tzai
69 SU, Jilan
70 SUN, Wen-Yih
71 TAO, Shiyan
72 TERNG, Chuen-Teyr
73 TSAY, Ching-Yen
74 TSE, Wai Ming
75 TSENG, Hsien-Yuan
76 WAI, Hon Gor
77 WANG, Bin
78 WANG, Jough-Tai
79 WANG Y.-J.
80 WU, Jin
81 WU, Ming-Chin
82 WU, Rongsheng
83 XU, Jianmin
84 YEUNG, K.K.
85 ZHANG, Zuojun
86 ZHOU, Xiaoping
87 ZHOU, Xiuji
88 ZHU, Qiangen
Royal Observatory, Hong Kong
Central Weather Bureau, Taipei, China
University of California, Davis, U.S.A.
Second Institute of Oceanography, State
Oceanographic Administration, Qingdao,
China
Purdue Unviersity, W. Lafayette, U.S.A.
Academia Sinica, Beijing, China
Central Weather Bureau, Taipei, China
National Taiwan University, Taipei, China
Royal Observatory, Hong Kong
Civil Aviation Administration, Taipei,
China
Royal Observatory, Hong Kong
University of Hawaii, Honolulu, U.S.A.
National Central University, Taipei,
China
National Taiwan University, Taipei, China
University of Delaware, Lewes, U.S.A.
National Taiwan University, Taipei, China
Nanjing University, Nanjing, China
Satellite Meteorology Center, State
Meteorological Administration, Beijing,
China
Royal Observatory, Hong Kong
Academia Sinica, Beijing, China
Academia Sinica, Beijing, China
Academy of Meteorological Science, State
Meteorological Administration, Beijing,
China
Nanjing Meteorological Institute,
Nanjing, China

XIII
INTERNATIONAL ORGANIZING COMMITTEE OF THE INTERNATIONAL CONFERENCE ON
EAST ASIA AND WESTERN PACIFIC METEOROLOGY AND CLIMATE
Chairman
: Mr. SHAM, Pak
Royal Observatory, Hong Kong
Vice-Chairman : Prof. CHANG, C.P.
Naval Postgraduate School, California, U.S.A.
Members : Prof. CHEN, Edward
University of Hong Kong, Hong Kong
Prof. DING, Yihui
Academy of Meteorological Science, Beijing, China
Prof. HONG, Siu-Shung
National Central University, Taipei, China
Dr. HUANG, Joseph Cht-Kan
Oceanic and Atmospheric Research/NOAA,
Maryland, U.S.A.
Mrs. KOO, Elaine
Royal Observatory, Hong Kong
Mr. LAM, Chiu-Ying
Royal Observatory, Hong Kong
Dr. LAU, William K.-M.
Goddard Laboratory for Atmospheres, Maryland, U.S.A.
Prof. LIN, Hai
Acadernia Sinica, Beijing, China
Prof. LIN, Ho
National Taiwan University, Taipei, China
Ms SAWYER-CROUCH, Karyn
University Corporation for Atmospheric Research
U.S.A.
Prof. TAG, Shiyan
Institute of Atmospheric Physics, Beijing, China
Prof. TSAY, Ching-Yen
National Taiwan University, Taipei, China
Prof. WU, Rongsheng
Institute of Atmospheric Sciences, Nanjing, China

XV
CONTENTS
Foreword
Photograph of Participants
List of Participants in the Photograph
International Organizing Committee
v
vi
viii
xiii
Session 1: Monsoon Meteorology (I)
The thermal structure and convective activities over Tibetan Plateau in
summer and their relations with large-scale circulation 1
Duzheng Ye
The changes in circulations during the transition period from winter
monsoon to summer monsoon in the northern hemisphere 2
Shiyan Tao
The coupling of upper-level and low-level jet streaks during Taiwan
heavy rainfall period in Mei-Yu season 3
Ching- Yen Tsay and Wen-Shung Kau
A numerical study on the effect of the Tibetan Plateau upon cold
surges in East Asia 4
Qiangen Zhu and Song Yang
Overview of Mei-Yu research in Taiwan 14
George Tai-Jen Chen
Session 2: Monsoon Meteorology (II)
On temporal variations of (ow level jets associated with the Asian
summer monsoon 38
Long-Nan Chang and Fang-Chuan Lu
Observed structure and propagation characteristics of summertime
synoptic-scale disturbances over the tropical western Pacific 48
Alexis-Kai-Hon Lau and Ngar-Cheung Lau

XV!
The mean heat sources over Asian monsoon region during the period
from April to October of 1980-1983 58
Huibang Luo
A numerical simulation of the Mei-Yu front 68
L C. Chou, C. P. Chang and R. T. Williams
Wind and moisture fields during the periods of enhanced and suppressed
convective activity over the Malaysia-South China Sea region during the
northern winter monsoon, November-December 1986 69
Boon-Khean Cheang and Prakash Sank a ran
A simulation of Lee-Cyclogenesis over Yun-Gui Plateau with the use
of a hemispheric spectral model 80
Wen-Shung Kau and Yi-May Lin
Seasonal and intraseasonal variations of the east Asian summer monsoon 94
K. -M. Lau
Influence of variations of the circulation system over the South Indian
Ocean on the East Asia summer monsoon and the northern hemispheric
general circulation of the atmosphere 105
Shisong Huang, Mingmin Tang and Xiuqun Yang
Some dynamic aspects of the equatorial intraseasonal oscillations 119
Bin Wang and Hualan Rui
Session 3: Remote Sensing and In-situ Measurements
Chinese polar orbiting meteorological satellite FY-1 131
Jianmin Xu
Applying tropopause data observed by VHP radar to improve satellite
temperature sounding 139
Gin-Rong Liu
The mesocale monitoring system of rainstorms and severe convective
weather events in China 140
XiujiZhou, Runsheng Ge, Da-an Ma, Conglong Zhao, Daren Lu and
Hal Lin

XVII
Structural features of a squall line over the Taiwan Strait revealed by
dual-doppler radar 150
Y. J. Lin, /. W. Pasken and H. Shen
Radar observation of precipitation systems in Taiwan 160
Pay- Liam Lin and Tai-Chi Wang Chen
Remote sensing of atmospheric compositions and optical characteristics 170
Hal Lin, Daren Lu and X/u/f Zhou
Comparison of radar estimates and surface rainfall during rainstorms in
1987-88 180
Hilda Lam and L. O. Li
Doppler weather radar observation and aviation weather service 190
Hsien-Yuan Tseng
Revival of the tipping-buket raingauge 199
Sheng-/ Hsu
Impact of hourly S-VISSR satellite imagery on operational forecasting
in Hong Kong 210
S. T. Lai, B. Y. Lee and H. K. Lam
Session 4: Climate and General Circulation
Predictability of low frequency modes 220
T. N. Krishnamurtl, S. Moten, D, Oosterhof and G. Daughenbaugh
A cloud wave theory and its application to the 30—50 day oscillation
in the equatorial atmosphere 226
Qfushi Chen and Kuo - Nan Liou
Normal modes of climatological mean flow and their roles in atmospheric
general circulation 240
Zuojun Zhang
yA study on the radiative balance simulated by a general circulation model 241
Jough-Tai Wang
Introduction to Heihe Basin Field Experiment (HElFE)-atmosphere-land
surface interactions program 251
Hal Lin

XVIII
, Jhe impact of urbanization on climate in Hong Kong and its
implications for human energy exchanges 261
William J. Kyle
Session 5: Mesoscale Meteorology
Recent applications of the Penn State/NCAR mesoscale model to
synoptic, mesoscale and climate studies 274
Richard A. Anthes
Baroclinic instability of modified Eady waves 284
Wen-YihSun
Influences of orography on flow in boundary layer 294
Rongsheng Wu
The microphysics of a Mei-Yu case: data analysis 304
Chung-Ming Liu and K. Kenneth Lo
On dynamical studies of orographically induced mesoscale phenomena 313
S/u-Shung Hong, Chung-Y/ng Hu and Fu-Shan Weng
Numerical simulations of topographical effects on airflow and
precipitation 323
Su-Tzai Soong, Mukut Mathur and Wei-Kuo Tao
Study on the frontal cyclone system in southern China and the vicinity
of Taiwan area during late-winter and early-spring 333
Huo-Ming Jiang
The microphysics of a Mei-Yu case: theory 343
K. Kenneth Lo and Chung-Ming Liu
Session 6: Air-Sea Interaction
Monthly and seasonal forecasts and tropical ocean-atmosphere
interactions 354
Jiping Chao f Zhengang Ji and Xiaoxi Wang
The teleconnections of equatorial SST in Taiwan area 364
Cho-Teng Liu

XIX
A three-dimensional model of the China Seas and aspects
of typhoon surge predictions 365
S. K. Liu, P. Wu, T. Y. Wu and S. T. Wang
Interannual variabilities of the western tropical Pacific Ocean
and low frequency response of the subtropical high over the
northwest Pacific Ocean 375
Shuzhen Pu and Hulling Yu
Large scale air-sea interaction in the western Pacific region 385
Joseph C. K. Huang
The relationship between currents and winds northeast of Taiwan 399
Wen-Ssn Chuang
China-Japan joint research program on the Kuroshio 400
Jilan Su
Influence of microscale air-sea interaction in climate research 403
Jin Wu
The variations of the SST in the eastern and western tropical Pacific
and their relationship with those in the world ocean 412
Yi-Hang Pan, Abraham H. Oort and William Richardson
On laboratory simulation of sea breeze 420
S. C. Kot
Session 7: Operational Weather Forecast and Numerical
Weather Prediction
The implementation and operation of an analysis scheme and a limited
area model for Hong Kong 429
Y. K. Chan
BMC limited area model: operational application and research 439
Xiaorong Quo, Zhihui Yan and Guoan Zheng
The operational global forecast system at Central Weather Bureau 449
Chuen-Teyr Terng

XX
The East Asia heavy rainfall numerical forecasting and the numerical
nowcasting of severe convective weather 450
Xiaoping Zhou
Numerical simulation of mesoscale meteorological phenomena in
Hong Kong 451
K. K. Yeung, W. L Chang and B. Wan
An overview of present typhoon forecast operation in Taiwan 461
Sh inn-L iang Sh ieh
The impact of the termination of aircraft reconnaissance on tropical
cyclone warnings and forecasts in western north Pacific 462
Johnny C. L. Chan and K. P. Wong
A spectral model for medium-range weather forecasts and its
performance 474
Liren Ji, Jiab/n Chen, Daomfn Zhang, Wanli Wu f Rujin Shen,
Hua Sheng and Boy in Huang
Long-range forecasting of Taiwan Mei-Yu 484
Ming-Chin Wu
The Royal Observatory long range rainfall forecast methods 485
Robert Lau and M. Y. Chan
Session 8a: Theoretical Studies
The nonlinear interaction of internal wave and turbulence 494
Sh/da Liu
Multiple equilibria of a thermally forced baroclinic atmosphere 502
Chi-Sann Liou
Effects of vertical wind-shear on Kelvin wave—CISK modes 515
Hock Urn, Tian-Kuay Urn and C. P. Chang
Analogous rhythm phenomenon of climatic anomalies on seasonal
scale 517
Jifan Chou
An inquiry into the nature of regional cyclogenesis 525
Mankin Mak

XXI
Session 8b: Typhoon Studies
Typhoon formation and development—an observational point of view 536
Cheng-Shang Lee
Dynamics of vortex motion on tropical /5-plane 537
Mefinda S. Peng and R. T. Williams
Effect of the thermal dynamic forcing on the secondary circulation
of typhoons 547
Yihui Ding, Yuezhen Liu and Ziping Bun
Recent results in limited-area numerical weather prediction 559
Simon Wei-Jen Chang, Rangarao Venkata Madala and
Keith Denis Sashegyi
Author Index 569

THE THERMAL STRUCTURE AND CONVECTIVE ACTIVITIES
OVER TIBETAN PLATEAU IN SUMMER AND THEIR
RELATIONS WITH LARGE-SCALE CIRCULATION
Ye Duzheng
The Chinese Academy of Sciences, Beijing
ABSTRACT
Before the 50's when people talked about the effect of the largescale
orography, they mainly concentrated their attention on the
dynamical effect. Since we discovered, during the middle 50's that
Tibetan plateau is a big heat source in summer and a cold source in
winter it was found that besides dynamical effect, Tibetan plateau
also produces big thermal effect on the atmospheric circulation. On
this thermal effect we have continuously been doing researches. In
this paper we review only part of the series ot these studies,
concentrating our attention only on the thermal structure and the
convectiye activities and their interactions with the large-scale
circulation in summer.
Because of its height and clear air above it and the scattering
of the short-wave radiation by convective clouds, the surface of the
plateau receives very large amount of solar radiation, causing in
summer a very steep, on the average very near dry adiabatic lapse rate
in the lower part of the atmosphere and the associated very widespread
and strong convective activities. The strong convective
activities produce a deep well mixed layer with its top on the average
reaching 40t) hPa. There are two regions, one on the eastern plateau
and the other on western, with relatively concentrated convective
systems. The wide-spread strong convections result in an average
upward motion over the plateau. There are also two centers of maximum
upward velocity roughly coinciding with the two regions of strong
convection.
On the whole the large amount of rising air in the upper
troposphere flows with the westerlies over the northern plateau far
eastward and descends in the east Pacific Ocean ana with the
easterlies over southern plateau far westward and descends iri Iran and
further west. This rising air over the plateau flows with the wellknown
monsoon Hadley circulation across the equator and descends in
southern hemisphere. But the rising air cannot flow far northward, it
descends immediately down along the northern periphery of the plateau,
causing very dry climate there.
In summer in the lower layer over the plateau on the average is a
relatively stable cyclonic circulation. In the upper layer is a large
fairly steady anticyclonic circulation with high temperature and high
humidity. Through strong non-linear interactions the wide-spread and
strong small-scale convective systems play an important role in
maintaining the said large-scale circulation in more or less ^steady
state. Besides, the convective systems have strong diurnal variation.
And this strong diurnal variation also has an important influence on
the large-scale circulation.

THE CHANGES IN CIRCULATIONS DURING THE TRANSITION PERIOD FROM
WINTER MONSOON TO SUMMER MONSOON IN THE NORTHERN HEMISPHERE
Tao Shiyan
Institute of Atmospheric Physics, Beijing
ABSTRACT
It is well known that in the monsoon climate zone of the Northern
Hemisphere, there is an abrupt change in general circulations during
the transition period from winter monsoon to summer monsoon. In this
paper, by using the ECMWF data (1980-1983), an analysis is made on the
characteristics of the changes in circulations in the monsoon area
(30-160E), non-monsoon area (160-20W) and the whole Northern
Hemisphere during this transition period. It is found that during this
period the zonal averaged planetary zonal wind systems have a rapid
change. The west wind belt of the middle latitudes withdraws northward
and the withdrawal occurs earlier in the upper troposphere than in
lower troposphere. In the tropics, in the lower troposphere the zonal
averaged easterlies change into westerlies, and in the upper
troposphere the intensity or the easterlies increased rapidly. In the
monsoon area the above mentioned changes in circulation occur much
earlies than that in the zonal averaged wind systems of the whole
Northern Hemisphere. The largest change in circulations of the
monsoon area is in the Indian summer monsoon system. We also find that
in the monsoon area the onset of the summer monsoon propagates from
the South China Sea westward. In the South China Sea it occurs in
middle of May, in the Bay of Bangal in the last ten days of May and in
the west coast of India in early or middle of June.
Analysis of the temperature field shows that, during the
transition period, the "thermal equator" in the zonal averaged
temperature field shifts northward rapidly, and direction of the
temperature gradient between 10-35N reverses. There are large
differences in the changes in the temperature field between the
monsoon area and the non-monsoon area. The changes in the temperature
field in the Northern Hemisphere reflects to a large extent the
changes in the monsoon area.

THE COUPLING OF UPPER-LEVEL AND LOW-LEVEL JET STREAKS
DURING TAIWAN HEAVY RAINFALL PERIOD IN MEI-YU SEASON
Ching-Yen Tsay and Wen-Shung Kau
Department of Atmospheric Sciences
National Taiwan University
Abstract
In this study, we analyzed three heavy rainfall cases in the Mei-Yu season of the
FGGE year, i.e. May and June of 1979.
Composite results of the three cases showed that the low-level jet streak (LLJ)
at 700 mb was formed to the south of the entrance portion of the upper-level jet streak
(ULJ) at 200 mb, 24 hours before heavy rainfall occurred in Taiwan. The vertical
motion at 500 mb was upward in the region south of the entrance portion of the ULJ
axis, and to the north of the LLJ. The maximum upward motion was located to the
north of the LLJ core. A strong mesoscale convective system (MCS) was found in the
upward motion region. The north-south temperature gradient in the upper
troposphere was very strong in the entrance region of the ULJ and very weak in the
region of the LLJ. The characteristics of the Mei-Yu system including the relative
positioning of the ULJ, LLJ, upward motion, MCS and temperature distribution are
in general similar to those of the unstable system described by Chen (1982). The ULJ,
LLJ and upward motion region moved eastward and caused heavy rainfall in Taiwan.
A detailed study of one of the three cases further showed that the LLJ was
accompanied by a shear line and easterlies to its north. A two-cell vertical circulation
system with upward motion located between the two jets was found to be a dominant
feature in the region. A direct circulation cell was located in the entrance region of
the ULJ, while an indirect circulation cell formed in the core region of the LLJ. The
north-south distance between the ULJ and the LLJ was found to be about 900 km.
This observed separation distance is slightly less than the upper limit value for the
gravity-inertia wave described by Chen (1982) to become unstable.

A NUMERICAL STUDY ON THE EFFECT OP THE TIBETAN PLATEAU UPON COLD
SURGES IN EAST ASIA
Zhu Qiangen
Yang Song
Nanjing Institute of Meteorology, Nanjing, JS 210044*
1. INTRODUCTION
The East-Asian winter monsoon represents an essential process
in the atmospheric motion of this season, and vigorous outbreaks of
the wind may extend its influence down to low latitudes and even
across the equator,The NE wind in the South-China Sea, when increased
to a certain intensity owing to the effect of the monsoon* is
called a cold surge '. The intensification of cold surges in Asia
brings about enhanced convection and disturbance development over
the Sea and around the equator, thereby invoking the reinforced
2 ll)
Hadley cell and E - W oriented circulations * J '. Cold surges show
characteristics of gravity waves during the southward travel
Theoretical studies indicated
that the dispersion by gravity waves
can induce multi-scale waves at equatorial latitudes ** * Murakami
et al. ' showed that cold surges are likely to be Kelvin-type waves
excited by large-scale mountain boundary. Sumi '^in his numerical
simulation indicated that the occurrence of two spells of strengthened
northerly wind in cold surges is due to the excitation of Kelvin
waves. The aim of our paper is to investigate in more detail
the effect on cold surges of the Qinghai-Tibetan Plateau, the
mechanism
for the surge excitation and the impact on low-latitude atmospheric
conditions. Employed for this purpose is a limited
area
P— f) five—layer primitive equation model incorporating more detail-

Q \
ed atmospheric processes *
2. MODEL'S INITIAL FIELD AND DETAILS OF THE NUMERICAL SCHEME
Fig,l illustrates the movement of a surface cold high's center
and cold front during a cold surge episode over the Sea during 23-
29 December 1982. The episode can be partitioned into three stages:
initial (23-24), active (25-2) and decaying (28-29). The initial
field was taken from ECMWF 5°X 5° grid data at 2000 BST, 24 December
of the year and the initial SST field from the monthly mean of
December 1982.
Designed were three experiments*ALL for the experiment involving
all physical factors of the model; DH for the adiabatic; and NM for
the non-mountain experiment*
3. RESULTS OF EXPERIMENTS
1) ALL Experiment
The 72-h integration indicates close similarity between the calculated
and observed situation on a large-scale basis,except that a
small low was not predicted to the NE of Japan.Horizontal grid spacing
of the model might be too large to show smaller-scale disturbance.
This, however, has no significant effect on the problems to
be discussed*
The near-surface flow evolutions in the integration between 12
and 48 hr are depicted in Fig.2. In Fi#.2d, solid circles (A^) denote
the position of the anticyclone's center and open circles (Vv)
of the maximum NE wind of J m/s, at the i-th hour of integration.
It can be seen that the high 1 s center is around 43°N, 95°® at 12 h
and then moves southeastward, reaching the middle and lower Chang-
Jiang at 60 h. The leading edge of the surge, at 12 hr integration,
is found at 15°N with the maximum NE wind center of 10 m/s located
in the middle Changjiang. At the 24 h, the robust surge crosses the
whole South-China Sea, its leading edge arriving at the equator,
with the maximum wind region > 8 m/s around 22°N off South-China
shores. Then a narrow belt of stronger cold air in the form of an

arc around the Plateau is seen between mid and lower latitudes of
the eastern Asian continent, with NW (NE) winds north (south) of
30°N,which is obviously in close relation to the flow-steering role
of the orography. The 36~h integration indicates that NE winds in
the southern part of the Sea keeps on intensifying and extending
westward to the northern Indian, with the maximum wind center of 8
m/s displaced southward to the central South-China Sea. At the 48 h
the surge has moved further southward, with the NE wind of 6 m/s
reaching equatorial latitudes and a reduced wind region of 8 m/s
travelling westward to the Bay of Bengal. The 60~h result indicates
that the maximum NE wind has dropped to 6 m/s, implying that the
surge is beginning to weaken.
It can be seen in Fig.3a that, north of 5°N, after the starting
of integration,the V - isopleths are getting N - S directed and become
closer together,suggest ing that the northerly wind strengthens
rapidly in the first 12 hr, whereas for 5 N and south of it the reinforcement
is initiated at the 12th hr. At the 72-h integration
no northerly wind is seen at 25°N, indicating the cutoff of cold
air supply, thus suggesting that the surge moves into the decaying
stage. A salient feature displayed in Fig.Bb is that temperature
drop south of 25°N occurs after the integration of 12 hr.lt follows
that the enforced northerly wind due to the surge is always ahead
of temperature drop caused by it, which may be explained by the
character of cold surges as gravity waves *^.
In the 120°E cross section at the 24-h integration (Fig.4a) the
local Hadley cell intensifies to a considerable degree, with the
ascending leg around the equator and in the Southern Hemisphere
and descending leg equatorward of 20°N. In the 48-h cross section
(Fig»4b) the surge-caused near-surface northerly flow extends to
equatorial latitudes and the ascending leg of the cell ahead of the
surge retreats to 5°S and south of it* Fig.4c gives the difference
in vertical speed (fl^gr- w pA^* which shows enhanced legs of the cell.
Hence, in the southward travel of the surge, the Hadley cell retreats
and intensifies.

2) DH Experiment
The horizontal flow and height fields given by DH are roughly
identical with those by ALL, except for a slightly weaker downdraft
in the local Hadley cell (figure omitted),which indicates that
diabatic
factors have little effect on the surge propagation but contribute
more or less to the reinforcement of the cell.
3) NM Experiment
Illustrated in Pig.5 is the near-surface horizontal flow field
given by NM, which greatly differs from that by ALL. At the 12 hr
integration, the high's center is around 42°N, 91°E, and the Plateau
is under the control of a strong NE wind with ]> 16 m/s at the
wind
center,and the southmoet limit of the northerly wind is around 20°N
in eastern Asia. After the 24 h integration, the anticyclone starts
moving towards the east and weakens.
The robust easterly wind center
over the Plateau
decreases in intensity to 12 m/s. The eastern
Asian cold surge spreads to the equator, and the South-China Sea
is covered by a weak NE flow, with no intense wind center. After 36
h, the high moves further eastward and weakens. The NE windspeed
center on its south side weakens to 8 m/s and the NE wind in the
Sea is of 6 m/s only. At 48 h, the high continues to weaken, with
the winds on its north and south sides reduced to 6 m/s, and the
winds in the Sea and low-latitude western Pacific remain at
roughly
6 m/s.
Prom the above analyses, it follows that even in the absence
of a large-scale orography cold surges can penetrate into
equatorial latitudes. The surge is however weaker, accompanied by
a weakened cold high travelling towards the east with no passage
of cold air on its east side that has veered to the easterly wind
on the south side.
The time variation in mean near-surface V component and temperature
in the 110-120°$ belt given by NM is shown in Fig.6. It indicates
much the same features as in Fig.3* wind refreshing, albeit
not too significant, occurs prior to temperature drop, leading
the conclusion that the surge has properties of gravity waves.
Pigs*a,b display height-varying distribution of difference
to
in

mean vertical speed (W^ - K NM ) between 110-120°E at 24 and 48
hr integration, respectively. From Pig.4 one can see that up- and
downdraft of the local Hadley cell shown in ALL are stronger than
in NM, which suggests greater contribution of the vigorous surge to
the excitation of the cell. Further, the cold front lifting is more
intense in the presence of a large-scale orography than without.
4. DISCUSSIONS AND CONCLUSIONS
During the cold surge episode the leading edge differs from the
vigorous NE wind center in the southward march as shown in ALL
(see Fig.2). The former moves on average at about 51 m / s ar *d "the
latter at 31 and 20 before and after arriving at 20 N, respectively,
and close to zero when reaching 10 N. Subsequently, the NE wind
center moves westward along the south side of the Plateau* On the
other hand, the edge of the surge marches down to the south at approximately
51 m / s when no such orography is available. For this
reason, we think that cold air forced to move down along the east
side of the Plateau may excite two types of waves. One is the fasttravelling
gravity waves that bear no relation to the topography
and so is probably excited by the cold front itself. It goes southward
ahead of the front, thus causing the strengthening of the
northerly wind at the leading edge of the surge to be prior to temperature
drop. The other has characteristics of Kelvin waves, travels
at a slow pace, depends on the orography for its existence
and becomes non-existent apart from the Plateau. The second type
occurring at the periphery of the topography with the wind direction
nearly parallel to it is identified as the vigorous NE wind
region behind the front.
From the foregoing discussions the following points are summarized
j
(1) The Qinghai-Tibetan Plateau has no significant thermal effect
on the propagation of eastern-Asian cold surges;
(2) The Plateau has, however, very pronounced dynamic effect on
the surge. It steers cold air to move southward along the east side

of the orography, establishing a narrow belt of cold air.During the
passage of air, the exciting effect of Plateau's boundary brings
about in the rear of the cold front an extremely high wind center
that moves along the fringe of the topography, at first down to the
south along its east side till it crosses 2Q°N and then turns westward
on its south side to the Indian. Such waves are found of the
character of Kelvin waves;
(3) Our simulation with and without the orography indicates that
in the frontal edge of a cold surge the northerly wind enforcement
is roughly 12 hr ahead of temperature drop, a result that may be
due to the southward propagation of fast-travelling gravity waves
produced by the cold front excitation;
(4) The surge goes southward under the Hadley cell, operating to
intensify the cell and to move it further southward.
REFERENCES
1 Chang,C.P. and Lau, K.M., "Short-term planetary-scale interaction
over the tropics and midlatitudes during northern winter, Part
I:Contrast between active and inactive periods", Mon.Vea.Rev.,
110,938-946 (1982).
2 Chang, C»P. and Erickson, J.E., "Northeasterly cold surge and
near-equatorial disturbance over the winter MONEX area during
December 1974 f Part Ii Synoptic aspects and conclusions", Mon.
Wea.Rev., K)J,8l3-829 (1979).
3 Chang, C.P. and Lau, K. M», "Northeasterly cold surge and nearequatorial
disturbances over the winter MONEX area during December
1974* Part IIi planetary scale aspects", Mon.Wea.Rev.,
108,298-312 (1984).
4 Chang, C.P. and Millard, J.E. "Gravitational characters of cold
surge during winter MONEX", Mon.Wea.Rev., iii»293-300 (1983).
5 Lim, H. and Chang, C.P., "A theory for midlatitude forcing of
tropical motion during winter monsoon", J. Atmos. Sci., j8
2377-2392 (1981).
6 Murakami,T« and Nakamura, H., "Orographic effect on cold surge
and lee-cyclogenesis as revealed by a numerical experiment,
Part 1$ Transient aspects'*»J.Meteor.Soc.Japan,61,547-567(1983).
7 Sumi,A.,"A study on cold surges around the Tibetan Plateau by
using numerical models",J .Meteor.Soc.Japan,jS3,377-369(1985)*
8 Kuo,H.L. and Qian,Y.F. "Influence of the Tibetan Plateau on cumulative
and diurnal change of weather and climate in summer",
Mon. Wea. Rev., 109, 2337-2356 (l98l).

10
Fig.l. Movement of the surface cold high's center and cold
front during the 23-29 December 1982 cold surge episode
(0800 BST).
•W- 60
*2. Time-dependent near-surface flow field and isolaches,
a, b, c and d are based on 12, 24, 36 and 48 hr
integration, respectively* For notations in (d), see
text.

11
a
o 12 21 36 4$ 60 fZ
Pig.3. Near-surface mean V component in m/s (a) and
temperature in °C (b) between 110-120°B during
the integration.
20 V
Pig.-4. The 120°B meridional circulation at the 24 hr
(a) and 48 hr integration (b')j (c) is the difference
in vertical speed W,g - W^.* Speed is magnified by a
factor of 1,000*

13
Pig.7* Difference in mean vertical speed betveen ALL and
NM (W ALT - W NM ) over 110-120 O B at the 12 h (a) and 48 h
fac-
integration (b). Unit: m/s. Speed is expanded by a
tor of 1,000.

14
OVERVIB/V CF MEI-YU RESEARCH IN TAIWAN
George Tai-Jen Qien
Department of Atanospheric Science
National Taiwan University
Taipei, Taiwan.
ABSTRACT
Mei-Yu (or Baiu) is a unique regional weather and climate
phenomenon over East Asia and the western North Pacific. It occurs in
the period of late spring to early surmer when the circulation regime
over the area changes from the northeast monsoon in winter to the
southwest monsoon in summer. The mean position of this phenomenon
migrates northward with time. It occurs over southern China and the
Taiwan area in the period of mid-may to mid-June. The main purpose of
this paper is to review and look ahead the Mei-Yu research work that
have been done and would be done by Taiwan meteorologists. Research
work for the Mei-Yu over southern China and the Yangtz River Basin was
mostly not included.
The primary focus of this paper is the basic and applied
research on various features of Mei-Yu on different time and space
scales. Oily a very small portion of this paper is devoted to the
aspect of forecast research. However, the current forecast skill of
the Central V\feather Bureau in heavy rainfall was evaluated and
research work needed for improving the skill were suggested.
The first part of this paper discussed the existence and
importance of Nfei-Yu in Taiwan, \\brk on synoptic and climatological
aspects of the Ivfei-Yu system was then reviewed. Studies of interannual
variability of Mei-Yu were also included. Investigations of the
characteristics of mesoscale convective systems (KCS's) as well as the
environmental conditions and mesoscale triggering mechanisms for the
formation and evolution of NCS's were swmarized. Research of
mesoscale circulation systems in the .Mei-Yu season, such as the Mei-Yu
front, low-level jet (LLJ), mesolow and outflow boundary, and
topographical effect was also discussed. Finally, the field-phase of
the "Taiwan Area Mssoscale Experiment (T/MEX)" were presented.

l.INIRCDUCTICN
Climatological data show that the annual rainfall distribution in
central and eastern China (i.e. along the Yangtze River) possesses a
relative maximum during the period of mid-June to mid-July.
Continuous or intermittent precipitation mixed with frequent
rainshowers and thunderstorms is the characteristic feature in this
rainy season. A similar phenomenon is also observed in Japan and
southeastern China where the rainy season occurs about one half to
one month earlier than that of the Yangtze River Basin. This rainfall
maximum is called "Ivfei-Yu"(plum rain) in China and "Baiu" in Japan.
Anong all the meteorologists in Taiwan, Chi(45) was probably the first
one to study the Taiwan Mei-Yu phenomenon. Regarding the occurrence
time of the Mei-Yu season in Taiwan, results of various
studies(39,59,77) suggested that the Taiwan Mei-Yu period occurs from
mid-May to mid-June (Fig.l). It was also found that the Taiwan Mei-Yu
has a more pronounced interannual variation than that of the Yangtze
River Basin, Due to the topographic effect of the Central Mountain
Range, which oriented in a north-south direction along the island, the
Mei-Yu regime is better defined over the west side than the east side
of the mountain(Fig.2,17).
The positive and negative impacts of the Mei-Yu rainfall have long
been recognized by the meteorological ccntnunity and the general public
in Taiwan. It is a well known fact that winter is a dry season over
Taiwan except the northeastern portion. Therefore, if the spring
rainfall is inadequate then drought will be underway. The rainfall in
the Mei-Yu season provides a most effective way to alleviate the
drought. If the spring rainfall is normal but there is a shortage in
the Mei-Yu rainfall, a serious drought can also be expected. It is
clear that rainfall in the Mbi-Yu season is an important water
resource. On the other hand, the negative aspect of the Mei-Yu cannot
be overlooked. Continuous rain in the Mei-Yu season is a disaster to
agriculture. It is harmful to the rice crop that is ripe for
harvesting in the fvfei-Yu season and is harmful to other agricultural
crops as well. In recent years, the growth of the economy in Taiwan
was tremendous. Therefore, property damages caused by heavy rainfall
and the associated flash flooding in the Mei-Yu season becomes much
more serious than agricultural losses due to the continuous rain. Each
of the heavy rainfall/flash flood events, such as the May 28 case of
1981, June 3 case and June 10 case of 1984, caused U.S. $100-300
millions in damage.
Trie main purpose of this paper is to review the studies of the
Taiwan Mfei-Yu. Studies focussed on the Mei-Yu over mainland China were
not included. Also, research in developing forecast techniques as well
as reviews on operational forecasts were not considered. The first
part of this paper discussed the ciimatological aspects and
interannual variability of the Taiwan Mei-Yu. Studies on the
synopticscale circulation systems as well as heavy rainfall and the
related mesoscale convective systems (MISs) were next reviewed. Then
the Taiwan Area Nfesoscale Experiment(X%EX) was described. Finally,
15

16
Fig.l Climatological daily rainfall (rrm) at Taichung in 1956-
1975 and the monthly mean daily rainfall (rnn) in 1951-1970.
The3Vfei-Yu season is indicated (39).
Fig.2 The ratio(%)• of the Mei-Yu rainfall in May 15 - June 15
to the-total Tainial I. in .May-June in 1950-1980'

the forecast capability of the Central Weather Bureau(OVB) of R.O.C
for heavy rain was assessed. For reference purpose, observational
aspects of Msi-Yu have been reviewed in an earlier paper by Chen (20)
and a complete publication list as well as brief summaries of all the
Taiwan Me i-Yu studies can be found in Chen( 18,23,26).
2. O.IM\TUjOGICAL /*N\LYS!S AND SlUtf CF THE l^^B^^NlM, VARIABILITY CF
TAIWAN 3VEI YU
The specific date of the Mei-Yu period varies from year to year
and varies among different researchers using different data sets as
well as different definitions. f-bwever, the mean Mei-Yu period in the
climato logical sense occurs in the period of mid-May to mid- June in
Taiwan. On the other hand, the Mei~Yu period at one location might be
quite different from those at others due to the complexity of the
terrain in Taiwan as well as local influence. For example, Taipei
and Taichung have the same Msi-Yu period of May 18 - June 19. Whereas
the Mei-Yu period at Keeiung, Tainan and Kaohsiung is May 19 - June
16, May 20 - June 15 and May 19 - June 19, respectively (9). A
climatological analysis by Qien(16) showed that the Msi-Yu is most
pronounced and less stable (i.e. with greater interannual variation)
over central and southern Taiwan to the west of the Central Mountain
Range. It is less pronounced to the east of the Central Mountain
Range. The interannual variation of Mei-Yu is minimum over
northeastern Taiwan.
The total rainfall in the Mai -Yu period is closely related to
large-scale circulations, especially the position and intensity of the
monsoon trough and the Pacific subtropical ridge(25,45). As pointed
out by Hsu and Chi (59), wet and dry Mei-Yu seasons were correlated to
negative and positive anomalies at 500mb, respectively. A similar
relationship also existed in the pentad rainfall and 500mb anomalies
(47). \%ng et al. (78) observed that the decrease of zonal index over
the Asian sector in May and June coincided with the time of the Taiwan
Mei~Yu. A lower zonal index was found to be accompanied by a more
pronounced Mei-Yu. Yu et al. (83) studied the relationship between
the spring Nfei-Yu and typhooon rainfall. They found that the Mei-Yu
season terminated earlier whenever a typhoon attacked Taiwan before or
during the Mei~Yu season. If the Mei-Yu season covered a longer
period, the first typhoon that attacked Taiwan would come later. iNo
relationship seems to exist between the durations of the spring and
Mei-Yu rainfall. The relationship between the Taiwan Mel-Yu and ENSO
was studies by VAi(80) and Chiang (50). They showed that the Taiwan
Mei-Yu was negatively related to the southern oscillation index and
was more pronounced in the El Nino year. Wi and Fu (81) used
empirical orthogonal functions to represent the monthly rainfall and
applied various statistical methods to analyze the monthly rainfall
data of all the QVB stations. They found that topography, latitude,
and blocking effects of the Central Mountain Range were primary
factors in determining the Mei-Yu rainfall. The spatial distribution
of the mean rainfall appeared to be rather uniform over the northern
and southern parts of Taiwan. However, the temporal distribution
17

18
showed a marked contrast between the earlier and later stages of the
Mei-Yu period.
Chen and Chi (29) analyzed 25 cases of fyfei-Yu frontogenesis in
1968 - 1977 and found that Mei-Yu fronts that affected Taiwan tended
to form in the region of 20~35°N and 100-130°E(Fig. 3). The life
time of these fronts ranged from 3 days to 22 days with an average
value of 8 days. Chi and Chen (48) used the same data set to analyze
the spatial distribution of the surface fronts. The maximum frequency
appeared to extend from the western Pacific to the Bashi Channel and
the South China Sea. A secondary maximum extended southwestward from
the East China Sea to northern Taiwan and southern China. For the 850
mb front, the maximun frequency extended fran the East China Sea to
northern Taiwan and southern China and a secondary maximim was
observed over the Bashi Channel and the South China Sea. The speed of
movement of the fronts at both the surface and 850 mb levels tended to
decrease as the Mei-Yu progresses.
IOO°E I!0°E 120°E I30°E
-J40°N
20°N h
IIO°E
I20°E
Fig.3 Frequency distribution of the surface frontogenesis during
Taiwan Nfei-Yu season (May 15 - June 15) of 1968-1977. The heavy
solid line indicates the boundary of polar front and Mei-Yu
front formations (29).

A marked interannual variation existed in both the Mei-Yu
duration and rainfall amount (25,50,59,81). Studying the circulation
systems associated with this interannual variation became more
attractive to the Taiwan meteorologists in recent years. Chen (16)
studied the differences in the monthly mean circulations in May and
June at the surface and 500mb for wet and dry Mei-Yu seasons. It was
found that the wet Ivfei-Yu season was accompanied by a weaker Pacific
ridge, a well-defined IVfei-Yu trough over Taiwan and its vicinity, and
stronger low-level southwesterlies. The dry Vfei-Yu season was
associated with an unseasonably strong Pacific ridge, a well-defined
ridge over Taiwan and its vicinity, and low-level southeaster 1ies.
Chen (25) also studied the mean circulations for the wet and dry
months of May and June. It was shown that an equivalent barotropic
warm core blocking over the Okhotsk Sea and a warm core monsoonal
circulation over northern India developed above 850mb in June. The
primary factors in determining the monthly rainfall in May and June
appeared to be the origin and the intensity of the lower tropospheric
flows. The wet month was dominated by stronger sou thwes ter lies
originated from the Bay of Bengal. Wiile in the dry month, low-level
flows were dominated by either the southeasterlies/southerlies/
southwes ter lies of the western Pacific high or the continental
norhtwesterltes to the west of the East Asian main trough. The origin
and the intensity of the low-level flows over the Taiwan area were
determined collectively by the monsoon low, the western Pacific ridge,
the East Asian main trough and the Okhotsk blocking.
Chen and Jou (33) found that the major differences in the
large-scale circulations for wet and dry Msi-Yu seasons were the
midlatitude blocking as well as the position and the intensity of the
western Pacific ridge. In the wet season, there was a blocking over
the Okhotsk Sea or the eastern Siberia. The western Pacific ridge was
weaker or shifted southward from its normal position. Also, the lowlevel
southwesterlies originated from the Bay of Bengal were stronger
over the Taiwan area and its vicinity. In the dry season, there was
no midlatitude blocking. The western Pacific ridge was stronger and
extended westward to southern China. The southwesterlies or
souther lies/southeaster lies of the Pacific ridge in the lower
troposphere dominated over the Taiwan area and its vicinity. It was
also found that the upper-level divergent outflow over the monsoon low
area and to the south of the Mei-Yu front contributed significantly to
the upper branch of the Hadley circulation over East Asia. The
northward branch of the divergent outflow from the Msi-Yu area
suggested that an important relationship exists between the Mei-Yu
activities and the midlatitude circulations. Although the eastward
branch of the upper-level outflow from the monsoon low area subsided
over the Pacific ridge area, no clear relationship between the
intensities of the monsoon low and Pacific ridge was found. This
suggested that monsoonal circulation is not the only factor that
determines the intensity of the Pacific ridge.
19

20
3. SWDPTiC ^mLYSIS WD DIA^DSTIC STUDIES
The detailed case study of the Mei-Yu frontal system in June 10-
15, 1975 presented by Chen and Tsay (35) was probably the first paper
in Taiwan belonging to the category of diagnostic study. The synoptic
conditions and the budgets on moisture, kinetic energy and vorticity
were studied to reveal the structure and dynamics of the Mei-Yu
system. It was found that cumulus convection played an important role
in the maintenance of the intensity of the Mei-Yu front (in terms of
vorticity) in addition to the mid latitude baroclinic processes. The
transport of the heat and moist rue by the low level jet was found to
be essential in maintaining the cumulus convection (7,12,27,36). The
mean vertical motion field for this case showed that the vertical
advection and twisting terms cannot be neglected in the vorticity
equation over mountain slopes and baroclinic zone (8). Results also
showed that the major moisture source for the Mei-Yu area was the Bay
of Bengal. The temperature and vertical motion fields showed that a
thermally direct circulation with ascending warm air and descending
cold air prevailed over the Mei-Yu area. This secondary circulation
was particularly strong over the baroclinic region especially in the
vicinity of Japan.
Chen (10) observed that there was a continuous cloud band
accompanying the Mei-Yu frontal system which extended from the
vicinity of Japan southwestward to Taiwan and southern China. Along
the Mei-Yu frontal system, a marked wind shear line at 850 mb and 700
mb coincided with the maximum gradient of mixing ratio, relative
humidity and equivalent potential temperature. To the south of the
wind shear line, a 700 mb low level jet was located over the area of
maximum mixing ratio and high equivalent potential temperature. Over
and to the south of the wind shear line, strong cyclonic vorticity,
horizontal convergence and upward motion prevailed. The strong
convection in the southern portion of the cloud band occurred over the
area of maximum value of these kinematic parameters. The vorticity and
kinetic energy budget studies by Chen and Tsay (37) showed that an
area of maximum vorticity at 850 mb was located between the Mei-Yu
trough and the area of maximum horizontal convergence (or maximum
upward motion). The generation of cyclonic vorticity due to horizontal
convergence was counteracted by negative vorticity advections and thus
led to a quasi -stationary state of the Mei-Yu front at surface and 850
mb, Kinetic energy budget showed that a major part of the kinetic
energy generated by cross-contour processes tended to dissipate in
situ over the Mei-Yu area while only a very small portion (15%) was
transported to the environment. The mesoscale analysis of this case
(13,36) showed that the mesoscale circulation system with horizontal
dimensions of 200-300 km was characterized by cyclonic vorticity,
horizontal convergence and upward motion in the boundary layer. This
mesoscale circulation system tended to organize and enhance the
mesoscale convective systems.

Diagnosis using the balanced o> equation (36,74) showed that the
vertical motion in the mid latitude were primarily due to vertical
differential of the vorticity advection and Lapiacian of the
temperature advection. The upward motion to the south of the Mei-Yu
trough was primarily contributed by the convective latent heating. To
the south of Japan and the vicinity of Taiwan, upward motion was
contributed by vertical differental of the vorticity advection,
Lapiacian of the temperature advection and convective latent heating.
On the other hand, the boundary layer frictional effect and convective
latent heating were primarily responsible for the upward motion over
southern China and northern Vietnam. Chen and Chang (27) studied the
structure and dynamics over different sections of the Msi-Yu front.
The results indicated that the structure of the eastern and central
sections resembles a typical midlatitude baroclinic front with strong
vertical tilt toward an upper level cold core and a strong horizontal
temperature gradient, \\faereas the western section resembles a
semi tropical disturbance with an equivalent barotropic warm core
structure, a weak horizontal temperature gradient, and a rather strong
horizontal wind shear in the lower troposphere. The vorticity budget
showed that generation of cyclonic vorticity by horizontal convergence
was counteracted by cumulus damping in the eastern section and by
boundary layer friction in the mountainous western section.
Analysis of moisture structure and rainfall for the sane case by
Chen (7) showed that low level southwesterlies, especially the low
level jet, tend to creat potential instability on the warm side of
the Mei-Yu front. Continued large-scale ascent then led to the release
of potential instability through organized mesoscale convective
systems. The convection produced radar echoes and cells of heavy
rainfalls. The moisture budget (12) showed that convergence of
horizontal moisture fluxes together with the divergence of vertical
fluxes in the lower troposphere below 700 mb were responsible for the
maintenance of the IVfei-Yu cloud band. It was found that the subgrid
convective process predominates over the large-scale process in
providing moisture in the cumulonimbus area, with the ratio of the two
terms being 1.5, 8.5, 5, and 5 at 850, 700, 500, and 400 mb,
respectively.
A composite study was carried out by Chen (14,15) using 8 cases
of Nfei-Yu frontal systems which affected the Taiwan area in 1975 and
1977. Results showed that the cloud band and convective area coincided
with the area of upward motion indicating that the large scale
circulation controlled the cloud development. In the later stage of
the Ivfei-Yu season, the Tibetan thermal low intensified and the
migratory high behind the Mbi-Yu trough weakened. Also, the
southerlies to the south of the Mei-Yu trough intensified and the
frontal system retrograded northward. Meanwhile, the convective
activities and rainfall amount over Taiwan area increased. Chen
(19,22) analyzed the composite structure and dynamics over
different sections of the Mei-Yu front using these 8 cases. Results
were consistent to those obtained in the individual case study by Chen
21

22
and Chang (27). It was also found that the secondary circulation cell
to the south of the IVfei-Yu trough was probably related to convective
latent heating. The low level jet to the south of the western section
trough was perhaps due to the lower branch of this secondary
circulation through the eastward Cor iol is acceleration. Whereas the
low level jet over the eastern section trough appeared to be
originated upstream through an advective process. The barocl inicity of
the 850 mb Msi-Yu trough was primarily maintained by an interplay of
frontogenetic forcing of the stretching deformation counteracted by
the frontolytic effect of the thermally direct circulation. In the
middle and upper troposphere, convective latent heating appeared to be
important in maintaining the barocl inicity.
Chen (21) analyzed the N&A 4 and NCKA 5 satellite IR imageries
over the area of 40-180°E and 40°N - 40 °S during the Vbi-Yu period of
1975 and 1977 to reveal the possible connection between the Mei-Yu
frontal system and planetary scale circulation. Results showed that
formation of the Pvfei-Yu frontal system was closely related to the
intensification of the Indian southwest monsoon and the northeast
trade winds over the Pacific (or ITCZ). • The formation and
intensification of the Mei-Yu front over southern China {western
section) and the East China Sea (central section) were closely related
to the intensification of the planetary scale circulation such as the
Pacific ridge, ITCZ and the southwest monsoon. Whereas the eastern
section of the front over the Japan area is related to the
intensification of the Pacific ridge but not to the southwest monsoon.
Besides the diagnostic studies discussed above, many researchers
have carried out synoptic and cl imato logical studies on the
relationships between circulations and rainfall during the Mfei-Yu
season. These include studies of the characteristics of circulations
and rainfall (4, 5, 47), studies of the variation of rainfall amount
in the Mei-Yu season ( 76) and individual case studies of the Mei-Yu
front (66). These studies are very helpful to better understand the
relationships between the circulations and JVfei~Yu rainfall. The
evaluation study of NAP products, such as that presented by Chen et
al.(38), also appears to be valuable in the better understanding of
the physcial processes of the circulation systems and model
performance.
4. HEAVY RAIM^ALL /ND.VESOSQMZ (INVECTIVE SYOTMS
4.1 Heavy rainfall
Fig. 4 shows the spatial distribution of heavy rainfall events in
the Msi-Yu season of -.May; and. June. (42 ). The pattern reflects the
influences of the Central .Mountain Range and local topography.
Chen(24) analyzed the heavy rainfall events over northern Taiwan in
the Ms i-Yu 'period of 1965-1984 and found an average of 1.8 event per
year. Analyses of the max imun hourly rainfall at all stations in each
event suggested that the topographic effect was directly or indirectly
responsible for heavy rainfall. Hsu (58) analyzed heavy rainfall

23
Fig.4
Distribution of the 326 cases of heavy rainfall events in May-
June, 1975-1984 (42).
events at Taipei in 1907-1970 and found that 7

24
triggering mechanisms. In addition to these case studies, results of
synoptic climatological studies of heavy rainfall also appeared to be
valuable to better understand the necessary conditions of the heavy
rainfall (24, 57, 67, 69).
To better define a heavy rainfall event, Wi et al.(82) analyzed
the characteristics of rainfall and damage patterns. Results
suggested that the criteria of 20 mn/3h or 30 nrn/Gh better define a
heavy rainfall event than the criterion of 100 mn/d does. Qii(46)
reviewed all the papers dealing with the synoptic and
climatological aspects of heavy rainfall and proposed several
techniques relevant to heavy rainfall forecast. Additionally, Chu and
Jen(56) analyzed the relationship of the surface frontal position to
the heavy rainfall over northern and southern Taiwan. It was found
that the Mei-Yu front moves slowly from southern China toward the
Taiwan area during the time of heavy rainfall occurrence for all the
cases analyzed. In consistency with the result of Chen and Chi(28),
the probability of heavy rainfall was found to reach a maximum
whenever the surface front moves over northern Taiwan or its vicinity.
4.2 aiMKIOjOOf OF THE JvESQGCALE CDNVE3CTIVE SYSTEMS
There are different sizes of mesoscale convective systems(MISs)
embedded in the Msi-Yu frontal cloud band. Chen et al.(40) studied the
climatology of the ivCSs over western Pacific and southern China in the
Nfei-Yu season of 1981-1983 using CMS satellite cloud pictures. Results
showed that the duration of the Mei-Yu IvCSs is similar to that
observed over North Anerica in warm season. The mean durations of the
meso-oc and meso-cc NCS are 14.6 h and 14.1 h, respectively. These
are somewhat shorter than that obtained by Maddox(70) of 16.5 h over
the U.S.. It was also found that the duration of the MIS is positively
related to the horizontal scale and increases as the season proceeds.
The MZS tends to move southeastward over land and then recurves
eastward or northeastward offshore. The afternoon maximum of the MIS
formation over land was apparently due to solar heating. The early
morning maximum of the MIS intensification was suggested to be due to
the differential radiation effect between the cloudy and cloud-free
area.
4.3 BWIRCNN€NTAL (INDITICNS
Since heavy rainfall in the Msi-Yu season is produced by the M2S
embedded in the frontal cloud band, the environmental conditions are
similar for the heavy rainfall, and for the NCS (7, 36, 49, 51, 52,
54, 68 ). Tsay and Chen(74) pointed out that the large-scale upward
motion to the south of the Msi-Yu front over Japan, Taiwan and
southern China was generated by vertical differential of the
horizontal vorticlty advection, Laplacian of the temperature
advection, low-level fricttonal effect, and convective latent heating;
The areas of large-scale upward motion and horizontal moisture flux
convergence were the areas favorable for the development of MISs (12,
27, 61).

Wu et al.(82) analyzed 43 cases of heavy rainfall in March-June
and found that there was often a distinct cyclonic shear or even a
closed circulation at 850 mb over the area of heavy rainfall. For all
the cases analyzed, a well developed low was usually located over
Manchuria, Korea or Japan with a deep trough extending southwestward
to the East China Sea or the Yangtze River estuary. For heavy
rainfall events in different seasons, a conrnon feature at 850 mb
and/or 700 mb is the existence of the meso-oc scale trough which
probably provides sufficient dynamical lifting for triggering off the
development of MIS (52, 53, 56, 63, 64, 68). Chen and Wi (41) studied
the relationships between trough and heavy rainfall using 35 cases of
heavy rainfall events in May-June of 1965-1984. It was found that all
the heavy rainfall events was accompanied by a short wave trough or a
shear line at 850mb and 700mb. The heavy rainfall tends to occur
when the short wave trough moves toward Taiwan from southern China.
Chen and Tsay(5) identified two types of synoptic conditions for heavy
rainfall based on the existence of the Manchuria low. Chu and
Jen(56) identified two types of surface Mei-Yu front during heavy
rainfall event. One is the front approaching Taiwan from the north,
and the other is the quasi-stationary front in the vicinity of Taiwan
with a wave formation.
Another feature accompanying heavy rainfall event in Taiwan is the
850 mb cold tongue which usually extends from the East China Sea
southwestward to southern China . Chu and Jen(56) observed that the
cold tongue over the southeastern China coast usually moves toward
Taiwan prior to the occurrence of the heavy rainfall. This is
accompanied by pronounced warm advect ion over the Taiwan area 12 h
before the heavy rainfall occurrence.
Potentially unstable condition favorable for the development of
MIS usually existed over the area to the south of thelVfei-Yu front (6,
28). Chu and Jen(56) observed the existence of a stable layer in the
lower troposphere (950-900 mb) for some heavy rainfall cases and
pointed out that the total index is more reliable than the K index in
identifying the heavy rainfall condition. Synoptic studies for the
MISs over Taiwan and southern China suggested that there are at least
six conditions favorable for the development of MIS. They are (1) warm
advect ion in the lower troposphere, (2) low-level convergence over the
low pressure and/or the front, (3) low-level jet, (4) short wave
trough in the lower to middle troposphere, (5) the middle and upper
troposphere diffluent flow and/or speed divergence, and (6) potential
instability in the lower and middle troposphere.
4.4 Triggering Mechanisms for the Mssoscale Convective Systems
Observational studies suggested that seme mesoscale circulation
systems are apparently responsible for the formation of the MISs over
Taiwan and southern China. These include the frontal secondary
circulation, low level jet, quasi-stationary mesolow, topography,
outflow boundary and local circulations^, 7, 11, 20, 28, 30, .32, 43,
51, 52, 53). The mesoscale circulation system appears to be an
25

26
important mechanism for creating greater instability over a smaller
area and for providing stronger lifting necessary for the mesoscale
convections. Some results of the mesoscale studies will be reviewed in
the next section.
5. N€SO- oc SCALE CIROULATICN SYSTEMS
5.1 Mfei-Yu Front
There are about 4-5 frontal systems affecting Taiwan during the
Ivfei-Yu period each year (5, 28). Chen and Chi (28) found that
rainfall due to the frontal system for both northern and southern
Taiwan covers an area of 700 km wide across the front. An interesting
feature of the interactions between the front and the topography is
that the portion of the front on the east side of Taiwan usually move
faster than that on the west side(ll).
A case study by Chen and Tsay (37) revealed that the Mei-Yu front
was characterized by strong cyclonic vorticity, horizontal
convergence, upward motion and abundant moisture in the lower
troposphere. A mesoscale analysis for the same case by Chen and Tsay
(36) showed an observed frontogenesis rate of i.5-2.0 °C/100km/3h.
Results of observational studies (27,74) and numerical
simulations(55,61) indicated the importance of latent heating in
maintaining the frontal circulation. It was also noticed that the
Mei-Yu front not only provides unstable atmosphere favorable for the
convection (28) but also serves as a triggering mechanism for the
convection. Chen and Chi (28) observed that there were two secondary
circulation cells across the Mbi-Yu front. Convective activities were
enhanced over the area of ascending part of the cell and were
suppressed over the area of descending part.
5.2 Low Level Jet
One of the very interesting features accompanying a Mei-Yu front
is the existence of a low level jet located to the south of the 850
mb/700 mb trough. It was found that organized convections and heavy
rainfall are closely related to the low level jet
(3,7,28,43,60,64,68,73).
Ageostrophic characteristics of the low level jet over Japan area
(72) led to a hypothesis of downward momentum transport in the
formation of a low level jet (1,71). Over the Taiwan area, on the
other hand, a low level jet was usually observed prior to the major
convections (28,43). Also, the area of MIS formation was usually
determined by the location and intensity of a low level jet (7,34,51),
It seems reasonable that at least some of the low level jets over
Taiwan area were the cause rather than the effect of the convection.
Simulation results by Chou et al.(55) suggested that Coriolis
acceleration is a possible mechanism for the formation of a low level
jet. Over the Midwest of the Uhited States, different mechanisms
have been proposed to explain the formation of a low level jet

(2,75,79). Over southern China, geostrophic forcing and lee
cyclogenesis process have been suggested as possible mechanisms for
the formation of low level jets in spring season(34). It still needs
more evidence to conclude that if the above ment ioned mechanisms are
responsible for the formation of low level jets in the Ivfei-Yu season.
Chen and Yu (43) analyzed 35 cases of heavy rainfall events over
northern Taiwan and found a close relationship between the low level
jet and heavy rainfall. They also found that convections tend to
smooth out the vertical wind shear. The low level jet appeared to
weaken over the convective area or shift southward away from the
convective area.
5.3 Mesolow
The shallow lower tropospheric mesolows often developed over
southeastern and/or northwestern Taiwan when the Ivfei-Yu front passed
over Taiwan or its vicinity. These mesolows were closely related to
the rainfall over the Taiwan area (11,30,36). Climatological analysis
by Chen (11) showed that the rainfall and me so low over western Taiwan
were positively correlated and the converse was true over eastern
Taiwan. The mean life span of mesolows was about 12 h. They persisted
longer in the later stage of the Mei-Yu season. Chen and Qii (30)
observed a close relationship between the heavy rainfall and mesolow
over northwestern Taiwan. It was also suggested that the mesolow over
northwestern Taiwan probably served as a mechanism for producing heavy
rainfall through enhanced southwesterlies. A case study by Qien (13)
showed that mesolow formation and the accompanied wind changes over
southwestern Taiwan were closely related to the enhancement of
convective rainfall.
Hsu (58) observed that there was about 7CP/o of heavy rainfall
events in Taipei accompanied by a migratory mesolow on the nearby
Ivfei-Yu front. Chen (13) showed that this mesolow had a horizontal
scale of 200-300 km. The mesolow was characterized by strong cyclonic
vorticity, horizontal convergerce and boundary layer upward motion.
The NCSs over the Taiwan Strait were observed to be enhanced by this
mesolow circulation. Chen and V\ti (41) observed that 19 out of. 35 cases
of heavy rainfall events over northern Taiwan in 1965-1984 were
accompanied by a surface mesolow circulation over the rainfall area or
the irrmediate vicinity. Besides the surface mesolow, seme of these
heavy rainfall events (8 out of 35 cases) were associated with a wel1-
defined mesoscale cyclonic circulation at 850mb.
5.4 Outflow Boundaries
It is a well known fact that the downdraft of a convective system
can generate an outflow current which will trigger new convective
activities in some circumstances. However, the dynamic and
thermodynamic structures of the outflow boundaries as well as their
interactions with the environment are less understood. Over the Taiwan
area, some of the heavy rainfall events were probably due to the
convections generated by the outflow boundaries from the pre-existing
27

28
mature convections. Satellite pictures showed that the heavy rainfall
event over southeastern Taiwan in the morning hours of May 13,1983,
with hourly rainfall of 60.4 rnn and daily rainfall of 173*9 urn, was
generated by the interactions of outflow boundaries from two preexisting
IvCSs (31). Also,the heavy rainfall case of May 28,1981 over
northwestern Taiwan, which had an hourly rainfall of 88.7 rrm, was
perhaps attributable to the interactions between the Ivfei-Yu front and
a gust front from a mature PvCS, Chiou and Liu (52) suggested that the
heavy rainfall event of June 3, 1984 was probably initiated from an
arc cloud along a gust front. Therefore, it seems clear that we have
to better understand the role of the outflow boundary in order to
improve the heavy rainfall forecast in the Msi-Yu season.
6. TOPOGRAPHIC EFFECTS
Taiwan is an island with complex terrain. The Central Mountain
Range oriented in a nouth-south direction across the island with an
average height of 2000 m (the peak height is about 4000 m). This
topography is responsible for generating various types of local
circulations affecting the local weather. Also, it interacts with
atmospheric flows and circulation systems. For example, the Mbi-Yu
front deforms over Taiwan and mesoscale disturbance often forms
primarily due to the blocking of the topography. Chen (11) observed
that theMei-Yu front over the east side of Taiwan moved faster than
that over the west side due to the topographic effects. A similar
situation was also indicated in the long-term mean frontal speed in
the Mei-Yu season which showed a faster speed to the east of Taiwan
and to the east of the southeastern China coastal mountain ranges.
Besides, the Mbi-Yu front often breaks into eastern and western
sections by the Central Mountain Range. As also pointed out earlier a
mesolow often forms over northwestern or southeastern Taiwan primarily
due to the topographic effect (Fig.5, 11).
Climatological data showed that the monthly rainfall in June
increases from the Taiwan Strait eastward and reaches a maximum over
the Central Mountain Range then decreases over the eastern side of the
mountain. This is also true in the tvfei~Yu season. As shown in Fig.6
(11) a marked contrast in rainfall existed to the west and east sides
of the mountain indicating the importance of topographic effect. The
frequency of heavy rainfall events over northern Taiwan reached a
maximum in the night time hours (24) suggested that the topography
possibly affect the local rainfall through the induced diurnal local
circulations. In addition, the mesolow formation over western Taiwan
enhanced the low level southwesterlies and thus a stronger topographic
lifting and more active convections over the area. (30,36).
7. OVERVIEW OF T/MEX
This section briefly describes the project of Taiwan Area
Mesoscale Experiment (T/MEX), more detailed information can be found
in Kuo and Chen (62), The field phase of X%£X extended from May 1 to
June 29,1987 covering thirteen Intensive Observing Periods (IGPs) and
ten P-3 aircraft flight missions. The participants from the United

29
Fig.5 Distribution of the mesolow formation (0.5° XO.5° longitude/
latitude grid square) in May 15 - June 18, 1972-1977 (11).
25° N -
24°N -
23°N -
22°N -
II9°E 12Q°E 12I°E I22°E I23°E
Fig.6 The iVfei-Yu rainfall (solid,cm) in May 15 - June 18, 1972
- 1977 and smoothed topography (dashed,m) (11).

30
States included 70 scientists and technical experts fron 10
universities and 3 research institutes. There were more than 1000
scientists and technical personnels from 4 universities and 11
governmental agencies of the R.O.C. The observational program
consisted of five components: an upper-air network, a surface network,
a radar network, an aircraft program, and a satellite program.
The goal of T/MEX is to improve forecasting of the heavy rainfall
events that lead to flash floods. The primary objective of the field
phase was to collect the data necessary for the study of: 1) the
mesoscale circulation associated with the M3i-Yu front, 2) the
evolution of the mesoscale convective systems in the vicinity of the
Msi-Yu front, and 3) the effects of topograghy on the Msi-Yu front and
mesoscale convective systems. The field program of T/MEX was an
operational success and an excellent data set was available for
studying various scientific problems relevant to heavy rainfall
events.
8. ASSES3v€NT OF HEAW RAINFALL FORECAST IN TAIWsN
The official heavy rainfall forecast issued by the Central Weather
Bureau in Taiwan in 1977-1986 was assessed using a scoring system used
by the National Meteorological Center (r
0.27
0.28
0.32
0.30
0. 15

31
9. CCNCLIDIN3 REMARKS
The Mei-Yu research in Taiwan in the last 20-30 years is reviewed
in this paper. The current understanding of various aspects of the
Taiwan Ivfei-Yu is presented and a relatively complete reference list is
given. It is hoped that the present paper will serve as a guidance for
future research of the Taiwan Mei-Yu.
The data collected in the field phase of T/MEXwill be a great
help to the mesoscale research on heavy rainfall events, mesoscale
convective systems, and mesoscale circulation systems. The much more
active research relevant to heavy rainfall will ensure better
understanding of these events and thus leads to improvement in heavy
rainfall forecast. The paper on the Taiwan Msi-Yu research during the
last 20 years (1968-1987) can be categorized by the different forecast
periods (Fig.7). About one half of these papers (54 out of 106) was
related to the 1-2 days short range forecast category. It is expected
that more research related to the categories of very short range
forecast(3-18 h) and nowcasting (0-3 h) will be carried out in the
next 10 years. The papers published in different categories of
forecast periods in the next 10 years are estimated in Fig.7. It is
expected to increase gradually from the category of interannual
variability towards the category of shorter range and reaches a
maximum at the category of very short range forecast. Apparently, the
results on mesoscale research will have a tremendous effect on the
very short range forecast and nowcasting.
Fig.7
interannual seasonal extended medium short very
short casting
(>month) (5 day- (2-5 (1-2 (3-18 (0-3
month) day) day) hour) hour)
Distribution of the Nfei-Yu papers in different categories of
forecast periods in •• .1968-1.987.( solid) and the estimated
distribution in 1988-1997(dashed).
now

1-4
M-H
C CO *+-« C CD 4-9 C
o
••-4
O O 42 to
CO
4_>
'O •o 42 ti>
42
«
o 0
8 I— H

for the growth of nocturnal inversions", Bu 11. Aner. jvfeteor. Soc.,
38,283-290(1957).
3. Chen, C.K., "An analysis of the relationship between the low level
jet stream and the heavy rainfall during the Msi-Yu season in
Taiwan", Atmos. Sci., (J,^,29-37(1979)(in Chinese with English
abstract).
4. Chen, C.K., and GS. Liaw, "The circulation features for "dry" Msi-
Yu "» Vfeteor. Bull.. 27,2,1-14(1981) (in Chinese with English
abstract).
5. Chen, C.K., and C. Y. Tsay, %fei-Yu systems which affect northern
Taiwan", Atmos. Sci. ,7,49-58(1980) (in Chinese with English
abstract).
6. Chen, C.S., T.K. Chiou and S.T. Wang, "An investigation of
mesoscale convective systems associated with Msi-Yu front in SE
China fran May 26 to 28, 1985", Papers Meteor. Res., 9,2,137-161
(1986).
7. Chen, G.T.J. , "An analysis of moisture structure and rainfall for a
Mei-Yu regime in Taiwan", Proc. Nati. Sci. Counc., 1,11,1-21
(1977a). ~~
8. Chen, G,T. J. , "A synoptic case study on mean structure of Ivfei-Yu in
Taiwan", Atmos. Sci., ^, 38-47(1977b).
9. Chen, G.T. J., "An analysis of climatological references for
subjective probability weather forecasting in Taiwan", Tech. Rep.
No. Prob-Fore-001, Dept. of Atmos. Sci., Natl. Taiwan Univ. ,85 pp
(1977c) (in Chinese with English abstract).
10.Chen, G.T.J., "The structure of a subtropical Nfei-Yu system in
Southeast Asia", Sci. Rep., Dept. of Atmos. Sci. Natl. Taiwan Univ.
,2,9-23(1978a).
11.Chen, G.T.J., "On the meso-scale systems for the Mei-Yu regime
in Taiwan", Proc. Conf. Severe Weather in Taiwan Area, N9C and
Academia Sinica,150-157(1978b) (in Chinese with English abstract).
12.Chen, G.T.J., "On the moisture budget of a Ivfei-Yu system in
southeastern Asia", Proc. Natl. Sci. Counc., 3,1,24-32(1979a).
13.Chen, G.T.J. , "Mesoscale analysis for a Mei-Yu case over Taiwan",
Papers Meteor. Res., 2,63-74(1979b).
14.Chen, G.T.J. , "On composite structure of the 3Vfei-Yu system near
Taiwan", Tech. Rep. Ivfei-Yu-004, Dept. of Atmos. Sci., Natl. Taiwan
Univ.,106 pp(1981a) (in Chinese with English abstract)..
15.Chen, G.T.J., "A preliminary analysis on the mean structure of 8
cases of Ivfei-Yu system", Atmos. Sci., 8, 43-52(1981b) (in Chinese
with English abstract).
16-Qien, G.T.J., "On the abnormal h/fei-Yu phenomenon of 1975 and 1977",
Proc. Symp. on Abnormal Climate, C.W.B., 111-130(1981c) (in Chinese
with English abstract).
17.Chen, G,T.J., "Analysis of the IVfei-Yu system and its application in
the aviation weather forecast (I)", Dept. of Atmos. Sci., Natl.
Taiwan Univ., Tech. Rep. NRKIM-1983-08,73 pp(1983a) (in- Chinese
with English abstract).
18.Chen, G.T.J., "Feasibi 1 i't'y studies on the app.lication of
scientific results in atmospheric science research to
Treteorological forecast operation: Part I", Natl. Sci, Counc.,
Science and Technology of Disaster Prevention Program, Tech. Rep.
72-09,113 pp (1983b) (in Chinese with English abstract).
33

34
19.Chen, G.T. J., "A study on synoptic and dynanic aspects of Mei-Yu
system over southeastern China, Taiwan, and Japan", Dept. of Atmos.
Sci., Natl. Taiwan Univ., Tech. Rep. NTIKIM-1983-06,84 pp (1983c) (
in Chinese with English abstract).
20,Chen, G.T.J., "Observational aspects of the Msi-Yu phenomena in
subtropical China", J. jvfeteor. Soc. japan, 61,306-312(1983d).
21.Chen, G.T.J., "A diagnostic study of circulation system on
different scales and evaluation of NAP products during the Taiwan
Mei-Yu period", Etept. of Atmos. Sci., Natl. Taiwan Univ., Tech.
Rep. NRMM-1984-06,88 pp (1984a) (in Chinese with English
abstract).
22.Chen, G.T. J., "Structure over one longitudinal wavelength of a
composite Mei-Yu trough in subtropical East Asia", Atmos.jSci.,
jj.,121-139(1984b) (in Chinese with English abstract).
23.Chen, G.T, J., "Feasibility studies on the application of scientific
results in atmospheric science research to meteorological forecast
operation: Part II", Tech, Rep. 73-16, Natl. Sci. Counc., Sci.
and Tech. of Disaster Prevention Program,376 pp (1985a) (in Chinese
with English abstract).
24.Chen, G.T.J., "Feasibility study of "A Severe Regional
Precipitation Observation and Analysis Experiment" ". Natl. Sci.
Counc., Sci. and Tech. of Disaster Prevention Program, Tech. Rep.
73-42,32 pp (1985b) (in Chinese with English abstract).
25.Chen, G.T.J., "Characteristics of the mean circulation patterns for
the wet and dry Mei-Yu seasons in Taiwan", Atmos. Sci., 15,17-30
(1987) (in Chinese with English abstract).
26.Chen, G.T.J., "An overview of the Nfei-Yu research in Taiwan",
Natl. Sci. Counc. Mon., j_6,239-266(1988) (in Chinese with English
abstract).
27.Chen, G.T.J., and C.P. Chang, "The structure and vorticity budget
of an early surrmer monsoon trough (Mei-Yu) over southeastern China
and Japan", Man. Wea. Rev., 108,942-953(1980).
28.Chen, G.T.J., and S.S. Chi, "On the meso-scale structure of Mfei-Yu
front in Taiwan", Atmos. Sci., 5^,^, 35-47(1978) (in Chinese with
English abstract).
29.Chen, G.T.J., and S.S, Chi, "On the frequency and speed of Msi-Yu
front over southern China and the adjacent areas",
Papers Msteor. Res., 3,l&2,31-42(198Qa).
30,Chen, G.T.J., and S.S. Chi, "On the mesoscale rainfal 1 and meso-low
in the Nfei-Yu season in Taiwan", Atmos. Sci., 7_, 39-48 (198Qb) (in
Chinese with English abstract).
31.Chen, G.T.J,, and S.S. Chi, "Case study of disastrous heavy
rainfall in Mei-Yu season over northern Taiwan-28 May 1981 case",
Proceedings of the ROC-JAPAN! Joint Seminar on Multiple Hazards
Mitigation, Taipei, Taiwan, ROC, 815-839(1985),
32.Chen, G.T.J., S.Y. Chen and M.H. Yan, "The winter diurnal
circulation and its influence on precipitation over the coastal
area of northern Taiwan" Mbn. Wea. Rev., 111,2269-2274(1983).
33.Chen, G.T.J., and B.J.D. Jou, "Interannual variations of largescale
circulations over East Asia during the Taiwan Ivfei-Yu season"
Dept. of Atmos. Sci., Natl. Taiwan Univ,, Sci. Rep., NRKTM-1986-
05,213 pp (1986).

34.Qien, G.T. J., and C.P. Pu, "A case study of the formation of a low
level jet over subtropical China and heavy precipitation in
northern Taiwan", Atmos. Sci., _12, 23-32(1985) (in Chinese with
English abstract).
35.Chen, G.T.J., and C.Y. Tsay, "A detailed analysis of a case of Ivfei-
Yu system in the vicinity of Taiwan", Tech. Rep. N/fei-Yu-001, Dept.
of Atmos. Sci., Natl. Taiwan Univ. ,249 pp (1977).
36.Chen, G.T.J., and C.Y. Tsay, "An analysis of mesoscale systems,
observational errors and balance w's for a Msi-Yu case in
Taiwan", Tech. Rep. Msi-Yu-002, Dept. of Atmos. Sci., Natl. Taiwan
Univ.,40 pp (1978a) (in Chinese with English abstract).
37.Chen, G.T.J., and C.Y. Tsay, "A synoptic case study of Msi-Yu near
Taiwan", Papers Ivfeteor. Res., i,25-36(1978b).
38.Chen, G.T.J., Y.J. Wang and C.P. Chang, "Evaluation of the surface
prognoses of cyclones and anticyclones of the JM\ and FNX models
over East Asia and the Vastern Pacific during the 1983 Ivfei-Yu
season", Mon. Vfea. Rev., 115,235-250(1987).
39.Chen, G.T.J., and C.C. \Mi, "On the cl imatological characteristics
at five cities in Taiwan", Atmos. Sci., 5,2,1-16(1978) (in Chinese
with English abstract).
40.Chen, G.T.J., C.W. Wu and S.S. Chi, "Climatological aspects of the
mesoscale convective systems over subtropical China and the
western North Pacific during Jvfei-Yu season of 1981-1983",
Atmos. Sci., 13,33-45(1986) (in Chinese with English abstract).
41.Chen, G.T.J., and T.Y. Wi, "Pilot study of "A Severe Regional
Precipitaion Observation and Analysis Experiment"", Natl. Sci.
Counc., Sci. and Tech. of Disaster Prevention Program, 108 pp (
1985) (in Chinese with English abstact).
42.Chen, G.T.J., and J.S. Yang, "On the spatial and temporal patterns
of heavy rainfall in Taiwan Msi-Yu season", Atmos. Sci., 16,151-
162(1988) (in Chinese with English abstract).
43.Chen, G.T.J., and C.C. Yu, "Study of low-level jet and extremely
heavy rainfall over northern Taiwan in the Ivfei-Yu season",
Mon. Wea. Rev., 116,884-891(1988).
44.Chen, L.F., "A synoptic scale diagnostic study of a heavy rain
event in northern Taiwan of 1984", Ivfeteor. Bull., 32,4,29-60(1986)
(in Chinese with English abstract).
45.Chi, C.H., "Plum rains in Taiwan", Ivfeteor. Bull., \Q_,2,1-12(1964)
(in Chinese with English abstract).
46.GU, C.H., "A guide for forecasting the heavy rain over Taiwan
during the passage of Kfei-Yu front", Nfeteor. Bull., 33,J,, 1-14(1987)
(in Chinese with English abstract).
47.Chi, S.S., "A preliminary study on the mean circulation of Mei-Yu
season in Taiwan", Atmos. Sci., £,2, 17-32(1978) (in Chinese with
English abstract).
48...Chi,. S.S., and G.T.J. Chen, "An analysis on the frequency and
movement of front over southeastern China and adjacent area during
Taiwan N/fei-Yu season", Proc. 2nd Cbnf. Atmos. Sci., Natl. Sci.
Counc., Natl. Taiwan Univ., 67-77(1980) (in Chinese with English
abstract).
49.Chi, S.S., and G.T. J. Chen, "A diagnostic case study of the
environmental conditons associated with mesoscale convective
35

36
complexes: 27-28 May 1981 case", Atmos. Sci., _16,14-30(1988) (in
Chinese with English abstract).
SO.Chiang, S.H., "Climatic fluctuations of Taiwan's Mei-Yu (Plum-
Rain)", J. Eng. Environ., _8,55-68(1987).
Sl.ChioUj T.K., and S.Y. Liao, "A study of mesoscale convective system
in the southern China and its vicinity", Atmos. Sci., jj_,85-100 (
1984) (in Chinese with English abstract).
52.Chiou, T.K., and F.C. Liu, "A case study of heavy rainfall in
northern Taiwan on June 3, 1984", Atmos. Sci., _[2,93-102( 1985a) (in
Chinese with English abstract).
53.Chiou, T.K., and F.C. Liu, "A mesoscale analysis of heavy rainfall
on June 3, 1984 and the discussion of flash floods in northern
Taiwan", Atmos. Sci., 31,.2,1-14( 1985b) (in Chinese with English
abstract).
54.Chiou, T.K., S.T. Wang, Y.W. Lin and C.S. Chen, "A study of the
movement of mesoscale convective syston from southern China to
Taiwan area", Atmos. Sci., j^, 121-132(1986) (in Chinese with
English abstractT
SS.Chou, L.C., C.P. Chang and R.T. Williams, "A numerical simulation
of the Msi-Yu front and the associated low-level jet", (To be
published in Mon. Vtea. Rev.) (1989).
56.Chu, K.K., and L.Y. Jen, "Synoptic method for predicting the
occurrence of heavy rain in Msi-Yu season", Atmos. Sci., 14,17-
32(1986) (in Chinese with English abstract).
57.Chu, K.K,, K.Y. Liu, I.F. Chang and W.C. Yeh, "The causes of
special distribution of heavy rainfall in Taiwan",
vjaart. J. Meteor., 97,1-15(1983) (in Chinese with English
abstract).
58.Hsu, C.H., "A study of heavy rain in the Taipei area ",
Meteor. Bull., _H,3,,49-71 (1971) (in Chinese with English abstract).
59.Hsu, M.T., and S.S. Chi, "On the analysis of "Msi-Yu" in Taiwan",
Meteor. Bull., 20,4.,25-44(1974) (in Chinese with English abstract).
60.Hwang, J.C., "The influence of lew level jet on the weather of
Taiwan", Quart. J. Meteor., 68,37-45(1976) (in Chinese with English
abstract).
61.Kuo, Y.H., and R.A. Anthes, "Numerical simulation of a Ivfei-Yu
system over southeastern Asia", Papers Meteor. Res., j>, 15-36(1982).
62.Kuo, Y.H., and G.T. J. Chen, "The Taiwan Area Vfesoscale Experiment (
T/M^X): An overview", Bull. Aner. Meteor. Soc., (in press) (1989 ).
63.Lee, C.W., andG.T.J. Chen, "Case study of heavy rainfall of 19
november 1980 over northern Taiwan", Atmos. Sci., _K),25-38(1983) (
in Chinese with English abstract).
64.Lee, H.D., "A preliminary analysis of heavy rainfall case in Mei-Yu
seasons over Taiwan", Quart. J. Meteor., 91,11-21(1982) (in Chinese
with English abstract).
65.Liang, J.C., T.Y. Shyu and D.G. Pan, "The investigation on severe
precipitation during Taiwan Msi-Yu season", Atmos. Sci., 13,109-120
(1986) (in Chinese with English abstract).
66.Lin, S.C., and T.K.Chiou, "Objective scale separation technique and
its application on the mesoscale convective system diagnosis",
Papers Msteor. Res., 8.,.2,, 69-94(1985).
67.Liu, K.Y., "A study of'.the heavy "rainfall in the Ms i-Yu seasons of
Taiwan", vjaart.. J, Ivfeteor., 88, 7-12 (1981) (in Chinese with
Engl i sh abstract)V"

68.Liu, K.Y. , "On the heavy rainfall of May 28, 1981", Quart. J.
Msteor. 91,1-10(1982) (in Chinese with English abstract).
69.Liu,K.Y., and W.C. Yen, "On the heavy precipitation over Taiwan and
the proposed observational project", Quart. J. Nfeteor. 105,1-18 (
1985) (in Chinese with English abstract) .
VO.Maddox, R.A., "Mesoscale convective complexes", Bull. Aner.
Meteor. Soc., 61,1374-1387(1980).
71.Matsunoto, S., and K. Ninomiya, "On the role of convective momentum
exchange in the mesoscale gravity wave", J. Nfeteor. Soc. Japan, 47
,75-85(1969). ~
72.Matsunoto, S., K. Ninomiya and S. Yoshizumi, "Characteristic
features of "Baiu" front associated with heavy rainfall", J.
Msteor. Soc. Japan, 49,267-281(1971).
73.Tsay, C.Y,, and B.F. Chain, "A composite study of low level jet and
it's relationship with heavy rainfall in Taiwan area during JVfei-Yu
season", Atmos. Sci., j_5,]_, 1-16(1987) (in Chinese with English
abstract).
74.Tsay, C.Y., and G.T. J. Chen, "Dynamic processes for vertical motion
in aM3i-Yu system", Papers. Meteor. Res., 3>, 67-77 (1980).
75.Uccellini, L.W., and D.R. Johnson, "The coupling of upper and lower
tropospheric jet streaks and implication for the development of
severe convective storms", Mbn. Vfea. Rev., 107,682-703(1979).
76.Wang, G.C.Y., "Mesoscale weather analysis and forecasting during
the Plun-rain season in the Taiwan area", Atmos. Sci., 5_,j_, 15-25(
1978) (in Chinese with English abstract).
77.Wang, S.T., "On the "Plum Rain" in Taiwan", Oiart. J. Msteor., 44,
12-20(1970) (in Chinese with English abstract).
78.Wang, S.T., P.L. Yu and U.M Lee, "The characteristics of Asian
sector circulation index for the stumer half year", Quart. J.
Nfeteor., 71,30-33(1977) (in Chinese with English abstract^
79.Wexler, H., "A boundary layer interpretation of the low-level jet",
Tellus, Jl,369~378(1961).
SO.Wii, M.C., "On the interannual variability of the Taiwan Msi-Yu",
Dept. of Atmos. Sci., NatK Taiwan Univ., Tech. Rep. NTLKM-1987-
03,41 pp (1987) (in Chinese with English abstract).
81.Wi, MLC , and T.Y. Fu, "The interannual variability of Taiwan's Mei
-Yu", Atmos. Sci., 15,1,31-42(1987) (in Chinese with English
abstract).
82.Wu, T.Y., G.T.J. Chen, S.L. Shieh, F.L. Qiiao, CK. Chen, CK.
Hsiao and S.G. Chu, "An analysis of the local heavy rainfall and
the accompanied rice crop damage in the spring and early summer
over Taiwan", Atmos. Sci., 11,29-44(1984) (in Chinese with English
abstract).
83.Yu, CS., K.Y. Liu and F,C. Lee, "On the relationship among the
Qiuen-Yu Mei-Yu and typhoon precipitation in Taiwan", Quart. J.
Mbteor,, 110,21-38(1987) (in Chinese with English abstract).
37

38
On Temporal Variations of Low Level Jets Associated with
the Asian Summer Monsoon
LONG-NAN CHANG AND FANG-CHUAN Lu
INSTITUTE OF ATMOSPHERIC PHYSICS,
NATIONAL CENTRAL UNIVERSITY,
CHUNG-LI, CHINA 32054
ABSTRACT
ECMWF FGGE level Illb data are used to evaluate the temporal
variation of the low-level summer monsoon in general and
the Somali jet in particular to identify the important features responsible
for the maintenance and variation of these circulation
systems. Throughout the summer season, the Somali jet remains
as quasi-Ekman boundary layer flow so that cross isobaric generation
of KE is balanced by frictional dissipation. The kinetic
energy of the Somali jet, the pressure gradient and the Rossby
wave source in the southwesterly monsoon region all exhibit the
same feature of having a low frequency oscillation of a period
of 40-45 days, which is in close relation to the onset, active and
break periods of the Indian monsoon. Temporal variation of the
energetics of the western north Pacific region is also discussed.
1. INTRODUCTION
The summer Indian monsoon is driven by large- scale differential heating.The
Asian continent (include the maritime continent) with more insolation as the season
progresses gradually develops thermal contrast against nearby ocean areas.
However, the largest heat source of the monsoon is associated with condensation

39
.0
heating which follows the movement of the maximun rainfall belt ( Krishnamurti
; Krishnamurti and Surgi 2^ ). The migration of condensation heat source and the
distribution of the secondary sources of a topographic nature are in close relation
to the onset and the maintenance of the Indian monsoon.
The Somali jet is an integral part of the Indian summer monsoon. It flows
across the equator carrying southern hemisphere air northward up the African
coast into the Arabian sea Findlater 3^. The Somali jet is responsible for almost 50%
of the total cross-equatorial transport Rao 4 L Anderson^ considered the Somali
jet as a western boundary current of the East African mountain chain forced by
heat source and sink. Stout and Young 6 ] using MONEX satellite wind data of
900 mb level concluded that the monsoon flow is in quasi-geostrophic balance
(Ho < 0.3) at locations further than 7° from the equator,and behaves as a quasi-
Ekman boundary layer flow except in the Somali jet entrance up to 9° N where
significant imbalance (RQ > 0.4) is observed. It is observed that a sharp rise in
the conversion from zonal available potential energy to zonal kinetic energy occurs
around the onset of monsoon rains over India (Krishnamurti and Surgi 2^).
These foundings and many others indicate that the pressure distribution as a
result of heating and the mutual adjustment between mass and momentum fields
is of vital importance in the onset and maintenance of the summer Indian monsoon
circulation. The FGGE Illb data allow us to pursue this problem in further detail.
In this study, ECMWF FGGE level Illb data are used to evaluate the temporal
variation of the summer monsoon in general and the Somali jet in particular
to identify the important features responsible for the temporal variations of these
circulation systems. Temporal variations of the low-level circulation in the western
North Pacific is also examined for comparison purposes. The western Pacific
anticyclonic flow is linked to the Indian summer monsoon through an east-west
circulation induced by the Asian heat low and through low-level transport from
Indian monsoon region to the western Pacific region as can be seen from Fig.l.
Finally, temporal variation of Rossby wave sources in the Indian monsoon region
is discussed.
2. DATA AND ANALYSIS PROCEDURES
FGGE level Illb data are used throughout this study. Horizontal velocity
and geopotential height values at 850mb at 1200 UT are used to represent the

40
low-level flows of the Asian summer monsoon. Consecutive five day averages of
the aforementioned data starting from May 1 through the end of August are then
used for pressure gradient and kinetic energy budget calculations for four different
regions. The kinetic energy (KE) budget equation can be written as:
Q TV- r\
-£• =-V-(K V)-JL( W *O- V-V4>-D(K) (1)
where V = ui 4- vj is the horizontal component of wind, u = ^ represents the
vertical motion, K = |(u 2 4- v 2 ) ,and the geopotential height. The vertical
fiux divergence is evaluated using the 700mb and 1000mb winds. The dissipation
of kinetic energy D(K) is evaluated as a residual of the energy budget equation. In
order to examine the relative contributions by rotational and divergent winds,the
velocity field is further decomposed into rotational and divergent parts according
to an iterative method by Endlich T L Four regions are selected for this study to
represent different low level flow situations during the summer monsoon season.
Longitudinal and latitudinal boundaries of various regions are given in Table 1,
and their geographical locations are represented in Fig.2,
3. RESULTS AND DISCUSSION
3.1 Temporal Variation of Kinetic Energy
Temporal variations of the kinetic energy and KE budget terms associated
with the low-level flow (850mb) during the Asian summer monsoon are examined
using consecutive five-day average FGGE Illb data starting from May 1 through
the end of August 1979. The results are discussed in this section for various
regions.
3.1.1 Region Al
This region covers most area of active south-westerly Indian summer monsoon.
Somali jet is an integral part of this southwesterly flow. Temporal variations of
the kinetic energy content and KE budget terms are presented in Fig.3. Prior to
the onset of the Indian summer monsoon, the wind speed is generally very small
in this region. The wind speed (and thus the KE) increases rapidly during the
onset period and reaches the peak intensity arround June 25 to June 29. The KE
drops again and reach a minimum at arround July 15 to July 19. A secondary
peak appeared about August 4 to August 8. The KE then decrease gradually

41
to the pre-onset level by the end of August. Thus,the kinetic energy undergoes
a low-frequency oscillation of a period arround 40 to 45 days. Murakami 8^ had
qualitatively related this oscillation to the alternate active and break periods of the
summer monsoon . Cross-isobaric generation of KE dominate the energy budget
terms. The inertia! terms are relatively small and cross-isobaric generation is
balanced by frictional dissipation throughout the summer monsoon season. Thus
a quasi-Ekman low-level monsoon flow persists during the Indian summer monsoon
as is pointed out by Stout and Young 6 }. Temporal variations of the cross-isobaric
generation of KE resemble that of KE and have a similar low- frequency cycle.
To examine the Somali jet further, we did similar calculations for regions Bl
and B2 which represent areas with wind speeds greater than 15 m/s and 10 m/s
respectively (Fig.4 and 5). Despite more intense kinetic energy and larger crossisobaric
generation near the jet core , the pattern of temporal variation is similar
to that of region Al. Contribution from horizontal flux convergence is significant
during the onset period while signifcant horizontal flux divergence is observed
after the peaks of maximum KE. The contribution of horizontal flux convergence
by the divergent wind is in general small as compared to that of the rotational
parts (figure not shown).
3.1.2 Region A2
Region A2 covers the low-level southeasterly summer monsoon flow of the
southern hemisphere. Temporal variations of KE and energy bueget terms are
shown in Fig.6. It is of interest. to note that, two KE peaks correspond to two
active periods of summer monsoon are again evident in this region. Prior to the
onset of northern hemisphere monsoon ,the intensity of the flow in region A2 is
larger than that in region Al.
During the onset of the India monsoon, a large amoumt of cross-isobaric
generation of KE was present but the intensity of the flow in this region did not
increase as did its northern hemisphere counterpart until around July 5 due to a
strong horizontal flux divergence of the rotational winds. This imply a strong crossequatorial
flux to the northern hemisphere during this stage. This corresponds to
the strong inflow which was observed in region A1 during the onset period.
3.1.3 Regions Dl and D2
For comparison purposes, we have also computed the energy budget for low
level flows in the western North Pacific region ( Fig.7a). In region Dl, that is,

42
the northern part of the Pacific high circulation, the flow is south-westerly. The
kinetic energy content shows a trend of decreasing throughout the summer period
examined. The temporal variation of KE shows higher frequency oscillation as
compared to region Al in the lower latitudes. Yet two distinguishable KE peaks
coinciding with those of the active Indian summer monsoon are evident in this
region. The KE for region D2 (Fig.7b) also shows a decreasing trend, so is the
cross-isobaric generation . Which even become negative during July and August,
and the flow is maintained by the subgrid-scale motion. Since this region covers a
large area of the I.T.C.Z., this may be interpreted as the contribution from cumulus
scale motion in cloud clusters. From Fig.7, we can see that the inertia! terms are
relatively large and both horizontal and vertical flux divergence (convergence) are
significant in the North Pacific region.
3.2 Temporal Variation of the Geopotential Field Associated with Somali jet
The pressure gradient resulting from the superposition of land-sea contrast
along the eastern Arabian and African coast and the large-scale north-south differential
heating determine the location and the strength of Somali jet. The formation
of a thermal trough along the eastern African coast is important for the establishment
and subsequent maintenance of the Somali jet . This is evident from
examining the sequence of five-day average geopotential fields (figure not shown).
The thermal trough forms along the eastern African coast during the onset period
set up the pressure gradient in the Arabian sea area which serves as pipelines
pumping and guiding the cross- equatorial flow along the coast up into the Arabian
sea. The trough remains stationary until about August 24 to August 28. The
disappearance of this trough marks the end of the summer monsoon season.
On the other hand, the temporal evolution of the south westerly monsoon in
general, and the Somali jet in particular, is controlled by the pressure gradient set
up by large scale differential heating. It is obvious that the thermal low pressure
area of the diabatic heating of Tibet-Himalaya Massif experiences a periodic oscillation
of expansion and contraction which is coincident with the low-frequency
oscillation of the monsoon winds. If we shade the band between 1425 to 1440 gpm
isolines in 850mb geopotential height field, we observe that this band oscillates
north-south around the Arabian sea and India with close relation to the variation
of KE of the Somali jet. For easier visual examination of the geopotential
height field variations, we calculated the pressure gradient along 56.25°E, between

43
5.625 N and 15 N (region Bl) and 61.875 E between 0 TV and 18.75 N (region
B2), and plot each in time sequence of five day average (Fig.8). It is very obvious
from Fig.8 that the temporal variation of the pressure gradient is in close
resemblance to that of KE. This result impies that the adjustment between the
wind field and the mass field is very important in determining the variation of the
monsoon circulation,
Although the increase of pressure gradient during the onset period can be
explained by differential heating due to insolation and condensation heating as
the season progress, the reason for the decrease of pressure gradient prior to the
break of the monsoon is still unresolved.lt may be due to a reduced differential
heating by the shading of insolation caused by the development of cloud convection
as suggested by Krishnamurti and Bhalme 9^ or due to the distribution of ground
hydrology in response to the northward migration of the rainfall system ( Webster
and Chou 10^). However, the Indian monsoon region in summer is perhaps one of
the few regions around the world that both baroclinic and barotropic instabilities
(as well as inertia! instability) are easy to occur. It is possible that the release of the
combined baroclinic and barotropic instability is the reason for this low-frequency
oscillation. This problem deserves further investigation.
3.3 Temporal Variation of the Rossby Wave Sources
According to Sardeshmukh and Hoskins 11^,the response of the large scale rotational
flow to diabatic heating can be written as:
(|+V-V)C = S'+f (2)
where V$ is the rotational wind associated with ( ,and ( is the absolute vorticity.
The source of Rossby waves is defined as:
V x is the divergent wind and D = V • V x while F is frictional dissipation. Fig.9
depicts the temporal variation of the Rossby wave source in a region bounded by
45°JEJ and 106.875° E and 5.625° N and 26. 25° AT . The source of the Rossby wave
shows a clear low-frequency cycle.
In- comparison to the pressure gradient variation of the Somali jet core region
(Fig.9a), we can see that the peaks of Rossby waves occur within five days immediately
after the peaks of pressure gradient and KE of the jet. Similar low-frequency

44
cycle is observed for the amplitude variation of both Rossby and Kelvin waves in
tropical region by Lee and Paegle 12^.
4. CONCLUSION
Temporal variations of the kinetic energy budget associated with low-level
flows of four different regions during the summer Indian monsoon of 1979 have
been examined. On the average, the Indian summer monsoon is characterized by
having an Ekman type boundary layer flow, with cross-isobaric generation of KE
balanced by frictional dissipation, while the low-levelflowin the North-west Pacific
Ocean have a more complicated behaviour,where both horizontal and vertical flux
divergence (convergence) of KE are important. And in the easterly flow region
where part of I.T.C.Z. is included, significant amount of subgrid-scale generation
of KE is required for KE balance.
Temporal variations of the kinetic energy of the Indian summer monsoon exhibit
a low-frequency oscillation which is closely related to the onset, active and
break periods of the Indian summer monsoon. The low-frequency oscillation of
low- level winds is found to be in close relation with the variation of pressure
gradient established by differential heating. The low-frequency oscillation of pressure
gradient is in turn caused by a sequence of expansions and contractions of
the thermal low located in the Asian continent.In a region of strong horizontal
and vertical wind shear, it is hypothesized that the release of combined bareclinic
and barotropic instablities might be the reason for this low frequency oscillation.
The resultant source of Rossby waves also possesses a similar low- frequency cycle.
The effect of the variation of the amplitude and the propagation of forced Rossby
waves might induce a similar low frequency response of the rotation flow in remote
regions.
Acknowledgments. We wish to thank Mr. Kung-Long Lin for his help in the
preperation of this manuscripts, Subtropical Meteorology Data Center of National
Science Council for kindly supplying of the FGGE data and Micro-Computer center
of the Department of Atmospheric Physic, NCU, for the use of computing
facilities. This work was supported by the National Science Council under Grant
NSC 78 - 0202 - MOOS-03.

45
REFERENCES
^Krishnamurti, T. N.,"Summer monsoon experiment - a review." Mon. Wea.
Rev., 113, 1590-1626 (1985).
2^Krishnamurti, T. N. and Surgi, N., "Observational aspects of summer monsoon.
In Monsoon Meteorology." Chang, C.-P. and T.N. Krishnamirti ed. Oxford
Univ. press. 3 - 25 (1987).
3 JFindlater, J., "A major low-level air current near the Indian Ocean during the
northern summer." Quart. J. Roy. Meteor. Soc., 25, 362-380 (1969).
4 lRao, Y. P., "Inter-hemispheric circulation." Quart. J. Roy. Meteor. Soc., 90,
191-194 (1964).
5^Anderson, D. L. T., "The low-level jet as a western boundary current." Mon.
Wea. Rev., 104, 907-921 (1976).
6^ Stout, J. E. and Young, J. A.,"Low-level monsoon dynamics derived from satellite
winds." Mon. Wea. Rev., Ill, 774-798 (1983).
7^Endlich, R. M., "An iterative method for altering the kinematic properties of
wind field." J. AppL Meteor., 6, 837-844 (1967).
^Murakami, T., "Monsoon (In Japanese) Tokyo press." 198 pp (1986).
9^Krishnamurti, T. N. and Bhalme, H. N., "Oscillations of a monsoon system.
Part I: Observational aspects." J. Atmos. ScL, 23, 1937-1953 (1976).
10 1 Webster, P. J. and Chou, L. C.,"Low frequency transitions of a simple monsoon
system." J. Atmos. ScL, 21, 368-382 (1980),
n^Sardeshmukh, P. D. and Hoskins, B. J.,"The generation of global rotational
flow by steady idealized tropical divergence." J. Atmos. ScL, 45. 1228-1250
(1988).
12^Lee, B.C. and Paegle, J. N., "Observed Tropical three-dimensional normal mode
energy spectra and low frequency oscillations based on the weekly averaged
gridded data during FGGE summer year." Papers Meteo. Res., 11, 148-157
(1988).

46
SUMMER (6-85 : 12 2 M979 P:850 MB
MEAN STREAMLINE
A
SOMALI JET [850MB] [V>15M/S3
Fig.l. Seasonal mean streamline at 850mb level for
FGGE year (1979) over (30E-180E,30S-i5N)
region.
PENTAD AFTER 1 MAY 1979
Fig.4. As in Fig.3. except for low level Somali Jet core.
(Bl) (Where mean wind speed is greater than
15M/S.)
r §
_ g
Fig. 2, Schematic geographical location of the six regions
for analysis.
Table 1. Domains for various region of analysis.
REGIOH
Al
A2
Bl
B2
01
D2
DOMAIN
38.375* E-105* E
0* H-20 .625* H
39.375* E-105* E
20.625* S-Q* H
45' E-65 .625* E
5.625* K-15* H
41.25* E-82.5* E
0* N-18.75* H
129.375' E-180* E
26,25* N-45* R
129.375' E-180* E
9.375* H -26.25* H
l 'i i i i
19 21 23
PENTAD AFTER 1 MAY 1979
Fig.5. As in Fig.3. except for low level Somali Jet core
(B2). (Where mean wind speed is greater than
10M/S.)
i i i i i i i i i i i i i i i
"I I I I I I I I I I I 1 I I I I I I I
5 7 9 11 13 16 17 19 21 23
PENTAD AFTER 1 MAY 1979
PENTAD AFTER I MAY 1979
Fig.6. As in Fig.3. except for region A2.
Fig,3. Temporal variation of KE content and KE budget
terms for region Al. 1-kinetic energy. 2:rate
of change of KE. 3:horizontal flux divergence of
KE. 4:vertical flux divergence of KE. 5:crossisobaric
generation of KE 6;dissipation.

47
s s :
SUBTROP. H. CIR. C850M81
] B
SOMALI JET [850MB3 [V>10M/S3
r"T r I i i 'i r T i I I""
7 9 11 13 IS 17 1
PENTAD AFTER 1 MAY 1S79
I 3 5 7 9 11 13 15 1 19 21 23
PENTAD AFTER 1 MAY 1979
Fig.8(b). Temporal variation of pressure gradient along
61.875E for Somali Jet core region.(O.N-18.75N
, Where mean wind speed is greater than 10 M/S.)
MONSOON CIRCULATION [200MB3
PENTAD AFTER 1 MAY 1979
Fig.7. As in Fig.3. except for Western North Pacific
regions Dl and D2.
fa,-
s;:
i i i i i i i i i i i i i i i i
V 1 3 5 7 S 11 13 IS 17 19 21 23
PENTAD AFTER 1 MAY 1979
Fig.9(a),Temporal variation of Rossby wave source for
upper (200mb) level region (45E-106.875E, 5.625N-
26.25N). l:source of Rossby wave, Rossby wave
source due to advection of absolute vorticity by
divergent wind (line 2.) and horizontal divergence.(line
3.)
3-
MONSOON CIRCULATION C850MB]
A
/ i
/ i
' 1 3 5 7 S 11 13 IS .17 19 21
PENTAD AFTER 1 MAY i979
Fig.8(&). Temporal variation of pressure gradient along
56.25E for Somali Jet core region. (5.625N-
15.N , Where mean wind speed is greater than
15M/S.) ~T~\ I t 1 r""T"T~ i i r rr-r r i
PENTAD AFTER 1 MAY 1979
Fig.9(b).As in (a) except for low level.(850mb)

Observed Structure and Propagation Characteristics of
Summertime Synoptic-scale Disturbances over the Tropical Western Pacific
Alexis Kai-Hon Lau
Atmospheric and Oceanic Sciences Program,
Princeton University, Princeton, New Jersey, USA
Ngar-Cheung Lau
Geophysical Fluid Dynamics Laboratory/NOAA,
Princeton University, Princeton, New Jersey, USA
ABSTRACT The observed properties of tropospheric transient disturbances
occurring over the tropical western Pacific are studied using global gridded analyses
for the 1980-1987 period. A region of enhanced transient activity is noted
along a zone extending from the equatorial western Pacific to the South China
coast The same region is also seen to be the preferred site for the propagation of
wavelike circulation features. Patterns of lagged correlation charts based on timefiltered
data indicate that the prevalent disturbances in this active zone have typical
wavelengths of 2500 km and periods of 6 days. These perturbations are elongated in
the southwest-to-northeast direction. Their characteristic shape and position
relative to the quasi-stationary flow field are suggestive of barotropic energy
conversion from the time mean circulation to the transient eddies. Computations
based on the lagged correlation charts indicate a general tendency for the disturbances
to propagate in a west-northwestward direction, with an average phase speed
of approximately 5ms" 1 . The temporal coherence of the perturbations is strongest
over the maritime areas, and weakens noticeably over land. Regression analysis
reveals that the disturbances over the western Pacific and South China Sea are
typically growing with time, whereas they dissipate rapidly after crossing the
coastline. Vertical cross-sections of different meteorological parameters, as
constructed using a composite technique, indicate that the passage of these
disturbances is accompanied by well-defined changes in vorticity, vertical motion,
temperature and humidity at various tropospheric levels. The phase relationships
among the fluctuations in these variables may be interpreted in terms of the
dynamical and physical processes operating within the disturbances.
1. INTRODUCTION
Tropical tropospheric disturbances with periods less than 2 weeks are known to be active in many
different geographical locations during the northern summer. As a result of the increased availability of
upper level observations in the late 1960s, these disturbances have been studied quite extensively using
synoptic, spectral, and composite methods [e.g., Carlson 1 ), and Reed and Recker 2 ) ]. Using observations

49
taken during the Global Atmospheric Research Program (GARP) Atlantic Tropical Experiment
(GATE) in 1974, Reed et. al. 3 ), and others have substantially improved our knowledge of the easterly
waves originating from central and western Africa. Much less is known about tropical transient phenomena
occurring elsewhere, especially in maritime regions where observations are scarce. In view of the
large differences in the ambient atmospheric flow pattern in various longitudinal sectors, the structure
of the African easterly waves is not expected to prevail throughout the tropical zone. Many studies [e.g.
Nitta et. al. 4 ) ] have already noted that atmospheric fluctuations at various tropical sites exhibit
different spectral characteristics. For instance, the dominant period of the African easterly waves (3-5
days) is known to be shorter than that found over the oceanic areas (5-7 days).
The availability of routinely produced global analyses after the First GARP Global Experiment
(FGGE) has revived the interest of the meteorological community in the behavior of tropical disturbances
at different locations. For example, Nitta et. al. 4 ) and Nitta and Takayabu 5 ) have conducted a
global survey of such phenomena using the FGGE dataset. These authors have identified several regions
of strong transient activities, and have described the characteristic spectra and vertical structure of fluctuations
in these regions. The present study has been launched with a similar goal in mind. By
diagnosing the global gridded analyses for an 8-year period, we attempt to document the three-dimensional
structure and propagation characteristics of summertime transient disturbances in different
tropical regions. In this particular report, the attention is focused on results pertaining to the tropical
western Pacific.
2. DATASETS
The primary dataset for this study consists of twice daily operational analyses archived at the
European Centre for Medium-Range Weather Forecasts (ECMWF) for the period 1980-1987. The data
grid has a uniform horizontal resolution of 2.5° by 2.5°. The variables examined include zonal and
meridional wind, pressure velocity, temperature, geopotential height and specific humidity at 7 standard
pressure levels (100,200,300,500,700,850 and 1000 mb).
Twice daily gridded data for the outgoing longwave radiation produced by the U.S. National
Oceanic and Atmospheric Administration for the period 1974-1987 have also been used to investigate the
relationships between tropical disturbances and convective activities. As an independent check of the
results derived from the ECMWF dataset, some of the diagnostics have been applied to a set of gridded
wind analyses compiled by the Royal Observatory of Hong Kong (ROHK) for the period 1981-1984.
3. TIME SCALE, HORIZONTAL STRUCTURE AND PATH OF MIGRATION
Following the earlier studies of Nitta and Takayabu 5 ) and others, we have chosen the relative
vorticity £ at 850 mb as the key meteorological parameter for identifying tropical disturbances. Power
spectra of £ at 850 mb have been computed at a large number of grid points scattered throughout the
tropical zone. These results (not shown) indicate considerable variations in the dominant time scales of
the fluctuations at different locations. Such time scales range from 7-8 days in northeastern India to 5-8
days in the western Pacific, and 3-4 days in western Africa. In view of the diverse temporal
characteristics in various regions, a 61-point digital filter has been designed to retain fluctuations
residing within a broad spectral band corresponding to the 3-10 day period range.
The summertime distribution of the root-mean-squares (rms) of 850 mb £ over the western
Pacific (not shown) is characterized by maximum values along a zone extending west-northwestward
from about 5°N 165°E towards the East and South China Seas. This region of enhanced transient activity
lies in the vicinity of the monsoon trough, and is displaced to the south/southwest of the mean

50
southeasterly current in the western Pacific.
The typical horizontal structure of the disturbances occurring in the region of maximum
variability in £ has been studied using lag-correlation maps with base points located in the active zone.
An example of such correlation charts is presented in Fig. 1, which shows the distributions of lagcorrelation
coefficients between the bandpass filtered fluctuations in £ at the base point 15°N 132.5°E
and the corresponding fluctuations at all other grid points. The temporal lags used in this example range
from -2 to +2 days. Here negative (positive) lags refer to computations with the time series at
individual grid points leading (lagging) the time series at the base point. A similar technique has been
used by Wallace et. al. 6 ) to illustrate the behavior of midlatitude baroclinic disturbances. These authors
have referred to the patterns in Fig. 1 as 'one-point correlation maps'. Each panel in Fig. 1 is indicative of
a wavelike feature with extrema of alternating polarities. Comparison between the patterns in different
panels reveals a systematic west-northwestward phase propagation.
The average wavelength of the disturbances depicted in Fig. 1, as estimated from the distance
between adjacent correlation centers with the same polarity, is approximately 2500 km. The typical
period, as given by twice the time interval between correlation patterns which are 180° out of phase with
respect to each other (e.g., see panels corresponding to lags of -2 and +1 days), is about 6 days. These
estimates yield a mean phase speed of 4.8 ms"*. A characteristic southwest-northeast tilt of the elongated
extrema is discernible in the patterns of Fig. 1. The placement of such tilted circulation features
to the south of a southeasterly current would imply a downgradient (southward) transport of easterly
momentum, and a barotropic conversion from mean to eddy kinetic energy.
The wavelike nature of the disturbances over the western Pacific may be further illustrated by a
teleconnectivity analysis similar to that performed by Wallace et. al. 6 ) The 'teleconnectivity' for any
point P is defined as the absolute value of the most negative correlation coefficient appearing in a simultaneous
one-point correlation map corresponding to base point P. A large teleconnectivity value for
point P implies that fluctuations at P are accompanied by strong fluctuations of the opposite polarity at
other locations, and is hence indicative of the prevalence of wavelike phenomena in the vicinity of P. Fig.
2 shows the teleconnectivity pattern based on bandpass filtered 850 mb £ data. A region of strong teleconnectivity
(larger than 0.4) is seen to extend west-northwestward from 5°N 16*0°E to South China.
The location of this maximum coincides with the region of enhanced rms of C mentioned earlier, and
with the propagation path of the wavetrain shown in Fig. 1. This suggests that, in addition to monopolar
vortex-like systems such as typhoons and monsoon depressions, a substantial fraction of the synoptic
scale variability in the tropical western Pacific is attributable to propagating, wavy disturbances.
4. PROPAGATION CHARACTERISTICS
By applying the analysis procedures outlined in Wallace et. al. 6 ), the phase velocity vectors and
the temporal coherence of the migratory signal have been computed from one-point correlation maps.
The distributions of these two quantities are displayed in Fig. 3 using arrows and stippling, respectively.
For a series of lag-correlation charts corresponding to the common base point P, the strongest positive
correlation centers appearing in the panels for lags of -2 and +2 days have been identified and labeled as A
and D, respectively (see example in Fig. 1). The line segments AP and PD represent the average spatial
displacement of the disturbance center 2 days before and 2 days after its passage over P, respectively. The
direction of the phase velocity vector at P may hence be inferred from the orientation of the line segment
AD, whereas the magnitude of the phase speed may be approximated by dividing the distance between A
and D by the elapsed time interval (i.e., 4 days). Furthermore, the temporal coherence of the migratory
signal for base point P is defined as the average of the lagged correlation coefficients at points A and D.

51
105°£ ISO'E 165°E
30°N
15 8 N
Equator
Equator
Figure 1. Distributions of
lagged temporal correlation
coefficients between bandpass
filtered fluctuations of
850 mb relative vorticity at
the base point (15°N
132.5°E) and the corresponding
fluctuations at individual
grid points. The time
lag relative to the time series
at the base point ranges from
-2 to +2 days, and is indicated
at the upper right hand corner
of the appropriate panel. The
position of the base point is
indicated in each panel by a
solid dot. Contour interval is
0.1. Values less than -0.1 are
indicated by stippling. The
zero contour has been omitted
for sake of clarify.
Extrema are expressed in
percent.
105°E 120°E 135°E 150°E 165°E

40°N
30°N
15°N
Equator
90°E 105°E 120°E 135°E 150°E 165°E 170°E
Figure 2, Teleconnectivity map of the bandpass filtered 850 mb relative vorticity field. Contour interval is 0.1. Only
values equal to or greater than 0.3 are contoured. The heaviest stippling corresponds to teleconnectivity values larger than
0.5. See text for delails of analysis procedure. Extrema and contours are expressed in percent

Scale: 8 °ld
40°N
30°N
15°N
Equator
105°E 120°E 135°E 150°E 165°E 170°E
Figure 3. Distribution of phase velocity vectors for disturbances in the bandpass filtered 850 mb relative vorticity field
(see arrows and scale). The stippling in this chart indicates those grid points which exhibit strong positive teleconnections at
lags of -2 and +2 days. Light stippling denotes mean lagged-correlation values of 0.3-0.4. Dense stippling denotes correlation
values exceeding 0.4. Phase velocity vectors are plotted only at grid points with mean lagged-correlation values larger than
0.3, See text for details of analysis procedures.

54
The pattern of arrows in Fig. 3 clearly portray the general west-northwestward propagation of
transient disturbances over the western Pacific. The average phase speed of about 5 m s" 1 is consistent
with the estimate based on the characteristic wavelength and period deduced from Fig. 1. The temporal
coherence of the migratory signal (see stippling in Fig. 3) generally exceeds 0.3 over the maritime areas,
and attains maximum values along a belt extending from the equatorial Pacific to the South China
coasts. The latter region is characterized by strong teleconnectivity (see Fig. 2) and enhanced variability
in £ (see Section 3). Very weak temporal coherence is seen over the Asian land mass.
The magnitude and spatial distribution of typical growth/decay rates of the disturbances have
been estimated using the algorithm developed by Wallace et al. 6 > The results, shown in Fig. 4, are again
based on one-point correlation maps. For a series of such maps corresponding to base point P, the
strongest positive correlation centers at lags of -1 and +1 day have been noted and labeled as B and C, respectively
(see example in Fig. 1). The regression coefficients between the normalized time series of
filtered C at P and the unnormalized time series of the same quantity at B and C were then computed at
time lags of -1 and +1 day, respectively. These regression values provide for a measure of the amplitudes
of the vorticity fluctuations at B and C one day before and one day after, respectively, the passage over P
of a disturbance with an amplitude of one standard deviation. The growth/decay rate at P was then computed
by subtracting the regression coefficient for point B at -1 day lag from the corresponding value for
point C at +1 day lag, and then dividing the result by the appropriate time interval (i.e., 2 days).
The general pattern in Fig. 4 is characterized by growth over the Philippine Sea, the Philippine
Islands and the South China Sea. The steepest decay rates are found along the South China coast. This
result is consistent with the land-sea contrast in the temporal coherence of the migratory signal shown
in Fig. 3. The large growth rates over the Bay of Bengal are associated with the developing monsoon
disturbances in that region.
5. VERTICAL STRUCTURE
The time-space behavior of the 850 rnb £ field has also been diagnosed using rotated extended
empirical orthogonal functions (REEOF). The horizontal structure of the leading pair of REEOFs (not
shown) is very similar to that displayed in Fig. 1. The temporal coefficients of these REEOFs provide
for a logical basis for constructing composites of the three-dimensional structure of the tropical disturbances
of interest. In Fig. 5 are shown the vertical cross-sections of perturbations in (a) relative vorticity,
(b) vertical pressure velocity, (c) temperature, and (d) specific humidity, as obtained from such a
composite procedure. The abscissas of these plots extend southeastward from South China (27.5°N
110°E) in the extreme left to the western equatorial Pacific (2.5°N 142.5°E) in the extreme right.
The pattern for £ [panel (a)] indicates a strong northwestward tilt with increasing height over
the equatorial ocean, where initial wave growth occurs (see Fig. 4). The vertical tilt is seen to decrease
gradually as we scan the panel from right to left, so that there is almost no vertical tilt as the disturbances
approach the South China coast. Perturbations in pressure velocity [panel (b)3 attain extrema at 300
and 850 mb. Comparison between panels (a) and (b) indicates the occurrence of maximum upward
motion at 850 mb (and hence near-surface convergence) at the trough. At the 200-300 mb level, the
upward motion tends to lag the positive £ fluctuations by a quarter wavelength. This phase relationship
may be attributed to the confinement of £ fluctuations in the middle and lower troposphere, and differential
relative vorticity advection by the prevalent southeasterly current in this region. The most notable
features in the cross-section for temperature [panel (c)] are the extrema at 300 mb. These upper tropospheric
warm and cold centers are located directly above the 850 mb trough and ridge, respectively.
The warm center may be associated with condensational heating accompanying the low level conver-

40°N
30°N
15°N
Equator
90°E 105°E 120°E 135°E 150°E 165°E 170°E
Figure 4. Distribution of the growth rate of disturbances in the bandpass filtered 850 mb relative vorticity field. Light
sdppling indicates positive tendencies in vorticity exceeding 2.5x10-7 s-l day-l (growth). Dense stippling indicates negative
tendencies less than -2.5x10-7 s-l day-1 (decay). Contour interval is 5x10-7 s-l day-1. See text for details of analysis
procedure.

56
Figure 5. Vertical cross-sections of (a) relative vorticity, (b) vertical pressure velocity, (c)
temperature, and (d) specific humidity, as constructed using a composite technique which is based on the
temporal coefficients of the leading pair of extended empirical orthogonal functions for 850 mb £, The
abscissa corresponds to a straight line segment extending northwestward from 2.5°N 142.5°E (extreme
right of each panel) to 27.5°N 110.0°E (extreme left). Contour intervals for panels (a), (b), (c), and (d)
are 3x10-6 s-l, 1x10-2 N m~2 s-1, 1x10-1 °C, and 2x10-4 respectively. Stippling indicates negative
values.

57
gence. This interpretation is supported by the almost in-phase relationship between the composite horizontal
charts (not shown) of outgoing longwave radiation, which is a good indicator of convective activity
in the tropics, and 300 mb temperature. The specific humidity centers [panel (d)3 are seen to lag
behind the corresponding £ centers [panel (a)] by one-eight of a wavelength. The moistening of lower
troposphere behind the trough may be associated with the increased evaporation resulting from the additive
effect of the southeasterly inflow in that region to the climatological southeasterly trades.
6. DISCUSSION
A composite method has been used to identify the characteristic large-scale circulation pattern
accompanying active episodes of the disturbances described above. It is seen that such active fluctuations
occur in conjunction with a southeasterly current towards the South China coast which is approximately
twice as strong as its climatological counterpart. The concurrent changes in both the level of synoptic
activity and the ambient flow field are suggestive of strong dynamical interactions among the high- and
low-frequency components of the circulation over the western Pacific. One of these modes of
interaction is the barotropic energy conversion from the mean flow to the transient eddies, as was
pointed out in Section 3 regarding the characteristic horizontal tilt of the disturbances and their
placement in the southeasterly shear zone of the quasi-stationary background flow. The relative
importance of this mechanism in the genesis and maintenance of the disturbances remains to be assessed.
This study has demonstrated that a considerable amount of usable information on the behavior of
tropical circulation systems may be extracted from the global 4-dimensional assimilation products currently
generated at various operational centers. The availability of these gridded analyses has substantially
expanded the data base for tropical empirical studies, which has heretofore been largely confined
to station data or observations taken within limited-area networks during intensive field experiments.
The principal findings of the present work have been verified using different independent datasets, such
as satellite measurements of the outgoing longwave radiation, and regional analyses produced by ROHK.
The global coverage of the ECMWF analyses also allows for the investigation of the nature of disturbances
occurring in other tropical regions, such as the Bay of Bengal/India area, the eastern Pacific, and
the Africa/Atlantic/Caribbean sector. It is seen that the transients over these regions exhibit interesting
similarities with, as well as differences from, the western Pacific features described here.
BIBLIOGRAPHY
*) Carlson. T.N., "Synoptic histories of three African disturbances that developed into Atlantic hurricanes",
Mon. Wea. Rev., 22,256-276 (1969).
2 ) Reed, R J. and E.E. Recker, "Structure and Properties of Synoptic-Scale Wave Disturbances in the
Equatorial Western Pacific", J. Atmos. ScL, 28. 1117-1133 (1971).
3 ) Reed, R.J., B.C. Norquist and EJE. Recker, "The Structure and Properties of African Wave Disturbances
as Observed During Phase HI of GATE", Mon. Wea. Rev., 105.317-333 (1977).
4 ) Nitta, T., Y. Nakagomi, Y. Suzuki, N. Hasegawa and A. Kadokura, "Global Analysis of the Lower
Tropospheric Disturbances in the Tropics During the Northern Summer of the FGGE Year. Part I:
Global Features of the Disturbances", J. Met. Soc. Jap., 122,272-292 (1985).
$ Nitta, T. and Y. Takayabu, "Global Analysis of the Lower Tropospheric Disturbances in the Tropics
During the Northern Summer of the FGGE Year. Part II: Regional Characteristics of the Disturbances",
Pageoph, 121272-292 (1985).
6 ) Wallace, J.M., G.-H. Lim and MJL. Blackmon, "Relationship between Cyclone Tracks, Anticyclone
Tracks and Baroclinic Waveguides", J. Atmos. Sci., 4$, 439-462 (1988).

58
TOE MEAN HEAT SOURCES OVER ASIAN MONSOON REGION
DURING THE PERIOD FROM APRIL TO OCTOBER OF 1980-1983
HUIBANG
LUO
Department of Atmospheric Sciences,
Zhongshan University,Guangzhou,China
ABSTRACT
An ECHIfF data set giving u,v, *, t and RH at seven levels
on a 2.5X2.5 grid iesh in the doiain 5°S-45°H, 50 0 -13G°E froi
April through October of 1980-1983 is used in this study.'*'
is recoiputed to satisfy such an upper boundary condition that
there are no heat sources other than the radiative heating in
the uppermost layer between 100 and 2QQhPa. The apparent heat
source Qi and the apparent loist sink Qz are coiputed for
each day. Only annual leans of Qi and Qz are discussed in this
paper! the inter-annual changes of Qi and Qz are discussed in
another paper.
The lain findings are :
1. The regions of the pronounced integrated apparent heat
source (Qi) are located over the continent and islands during
the period of suner half-year, with a laxiiui of 20Q*/! 2
located over the Buna-Yunnan region.
2. The integrated apparent heat source (Qi) is relatively
siaii over the oceanic regions. Over the South China Sea region,
the values of (Qi) are relatively large along the latitude of
17.5°K with a laxiiui not exceeded lOO^/i 2 ,which is far sialler
than that over the continent.
3. The sensible heat flux is very iiportant in taking
Buna the laxiiui heat source center. Cuiulus convection is
also vevy strong over this region after the onset of the
suner lonsoon.
4. This study verifies that dry therial convection
originating near the heated surface is responsible for the
deep tropospheric heating over the Plateau during the preonset
phase even for the interannuai lean results.
1.Introduction
The distribution and the heating mechanism of the atmospheric
heat source over the Tibetan Plateau and surronding
areas, and their importance on the Asian summer monsoon
circulation have been discussed by many authors (e,g.» Flohn,
1968; Yeh and Gao et al., 1979i Chen and Li,1983; Luo and Yanai,
1984; Johnson et al, 1985: He at ah, 1987).. But there
exists discrepancy among different results, for instance,

59
exists discrepancy among different results, for instance,
about the location of the heating maxiiui. In addition, discussions
adout the distribution and the heating mechanism of
atmospheric heat source were mostly based on daily data of
individual years, or on Monthly mean data of several years. In
this study we calculate heat sources for each day of the years
of 1980-1983 to obtain the 4-year mean characteristics of
their spatial and temporal distribution over the Asian monsoon
region.
2. Data and Method
An ECMWF data set u,v, « ,T and RH at seven standard pressure
levels (1000, 850, 700, 500, 300, 200 and 100 hPa) on a
2.5°X2.5° grid lesh in the doiain 5°$-45 0 N»60 0 -130°E from
April through October of 1980-1983 are used in this study. The
apparent heat source Qiand the apparent Moisture sink Qz
and their vertical integration are calculated from (e.g. Luo
and Yanai,1984):
(2)
where q and are respectively the lixing ratio of water
vapor and the potential temperature, and
=+LP+$, (3)
=L(P-EK n (4)
where < >=(l/g)J fob( )dp, (5)
P a and QR are respectively the surface pressure
and the radiative heating rate, P, S, and E are respectively
the precipitation rate, the sensible heat flux and evaperation
rate per unit area at the surface.
¥e recoipute the w to obtain more reasonable Qi in
the upper troposphere. The recomputed o> is obtained using
kinematic Method by integrating the continuity equation
(1/acoscj) 0u/aA + vcoscp/3cp) + 3»/3P=0, (6)
with the upper boundary condition
u=-(a&>9t+V^)/(a&0P). at P=150hPa» ()
which means that the lotion is approximately adiabatic in
the uppermost layer between 100 and 200hPa. The lower boundary
condition is obtained from
o.= o,,-D(P.-PN)
ECMWF data
where the suffix N denotes the value at the standard pressure
level nearest to the ground surface. D is the divergence
of the layer between PH and Ps.
In this study, (6) is integrated downward with the boundary
condition (7). The correction to the integrated divergence
is found by requiring « (Ps) = »s. The estimates^ of the
divergence at all layers are adjusted then by distributing
the required correction vertically with more weights at
upper layers.
3. Monthly Distributions Of The Vertically Integrated
Heat Source and Moisture Sink
Figures la-g show the horizontal distributions of .
(left side) and (right side) for the seven laonthes from

60
over tost parts of India, Southeast Asia (Burma-Thailand-
Indochina) and North Chin during the monthes of April and
May are ^ accompanied by negative moisture sinks, indicating
sensible heat flux and evaporation from the ground
surface. Besides, the areas of strong positive heat source
over Burma and southern part of India along 75°E are
acconpanied by weakly positive moisture sink, indicating
that both condensation heating and sensible heating are
iaportant.
As shown by Zhu et ah (1980), the South Asian anticyclone
tigrates from the western Pacific to Southeast Asia
in spring and moves towards the Tibetan Plateau in summer.
From the 4-year lean flow fields.we see ( figures not shown)
that the center of the anticyclone at 200hPa is located at
about 10°N» 110°E in April, moves continously towards the
Plateau and reaches about 30°M, 89°E in August. On the other
hand, we only see the intensification of the heating in the
region of Buna-Yunnan-Sichuan and the expansion of the domain
of this strong heating region. We do not see the corresponding
ligration of the center of heat source during the same
period. It confirms that there is no immediate association
of the migration of the anticyclone with the seasonal change
of heating over Burma-Yunnan-Sichuan (cf. He et al.» 1987).
4. 7-Month Mean Distributions Of Heat Sources and Moisture
Sinks
a.Mean horizontal distributions of and Figures
2a~b show the horizontal distributions of the vertically integrated
heat source and moisture sink averaged
over the seven monthes. The pronounced heat source of 25GW/I
along the border between China and Burma Kj°2 si £22|j, ¥ i{\ n( \
the moisture sink of the sane order of magnitude «00¥/» ) and
is related to the rain in this region. Heat sources MOOi/ffl
in a belt extending from northeastern India .and Bengal to
South China also corresponds well to miosture sinks of similar
magnitudes. The saae is true over the island of Borneo. He
find heat sources of 150-20QW/ 2 over southern Burma, which
are accompanied by weak moisture sinks of.50S/»«, indicating
that sensible heating is also very important in this region.
There exists a region of relatively large values of heat source
( .in the vertical plane along 110 E.

61
The heating over the South China Sea appears to consist of
two different types. The heat source over its northern part
W u . P f cM 1 " 68 ° f >! k/day in the 300-500hPa layer at
about 17.5 N is accompand by a Moisture sink with values >1
K/day in the 500-700hPa layer. Both heat sources and moisture
sinks over its southern part have small negative values
in the upper layers and small negative values in the lower
layers, with I Q 2 I > I Qi I in the lower layers where the
downward motion is dominant. This figure indicates again
that heat sources and moisture sinks over the continent is
stronger than those over the oceanic region during this
period of the year. There are two peaks of heat source in
the layer of 300-500hPa at 22.5°H and 30°H. There exists a
very strong moisture sink (>3 K/day) in the layer just above
the ground at 22.5 N, probably due to topography-induced
strong upward motion along mountain slops.
Figure 4 shows the 7-month mean vertical distributions of
w » Qi/Cp and Qa/Cp in the vertical plane along
20°N. The pronouced heat sources located over India, Burma and
northern Indochina are accompanied by weak moisture sinks indicating
again that sensible heating is very important in
these areas. On the other hand* the separation between the
levels of the Qi and Qz maxima suggests that cumulus convection
may be important for the heating in the Bay of Bengal at 85°
-90°E,northern Indochina and Hainan island.
5. Regional Characteristics of Heat Sources and Moisture
Sinks
To identify the principal heating contributors in different
parts of the domain under discussion, we examine the
areal mean values of Qi and Q 2 (denoted by [Qi] and
[Q 2 ] (over selected 7.5 X7.5 regions). These regions are
(1). middle India, (2). Bengal, (3). Burma, (4). middle South
China Sea, (5). South China, (6). middle Yangtze River, (7).
North'China, (8). eastern Tibetan Plateau.
a. Mean vertical profiles The 7-»onth mean vertical
profiles of [« ]* [Qiland [Qz] are shown for the eight
regions in Figs. 5 a-h. tfe divide
them into 4 types. The first one consists of Figs. 5^a, b
and c. It is characterized by very strong upward motion in the
lower levels and the heat source profile shows a peak value in
the lowest layer, which may be caused mainly by sensible heat
flux from the ground. Differences in moisture sink profiles
in Figs. 5 a, b and e show that cumulus convection over the
region of Bay of Bengal may be stronger than that over the
other two regions. On the other hand, sensible heating over
the region of Bengal may be stronger than that over the other
two regions, . ,
The second type of profies consists of Figs. 5 d,e and
f with similar w profiles and heat source peak appearing
in the 300-500hPa layer. Differences occur in the moisture
sink profiles, with the strongest cumulus convection occurring
over South China region. .
The other profiles appear to belong to two other dist

62
inctly different types. One is the type for North China
( Hg.bg) which shows downward motion and a weak positive
heat source in the middle troposphere due to the frontal
rain during summer monthes. The other is the type for eastern
Tibetan Plateau ( Fig.Sh). In this region, the heat
source peak is located in the lowest layer, while moisture
sink has large value up to the level of 300hPa, indicating
that condensation heating is important over the
eastern Tibetan Plateau. Dry thermal convection suggested
by Luo and Yanai (1984) is confirmed even using 4-year
mean data during late April and early May in this region
(not shown).
b * Vertical time sections of [Qi] and [Q g ] The 5 day
nean values of [w] ,[Qi] and [Q 2 J are computed to get the
height-time sections. Here we only show the results for middle
India region (Fig. 6) due to limited space.
As shown in Fig.6 (middle), [Qi] is positive in the lowest
layer and [Q 2 ] is negative (Fig.6, upper) in almost the whole
troposphere before later May, indicating that evaporation
and sensible heat flux are evident during this period. The
correspounding [«] (Fig.6, lower) appears to be upward
beneath the 700hPa level and downward above the level. After
a transition period from later May through middle June, the
vertical distributions of [w] [Q a ] and [Q 2 ] show completely
different characteristics. Upward motion prevails in deep
troposphere with the peak shifting from the 850hPa level to
the 700hPa level. The weak heat source of 1 k/day with a peak
in the 300-500hpa layer is accompanied by a strong moisture
sink of >2 K/day in the 850-700hPa layer, suggesting the
presence of strong cumulus convection after the onset of the
summer monsoon in this region.
6. Summary and Future Work
The main findings in this paper are:
1) The time sequence of does not show a migration
of heat source center from April to August corresponding to
the South Asian anticyclone. It confirms that the direct association
with the seasonal change of heating over Buna-
Yunnan region does not seem to be adequate.
2) The regions of pronounced integrated apparent heat
source are located over the continent and islands during
the period of the summer half-year, with a maxiiun of >200
V/i 2 over the Burma-Yunnan region.
3) The apparent heat sources are far smaller over the
oceanic regions than those over the continent and islands, .
Over the South China Sea region, relatively large values ot
are found along the latitude oM7.5°N with maxima not
exceeded 100V/m 2 . . ' ...
4) The sensible heat is flux very important in maKing
Buraa the maximum heat source center. Cumulus convection is
also very strong over this region after the onset of the
suner monsoon^ verifies again that dry thermal covection
originating near the heated surface is responsible for

63
the deep tropospheric heating over the Plateau during the
pre-onset phase even for the interannual mean results.
The agreement of the results over China between this
work and that of heat budgets from climatological data (Gao
and Lu, 1981) Is good. But for the oceanic region of the
domain^under discussion, further work is required before a
comparison can be made because no significant divergent winds
are retained in the tropics in the ECMVF analyses (Chen, 1987)
7.Acknowledgement.
This work was supported by the National Natural Science
Foundation of China under grant 4870262 and was also a part of
the Monsoon Research Project, State Meteorological Administration
of China,
REFERENCES
Chen, L. -X. and V.-L. Li, "The Atmospheric Heat Budget of
the Asian Monsoon in July." Proceedings of the Symposium
on the Sumer Monsoon in South East Asia, 86-101. (1983)
(in Chinese).
Chen, T-C., *Interannual Variation of the Tropical Easterly
jet" , Mon. ¥ea. Rev., 115, 1739-1759 (1987).
Flohn, H., "Contribution to a Meteorology of the Tibetan
Highlands" ,Atnos. Sci. Paper No. 130, CSU, 120pp. (1968)
Gao, G. and Y. Lu, * Atlas of Physical Climatology of China" ,
Agriculture Press, 183pp (1981). (in Chinese).
He, H., J. tf. McGinnis, Z. Song, and M, Yanai, * Onset
of the Asian Suuer Monsoon in 1979 and the Effect of the
Tibetan Plateau" . Mon, ¥ea. Rev., 145, 1966-1995.(1987)
Johnson. D. R., r, d. Tovnsend and M.-Y. ¥ei, * The Thermally
Coupled Response of the Planetory Scale Circulation
to the Global Distribution of Heat Sources and Sinks" .
Tellus, 37A, 106-125(1985).
Luo, H., and M. Yanai, * The large-Scale Circulation and
Heat Sources over the Tibetan Plateau and Surronding Areas
during the Early Suuaer of 1979. Part II** Heat and Moisture
Budgets" . Hon. tfea. Rev., 112, 966-989 (1984)
Ye, D, and Y.-X. Gan etal., ''The Meteorology of the Qinghai-xizang
(Tibet) Plateau" . Science Press, Beijing, 278pp
(1979) (in Chinese).
Z,hu, F.-K,,L.-H. Lu.X.-J. Chen and ¥. Zhan, " The South Asian
High" . Science Press, Beijing, 95pp (1980) (in Chinese)

64
60 70 too no 120 00

65
. 3
Fig.l. The monthly mean values of vertically integrated
apparent heat source (left) and apparent moisture
sink (right) in units of 50¥/m 2 for (a)
April.(b)May.(c)June.(d) July.(e) August.(f)September.(g)
October.(from top to bottom)

Fig.2.The 7-month mean
values of (a)vertically
integrated apparent heat
source and (b)apparent
moisture
50W/m 2 } .
sink (units
60 TO 90 100 HO 120 13060 70 90 «00 HO 120 (30
is—isr
100 \
700 ;
woo;
IT
i5o
100
zoo
303
50P
6 5 16 ft X) Ti ft & &
Fig,3. North-south vertical cross
sections showing the 7-month mean
vertical p-velocity G>{lower,10~ hpa/s),
heating rateQ /C (middle,K/day) and
* p
drying rate Q /C {upper,K/day) along
2 p
5 ^ £ Is 85 iw » iw w no ii5 50 IS
Fig.4. East-west vertical cross
sections showing the 7-montlj mean
vertical p-velocity o>(lower,10~ hpa/s) ,
heating rateQ^/C^(middle,K/day) and
drying rate Q 2 /C (upper,K/day) along
20°N.

Fig.6. Time sections of the areal mean
vertical p-velocity -.[wi (lower, 10" hpa/s
),heating rateCQ^l/C (middle,K/day) and
drying rate iQ 2 l/C (upper,K/day)for
middle India.
Fig.5. The 7-month mean profiles of
the areal mean vertical p-velocity
iwHthin solid line, 10" hpa/s},heating
rate [Q^/C (thick solid line,K/day) and
drying rate [Q 2 "|/C^(dash llne,K/day) for
eight regions(see text for details).
200-
500-
500-
700-
850
tooo
loo
200
300
500
700
B50
/WO -0 -8 -6 -* -2 0 2 4 6 8 fo -/O -8 -6 -* •* 0 2 A- 6 8 10

68
A NUMERICAL SIMULATION OF THE MEI-YU FRONT
L.C. Chou, C.P. Chang and R.T. Williams
Department of Meteorology, Naval Postgraduate School,
Monterey, CA, USA
ABSTRACT
A two-dimensional frontal model was used to study the structure
and behavior of the Mei-Yu front, with special emphasis on the
dynamics of the low-level jet (LLJ) that is frequently observed ahead
of the front. This study is motivated by the unique mixed midlatitudetropical
properties of the Mei-Yu front, where heavy convection and
strong low-level horizontal shear are associated with the nearly
zonaliy-oriented front. The occurrence of the LLJ is highly correlated
with heavy convective rainfall.
The quasi-steady state responses to a large-scale stretching
deformation forcing were obtained by integrating the perturbation
equations from an initial state of seasonal-mean zonal^flow. Two major
sets of experiments, based on midlatitude and subtropical conditions,
were conducted. For. each set of initial conditions, a number of
supplemental cases were integrated to study the effects of varying
initial humidity values and surface fluxes of heat and moisture.
The midlatitude front extends deeply into the upper troposphere
with a strong poleward tilt, whereas the subtropical front is confined
to the lower troposphere with less tilt, in good agreement with
observations. Along the sloping front, slantwise updrafts develop with
a multi-band structure. This updraft is more evident in the
subtropical cases and in the more moist midlatitude cases.
A westerly jet In the upper troposphere and an easterly jet in
the lower troposphere develop in both midlatitude and subtropical
cases, with smaller intensities in the subtropical cases. The
inclusion of more humidity or surface heat and moisture fluxes leads
to stronger upper westerlies, apparently as a result of stronger
cumulus convection. For the subtropical cases, concurrent development
of upper-level easterlies and low-level westerlies equatorward or the
front Is observed. The low-level westerly maximum at z = 3-4 km
resembles a LLJ, whose Intensity increases when more moisture is
Included. The concurrent development suggests that the LLJ may be the
result of a thermally-direct secondary circulation that resembles a
"reversed Hadley" cell. This circulation is revealed by a meridionalvertical
streamfunction, with a strong lower branch return flow
coinciding with the development of a LLJ in the more moist,
subtropical cases. The Coriolis torque of the meridional circulation
can develop and maintain the upper easterlies and the LLJ. Importance
of cumulus convection and especially slant-wise convection in
developing the reversed Hadley cell and the LLJ is suggested.
These conclusions are consistent with the observed intense
convection and heavy rainfall in the Mei-Yu front, and a sinking
region south of the Baiu front as revealed by Matsumoto's (1972)
moisture analysis.

69
WIND AND MOISTURE FIELDS DURING THE PERIODS OF ENHANCED AND
SUPPRESSED CONVECTIVE ACTIVITY OVER THE MALAYSIA-SOUTH CHINA SEA
REGION DURING THE NORTHERN WINTER MONSOON, NOVEMBER-DECEMBER 1986
ABSTRACT
By
Boon-Khean Cheang and Prakash Sankaran
Malaysian Meteorological Service
From the 1986 upper air streamline analyses, two synoptic
patterns for the interaction of cold surges with the tropical
circulation during the northern winter monsoon were identified.
The
first pattern is associated with the occurrence of widespread
heavy rain in the Malaysia-southern South China Sea region.
second pattern is associated with the occurrence of widespread
heavy rain in Philippines-West Pacific region and suppressed
condition in Malaysia-southern South China Sea region. The main
distinction between these two patterns is the presence (absence) of
deep and broad zonal easterlies In the tropics in the first
(second) pattern. The tropics in this report is referred to the
region from the West Pacific to Southeast Asia. Moisture flux was
computed for the period, 15 Nov - 15 Dec 86.
The
Latitude-time section
of the flux at 850, and 500 hPa levels are presented and discussed
in the report.
1.0 INTRODUCTION
Malaysia-southern
Enhancement of eonvective activity over the
South China Sea region due to interaction of cold
surges and tropical circulation during the northern winter monsoon
has been documented In many studies (Ramage 1971, Cheang 1976,
Chang et al 1979 and others).
Nevertheless supressed convective
activity over the same region due to interaction of cold surges and
tropical circulation has yet to be documented.
This report presents two examples of interaction of cold
surges and tropical circulation bringing about different weather
conditions in Malaysia December 1983 and November and December
1986.

70
Synoptic circulation patterns for the two examples were
compared to bring out the differences. Fields of specific humidity
and water vapour flux were computed and compared for the two
examples. In this report only the latitude-time cross-sections of
specific humidities and moisture fluxes along longitude 115°E (over
the South China Sea) are presented.
2.0 DATA
The season 15th November - 15th December 1986 was
examined in this investigation. Operational charts in the
Malaysian Meteorological Service were used to study the synoptic
circulation.
For the computation of the moisture parameters (specific
humidity and water vapour flux), ECMWF (European Centre For
Medium-Range Weather Forecast) 2.5° grid data for the 1986 season
were used.
3.0 SYNOPTIC SITUATION
Figure la depicts the time series of surface pressure
over northeastern China and Hong Kong. Below the time series are
the following rainfall histograms (Figure lb):-
(i) The total daily rainfall of 4 stations along the
east coast of Peninsular Malaysia,
(histogram-striped)
(ii) The total daily rainfall of 3 stations along the
north-western coast of Borneo Island,
(histogram-shaded)
There were altogether 5 cold outbreaks (intensification
of surface pressure) in China from the 15th November through 15th
December 1986. Heavy rain was recorded only during the short
period from the 25th November through 1st December. Heavy rain was
not recorded in Malaysia during the 15th-24th November 1986 and the
2nd~15th December periods. Heavy rain was recorded along the east
coast of Peninsular Malaysia during the spell from the 25th
November through the 1st December 1986 with peak of 919.5 mm

71
recorded on the 27th. The period (15th-24th November) did not
experience any heavy rain in Malaysia because the winter monsoon
was not established yet since the near-equatorial trough was
located north of Malaysia,
The period (15th November - 15th December) can be divided
into 2 periods of different convective activities over Malaysia;
one (15th November - 1st December) with enchanced convective
activity and the other (2nd-15th December) with suppressed
convective activity.
The period of enhanced convective activity is
charaterised by the following synoptic features:-
(i) There were 4 cold outbreaks in China.
(ii) The near-equatorial trough axis was nearly parallel
to the equator from the West Pacific to the South
China Sea with broad zonal easterlis north of it in
the lower and middle troposphere (850 hPa to 500
hPa levels).
(iii) At the 500 hPa level the mid latitude westerlies
over the Asian continent extended southwards to
only 20°N and the subtropical ridge axis was
located between 20° and 15°N.
Figure 2a and b are the streamline analysis respectively
for the 850 and 500 hPa levels for the 26th November 1986 when
heavy rain began to fall in Peninsular Malaysia.
The period of suppressed convective activity is
characterised by the following synoptic features:-
(i) There was only one clod outbreak on the 7th
December 1986.
(ii) During most of the days, the near-equatorial trough
was observed to disappear from the West Pacific -
South China Sea region. Instead, the northeasterly
cross-equatorial flow prevailed over this region.

72
(iii) The mid-latitude westerlies over Indo-China at the
700 and 500 hPa levels extended as far south as
12°N. The subtropical ridge axis also progressed
southward to a position between 15°N and 10°N.
(iv) Tropical depressions/tropical storms/typhoon
occurred more frequently over the West and central
Pacific during the suppressed period than the
period of enhanced convective activity. They were
present for 12 out of 14 days of the suppressed
period.
(v) There was a 500 hPa north-south oriented westerly
wave trough moving eastward from India across Indo-
China to the South China Sea.
(vi) As a result of the conditions stated in (ii), (iii)
(iv) and (v) 3 there were no broad easterlies from
the West Pacific to the South China Sea at the 700
and 500 hPa levels during most of the days of
suppressed convective activity over Malaysia.
Figures 3a and b are those for the 7th December 1986,
when there was a cold outbreak over China but there was no heavy
rain in Malaysia. There were one typhoon (KIM) and one tropical
depression over the West Pacific on the 7th December. Both of them
were located in the region north of latitude 10°N.
4.0 SPECIFIC HUMIDITY AND WATER VAPOUR FLUX
Figures 4a, b and c and 5a, b and c are the latitude -
time cross-sections along 115°E for the upper winds, specific
humidity and water vapour flux for the 850, and 500 hPa levels
respectively for the above-stated period. In the figures, the
subtropical ridges are represented by bold lines whereas the
near-equatorial troughs, by dashed lines. The bold arrows below
the date axes separate the period of enhanced convective activity
before the 2nd December 1986 from the following period of
suppressed convective activity. The letters M and D in the

73
cross-sections for specific humidity represent maximum specific
humidity (mosit) and minimum specific humidity (dry). The
following features can be noted from the figures:-
850 hPa level (Figures 4a, b and c)
(i) All the f M ! (moist) periods were confined to the
equatorial region where the near-equatorial trough
was located,
(ii) The atnosphere during the suppressed period was
dry.
(iii) There was a weakening of easterly water vapour flux
across 115°E between 10° and 20°N from the 2nd to
15th December (suppressed period). It coincided
with the southward progression of the subtropical
ridge.
500 hPa level (Figures 5a, b and c)
(i) The most significant feature is the appearance of
drier air associated with the southward progressing
of the subtropical ridge and westerlies to the
equatorial South China Sea.
5.0 SUMMARY AND DISCUSSION
(i) In the presence of deep and broad zonal easterlies from
West Pacific to South China Sea, cold surges interacted
with near-equatorial vortex embeded within the
near-equatorial trough and brought about heavy
precipitation to the east coast of Peninsular Malaysia
during the last few days of November 1986. The local
Hadley circulation can be considered to be active over
the China •- South China Sea region.
(ii) In the absence of deep and broad zonal easterlies at
the 850, 700 and 500 hPa levels, cold surges did not
bring about heavy rain to Malaysia. The absence of
deep and broad zonal easterlies occurred simultaneously
as the southward progression of the subtropical ridge

74
and the temperate latitude westerlies to Indo-China and
also the occurrence of tropical storms over the West
Pacific.
(iii) The atmosphere was moist at the lower level (1000 hPa
level) throughout the two periods (Figures for moisture
field not shown). On the contrary at higher levels,
850 hPa and particularly 700 and 500 hPa, the variation
in the circulation pattern and moisture content in the
atmosphere was more prominent and can be considered to
play more important roles in controlling the weather
pattern in the South China Sea - Malaysia region,
(iv) The deep and broad zonal easterlies from West Pacific
to South China Sea were moist and not associated with
subsidence. On the contrary, the southward progression
of the subtropical ridge and the westerlies to about
10°N must have resulted in large-scale subsidence over
the South China Sea-Malaysia region. That is the main
reason why cold surges occurred over the South China
Sea with that upper air environment did not result in
the occurrence of heavy rain to Malaysia.
(v) It is noted that tropical storms were present more
frequently over the West Pacific during the period of
suppressed convective activity. There is a possibility
that the tropical storms helped to draw the
mid-latitude westerlies southwards changing the
regional circulation. Furthermore, since the West and
central Pacific experienced frequent tropical storm
activity during the suppressed conditions (2nd-15th
December) an east-west zonal circulation might have
existed with the upward (ascending) branch over the
West and central Pacific and the downward (sinking)
branch over the South China Sea.

75
(vi) It will be very useful and interesting to determine
whether advection of dry air from Indo-China to
Malaysia because of the southward progression of the
subtropical ridge and westerlies, was occuring during
the suppressed condition besides the large-scale
subsidence suggested in paragraph (.v). A more detailed
study will be conducted for the purpose.
In conclusion, it is felt that the prediction of heavy
rain over the Malaysia-South China Sea region should not be based
solely on the monitoring of the occurrence of cold surges but
should also monitor the day-day variation of the large-scale
circulation.
REFERENCES:
Ramage, C.S.,: Monsoon Meteorology, Academic Press, New York,
1971, pp 158-160.
Cheang, B.K.,: Synoptic features and structures of some equatorial
vortices over the South China Sea in the Malaysian region during
the winter monsoon December 1973, Pure Appl. Geophs. 115, 1303-1333
(1977).
Chang, C.P., J,E. Erikson and K.M. Lau: Northeasterly cold surges
and near-equatorial disturbances over the winter MONEX area during
December 1974, Part I; Synoptic aspects, Mon.Wea. Rev., 107,
812-829 (1979).
Cheang, B.K,,:
579-606.
Monsoon, John Wiley & Sons, Inc, New York, 1987, pp

76
O 1 -
(b)
Total I
Rainf a"
4 0 J
i -» » * 4 4
tet i i i i i
!'•y^..... .. ^fU.J. 0 ,.
E^-I-I ^o as "Jo « so B is
NOV DEC 1986
Figure 1 (a) Surface Pressure over North-eastern
China and Hong Kong
(b) Total Daily Rainfall in Malaysia

77
Figure 2 Streamline Analyses
(a) 850 hpa
(b) 500 hpa
Figure 3 Streamline Analyses
(a) 850 hpa
(b) 500 hpa
T = typhoon

78
LONGITUDE TIME WINOCROS5 SECTION 850 HPR
40
35
30
25
20
35
10
5 •
0 -
-5 •
-10 •
-,15 -
-20 -
SPECIFIC HUMID!TT TIME CROSS SECTION 850 HPfl flT LONGITUDE 'US DEC ERST
I I 2 l l l 2 J l l l l l | l l l l l t l l l l l 2 I i
35 -
30 -
25 -
20 -
15 -
10 -
S -
0 -
-5 -
-10 -
-15 -
-20 -
10 •
35 •
30 •
25 -
20 •
IS •
10 -
5 •
0 -
-s -
-10 -
-is -
-20 -
-T—i—i—i—i—i—i—i—i—i—i—i—i—j—i—i—i—i—i—i—r~
WRTER VHPOUR FLUX TIME CROSS SECTION 850 HPfl
® s. A. N, ,- S,
IS IS 17 18 |<
MOV-DEC 86
tt-t-w^;-
«-t V, V V v V-*
—->n^J\'
""^ "^ C.-' •w'•*/ ** s
^ ^ «u J v, v_ v 1
V s * i" T ** r ^ Y f
Figure 4

79
N
uo
35 -
30 -
25 -
20 -
IS -
10 -
5 -
- 0 -
-5 -
-10 -
-IS -
-20'-
10
35
30 •
25
20 -
IS •
10 •
S -
0 -
-S -
-10 -
-15 -
-20 -
LONGITUDE TIME WJNDCROSS SECTION 500 HPR
i V f *• r
f r - r r"
r r c r r '
SPECIFIC HUHJDITT TIME CROSS SECTION
r* r f t*
r r r r
_ - - ^ > v > A
sy «v/ %x •w' i^. —/ •* «J~/A
•""y-'lll*
V-/
v > . ^ > .* > n
»-A, "V A 7 A -\
SOO HPfl flT LONGITUDE IIS DEC EflST
i>;; ; re
; s;o^
10
35
30
25
20
IS
10 •
5 •
0
-5
-10 -
-15 .
-20 -
- V-
HRTER VflPOUfi FLUX TIME CROSS SECTION
500 HPfl
Figure 5

A simulation of Lee-Cyclogenesis over Yun-Gui Plateau
with the use of hemispheric spectral model
Wen-ShungKau
Yi-May Lin
Department of Atmospheric Science
National Taiwan University
ABSTRACT
In East Asia, the large-scale circulation is influenced by the Tibetan
Plateau, especially in the lee side. The area of Szechwan and Yun-
Gui is the highest frequency of cyclogenesis. The development of
Yun-Gui cyclones affect not only the weather system of western and
southern China but also that of Taiwan, when the cyclones move away.
This report will simulate the case found during February 19-21,
1979 by using a hemispheric spectral model and the data of FGGE level
Illb. The results show that the model can predict the tendency of the
development of cyclogenesis.
Different topographies will give different intensities of the low
pressure system. Comparing the result using the controlled topography
to that using the intensified topography, the former is much weaker in
the poorer prediction of low. The forecast intensity of the Yun-Gui low in
the small topography case is also much weaker than that in the controlled
topography case. All in all, the effect of topography is obviously
important. To get a further understanding of the whole process of
cyclogenesis, the vertical profile of vorticity field of cyclogenesis, the
vertical profile of vorticity field of cross-section on 27.6° N, is studied.
1. Introduction:
Orography plays an important role in the atmospheric circulation, because mountain
ranges modify and generate weather systems of all scales of atmospheric motion. Thus,
for example, the lee slopes of the Rocky mountains, the Alps and the Tibet plateau are
favored cyclogenetic regions. However, different orography configuration and different
large-scale weather patterns will cause different weather systems. In our area, leecyclogenesis
over the Yun-Gai (Tibetan) plateau is the major concern. Chung et al. (1976)
point out that most lee cyclones occurring in the lee of the Tibetan plateau do not acquire

81
appreciable intensity throughout their life span. They regard this to be due to weaker
upper-air flow in this region. Besides they also point out that the size and orientation
of the mountains are important for the generation and modification of lee cyclones, as
well as the motion and intensification of cyclones. Using the Winter Monex data to
study the period of 19-21 February 1979 over the lee side of the Tibetan plateau, Chen
(1985) points out that Yun-Gui lee cyclogenesis of Feb. 19 was a hydrostatic response to
the subsidence warming in the lower-and mid-troposphere when the temperature tendency
by warm advection was roughly counteracted by the diabatic cooling process. At the
same time, the secondary vertical circulation associated with the upper-level jet streak
was suggested to have some contribution to the subsidence over the cyclogenesis area
in addition to the leeside orographic effect. Jen (1987) studies the same case with the
use of isentropic potential vorticity (IPV) analysis. He points out that the stretching
of IPV in the lee side of the Tibetan plateau, if in phase with low-level warm vorticity,
can cause a phase-lock and intensify the cyclogenetic process. The purpose of this study
is to use a numerical model to study the Yun-Gui lee-cyclogenetic mechanisms. The
hemispheric spectral model and its physical processes will be briefly described in section
2. Section 3 presents the synoptic analysis of Yun-Gui lee-cyclogenesis during 19 to 21
Feb. 1979, and the results of numerical model simulations. Section 4 provides the
discussion of the results.
2. The hemispheric spectral model
The equations and framework of the numerical model used in this study are described
in Brenner et al. (1982) and Brenner et al. (1984). A brief description will be provided
here as background. This is a spectral model with 12 layers and rhomboidally truncated
at wavenumber 30.
The prognostic equations are vorticity, divergence, temperature,
specific humidity, and the logarithm of the surface pressure (continuity equation). The
vorticity and divergence are given conventionally as Laplacians of a streamfunction and
velocity potential respectively. The velocity components and geopotential are obtained
diagnostically with the aid of the hydrostatic relation.
The vertical coordinate of the
model is cr=P/P*, defined by Phillips (1959), with a specification of layer locations
described by Brown (1974) and Phillips (1975). The numerical methods in the model
include spectral representation in the horizontal, the Arakawa quadratic conserving finite
differencing in the vertical, and the semi-implicit time integration scheme. The nonlinear
normal mode initialization technique developed by Machenhauer (1977), which is

82
implemented to the global model by Ballish (1980), is applied to adjust the input data.
In this procedure, the data are projected onto the normal modes of the model and
separated into the rotational (or Rossby) modes and the gravity modes. The gravity
modes are adjusted in such a way that the linear tendency terms are balanced by the
adiabatic nonlinear advective terms, thus resulting in a net initial gravity wave tendency
of zero. The physical effects included in the model are the influences of orography,
position-dependent surface friction, moisture physics, and sub-scale horizontal dissipation,
parameterized by diffusion. Evaporation and sensible heat flux from the oceans are
also included. The application of the moisture physics consists of a sequence of three
steps to adjust the temperature and specific humidity. Each step possesses a characteristic
spatial scale: (1) cumulus convection in a conditionally unstable, generally unsaturated
large-scale flow, (2) large-scale condensaton in stable, saturated large-scale flow, and (3)
dry convection in unstable, unsaturated large-scale flow.
3. The simulations of Yun-Gui cyclogenesis:
3.1 Data:
The data used in this study are taken from the FGGE Level III-B data from 12 GMT
Feb. 19 to 00 GMT Feb. 21, 1979. The input data for the numerical model includes
winds and geopotential height (from 1000mb to 500mb) at 12 layers and moisture
field (from 1000mb to 300mb) at 6 layers at 12 GMT Feb. 19 and 00 GMT Feb. 20
respectively.
3.2 Synoptic situation:
The synoptic situation during this period will be described briefly in the following:
In the 850mb analysis (Fig. 1) we can find a low pressure trough around the Yun-
Gui area at 12 GMT Feb. 19. This trough the intensified for the next 36 hours. At the
same time, southwesterly warm advection prevailed over southern China which caused
unusually warm weather over the Taiwan area. For the 500mb synoptic chart (Fig. 2) we
can see a warm ridge located around the Yun-Gui area at 00.GMT Feb. 20 and a weak
trough around 105° E, 23 0 E-30°N. This trough continued to deepen for the next 12
hours. At 300mb the flow was generally from the west-northwest over the Yun-Gui
plateau during this period. We also do the vorticity analysis for this case. Fig. 3 shows
the vorticity field at 850mb. At 12 GMT Feb. 19 there exists a weak positive vorticity

83
center over the Yun Gui area. This center started to grow and intensify for the following
36 hours. Fig. 4 shows the 500mb vorticity field. At 12 GMT Feb. 19 there has no
vorticity center around the Yun-Gui area. However, 24 hours later (12 GMT Feb. 20) a
positive center at this area can be seen. This center almost coincided with the vorticity
center at 850mb except it was towards the west side of the latter.
3.3 Numerical simulation:
Four simulations were performed (Table 1) in the study. They are:
(A) The control orography experiment, referred to as CON, includes all the model
physics and the orography derived from the U.S. Navy data, which has a resolution
of 10' of arc. Two simulations with different initial data were included (i.e.
simulation AA and A).
(B) The significant orography experiment, referred to as SOE. This experiment is similar
to CON but the orography is changed to significant height (Pfaendtner et al. (1985))
(i.e. simulation B).
(C) The smaller orograph experiment, referred to as SMB. This experiment is similar
to CON but the orography reduced to 1/10 of CON (i.e. simulation C).
3.3.1 Simulation AA:
Figs. 6 a~e depict the successive development of the simulated 850mb geopotential
height, temperature and wind fields from the CON experiments for the initial data at 19
GMT Feb. 20. It can be seen that the trough keeps deepening for this 36 hours simulation.
Compared to the analyses, the trough, warm ridge and geopotential height are similar
to those observed. Another notable feature of the CON is the correct prediction of the
enhancement of the thermal ridge around Yun-Gui area. The 500mb wind, temperature
and height fields from 6 to 36 hours simulated in the CON experiment are shown in
Figs. 7 a-e. A trough starts to grow after 12 hours and then keeps deepening for the
next 24 hours around the Yun-Gui area which is in good agreement with analysis,
3.3.2 Simulations A to C:
The initial data for the simulations A to C is at 00 GMT Feb. 20. Figs, 8 a-c depict
the 12 hour simulated 850mb flow patterns for the CON, SOE and SME experiments. ;

84
In all three experiments, a deep trough over the Yun-Gui area is simulated. However,
their intensities are quite different. In CON, the trough is similar to that observed. In
SOE, the trough is much deeper than observed but in SME it is much weaker.
Figs. 9 a-c show the 12 hour simulated 500mb wind, temperature and height fields
for the above three experiments. The trough over Yun-Gui area in the SOE experiment
is deeper than that in CON but in SME this trough is weaker.
3.4 The vertical profile of vorticity field :
The vertical cross-section along 27.6°N for the observation and model simulations
are shown in Figs. 10 to 12 (The cyclogenesis in this study is around 105°E±3.75°E).
Fig. 10 is the evolution of the observed vorticity field. From this figure we can get:
(A) At 12 GMT Feb. 19, there exists a weak positive vorticity area at lower level around
105°E.
(B) At 00 GMT Feb. 20, the positive vorticity area at 105°E has intensified and extended
to the upper level. The maximum center is at around 600mb.
(C) For the next 12 hour, the vorticity at all levels around 105°E is much stronger
than that at 00 GMT, and the maximum center has ascended to 450mb and is moving
towards the east.
(D) At 00 GMT Feb. 21, this vorticity is much weaker than 12 hour before and the
maximum center has descended to 750mb.
Fig. 11 is from simulation AA. In general, the predicted vorticity field around
105°E is about the same as that observed. However, its intensity is weaker.
Fig. 12 is from simulations A, B and C. We can find the following features when
compared with the observations at 12 GMT Feb. 20:
(A) From simulation A, we find the predicted vorticity profile to be very close that
observed. The maximum vorticity is around 108.75°E which is the same as observed.
However, the maximum center is at around 650mb instead of 600mb.
(B) From simulation B we find the vorticity field around 108.75°E has become more
extensive and stronger than that in simulation A. This is apparently due to a charge
a the orography.

85
(C) With the reduction of the terrain height we find that the vorticity has become weaker
and its maximum center has moved faster toward the east than simulations A and B.
4. Discussion:
The results in the preceding section have led us to the following conclusions:
(1) From the 36 hour simulation of control experiment, we find that the hemispheric
spectral model can predict the tendency of development of the Yun-Gui cyclogenesis.
The deepening of the trough and the enhancement of the thermal ridge are also
simulated.
(2) Different orography configuration tests show that the orography is important for the
generation and modification of lee cyclones as well as the motion and intensification
of cyclones. This is in agreement with the results of DelPosso & Chen (1985).
(3) The low-level southwesterly warm advection provides the development of the lowpressure
system with a warm core and the high-level (500mb to 300mb) positive
vorticity advection in the western side of the plateau also plays an important role
in lee-cyclogenesis.
(4) The preferred propagation tracks of Yun-Gui cyclones have not been addressed in
this study. Also the sensitivity of the upper tropospherical jet and the surface
sensible heat flux to the lee-cyclogenesis need more studies.

86
References:
Ballish, A.B., 1980: Initialization theory and application to the NMC spectral model,
Ph.D. thesis, U. of Maryland.
Bourke, W., 1974: A multi-level spectral model. Formulation and hemispheric integrations.
Mon. Wea. Rev., 102: 687-701.
Brenner, S., Yang, C., and Yee, S., 1982: The AFGL spectral model of the moist global
atmosphere:
128233.
Documentation of the baseline version, AFGL-TR-02-0383, Ad A
Brenner, S., Yang, C, and Mitchell, K., 1984: The AFGL Global spectral model: Expanded
resolution baseline version, AFGL—TR-840308.
Brown, J.A., 1974: On vertical differencing in the sigma system. NMC office note 92,
13pp.
Chen, G.T.J., and F.W.C. Yeh, 1982: The climatology of winter cyclones over subtropical
china and adjacent oceans. Paper Meteor. Res., _5_, 85-98.
Chen, G.T.J., 1986: A diagnostic case study of Tibetan lee cyclogenesis during winter
Mon ex, Monsoon and meso- scale meteorology, 90-94.
Chung, Y.S., K.D. Hage and E.R. Reinelt, 1976: "On Lee cyclogenesis and airflow in
the Canadian Rocky mountains", Mon. Wea. Rev., 194, 879-891.
Cressman, G., I960: Improved terrain effects in barotropic forecasts. Mon. Wea. Rev.,
88, 327-342.
Dell'osso, L. and S. J. Chen, 1985: "Numerical experiments on the genesis of vortices over
the Quinghai-Tibet plateau*', Tellus, 38A, 236-250.
Jen, F. C., 1987: The isen tropic analysis of the lee-cyclogenesis. National Taiwan
University master thesis. 83pp. (in Chinese)
Machenhauer, B., 1977: On the dynamics of gravity oscillations in a shallow water model
with applications to normal mode initialization. Beit. Phys. Atmos. SO, 253-271.
Phillips, N.A., 1959: "Numerical integration of the primitive equations on the hemisphere",
Mon. Wea. Rev., 88 ; 333-345.
Philips, N.A., 1975: Application of Arakawa's energy conserving layer model to operational
numerical weather prediction. NMC Office Note 104 S 40pp.

92
10-.. 10-c
S1.S S. B2.S M. IBS. U2.S 120. 121,5 US
1 O -b
Fig. 10 The cross-section along 27.6°N produced
from analyzed data, showing
the development of the vorticity at
about 105°E in 19-21 Feb. 1979. (a)
12Z, 19 Feb. (b) OOZ, 20 Feb.; (c) 122,
20 Feb.; (d) OOZ, 21 Feb. (unit in 10*

93
1.2-a
I2.S 120. 127,5 US.
11.5 75. M.S W. ITS IOS.
1 1-C 12-c
Fig. 11 The cross-section along 27.6°N produced
from the experiment CON with
the initial data at 12Z 19 Feb,, showing
the development of the vorticity at
about 105*E, (a) 12 hour prediction,
(b) 24 hour prediction, (c) 36 hour
prediction, (unit in l$ r s~ l ).
Fig. 12 The cross-section of the vorticjty along
27.6°N produced from different experiments
with the initial data at OOZ 20
Feb. for 12 hour prediction, (a) CON
run, (b^SOE run and (c) SME run
(unit in ids' 1 )-

94
SEASONAL AND INTRASEASONAL VARIATIONS
OF THE EAST ASIAN SUMMER MONSOON
K.-M. Lau
Laboratory for Atmospheres
NASA/Goddard Space Flight Center
Greenbelt,MD 20771
U S. A.
ABSTRACT
Large scale features associated with the seasonal and intraseasonal variations of the
East Asian summer monsoon are reviewed. It is shown that the onset of the Mei-Yu in
southern and central China is coincident with the arrival of the first wave of organized
convection/rainfall associated with the 30-60 day oscillation. The latter propagates
eastward along the equator with an average speed of about 3-5 m/s over the Indian
Ocean/Western Pacific region. An east-west and a north-south seasaw are foundbetween
the Indian Ocean and the Western Pacific, and between the equator and the monsoon region
(15-20°N, 60 -150°E). There seesaws are associated with the northward and eastward
propagation of the rising and sinking branches of the divergent circulation associated with
the 30-60 day oscillations. Over the monsoon ITCZ region, westward propagating cyclones
with periods of four to five days and wavelengths of 1000-2000 km are predominant. Most
interestingly, over the equatorial Western Pacific, westward propagating cloud clusters
with periods of 2-3 days and length scales of 500 - 1000km are found to be embedded in
eastward propagating superclusters associated with the 30-60 day oscillations. Some of
the clusters are seen to develop into tropical cyclones over the northern South China Sea
and the East China Sea and then follows a northeastward path to reach Japan. Others
may propagate further westward across IndoChina and subsequently develop into monsoon
depression over the Bay of Bengal.
1. Introduction
The East Asian summer monsoon possesses a high degree of spatial and temporal
variability ranging from daily to interannual time scales. These have profound effects on
the weather and climate over countries of East and Southeast Asia including Indo-China,
China, Korea and Japan (Lau and Li, 1984, and Lau et al, 1988). In this paper, a review
based on some current and previous work by the authors and his collaborators will be
presented. The emphasis of this paper is on the connection of the regional aspects of the
East Asian monsoon to the global scale circulation.

95
2. Climatology of East Asian monsoon transitions
Figure 1 shows the time-latitude section of a 10-year average of 10-day rainfall
based on 50 stations over China east of 100°E. Rainfall amounts shown are averages from
stations with a latitude band 5 degree wide centered at 25°N, 30°N etc.
45N -
10 5040 30 20
40N
35N
SON
25N
APR i MAY I JUN I JULl AUG I SEP!
Figure 1. Time-latitude section of 10-day rainfall total averaged over stations within a 5°
latitude band and between the coast of China and 100°E from April to September. Units in
mm.
Features shown in this figure are representative of the multi-scale nature of the
climatology of monsoon rainfall transitions over the most populous region of China from the
Yangtze to the Yellow River. The abrupt increase in rainfall in southern China (25°N) near
the first 10-day of June marks the onset of Mei-yu (Plum rainfall). Following the onset, the
Mei-yu rainband migrate northward steadily at a rate of about 3 to 10 days. By the first
10-day of July, the rainfall maximum reaches between 35°N and 40°N. At this time
southern China undergoes a dry period or monsoon break while northern China

96
However, not all subseasonal scale variations are necessarily tied to the 30-60 day
oscillations. Recent results (not shown) from examination of 5-day averaged rainfall over
China indicates that there are higher frequency fluctuations that may not be related to the
equatorial 30-60 day oscillations.
3. East Asian monsoon and tropical convection
The East Asian monsoon rainfall variability discussed above is in fact part of a larger
variation involving the 30-60 day oscillations and the seasonal variation of tropical
convection in the global tropics. Fig. 2 shows the time-latitude section of satellite-derived
D J J jr r F
S SOOONKNDOD
Figure 2. Time -latitude section of 10-day satellite-derived rainfall total (mm) averaged
over a 5° wide latitude band and between 100°E to 115°E from January to December,
1979. The small rectangle covers the domain approximately the same that use in Fig. 1.
rainfall estimates based on High Resolution Infra-Red Sounder (HIRS) for the entire year
of 1979. The region marked in the small rectangle represents the variation over East Asia
with the same spatial and temporal domain as in Fig. 1. First, it is noticed that within the
small rectangle, the satellite derived rainfall features are remarkably similar those in
Fig. I, suggesting that these features are representative of the basic variations in this
region. However, the larger domain in Fig.2 reveals that the East Asian rainfall
variations is part of a pattern of global rainfall fluctuation. For example, the northward
extension of the Md-yu is now seen as northward penetration of a rainfall pattern that is
centered near 10°N that appears to have its origin near 5°S. This pattern appears to be
part of a larger pattern of rainfall fluctuation between 10°N and 10°S at intervals of 30-60
days. On an even large scale, the East Asian monsoon is seen as part of the northward
migration of the entire tropical convective zone from 10°S to 10N. It appears that the
monsoon onset of China may be traced to the first appearance of a rainfall anomaly
associated with the 30-60 day disturbance in the equatorial region during the month of June.

97
850 MB WIND (U*. V) S. 0
30N i~^~>iri*—.--
IV
-/v
1^."
\
_;.
— r— -<
X"
\
\ N
/'/
\
\ "X
' *\
\^
J\ V J
30 £ 60 ISO
Figure 3, Observed 850 mb wind fields showing the horizontal structure and propagation
of synoptic scale disturbances (Q, C2>...) over the monsoon region at two-day interval
during June 11-11,1979, Full barb is 5 m/s. (adopted from Murakami et al, 1984).
4. Monsoon Disturbances
Embedded in the East Asian monsoon trough are active synoptic scale disturbances that
generally propagates westwards threading the regions of Indo-China, southern China and
the Bay of Bengal Some of these disturbances can be traced to the western Pacific which
appears to be a spawning region for westward propagating disturbances such as easterly

98
waves. Fig. 3 adopted from Murakami et al (1984) shows a sequence of synoptic scale
disturbances that traverse the East Asian monsoon region during June 21-27,1979 at 2-day
interval. These disturbances all have low level cyclonic circulation with a westward
phase speed of about 5 m/s and periods of about 4 to 5 days. A number of these cyclones
have been found to develop into the well-known monsoon depression over the Bay of
Bengal. It is also noted that when the cyclones are well developed in the monsoon region
June 21-23, the large scale flow over the Western Pacific is strongly easterly and the mean
westerly over Indian (15 -20°N) is weak. This suggests an intensification of the
Subtropical High in the western Pacific concomitant with a weakening of the monsoon
westerly at the time of strong monsoon cyclone development. On the other hand, from June
23-25, the Western Pacific is highly disturbed when the monsoon cyclones are relative
weak and the monsoon westerly flow is strong. These apparent inverse relationship
between the western Pacific convection and the Indian monsoon have been noted in a number
of previous studies (e.g. Lau and Chan, 1986).
A recent theoretical study has linked the westward propagating monsoon
disturbances to Rossby waves generated by the mobile wave-CISK in the presence of a basic
planetary scale monsoon circulation (Lau and Peng, 1989). Fig. 4 shows the vertical
structure of model disturbances identified with these westward propagating tropical
cyclones. All the disturbances are found to have a distinct eastward tilt with height and
are generated as the convection associated with the 30-60 day oscillations approach the
monsoon region near the equator. From a simple stability analysis, Lau and Peng
tentatively concluded that the low level monsoon westerly flow and the horizontal
vorticity gradient of the three-dimensional monsoon basic flow provide favorable
conditions for these disturbance to extract energy from the latent heat release.
Fig. 5 shows a schematic diagram of the preferred region of growth in relation to the
monsoon basic flow based on their analysis. This is generally in agreement with that
discussed in Fig. 3. Lau and Peng's study suggests a link between the eastward propagating
30-60 day oscillation in the equatorial regions and the sudden development of the monsoon
trough at 15-20°N including the geneses of the westward propagating monsoon
disturbances. This supports the above-discussed notion that the Mei-yu onset-break-revive
sequence in central and southern China may be associated with the different phases of the
30-60 day oscillation in the equatorial regions.

99
120 ISO ISO 210 2VJ 770 ISO 180 210 2*0 270
Figure 4. Longitude-height cross-section of meridional wind at two day interval showing
the vertical structure and propagation of monsoon disturbances generated by the approaching
30-60 day oscillations.
20 N
10N
V>0
* t
U>0
Figure 5 Schematic showing the preferred region for development of westward propagating
monsoon disturbance relative to the monsoon heat source and the low level flow.
5. Western Pacific disturbances
In the previous discussion, we have linked the equatorial 30-60 day oscillations to the onset
of the summer monsoon over East Asia. However, the 30-60 day oscillations are found
(with varying amplitudes) all year round in the equatorial regions especially over the

100
Indian Ocean and the western Pacific and are known to possess a hierarchy of sub-structures
associated with cloud cluster and superclusters (Nakazawa, 1989, Nakazawa and Lau,
1989 and Lau et al, 1989). During the northern summer, a large number of westward
propagating disturbances associated with easterly waves traverse the western Pacific at
10-15°N. Many of these disturbances reaches East Asia and the South China Sea. It is
therefore necessary to examine the small scales structures embedded in the 30-60 day
oscillation and their possible relationship with the monsoon disturbances. Because
intermediate scales such as 10-20 day fluctuations are often found in groups within the 30-
60 day oscillations, we shall use the term 30-60 days oscillation in its generic sense in the
following to include these intermediate scales.
Figure 6 shows the time-longitude section of 3-hourly CMS convective index
averaged between 1°N and 1°S from March 14 to April 12,1984. During the first half of
the period, the convection over the region is relatively sparse consisting mainly of
westward propagating transients. During the second half, the convection in this region
becomes very active. Two supercluster complexes (labelled A and B) can be identified
during this period. The eastward propagation is seen to be made up of successive formation
of new clusters to the east while individual clusters propagate westward. The detailed 6-
hourly synoptic sequence during the period 12 Z, March 29 to 06 Z, March 31 is also shown.
The spatial scale of the superclusters is of the order of 1000 to 2000 km whereas smaller
clusters of the order of 500-1000 km can be found in the vicinity. The period represents the
beginning of the convective phase of the 30-60 day oscillations in the Western Pacific.
The above is a rather typical situation for the propagation of organized convection
associated with the westward and eastward components of the supercluster. In many noted
synoptic situations, a pair of cyclones on each side of the equator is found associated with
the equatorial supercluster. During northern summer, not infrequently, the northern
cyclones develop into tropical cyclones that are (a) either advected by the mean flow first
northward and then northeastward across Japan or (b) drifted across Indo-China to the Bay
of Bengal giving rise to monsoon depression there. The tropical cyclones in the former
categories can further develop into typhoons especially during the late summer season.
The discussion in this section and in section 2 is also consistent with the classical
zonally symmetric description of the equatorial ITCZ and the monsoon ITCZ. When
averaged over the equatorial belt between the Indian Ocean and the western Pacific, the
arrival of the rising branch of the 30-60 day oscillation feature enhanced convection and

Figure 6 (a) Time-longitude section of 3-hourly GMS convective index averaged between
1°N and 1°S showing the propagation of superclusters (labelled A and B) and their
embedded high frequency components, (b) Synoptic sequence indicated by GMS convective
index from 12 Z, 29 March to 06Z, 31 March, 1984, Heavy dots denote region less than
200°K and light dots between 200° to 225° K.
101

102
the development of an equatorial ITCZ within this region. However, the equatorial ITCZ
subsequently gives way to a monsoon ITCZ that develops near 15 -20°N by the interaction of
the 30-60 day oscillation with the monsoon circulation. Thus there will be an inverse
relationship between the monsoon ITCZ and the equatorial ITCZ. The abrupt northward
movement of the monsoon trough is manifested as a meridional see-saw between the
equator and the monsoon region. Fig. 7 shows the result of a model simulation of the
interaction between the equatorial and the monsoon ITCZ based on the interaction of the
30-60 day oscillations and the three-dimensional monsoon circulation (Lau and Peng, 1989).
The approach of the 30-60 day oscillation to the monsoon region along the equatorial is
signalled by the subsidence motion ahead of the concentrated rising branch of the
oscillation. At this time the whole monsoon region remains quiescent. The second panel at 6
days later, captures the double ITCZ that forms as the rising branch of the the equatorial
30-60 day oscillations is well within the monsoon domain and a monsoon ITCS is generated
at around 15-20°N. Another six days later, the monsoon ITCZ has completely dominated
the convection within the monsoon domain. The equatorial ITCS is suppressed and weak
subsidence appears over the equator.
6. Summary
Based on the above analyses and the examination of a video of GCM 3-hourly IR
images from 1978-1987 (courtesy of the Meteorological Research Institute of Japan), key
features of the East Asian summer monsoon are summarized in Fig. 8.
DRY 2O
0.
\
Q. 0-6
n v
-30 -IS
LRTITUOE
...ft^T/..,
, ! i; i ,
% . „ ... / / • / ^' / '
Figure 7. Vertical streamline pattern showing the vertical structure of the local Hadley
circulation before (Day 6), during (Day 6) and after (Day 12) the passage of the equatorial
intraseasonal disturbances.

103
Monsoon cyclones
/-~N I Subtropical High
^O
30-60 day oscillations
/-"\
double cyclones
Supercloud clusters
Figure 8 Schematic showing the relative motion of the monsoon disturbances, 30-60 day
oscillations and West Pacific disturbances with respect to the monsoon heat source and the
Subtropical High. The directions of motion of the synoptic scale disturbances, the 30-60
day oscillations are indicated by dark and shaded arrows respectively. Thin arrows indicate
the direction of the mean flow.
The key features are :
(1) The East Asian monsoon rainfall variation is closely linked to the planetary scale
30-60 day oscillations. The onset of the Mei-yu in southern and central China
may be related to the arrival of the first wave of organized convection/rainfall
associated with the 30-60 day oscillations in June. However, there are regional
scale features, mostly high frequency fluctuations, that appear to be quite
independent of the 30-60 day oscillations.
(2) There appear to be both an east-west and a north-south seesaw between the
Indian Ocean and the Western Pacific, and between the equator and the monsoon
region (15-20°N) in the 30-60 day time scales.
(3) The monsoon ITCZ is composed of westward propagating monsoon cyclones that
have periodicity of four to five days and wavelength of 1000-2000 km. They are
likely to be unstable heat-induced Rossby waves,
(4) Over the equatorial Western Pacific, superclusters associated with the 30-60 day
oscillations may give rise to westward propagating synoptic scales clusters. Some

104
of these clusters may develop into tropical cyclones that reach the South China
Sea and then follows a northeastward path to the North Pacific through Japan.
Others may also propagate into the monsoon region and subsequently developed
into monsoon depressions.
Reference
Lau K. M. and M. T. Li, 1984: The monsoon of East Asia and its global associations. Bull.
Am. Meteor. Soc, 65,114-125.
Lau, K. M and G. J. Young, and S. Shen, 1988: Seasonal and intraseasonal climatology of the
East Asian summer monsoon. Mon. Wea. Rev., 116,18-37.
Lau, K. M and P. H. Chan, 1986: Aspects of the 40-50 day oscillation during the norther
summer as inferred from outgoing longwave radiation. Mon. Wea. Rev., 114, 1354-
1367.
Lau, K. M., L. Peng, C. H. Sui and T. Nakazawa, 1989: Dynamics of westerly wind burst,
supercloud cluster, 30-60 day oscillation and ENSO : a unified view. J. Meteor. Soc.
Japan (in press)
Lau K, M. and L. Peng, 1989: Origin of low frequency (intraseasonal) oscillations in the
tropical atmosphere. Part III: Monsoon dynamics. J. Atmos. ScL, (to appear).
Murakami, T., T. Nakazawa and J. He, 1984: On the 40-50 day oscillations during the 1979
northern hemisphere summer, Part I: Phase propagation. J. Meteor. Soc. Japan, 42,
1107-1122.
Nakazawa, T., 1988: Tropical super clusters within intraseasonal variations over the
western Pacific. J. Meteor. Soc. Japan, 66,823-839.
Nakazawa, T and K. M. Lau, 1989: Super cloud clusters over the western Pacific from CMS
data. Proceedings of the 18th Conference on hurricanes and tropical meteorology.
San Diego, CA, 127-128.

105
INFLUENCE OF VARIATIONS OF THE CIRCULATION SYSTEM OVER THE
SOUTH INDIAN OCEAN ON THE EAST ASIAN SUMMER MONSOON AND
THE NORTHERN HEMISPHERIC GENERAL CIRCULATION
OF THE ATMOSPHERE
Huang Shi-Song, Tang Ming-Min, Yang Xiu-Qun
(Department of Atmospheric Sciences; Nanjing University)
ABSTRACT
This paper aims mainly to investigate the influence of
variations of the South Indian subtropical high on the East Asian
summer monsoon and the Northern Hemispheric atmospheric circulation.
Observational studies show that a large number of circulation
systems in low latitudes, such as the subtropical highs, equatorial
westerlies, cross-equatorial currents and etc. exhibit an oscillatory
character in their intensities and positions during May through
September. The oscillation occurs first over the southern Indian Ocean
and then the one over the northwestern Pacific Ocean follows. The
teleconnection is evident and clear. The physical processes of
teleconnection are elucidated.
Based on the daily, pentad and monthly low and high levels
average stream fields, the 850 hpa air trajectories entering China
during the Mei-Yu period and the above results relating to the
teleconnection processes, the principal components of the summer
monsoon regime of East Asia are reviewed. It is pointed that, as far as
the low level concerned, the key components should be the NW Pacific
High and the Mascarene High. A new schematic model of the summer
monsoon regime is then proposed.
In order to verify the above findings, a global spectral
general circulation model is used to demonstrate the influence of
intensity changes of the Mascarene High on the variations of the
atmospheric general circulation. The model is truncated at wave number
15 and variables are represented at nine sigma levels in the vertical.
It is found that the simulational results are quite consistent with the
observations. Further, an anomalous intensification of the Mascarene
High will also have a substantial contribution to the development of
global low frequency fluctuations of atmosphere. Three wave trains can
be found, one in the Southern Hemisphere and two in the Northern
Hemisphere.
So we may conclude that, during the northern summer, the
Mascarene High over the South Indian Ocean acts as a key system in the
interaction between the Northern and Southern Hemispheres, ft is an
indispensable constituent of the summer monsoon regime of East Asia. It
influences the variation of general circulation of the atmosphere far
into the high latitudes.

106
I . INTRODUCTION
The structure of the East Asian summer monsoon regime is of
primary importance to the rainfall during rainy season of China. Tao
and Chen(1987) considered,, as far as the low leyel is concerned, the
Australian high to be the key component system of the East Asian summer
monsoon regime and excluded the Mascarene high from it. In this paper
it is argued that the key components should be the Mascarene high and
the NW Pacific high. A new schematic model of the summer monsoon regime
is then proposed.
The periodic variation of the atmospheric circulation with time
scales larger than a week at middle latitudes, namely, the cyclic
change of the westerlies index, has been described as early as in
forties (Starr 1942, Namias 1947). But for low latitudes, it was not
until the seventies, the 40-50 day oscillations of sea level pressure
and wind over the equatorial region was firstly noticed (Madden and
Julian 1971,1972). Thenceforth, many investigations have been
undertaken (e.g. Krishnamurti et. al., 1976, 1982/ 1985; Yasumari,
1980, 1981^Murakami and Nakazawa, 1985, and Lorence 1984). All these
investigations were mainly based on spectral analsis. This paper is
involved in investigating the substantial variations in intensity
and/or position of the circulation systems in low latitudes, such as
the equatorial westerlies, cross-equatorial currents, the NW Pacific
subtropical high, the Mascarene high and the Australian high. The
teleconnection between the variations of circulation systems over the
South Indian Ocean and the Northwest Pacific Ocean and its physical
processes are elucidated.
In order to verify the above findings, a global speatral
general circulation model is used to demonstrate the influence of
intensity changes of the Mascarene high on the variations of the
atmospheric general circulation. It is found that the experiaent
results are quite consistent with the observations. Further, an
anomalous intensification of Mascarene high will also have a
substantial contribution to the development of global low frequency
fluctuations of atmosphere. Three wave trains can be found, one in the
Sounthern Hemisphere and two in the Northern Hemisphere.
II. OBSERVATIONAL STUDY
(1) Structure of the Summer Monsoon Regime
The monsoon study is generally focused on the seasonal
variation of the low layer winds. But its activity should be considered
in accordance with two factors, the prevailing air currents and the
thermal properties of the air. In other words, the source regions
should be considered.
The summer monsoon of China usually develops at its culmination
in July each year. The monsoon currents penetrating into China can be
either the tropical maritime air from the NW Pacific Ocean or the one
from the southern Indian Ocean, passing over the Bay of Bengal. So the
monsoon can be southeasterly winds or southwesterly winds as shown in
Fig.Cl), and Fig.(2) protrays several air trajectories at the 850 hpa

surface during the rainy period of Huiahe River-Yellow River reaches.
They distinctly demonstrate the origin regions of the monsoon air.
The establishment of the low layer monsoon is attributed to the
cooperation of a lot of circulation systems, mainly they are the NW
Pacific subtropical high, the low pressure trough over northern
Indian-Bay of Bangal-South China Sea, the southern Indian Ocean
subtropical high (Mascarene high) and the Australian high.
It is found that the tropical maritime air on the south side of
the NVf Pacific high , appearing as a southeasterly current, begins to
influence the South China Sea in March, and then reaches Southern China
in April. The tropical maritime air from the Southern Hemisphere, after
.crossing the equator and turning into a southwesterly current, begins
to influence China in May. From then onwards, the southeasterly
currents or the southwesterly currents alternately appear over China,
but the southwesterly currents predominate. Meantime the southeasterly
currents from the NW Pacific Ocean also turn into southwesterly while
advancing on China (Tang and Huang 1984).
There are two main passages, one at around 4Q a -&0°E and another
at around 100* -110° E, through which the air from the Southern
Hemisphere crosses the equator and then turns to east, flowing into
East Asia. The southerly current passing through 4Q S) ~6Q' E, called the
Somalian cross-equatorial current which is linked with the activity of
Mascarene high, is much stronger than the cross-equatorial current at
100° -110 a E which is linked with the activity of the Australian high
(Tang Huang and Zhou, 1985). But, as discussed in the next section, the
development of Mascarene high usually takes place earlier than the
development of the Australian high. Only when the Somalian
cross-equatorial currents intensify and the equatorial westerlies over
northern Indian Ocean strengthen, then the cross-equatorial current at
100 a -IIO^E intensifies and turns to east, joining the former one. The
southwesterly currents entering East Asia are mainly associated with
the Soaatian cross-equatorial currents.
Since the summer monsoon of China originates from the tropical
oceans, the air is warm and moist. The 8se value of monsoon air over
China is found to be that fi»>344 (or 348) K on the 1000 hpa level and
0$e > 336 (or 340) -K on the 850 hpa level. If taking the isoline of
&$e -348.K as the boundary of the summer monsoon on 1000 hpa surface
and putting the boundaries of each day (or each jjentad) on one chart,
we can obtain a chart showing the trends of monsoon's advance and
retreat processes (Tang and Huang, 1984).
The isolone of &se = 336K in Fig.(3) can be considered as the
temporal variation of the boundary position of the summer monsoon along
the 120° E meridian. It can be clearly seen that the latitudinal
fluctuation of monsoon boundary over China almost marches in step with
the displacement of the NI Pacific high ridge, but with a lag of about
one pentad. When the NI Pacific high moves northward, the southeasterly
currents on the south of the high wiIt strengthen, broaden and turn
into southwesterly currents. As the SI or SE winds to the south of
boundary increases, the monsoon pushes rapidly northward.
107

108
But, as discussed in the next section, the displacement of the
NW Pacific high ridge is closely teleconnected with the activity of
the Mascarence high. Therefore, the activity of Mascarene high seems
to play a leading role in the function of East Asian summer monsoon
regime. Any idea of the summer monsoon regime in which the Mascarene
high is excluded is certainly questionable. Accordingly, a schematic
modle to consolidate the structure characteristic's of the East Asian
summer monsoon regime can be constructed as shown in Fig.4(a). The
principal components of the monsoon regime in the low levels of the
troposphere are the NW Pacific High, Mascarene high, Australian high
and the Indian monsoon depression. The first two subtropical highs are
of primary importance. If we consider the three dimensional integrated
structure of the summer monsoon regime, we can construct a block
diagram illustrating the interconnection and interaction properties
among all the components of the monsoon regime, see Fig.4(b).
When one is interested in the Mei-Yu problem, one must take
into consideration the activity of the cold air from the north.
However, it is shown by Fig.5 that the characteristics of the variation
of the pented mean latitude position of the subtropical high ridge over
NW Pacific in the years of flood summer (e.g. 1954 and 1980) are
entirely different from those in the years of drought summer (e.g. 198
and 1981). In addition, we find froia Table 1 that during the flood year
(e.g. 1980 and 1983) the summer average moisture flux across the
equator between 40 P E and 70°E and the flux across the 20°N circle
between 3Q*E and 110°E are all much larger than the corresponding flux
during the rather drought years (e.g. 1981 and 1982). Therefore, the
summer precipitation in the middle-lower reaches of Yangtze River seems
still to be determined, to a great extent, by the influence of the
activity of the NVf Pacific subtropical high and the Mascarene high.
(2) Fluctuation of the Circulation Systems
It has been shown that a large number of circulation systems in
low latitudes, such as the subtropical highs, equatorial westerlies,
cross-equatorial currents and etc. exhibit an oscillatory character in
the variation of intensity and position during the months May through
September. The oscillation period may range from 10 to 50 days. Certain
20-40 day oscillations occur almost in the same period of time in both
hemispheres, but it occurs first over the southern Indian Ocean and
then the one over the north-western Pacific Ocean follows (Huang and
Tang 1982, 1987).
From Fig.(6), in which the curves a, b, c and d represent
respectively the pentad-by.-pentad variations of the intensity of the
Mascarene high, the intensity of the Australian high, the speed of
cross-equatorial current at (0*N, 105*E) and the westerly speed at (10*
N, 115*E), we find that the Australian high oscillates with a phase lag
of about 1-2 pentads relative to Mascarene high, but with a lead of
about 1-2 pentads over the V and U oscillations. As strong westerlies
prevail on the South of the Mascarene high and the Australian high,
so the intensification of the Mascarene high may, through the energy
dispersion, cause intensification of Australian high which is just

located at the position of the first down-stream ridge of the
quasi-stationary long wave. Intensification of the Australian high then
gives rise to strengthening of the cross-equatorial current at about
105°E, which turns to east and speeds up the equatorial westerlies. And
from Fig.(7) in which the curves a, b, c, d and e represent the
pentad-by-pentad variations of the intensity of the Mascarene high, the
speed of the Somatian cross-equatorial currents, the speed of the
equatorial westerlies, the longitudinal position of boundary of the
equatorial westerlies and the latitudinal position of the NW Pacific
high ridge respectively, we find that there exist one-to-one
corresponding oscillations, as denoted by ordinal numbers 1, I, 3, 4
and 5, with period of about 4-6 pentads. In general, the Mascarene high
intensifies first, then the following sequence occurs, the
strengthening of the Somalian cross-equatorial currents, the
strengthening of the equatorial westerlies and the eastward extension
of its boundary. And finally the rapid advancing northward of the NW
Pacific high takes place. The physical processes of teleconnection
can be elucidated as follows by a case in July 1979.
The solid curves in Fig.(8) represent the 3-day averaged
position of the ridge line of NW Pacific high for 13-15, 16-18, 19-21,
22-24, 25-27 and 28-30-31 of July, which are respectively denoted by
the number 5, 6, 1, 9 and 10. Each small circle on the ridge line
indicates the place at which a high center is located. So Fig.(8) shows
that the western part of the NW Pacific high ridge line moved to north
most rapidly during the period from 16-18 to 25-27 of July (i.e. from 6
to 9 in ordinal number) and new high centers developed to the south of
Japan while the old centers disappeared.
The corresponding wind fields at 850 hpa level are shown
in Fig.(9). On July 19-21, the anticyclonic circulation over South
Indian Ocean was rather strong, the Somalian cross-equatorial currents
and the equatorial westerlies were quite distinct, and the tropical
convergence zone between the equatorial westerlies and the easterly
trades was located at around H0*E. On July 22-24 and 25~27, the
tropical convergence zone was pushed eastward to about 135*E or even to
about ISO^E and moves northward as well due to steady strengthening and
further eastward extension of the equatorial westerlies which resulted
froa the strengthening and developing of the cros^-equatorial currents
at around 40° -60° E and 105*-120 P £. On July 28-31, the tropical
convergence zone got further development.
Because of intensification of the horizonal convergence between
the lower layer's equatorial westerlies and easterly trades, strong
ascending motion and upper-level horizontal divergence are produced.
Fig.(10) displays the variation of the horizontal velecity divergence
fields on 200 hpa surface over the South China Sea and Ni Pacific
region. The intensity of divergence on 28-31 chart increased to three
times as that on 16-18 chart, and its center moved quite a distance to
northeast. At this time, the associated strong ascending motion then
leaded to developing of Hadley circulation as depicted in Fig.(11). It
can be seen that during the period 16-18 and 19-21, the Hadley cell was
located south of 15°N and 20°N respectively, During 22-24, the Hadley
109

110
cell extended northward to 30 N. And during 28-31, the Hadley cell as a
whole moved to north, its descending brench extended to north of 3G°N.
Following the establishment of Hadley circulation, the NW Pacific
subtropical high was bounded to abjust itself properly in position and
intensity. The ridge, line moved rapidly northward and new centers
formed while some old centers disappeared.
In addition,, it is also pointed out that*the seasonal variation
of the latitudinal position of the westerly jet of temperate latitudes
is closely associated with the subtropical high ridge, but with a phase
lag (Huang and Tang 1962). This phenomenon is extremely prominent in
1979 too. Therefore we may consider that the intensity change of the
Mascarene high can affect the variation of the mid-latitude atmospheric
circulation over the Asia-Pacific region. Recently Huang and Li(l987)
and Nitta (1987) find that the response to strong convective heating
over the Fhillipine Sea region can induce low frequency wave train as
two dimensional Rossby wave propagating to high latitudes, through the
northern part of Pacific Ocean to eastern United States. We consider
that the strong convective heating over the Phitlipine Sea is caused by
the intensification of horizontal velocity convergence which results
from the strengthening of the equatorial westerlies and its eastward
extention into the NW Pacific. Therefore, the formation of the low
frequency wave train should be finally attributed to the intensity
variation of the Mascarene high.
I. NUMERICAL EXPERIMENT
In order to verify the above findings, a global spectral
general circulation model, designed by Bourke (1977) and improved by
Simmonds (1985) and Lin (1987), is used to demonstrate the effect of
intensity changes of the Mascarene high on the variations of the
atmospheric general circulation. The model is truncated at wave number
15 and variables are represented at nine sigma levels in the vertical.
Two experiments, a control one (Exp.l) and an anomaly one (Exp,2) are
performed. The climatological normals for SST, water vapor, C0 2/ 0$ /
snow cover and polar ice of July are used, and the sun's altitude is
fixed at the mid July position. In the anomaly experiment, an anomaly
geopotential height, with maximum value of 100 gpm centered at 35 a S,
60° E, is superimposed on the Mascarene high at 850 hpa level. The
initial field is the one of a certain day extracted from the results of
integrating the model for hundreds of days. The 850 hpa initial field
and the geopotential height anomaly field are respectively shown in
Fig.(12) and Fig.(13). Since it is found that the global averaged
kinetic energy at 850 hpa becomes steady just only one day after
integration from the initial state for Exp.2, analysis of the
Integrated results can start with the first three-day average field.
Fig, 14 (a) and (b) show the 850 hpa wind anomaly fields of the
first and the second three-day period respectively. A strong
anticyclonic circulation exists over the region where the Macarene high
is located. It then results In a strengthening of the Somalian
cross-equatorial currents and formation of a cyclonic vortex over the
central South Indian Ocean and an anticyclonic vortex over the coastal

egion of western Australia. These two vortices, especcially the
anticyclone, develop further in the second three-day period. At the
same time, several vortices also appear in the Northern Hemisphere^
cyclonic vortices over Arabian Sea, Indo-China peninsula and
southeastern part of the South China Sea, and anticyclonic vortices
over India and Western Pacific Ocean. This situation is entirely in
accord with the precipitation distribution depicted in Fig.15. A
well-developed Hadley circulation cell over the northwestern Pacific
Ocean is clearly shown in, for example, Fig.16. Evidently the
experiment is extremely consistant with the observational studyCYang
and Huang, 1989).
Besides, an anomalous intensification of the Mascarene high
will also make a subtantiat contribution to the development of global
low frequency fluatuations of atmosphere. As shown in Fig. 17 and 18,
three wave trains can be found. The first one is a zonal wave train of
k-3
to
in the Southern Hmisphere. The second is the one with a tow center
south of Japan, a high center over Okhotsk Sea, a low center over
North Pacific, a high center over alaska and a low center over the
southern part of North America. This wave train is just the one
described by Huang and Li (198) and Nitta (1987). The third is the one
with a low center over North Atlantic, a high center over Europe and a
low center over Asia.
IV. CONCLUSION
From above investigation, we may conclude that, during the
northern summer, the Mascarene high over South Indian Ocean seems to
act as a key system in the interaction between the Northern and
Southern Hemispheres. It is an indispensable constituent of the summer
monsoon regime of East Asia. It can effect the variation of the general
circulation of the atmosphere far into the high latitudes.
111

112
References
Bourke, W., McAvaney, B., Puri, K. and R. Thurling (1977). Global
Modeling of atmospheric flow by spectral methods. In "Methods in
Computational physics", Academic PPress.
Huang, R. H. and W. Li, (1987), Collected Papers of the Internationnal
Conference on the General Circulation of East Asia. Chengdu,
China. 40-51.
Huang, S. S. and M. M. Tang (1962), Jour, of Nanjing University
(Meteorology). No. 2, 41-56. (in Chinese with English abstract)
Huang, S. S. and M. M. Tang/1982), Jour, of Nanjing University
(Meteorology). 1-16. (in Chinese with English abstract)
Huang, S. S. and M. M. Tang (1987), SCIENTIA METEOR. SINICA, No. 3.
1-14. (in Chinese with English abstract)
Krishnamurti, T. N. and H, N. Bhalme (1976). Jour. Atmos. Sci., 33,
1937-1954.
fCrishnanmrti, T. N. and D. Subrahmanyem (1982), Jour. Atmos. Sci.,
39, 2088-2095.
Krishnaraurti, T. N. and S. Gadgilt (1985), Tellus 37A, 336-360.
Lin, Y. B. (1987), "Experiment of General Circulation of Atmosphere,
Department of Atmospheric Sciences, Nanjing University", pp.14.
(in Chinese)
Lome, A.C. (1984), Tech. Notes 11/210, Meteor. Office, Bracknell,
Berkshire, England.
Madden, R. A. and P. R. Julian (1971), Jour. Atmos. Sci. 28, 702-708.
Madden, R. A. and P. R. Julian (1972), Jour . Atmos. Sci. 29, 1109-1123.
Murakami, T., Nakazawa, T. and J. He (1984), Jour. Meteor. Soc. Japan,
63, 250-271.
Murakami, T, and T. Nekazawa, (1985), Jour. Atrnos. Sci. 42, 1107-1112.
Namias, J. (1947), Extended Forecasting by mean circulation method,
Washington D. C., U. S. Department of Commerce, leather Bureau.
PP. 89.
Nitta, T. (1987), Collected Paper of the International Conference on
the General Circulation of East Asia, Chengdu, China, 121-126.
Simmonds, I. (1985), Jour. Geoph. Rev., 90, 5637-5660.
Starr, V. P. (1942), Basic Principles of Weather Forecasting, Harper
Brothers Publishers, pp. 299.
Tang, M. M. and S. S. Huang (1984). Proc. of the Symp. on the Summer
Monsoon in Southeast Asia, People's Press of Yunnan Province,
15-29. (in Chinese with English abstract)
Tang, M. M. Huang, S. S. and S. E. Lu (1984), Proc. of Conf. on
Tropical Circulation and Systems, China Ocean Press, 81-94. (in
Chinese with English abstract)
Tang, M. M., Huang, S, S. and T. P. Zhou (1985), Jour, of Tropical
Meteorology, I, 287-296. (in Chinese with English abstract)
Tao, S, Y. and L, X, Chen (1987), A review of recent research on the
East Asia summer monsoon in China, 3rd Chapter in Reviews in
Monsoon Meteorology, Oxford University Press.
Yang, X. Q. and S. S. Huang (1989), SCIENTICA METEOR. SINICA, No.2,
125-138. (in Chinese with English abstract)
Yasunari, T; (1980). Jour. Meteor. Soc. Japan, 58, 225~229.
Yasunari, T. (1981). Jour. Meteor. Soc. Japan, 59, 336-354.

115
Fig.^57 Characteristics of variation of ihe pentad mean latitude >.iiion of the subtropical
high ridff over Hie Northwest Pacific O2S*-140*£) on ihe 500 i» level during ilic period
from April through October.
Curve N: 31 years (1954-1984) average pentad mean latitude position (normal).
Other curves: Departure tUt.) from normal of the pentad meat, latitude position for four
sample years, summer Hood years (1954, 1980) and drought years -,1978. 198!) in the middlelower
reaches of the Changjiang Rivet
»!«« (6) T*ot«d-by-j>»nt*

^
- ^^KiKS^;--^Sxy^^
. ^ &^>sx^>^.--
Fig. {15} Precipitation anomaly distribution for the
second 3-day period.
y i«.'-*•>•«•- «.. » >
4 * *
f V I *-y;.». * : |
Fig, (18;
Anomaly meridional oirculation. on the
iao°-lE8°E mean cross-section for the
second 3-day period.
SOU 30U OH
geopotential height anomaly distribution for

119
Some Dynamic Aspects of the Equatorial
Intraseasonal Oscillations
Bin Wang and Hualan Rui
Department of Meteorology, University of Hawaii
ABSTRACT
The slowly eastward-moving equatorial convection and
circulation anomaly is a major mode of the tropical
intraseasonal oscillations.
The dynamics of this mode may
be understood in terms of moist equatorial wave dynamics.
A linear semi-geostrophic model on equatorial Beta~pl& ne is
used to study the behavior of equatorial low-frequency
motions.
The unstable interaction of boundary layer frictional
moisture convergence with condensational heating could
generate efficiently eddy available potential energy for
moist Kelvin mode but not for long Rossby modes.
growing mode is rooted in a moist Kelvin wave but modified
through
The
coupling with a long Rossby wave of the lowest
meridional index.
The horizontal mode coupling makes the growing mode
have a horizontal structure bearing similarity "to both
Kelvin and Rossby modes.
It favors the amplification of
long planetary scales and slows down the eastward movement.
It also suppresses unrealistically fast growthof the
uncoupled Kelvin waves by creating substantial meridional
flows which induce kinetic energy destruction.

120
The model results also demonstrate that when maximum
SST moves from the equator to 7.5°N, the growth rate of the
unstable wave is significantly reduced, suggesting that the
annual march of the "thermal equator" and associated
convective heating is likely responsible for annual
variations of the equatorial intraseasonal wave activities.
1. INTRODUCTION
The slowly eastward moving equatorial circulation
anomaly first described by Madden and Julian (1972) is
believed to be a dominant mode of tropical intraseasonal
variations. In the eastern hemisphere, the circulation
anomalies are accompanied by convection anomalies, migrating
due east at a speed of about 5 ms" 1 (e.g., Lau and Chan,
1985; Weickmann et al. , 1985; Murakami et al., 1986). After
crossing the date line, the convection anomalies decay and
emanate away from the equator towards North America and the
southeastern Pacific (Rui and WAng, 1989), while the
circulation anomalies can transverse the equatorial western
hemisphere at a much faster speed of about 15-20 ms~ 1 (Knutson
et al., 1986).
Numerical studies which rule out topographic and
thermal land-sea contrast effects demonstrate that the
precipitational heating interacting with equatorial wave
motions may maintain long lasting transient planetary scale
disturbances with eastward propagation speed of 10-20 ms" 1
(Hayashi and Sumi, 1986; Lau and Peng, 1987; Swinbank et
al., 1987; Lau et al., 1988).
Attempts have been made to understand the oscillations
in terms of moist equatorial wave dynamics. A number of
recent theoretical model studies have provided useful
insights into moist Kelvin wave dynamics (e.g., Lau and
Shen, 1988; Chang and Lim, 1988; Wang, 1988) . Yet these
analyses are confined to two-dimensional motion in the

121
equatorial zonal plane. In this paper, we augment the moist
Kelvin wave model by including meridional wind component,
boundary layer frictional effect, and latitude-dependent SST
for the basic state.
It allows investigation of the impacts
of horizontal-mode coupling and annual march of SST on low
frequency equatorial wave dynamics.
2. SIMPLIFIED DYNAMIC FRAMEWORK
Consider perturbation motion on an equatorial Betaplane.
The primitive equation in p-coordinates are

122
The diabatic heating rate Q includes only longwave radiation
cooling Qi=M (c p p/R) 8/dp, and condensational heating Q 2 which
is related to precipitation rate P r by
Ps
J Q2^T = LcPr • (5)
y
o
A crucial assumption for moisture conservation equation is
that the precipitation rate is linearly balanced by the
moisture convergence. This leads to
3. MODEL EQUATIONS
Ps
Ps
J Q2(P)

123
where N is nondimensional Newtonian cooling coefficient, S
measures mean static stability of the basic state; I and B
represent ratios of latent heating to adiabatic cooling due
to vertical motion at P2 and p e/ respectively. We have used
c 0 / (c 0 /B) 1/2 , (6c 0 )~ 1/2 , GO, and 2Ap(Bc 0 ) 1/2 as horizontal
velocity, length, time, geopotential, and vertical p-
velocity scales, respectively, where c 0 is long gravity wave
speed in 2-level model.
The boundary layer friction-induced vertical velocity
OJ e can be expressed in terms of geopotential at p e and 0 e
(Wang and Chen, 1988):
^2 A S (E 2 + 2 ) ^"^271^EVy a x)
where E=p e gAz/ [ (p s -p e ) /Bc 0 -h- In(h/z 0 ) ] is the Ekman number
determined by turbulent viscosity A z , depth of the surface
layer h, the surface roughness length z 0 and density p c .
The upper boundary and lateral boundary conditions are
c

124
MODE SELECTION (WL=200QQ KM) MODE SELECTION (WL=2QOQO KM)
. -0.3
X
a
O -0.4
-0.5
12 H JB 18 20 22 24 26 28 30
SST (°C1
(a)
10 12 14 16 18 20 22 24 26 28 30
SST (°C)
(b)
Fig. 1, (a) Phase speed and (b) growth rate of the moist Kelvin mode (K),
m=*l Rossby mode (Rl), and m=2 Rossby mode (R2) as functions of SSTM. The
wavelength is 20,000 km.

125
moisture concentration gradually increases, the moist Kelvin
mode becomes progressively less damped and finally begins to
grow when SST exceeds an critical value. On the other hand,
moist Rossby waves are always damped.
The unstable mode
selection in the present model can be explained in terms of
wave energy generation due to latent heating induced by the
boundary layer frictional moisture convergence.
Because of
the large concentration of moisture in the boundary layer,
the latent heating associated with frictional moisture
convergence produces a substantial portion (about 1/3) of
the wave energy.
Since the availability of basic-state
moist static energy is highest at equator, the equatorial
warm region is most conducive (or destructive) for wave
energy generation.
In this region, both Kelvin and Rossby
wave-induced boundary layer convergences reach their maxima
(Figures not show) ; however, the frictional upward motion is
positively correlated with temperature in the Kelvin mode,
while the negative covariance between them is found for the
Rossby mode. Thus, available potential energy is generated
efficiently by the frictional convergence-induced latent
heating in the moist Kelvin mode, but is destroyed in the
moist Rossby modes.
5. THE EFFECTS OF THE HORIZONTAL MODE COUPLING
Although unstable modes appear to be rooted in Kelvin
waves, they are modified by the dynamic coupling with Rossby
waves. The horizontal structures of, the zonal wind and
geopotential of the unstable mode shown in Fig.2 resemble
those , of Kelvin waves but exhibit significant meridional
wind components, which are asymmetric about the equator and
similar to those of the lowest meridional mode of Rossby
wave.
This suggests that the unstable mode is in nature a
moist Kelvin mode modified through coupling with a long
Rossby wave.
The coupling between moist Kelvin and Rossby modes via

126
(a)
OEOPOTEHTIflL FRED
(A]
VEKT1CRL MOTION FIELD AT HlOTKOPOSfHEKE
W..10000 Htt. MT.M.S *C
MtitfOOOO HK. tSUri.K *C
B.w/o.O
ZONAL CURSE
tDECREE)
(b)
ZONRL MIND FIELD
(t) VERTICflL MOTION FILED »1
HL*DOOO KM. SST=«.S *t
W-tlOOOO KR. SSUZS.S *C
£ «
%
I.,
ZONAL fHftSE I DECREE I
ZONAL CMBSE fDECREE I
Fig. 2. Horizontal structure of the
unstable wavenuniber two at SSIM=29.5°C:
(a) geopotential ., (b) zonal wind,
u , (c) meridional wind, v., (d) 0)2>
and (e) w e . Solid and dotted lines
denote positive (or zero) and negative
contours, respectively. Contour intervals
for 4>_ and u. are 10% of the
maximum value and those for v_, o} 2 , and
o) e are 201 of the maximum value.

127
boundary layer frictional convergence and associated latent
heating has fundamental impacts on wave instability. Fig.
3 compares growth rates and phase speeds computed from noncoupled
moist Kelvin model (dotted) and moist coupled
Kelvin-Rossby models under constant SST (dashed-dotted) and
a latetude-dependent SST (solid) . It is seen that the
horizontal mode-coupling acts as an efficient brake to
reduce the eastward propagation speed (Fig.3b), suppresses
unrealistically fast growth of the uncoupled moist Kelvin
mode by creating significant meridional flows (Fig.Sa).
More importantly it favors the amplification of long
planetary waves, rather than short waves, providing a
longwave selection mechanism (Fig. 3a).
6. THE SEASONALITY OF THE INTRASEASONAL OSCILLATION
By analyzing 19 near equatorial station rawinsonde
data, Madden (1986) showed that the intraseasonal
variability in zonal wind exceeds that in adjacent lower and
higher frequency bands by the largest amount during
December, January, and February. Using ten years of
outgoing longwave radiation (OLR) data, we investigated
temporal variations of 77 low frequency equatorial eastwardmoving
events and found that the overall intensity of
intraseasonal convective anomalies are significantly
stronger in boreal winter (from November to April) than in
boreal summer (Wang and Rui, 1989).
The annual variation of equatorial intraseasonal wave
activity may be caused by the annual march of the "thermal"
equator where the highest SST is located. This notion is
confirmed by an experiment in which the maximum SST is
shifted to 7.5° N, a situation occurring during boreal
summer. In this case, the growth rate of the unstable
coupled Kelvin-Rossby mode is substantially reduced (Fig.4).
The separation of the thermal effect of warm ocean water

128
CQMPRRISON OF GROWTH RflTE
KELVIN WflVE CSST=29.Q °C)
0.40
•
COMPARISON OF PHflSE SPEED
KELVIN HftVE (SST=29.0 °C)
35
0.35
0.30
\
\
25
0.25
0.20
\ x
**"*••
•
- 20
K.
O
0*15
0.10
0.05
_______^ —-^~
15
10
V
\
0*00
_---•** — ""*"""
-0.05
SO 15 20 25 30 35 40
HfiVELENGTh J10 3 KM)
(a)
10 15 20 25 30 35 40
WflVELENOTH U0 3 Ktt)
(b)
same for three cases.
'

129
000 -
20 30
WAVELENGTH (10 3 KM)
(a)
20 30
WAVRENCTH (10* KM}-
(b)
Fig. 4. Comparison of (a) growth rate and (b) zonal phase speed of the
unstable mode computed using SST profile with maximum at the equator (solid
lines) and a profile with maximum at 7.5°N (dashed lines). In both cases
the maximum SST is 30°C.

130
from the dynamic effect of the equator stabilizes the
tropical atmosphere for planetary scale low frequency
disturbances.
REFERENCE
Chang, C.-P. and Lim, H., J. Atmos. Sci., 45, 1709 (1988).
Hayashi, Y. Y. and Sumi, A., J. Meteor. Soc. JApan, 6.4,
451(1986).
Knutson, R. and Weickmann, K. M., J. E. Kutzbach, Mon. Wea.
Rev, 114, 605 (1986).
Lau, K.-M. and Chan, H., Mon. Wea. Rev., 113, 1889 (1985).
Lau, K.-M. and Peng, L. v J. Atmos. Sci., 44, 950 (1987).
Lau, K.-M. and Shen, J., J. Atmos. Sci., 45, 1781 (1988).
Lau, N.-C. and Held, I. M. , J. D. Neelin, J. Atmos. Sci.,
45, 3810 (1988).
Madden, R. A., J. Atmos. Sci., 43, 3138 (1986).
Madden, R. A., P. R. Julian, J. Atmos. Sci., 29., 1109
(1972) .
Murakami, T. and Chen, L.-X., A. Xie, M. L. Shrestha, J.
Atmos. Sci., 43, 961 (1986).
Rui, H. and wang, B., J. Atmos.Sci., (in press) (1990).
Swinbank, R. and Palmer, T. N., M. K. Davey, J. Atmos. Sci.,
15, 774 (1987) .
Wang, B. , J. Atmos. Sci., 45, 2051 (1988),
Wang, B. and Chen, J.-K., Quart. J. Roy. Met. Soc. (in
press) (1989) .
Wang, B. and Rui, H. , Meteor. Atmos. Phys. (in press)
(1990).
Weickmann, K. M. and Lussky, G. R., J. E. Kutzbach, Mon.
Wea. Rev., 112, 941 (1985).

131
CHINESE POLAR ORBITING METEOROLOGICAL SATELLITE FY-1
XU JIANMIN
Satellite Meteorological Center,
State Meteorological Administration ,
Beijing.
1. INTRODUCTION
China successfully launched its polar orbiting satellite, named
FY-1, from a satellite launching Center in Taiyuan, Shanxi province,at
5:30 Beijing Summer time on 7th September, 1988. The satellite is a
three-axis stabilized polar orbiting platform in a sun-synchronous orbit.
It is the first Chinese experimental and trial meteorological satellite.
The principal payload of the satellite is a multi-spectral scanning
radiometer. This provides the basic data of the FY-1 system as
three visible channels, one near-infra-red channel and one infra-red
channel product images. Various image products, land and ocean digital
remote sensing products are derived through the ground data-receiving
and processing system. The communication equipment aboard the satellite
in real-time broadcasts High Resolution Picture Transmission (HRPT) and
Automatic Picture Transmission (APT) data over the world. The data
format of HRPT and APT are compatible with the NOAA satellite.
2. THE SATELLITE
2.1 Orbit
FY-1 satellite is in a sun-synchronous orbit. The orbit parameters
are:

132
Altitude:
Inclination:
Period:
Eccentricity:
900 Km
99°
102 minuts
ejm
0.53 - 0.58 jam
10.5 - 12.5jjm
Scanning rate: 360/minute
Resolution: HRPT 1.1 Km (at SSP)
APT 4 Km (uniform)
2.3 Image Dissemination
The characteristics of image dissemination are as follows:
HRPT/APT Transmission Channel Characteristics
carrier frequency code rate modulation code power
HRPT * MHz 0.6654Mbps PCM/PSK 5W
APT
"oasMfe
2 " 4 mz AM / FM 8W
3. GROUND RECEIVING AND PROCESSING SYSTEM
The ground receiving and processing system, which consists of
three ground stations located at Beijing, Guangzhou and Urumqi and the
Data Processing Center in Satellite Meteorological Center, undertook the
tasks of satellite data receiving, transmission, processing, products
distribution, archiving, scheduling, exploitation, some application and
research. This system is also capable of giving consideration to process
NOAA polar-orbiting satellite and GMS S - VISSR data. Besides,

some middle and small receiving station inside China can receive and
use HKPT and APT data from FY-1 satellite.
The Data Processing Center is located at the Satellite Meteorology
Center (SMC), State Meteorological Administration (SMA), Beijing. The
data received in Guangzhou and Urumqi ground stations are relayed to
Beijing Data Processing Center via an international communication satellite
while the data received in Beijing ground station are relayed to the
Beijing Data Processing Center through microwave circuits. Figure 1 outlines
the FY-1 system.
Each ground station is equipped with an S-band antenna and UHF
antenna receiving system, IBM Series/1, PC/AT computer and other equipment.
It accomplishes the tasks of tracing, reception, compilation,
storage, monitoring and dissemination.
The Data Processing Center is composed of a multiprocessor system
which includes communication computers, communication controllers, three
main computers (IBM 4381 - P03) and corresponding I/O equipments, disc,
tape device, interactive picture processor etc. The computer system
with corresponding equipments are outlined in Figure 2.
The software of the system includes computer operating system,
I/O communication software, data processing and application software,
operation control and scheduling software, products dissemination software,
products archiving and retrieving software. The structure of these
softwares, is outlined in Figure 3. Uncfer the support of the hareware,
the system undertakes data input, pre-processing, image and meteorological
parameter processing, data archiving and dissemination etc.
133
4. EXPERIMENTAL PRODUCTS
Center:
Following products can be derived through Data Processing
4.1 Image Products
Single orbit image display in real - time
Single orbit stretched image
Local multi - spectral composite image

134
Local enhanced cloud image
Cloud analysis image
Polar projection and Mercator projection mosaic
4.2 Products of Meteorological Parameters
Sea surface temperature
Cloud top height and temperature
Typhoon analysis report
4.3 Vegetation index, snow cover, sea ice and other remote sensing
products.
Besides, this system also provides the information of satellite
orbital prediction and has the capability to exploit other products
from NOAA and CMS data.
5. SATELLITE DATA ARCHIVE, DISSEMINATION AND APPLICATION
5.1 Archive
In SMC, meteorological satellite data are archived in three
modes: tape, picture and microfilm.
The archiving tape is raw digital tape (recorded on 38000 bpl
tape decks), intermediate result tape, product tape in digital form
and videotape.
The picture has photograph (in paper) of image and graph, film
and microfilm. In order to preserve satellite picture dat for a long
time, various picture data are miniaturized in microfilms.
5.2 Dissemination
Same as the NOAAseries satellites, the FY-1 satellite broadcasts
HRPT and APT data in real - time. Ground stations can directly
receive the information and use the information/The nearby domestic
user can get meteorological satellite products by, line on line, conventional
analogue transmissions (WEFAX), micro - computer, interactive
terminal and other ways. Some image products are broadcast via television
and broadcast. Also these image and graph products can be sent to

135
local meteorology units through communication circuits. The satellite
orbit prediction information and some meteorological parameter products
are transmitted via GTS by the Beijing National Meteorology Center.
5.3 Application
The Satellite Meteorology Center has provided their products
for the native user in various ways. These products have played an
important role in meteorological and other departments. It has evidently
gained economic efficiency.
In order to enlarge the application fields of meteorology satellite
data and to improve application quality, SMC is also engaged in
experimental research work under the cooperation with other units.
These researchs have applications of satellite data in weather, numerical
forecast, climate, ocean, agriculture and disaster monitoring.
6. DEVELOPMENTAL PROJECTS OF POLAR ORBITING SATELLITE IN FUTURE
China puts the improvment of polar orbiting system into schedule
of 90 th developmental projects. The projects include:
Adding infrared channels.
For example, 3.5 - 4.0^um channel
10.5 - 12.5 urn split window channels.

139
APPLYING TROPOPAUSE DATA OBSERVED BY VHF RADAR TO IMPROVE
SATELLITE TEMPERATURE SOUNDING
Gin-Rong Liu
Center for Space and Remote Sensing Research
National Central University
ABSTRACT
In using satellite remote sensing data to inverse atmospheric
vertical temperature profiles, because of the limitation of channel
selection, there are relatively large errors in two layers. One is in
the surface, the other is in the tropopause. For the surface
temperature error, one can use the surface hourly report data observed
close to the satellite passing time to the retrieval algorithm to
improve it. But for the tropopause temperature error, so far no
technique is developed to correct it. The purpose of this study is to
develop a new technique to improve this basic difficulty of satellite
temperature soundings.
The normalized power of the received signal of VHF radar has very
high correlation with the atmospheric stability. So one can use VHF
radar data to find out the tropopause position accurately (Gage et.
al.. , 1986). Therefore from the climatological relationship of
tropopause temperature and tropopause height, the tropopause
temperature can be determined from the VHF radar derived tropopause
height. Applying these tropopause data to the "Simultaneous Physical
Retrieval Algorithm" (Smith et. al., 1985), more accurate satellite
temperature sounding can be expected.
The results obtained from this study are compared with the
radiosonde observations and sounding determinations by using NOAA
satellite TOVS data alone. The intercomparisons reveal improvements in
accuracy in atmospheric vertical temperature profile retrieval using
this new technique.

140
The Mesoscale Monitoring System of Rainstorms and Severe
Convective Weather Events in China
Zhou Xiuji
Ma Da-an
Ge Runsheng
Zhao Conglong
(Academy of Meteorological Science, Beijing)
Lu Daren
Lin Hai
(Institute of Atmospheric Physics, Beijing)
ABSTRACT
In this paper the development of weather radar, UHF
wind profile radar and microwave radiometer and its
application in China are described. The theoretical studies
on radiation transfer in atmosphere provide some new methods
to retrieve the temperature, humidity, column liquid water
in cloud and rainfall rate distribution. On the basis of
these results the mesoscale monitoring system of rainstorms
and severe convective weather events has been established in
Beijing Area.
Among numerous disastrous weather events in China,
heavy rain hailstorms, local strong winds and severe
thunderstorms mainly caused by small and mesoscale weather
systems are most destructive. During the last three
decades, although the capability and skill of large-scale
weather prediction have been brought to a high state of
development in China, the monitoring and forecasting the
small-and meso-scale weather systems producing major
damaging weather events has been improved only with limited
success. As the economic activities of all aspects rapidly
expand during the process of implementation of modernized
construction of China, the Chinese meteorologists already
made a plan 25 years ago to develop a warning system to
effectively monitor and predict the mesoscale disastrous
weather.
In 1985 a special Scientific Organizing and Advisory
Committee headed by the State Meteorological Administration
(SMA) was established to formulate a detail research plan
which is designated as the National Research and Operational
Program of the Monitoring System and the Very-short-range
prediction of the disastrous Weather in China (1986-1990)
[1]. According to this program four experimental areas have
been selected with high priorities for developing mesoscale
meteorology: Beijing-Tianjin-Hebei (Beijing Area), Yangtze
River delta area (Shanghai area), Pearl River delta area
(Guangzhou area) .and.Sanxia'or Three Gorges district on
Yangtze River (Wuhan area) (Fig.l) These areas cover roughly
160,000, 70,000, 60,000, and 50,000 km respectively.

The main objectives of the Program are:
In operation:
(1) to establish an operational disastrous weather
monitoring system,
(2) to establish an operational nowcasting and veryshort-range
forecasting system,
(3) to establish a warning system for rapid and timely
delivery of monitoring and forecasting products to different
users
İn research:
(1) to work out a conceptual model for mesoscale
weather system on the physical basis of observetions and
diagnostic analyses.
(2) to develop a mesoscale numerical prediction model.
(3) to develop new observing technologies, study the
optimized mesoscale observing network and method of
assimilation of data from this network.
(4) to carry out theroretical and fundamental studies
on mesoscale meteorology.
(I) The Experimental Area of Beijing
The size of the Experimental area of Beijing is
designed as 400km by 400km (Pig.2). In this area, the
effects of mountains in the western and northern part and
land-sea boundary in the eastern part are very important for
the occurrence and intensification of some significant
mesoscale weather events such as severe local storms and
heavy rain in late spring and summer. So, the main objective
of the program is to study hailstorms, localized rainstorms
and destructive strong winds caused by squall lines, lowlevel
shear line and upper-air vortex, and to establish a
monitoring and prediction system for 0-12 hours.
According to this plan, the first step which consists
of the following four subsystems will be completed in 1990.
In the following five years (1991-1995), these subsystems
will be extended and improved.
(1) Observing subsystem, including surface
observations, upper-air observations, weather radars,
lightning location and satellite soundings(Fig. 2).
The 20 automatic surface weather stations have been
deployed to constitute a mesoscale surface observation
network together with existing conventional stations.
Average spatial separation of the stations in this area is
approximately 30km. The network data of 6 surface
meteorological elements (temperature., -.humidity, 'pressure,
wind direction and speed, rainfall) is collected by VHF and
UHF radio communication every 10-30 minutes.
The upper-air sounding consists of 3 radiosonde
stations, a pibal stations and 3 profiler station. A UHF
Doppler radar and microwave radiometers for remotely sensing
wind, temperature humidity profiles and column liquid water
content in cloud have been established.
141

142
The weather radar network is composed of two kind of
radars which have two capabilities. First, three digitized
weather radars at wavelengths 5cm and 10cm are capable of
providing echo intensity information with a horizontal
spatial resolution of 1x1 km . Second, two Doppler weather
radars at wavelengths 5cm and 10cm can provide data of wind
and turbulence intensity.
The lightning location system consists of three
detection sites and one central processing station. This
system is capable of monitoring the activity of lightning
and determine its location in the experimental area.
The digited data from the NOAA polar-orbiting satellite
and the Japanese QMS are provided by the National
Meteorological Satellite Center of SMA for mesoscals
analysis and forecasting applications.
(2) Data collection, processing and computer-operated
communication subsystem. This subsystem is used to timely
collect and process all data from the observing subsystem
and to rapidly transmit and disseminate data and forecasting
products to each station, local forecasting center and other
special users. The high-speed communication techniques by
satellite, VHF and UHF radio and telephone line are used to
make the rapid transmission possible. The collection of all
kind of data in this network can be accomplished within 30-
40 minutes.
(3) Nowcasting, very-short-range forecasting and
warning dissemination subsystem. The main task of this
subsystem is to rapidly compile, analyse and display data
and information collected from the different sources. At the
same time, this subsystem can produce, integrate and display
a great variety of forecasting (including numerical
mesoscale prediction) and warning outputs that will be
disseminated to local centers and special users via the
high-speed communication network. To accomplish this goal,
the computer network, data bank, software packages and
plotting bank that can meet the need for high-speed
computation and processing have been developed. Of course,
in this subsystem it is neceessary to involve the diagnosis
of synoptic and dynamic quantities, the methods and software
packages of nowcasting and very-short-range forecasting, the
results of numerical prediction, conceptual forecasting
models, the expert system for forecasting and forecasting
flow charts.
(4) Research and development subsystem. Firstly, based
on the large amount of observed data, the law of formation,
movement, intensification and decay of local and regional
disastrous weather phenomena and their three dimensional
structure will be studied. Secondly, a fine-mesh mesoscale
numerical model will be developed to produce useful
prediction for 6-12 hours. For this goal f the parametrization
of boundary layer, cloud and radiation processes in the
mesoscale will also be studied. The third aspect of the
subsystem is to develop and test the new optimized sounding

143
system, method of assimilation of observed data from
different sources and its application in monitoring and
forecasting.
(II) A New Atmospheric Remote Sensing System
During the last 15 years, we made every effort to
develop the theory and technology of ground-based remote
sinsing profiles of dynamic and thermodynamic parameters of
the troposphere. Since 1989,.a profiler station installed
with microwave radiometers and UHF Doppler radar has been
put into operation in the Beijing area.
(1) Thermodynamic profiling
Passive radiometers operating in the microwave
absorption bands of oxygen and water vapor have been
developed for sounding the thermodynamic prifiles such as
temperature, humidity and column liquid water in cloud.
Firstly, a new type of dual-channel microwave radiometer was
set up for providing humidity profiles and column liquid
water content in cloud: (Fig.3) [2]
Table (1) Dual-Channel Radiometer Characteristics
Parameters
Operating Frequencies
If Band width
Receiver Noise Figure
«. —
—— Integration
— — -
Time
—
___ _„_ _,_
- -
Antenna Beamwidth
Specification
_ — — __ — _ _
20.6, 31 .65 GHz
50-550 MHz
5.5 dB
Software selectable
2.5 deg
New features of this radiometer include a steerable
offset dual-paraboloid antenna producing optimum antenna
performance at each frequency, using noise diodes for gain
calibration and using voltage to frequency converters for
signal intergration and digital demodulation. The instrument
has a built-in-microprocessor to ensure system reliability.
In deriving total precipitable water vapor and path-
Intergrated liquid water content in cloud from radiometer
observations, linear statistical retrieval methods are
employed. For the Beijing area, the path-integrated
precipitable water vapor V and liquid water content in cloud
L can be calculated in realtime from the following formula:
V»-0.037+30.41 * (20.6 GHz > "" 12.86 T(31,6 GHz) (1)
Er=-0. 017-0.214 T (20.6 GHz) + 0 . 663 t (31. 6 .GHz-) (2)

144
where V and L are in centimeters and the radiometer measured
quantities in nepers. Root mean square (RMS) difference of
total precipitable water vapor measured between the
radiometer and the radiosonde is about 0.3 mm (7*).
A nonlinear iteration method is used to retrieve the
water vapor profile in the troposphere from microwave
brightness temperature measurements at different elevation
angles. Some of the retrieved profiles have been shown in
Fig.4-5. It is found that the inversion of humidity in lower
troposphere can be detected by our method. Comparing with
radiosonde data, the relative errors of radiometer
measurement at different levels are shown in Fig.6 [3]-[7]
Secondly, another two microwave radiometers operating
at frequencies 52.8 and 54.4 GHz also have been developed
and applied by the Department of Geophysics, Peking
University to obtain the temperature profiles [8]-[9], Fig.7
show the RMS errors of the derived temperature at different
levels of troposphere by using radiometers measurement in
comparison with radiosonde data.
(2) Wind profiler
The first UHF wind profile radar in China has been
installed and tested in Beijing (Fig.8} [10]. The major
specifications of the radar are listed as following:
Radar:
Wavelength: 0.82 m (365.85 MHz)
Peak Power: 25 kw
Antenna: 108 Yagi elements, 3 beams (one is
zenith, two are 15 degrees off zenith)
Beamwidth: 6 degrees (3 dB)
Signal processing:
Preprocessing: time and spectrum averaging
Processing mode: 3 (High, Mid., Low altitudes)
Pulse width: 1-9 microsec.
PRT: 50-300 microsec.
Spectrum resolution: 128 points
Number of heights: 30 max for each mode
Estimated parameters: first 3 of Doppler mements
horizontal wind velocity,
horizontal wind direction,
vertical wind velocity
Fig, 9 is an example of a time series of horizontal
wind profiles given by UHF Doppler Radar. Every profile was
produced at an interval of 30 minutes. The height resolution
is 100 m at the height coverage from 0.5 to 4.1 km and 200 m
at 5.0-12.5 km. There are 2 additional profiles of
horizontal winds ploted with rawinsonde data time 1900 and
0100 as compared to those profiles measured by profiler
radar at the same moments. It can be seen that profiler's
measurements a well agreed to rawinsondeVs ones but show the
presence of a smaller scale system at the height from 3 to 4
km.

145
Some radiosonde data were used to compare with the
radar data. The rawinsonde site is located about 30 km apart
from the radar site. Fig. 10 is an example of the profile of
horizontal wind observed by the UHF radar (crosses) and the
rawinsonde (solid line) at 19:00 May, 11 1989. Fig. 11 is
the scatter plot of the horizontal velocity data observed by
UHF radar and the rawinsonde in the period of Jan. 1989. A
calculation of the correlation and deviation from data in
Fig. 11 gives the values of 0.96 and 2m/s respectively.
(Ill) Weather Radar Network
Weather radar technology and radar meteorology studies
have been developed in China rapidly. Science 1969, three
types of conventional weather radar (711, 713, 714) at X-
band (3.2 cm wavelength) C~band (5.7 cm) and S-band (10.7
cm) have been manufactured locally and installed to form the
weather radar network. Currently, there are 187 operational
weather radar stations all over the country, with which the
eastern parts, where severe convective weather often occurs,
are well covered. The basic technical characteristecs of
these radar are shown in Table (2).
Table (2) Technical characteristics of
weather radars in China
Types|WavelengthjPeak PowerjDiameter of[Data Processing
(cm) _ (km) Antenna (m) |
"
j
711 3.2
75
1.5 JPPI, RHI,A
713
714
5.7
_ __ «.__ _ «_
10.7
250
— _. «. «._
600
— —i
1
— __«._ «. _
3.8 | Degital Data Proces
(sing, Color Display
.__«. t j — — — —„ — — —.— — ~
4.0
j Degital Data Proces
jsing, Color Display
In the Beijing experimental area, the weather radar
network consists of three digitized conventional weather
radars at 5 cm and 10 cm and two Doppler Weather Radars at
5 cm and 10 cm. All radar data are collected in real-time by
satellite communication and synthesized to retrieve the
spatial distribution of the related meteorological
parameters at workstations of the Beijing Center.
The performance of the 10 cm Doppler Weather radar
applied in network is described as follows:
Wavelength: 10,7 cm
Pulse power: 500 KW
Diameter of Antenna: 8.23 m
Beam width: 0.92
Stability of phase: 0.1
Errors of radial velocily: 0.2 m/s
Pulse length: 0.5, 1.0, 2.0, ralcrosec.

146
PRF: 80-2000 Hz
Minimum detectable power: -112 dBm
Prom the intensity and spectrum of radar echoes, rainfall
rate, radial velocity/ wind shear and turbulence intensity
have been derived and identification of hail also will be
tested. In the summer time, radar echoes from clear-air of
lower troposphere below 3 km often were detected.
The type 713D radar that is the first Doppler radar at
5 cm wavelength manufactured in China will be operated in
1990 in the Beijing area.
Accurate estimation of the rainfall rate from radar
measuremant is an attractive issue for meteorologists and
hydrologists. Though many years of study, we have found a
usual formula that relates rainfall rate R to the radar
echoes intensity Z in the Beijing area as follows: [11]
b
Z=a R ( 3 )
where in averaging a=236, b=1.46. However, the parameters a
and b significantly depend on the raindrop-size
distribution. The standard deviation between radar and
raingauge measurement is about 40&.
On the basis of theoretical calculation and comparison
of radiometer data with the raingauge measurement, the
relation between extinction coefficient and rainfall rate R
has been found as follows. [3]
d
a =C R (4)
It is shown that the parameters C and D are insensitive
to the raindropsize distribution. The RMS deviation of the
relation a-R for different size distribution is reduced to
15 &. Therefore, more accurate estimation of path-integrated
rainfall from microwave radiometer measurements might be
expected. Unfortunately, the spatial distribution of
rainfall rate along the sounding path cannot be retrieved
from microwave radiometer data. To solve this difficulty, a
synthesized system consisting of radar and microwave
radiometer with a common antenna has been developed and
tested.

147
[13
[2]
[3]
[43
[5]
[6]
[73
[8]
[9]
[10]
[11]
REFERENCES
Annual Report 1987-1988, Academy of Meteorological
Science, 1988, China Meteorological Press.
Zhao Conglong, Zhou Xiuji, Ma Da~an. Liang Qixian,
1988, Lower Tropospheric Profiling: Needs and
Technologies, pp 257-258. AMS, NCAR, NOAA .
* H W * IS A^feJIiF&Ml.tSMJiMir , tHWSffSB*
^ftS«*»»W»»«ttr 1982 . ® ft x it Hi IK II -
Wei Cong et al., 1984, Advances in Atmospheric Sciences,
Vol. 1 pp 119-127.
Huang Runheng and Wei Cong, 1986, Advances in Atmospheric
Sciences, Vol. 3, pp 86-93.
JtffiS, 1987, *^f*$. Vol. 11 No. 4.
Zhao Bolin et al. 1981, Scientia Sinica, Vol. 24, No 3,
pp 36-373.
Zhao Bolin et al. 1985, Scientia Sinica, Vol. 28, No 5,
pp 528-536.
Ma Da-an, Zhou Xiuji, Zhao Conglong, Liang Qixian, 1988,
Lower Tropospheric Profiling: Needs and Technologies,
pp 127-128, AMS, NCAR, NOAA.
* % ft m £ ft , 1983 . ^ £ tfi JK f± .

148
[UHF Doppler RadaT]
^Pibalk
' ZH^^——• —"I
Doppler Weather Radar
. i J^S> Chengde
Weather IT
' • Radar
Zhang jiaKou
Fig,
1 Four Experimental Areas of Monitoring
and Forecasting System of Mesoscale
Disastrous Weather in China.
Fig. 2 Monitoring Network of Mesoscale Weather Systems
in the Experimental Area of Beijing
1982.4.16
7' 00
: eoo
Fig. 3 The New.Dual-channel Microwave RadiomH"fcr
19'00
1982.7.16
1000 ^0 5 10 Q (g/kg)
Fig. 5 The Variation of Humidity
Profiles by Radiometer
1000
0 1 2
q(g/kg)
Fig. 4 The Inversion of
Humidity by Radiometer
|P< hPa
• 9r>n
_ , i
I
j ".: i ,
>
J
/
^
1 "
^
/
x^
z
10 20 30 40 5
| | 1 1 l 0 Aq/q (%)
i i j
1982 Jan. | Apr. July Aug Oc i^
i'i

149
0 1 2 3 4 o T(°K)
FLq. 7 The RMS Errors of the
Desived Tenperature by
Radiometers
Fig. 8 UHF Profiler's Phase array antenna
(km)
12.2
10.6
9.0
7.4
5.8
4.7
4.1
3.5
2.9
2.3
1.7
1.1
0.5
23 23 23
22 22
4^% 4^r.
IT/^j4fo
0100 0100 0030 0000 2330 2220 2130 2100 2030 2000 1930 1900 1900
>50 50—25 25 — 17 17 — 10 10—5 5 —1

150
STRUCTURAL FEATURES OF A SQUALL LINE OVER THE TAIWAN
STRAITS REVEALED BY DUAL-DOPPLER RADAR
Y. J. Lin, R. W. Pasken and H. Shen
Saint Louis University
St. Louis, MO 63103
U. S. A.
ABSTRACT
In this study, structural features of a subtropical
squall line, which occurred on 17 May 1987 over the
Taiwan straits, were investigated using the dual-Doppler
data collected during the Taiwan Area Mesoscale Experiment
(TAMEX). Fields of the storm-relative wind and reflectivity
were derived in the horizontal domain of 45 km by
25 km using an objective analysis scheme with 1 km grid
spacing in all three directions. Vertical velocities were
computed from the anelastic continuity equation by
integrating downward with variational adjustment. Fields
of pressure and temperature were then retrieved from the
Doppler derived winds using the three momentum equations.
Results show that many structural features of a
subtropical squall line are similar to those for a fastmoving
tropical squall line. A low-level jet (LLJ) associated
with the frontal system provides the necessary
strong shear at lower levels. On the front side of a
squall line, the front-to-rear environmental flow prevails
at all levels and is accompanied by the shallow rear-tofront
flow on the back of the line. There are many individual
cells embedded within the squall line. Relatively
weak convective downdrafts occur between the cells and
behind the main cells. The retrieved pressure and temperature
deviations are in good agreement with the
updraft-downdraft structure. In general, high pressure
forms on the upshear side of the updraft with low pressure
on the downshear. The temperature excess (deficit)
corresponds well to the updraft (dowiidraft). The interaction
between the convective updraft and downdraft plays an
important role in maintaining the three-dimensional circulation
within a squall line.

151
1. INTRODUCTION
In this study, structural features of a squall line, which
occurred on 17 May 1987 in IOP-2 over the Taiwan straits, were investigated
using the thermodynamic retrieval method of Gal-Chen (1978).
Dual-Doppler data collected from the CP-4 and TOGA radars (Fig. 1) at
three analysis times were objectively analyzed in the horizontal domain
of 45 km by 35 km (see the shaded area in Fig. 1) using a 2.5 km scan
radius. There were 10 analysis levels in the vertical ranging from
0.35 to 8.75 km. The grid spacing was 1 km for all three directions.
Vertical velocities were calculated from the anelastic continuity equation
by integrating downward with variational adjustment. Subsequently,
fields of deviation perturbation pressure and temperature were
retrieved from the Doppler derived winds using the three momentum equations.
These fields were then employed to investigate some kinematic,
dynamic and thermodynamic structures of a squall line. The goal is to
better understand the structure and internal dynamics of a squall line
which produced heavy rain in the Taiwan straits.
2. DATA
At 2000 LST 16 May, a small squall line had organized in a
north-south direction over the Taiwan straits, and was moving eastward
at about 25 to 30 kts. The conventional 10-cm radar located in Kaohsiung
(Fig. 1) showed the intensification of a squall line after 2000 LST
as depicted in Fig. 2. The northern portion of the squall line moved
into the SW lobe of a triple Doppler network, Dual-Doppler data were
collected from the CP-4 and TOGA radars after the line entered the coverage
area near the 0030 LST 17 May 1987. The data considered in this
study centered at 0043 LST.
Figure 3 shows the 2000 LST 16 May sounding released at Makung
over the Taiwan straits (see Fig. 1). Note that moisture was available
in the layers from the surface to the upper troposphere. The sounding
exhibited conditional instability. A moist adiabat (dotted line)
denotes the ascent of an undiluted parcel from the cloud base. In the
absence of mixing with the environment, this air parcel would ascend
freely above 200 mb with the maximum temperature excess (* 4 C) near
600 mb. Winds were from the southwest throughout the whole tropor
sphere. A low-level jet (LLJ) with a maximum speed near 20 m s
occurred in the layer between 3 and 4 km. Chen and Yu (1988) found
that extremely heavy rainfall in Japan and China is associated with a
LLJ. A composite, vertically smoothed hodograph, based on the Makung
soundings released at 14, 17 and 20 LST, is presented in Fig z .4. Storm
motion during the analysis times was from 250° at 16.5 m s (see the
dashed line in Fig. 4 ). Note that veering occurred in the lower and
upper layers, while backing occurred in the middle layer. The strong
winds observed in the lower levels near 4 km were associated with a LLJ
mentioned previously. Using the environmental soundings of tropical
mesoscale cloud lines in GATE, Barnes and Sieckman (1984) found that
the environment of the fast-moving line was characterized by strong
low-level shear, and the principal component of the shear is perpendicular
to the line.

152
25
2200L
16 MAY
0000 L
17 MAY X- ,
23
I 19" ,
12!'
Fig. 1. A TAMEX tripple Doppler Fig. 2. PPI displays revealed by
network showing the locations the conventional 10-cm radar
of CE-4, TOGA and CAA radars. in Kaohsiung. A black area
The shaded box signifies the signifies the radar reflecdomain
of interest.
tivity > 30 dBZ.
/ \
x
Y /
XV/A
\V\ \/// /^/
/ -. \ / ,«ky \ , \' f
600
X\/
^
800
1000
Fig. 4. A wind hodograph showing
Fig. 3. The Markung sounding at winds at various heights (km)
2000 LST 16 May. in the prestorm environment.

153
3. METHODOLOGY
The data analysis and reduction procedures were detailed in Lin
et aJ. (1986). We employed the Euler-Lagrange equations and the constraint
equation to derive three variationally adjusted wind components
(u, v, w) at each analysis level. The derived wind field is subject to
both random and nonrandom errors. Following Lin et al. (1986), an
error analysis was conducted. Our finding shows that the combined
errors due to statistical uncertainty in the radial velocity estimates
and geometrical considerations are 1-2 m s for the horizontal derived
winds.
Once the detailed wind field was obtained, fields of deviation
perturbation pressure and temperature were recovered from the derived
three-dimensional wind field via a thermodynamic retrieval method
(Gal-Chen, 1978). The retrieved fields, together with the derived wind
field, were then used to study some structural features of a squall
line.
4. DISCUSSION OF RESULTS
4.1 Horizontal View at 0.73 km
The horizontal storm-relative wind field with radar reflectivity
contours superimposed is shown in Fig. 5a. Distances are in
kilometers from the TOGA radar. At the time of analysis, the leading
edge of the squall line was located about 10-20 km west of TOGA. There
are many cells embedded within the squall line. The leading edge is
almost in a north-south direction and moves from 250 at 16.5 m s
The maximum reflectivity is less than 45 dBZ. Several new cells occur
along the eastern edge of the squall line approximately 5-10 km to the
east of the main cells. The low-level convergence line is evident
especially over the northern portion of the domain. On the east side
of the convergence line, storm-relative winds are mainly from the
southeast. The southeast winds transport high 6 environmental air
toward the leading edge of the squall line. Such air feeds the convective
updrafts, resulting in a broad area of high reflectivities in the
vicinity of the convergence line. Pronounced upward motion (Fig. 5b)
prevails at the leading edge due to the strong low-level convergence.
Conversely, downward motion dominates in a broad area west of the convergence
zone. There are many cells in the squall line with new cells
to their east. Most convective updrafts are accompanied by relatively
weak convective downdrafts on their west side. This finding is consistent
with that reported in Roux (1988) for a tropical squall line.
Using dual-Doppler data obtained in the northern Ivory coast, Roux
(1988) showed that the convective region has many short-lived cells of
intense updrafts and high reflectivities with downdrafts in between and
behind.
Figure 6a displays the retrieved P• ' field at the same level*
Low pressure occurs on the downshear side (east) of.the.updraft (U)
with high pressure on the upshear. Notice that high pressure observed
on the west side of the updraft corresponds to the near-saturated
convective-scale downdraft behind the leading edge. The density of the
downdraft air is increased due to precipitation loading, resulting in a

154
0*73 KN
0,73 KM TAMEX IOF2 00*131.
& 0 , --J^-—~T ^f. ,
KM EAST OF TOGA
-23 -20 -13 -10 -3
Kft £AST OF TOGA
Fig. 5. Horizontal distributions of the (a) storm-relative wind with
reflectivity (Z) contours superimposed, and (b) vertical
velocity (w) at 0.73 km for 0043 LST 17 May._.Contour intervals
for Z and w are 10 dBZ and 2ms , respectively.
Reflectivities > 40 dBZ are shaded.
shallow layer of the negatively buoyant outflow toward the rear (west).
This finding is in qualitative agreement with those presented in the
GATE study by LeMone et al. (1984). Using the aircraft and rawinsonde
data, gathered in nine tropical meso-scale convective line cases,
LeMone et al. (1984) showed that a mesohigh centered at the surface
with a mesolow just above the cloud base. On the other hand, low pressure
on the downshear (east) side of the updraft appears to be related
to the buoyancy term since the lifted environmental inflow is warmer
(lighter) than the surroundings. To the east and southeast of the
leading edge over the west coast of Taiwan, a mesohigh is observed
(outside of Fig. 6a). This mesohigh is dynamically induced due to
interactions between the southwest monsoon and the Central Mountain
Range (CMR). Such interactions often produce a mesolow on the east and
southeast side of the CMR with a mesohigh on the west and southwest
side (not shown). A pronounced horizontal pressure gradient develops
between this high over the west coast of Taiwan and low pressure near
the leading edge. In the convective region, such a pressure gradient
accelerates the convective updrafts front-to-rear, resulting in a west-

155
0*73 KM TAMEX IOP2 00*31
a
P » 0,01 «S
0.73 KH TAMEX IOFjg_ QM3L
' T * ovi ue6 J '
-23 -40 -13
KM EAST OF TOQA
-25 -20 -15
KH EAST OF TOGA
Fig. 6. Horizontal distributions of the (a) deviation perturbation
pressure (P '), and deviation virtual temperature (T ') at
0.73 km for 0043 LST 17 May. J Contour intervals for P"/ and
rs
A
are 0.1 mb and 4 C, respectively.
vd
ward tilt of the updrafts into the shear (LeMone et al., 1984). At
this level, the vertical pressure gradient and buoyancy terms have the
same order of magnitude but opposite sign, resulting in a much smaller
value of vertical accelerations. As a result, the high and low pressure
centers seen in Fig. 6a are largely caused by the hydrostatic
effect. The retrieved T 1 field is shown in Fig. 6b. Examination of
Figs. 5b and 6b reveal tKat the deviation temperature is closely
related to vertical velocity. In general, warming is found in the area
with upward motion, while cooling is associated with a downdraft. A
cross-comparison between the reflectivity field (Fig. 5a) and the T .'
field shows that centers of the maximum temperature deficit coincide
well with high reflectivities, indicative of the significant effect of
precipitation loading on making the downflow air negatively buoyant.
As a consequence, the downdraft air on the west side of the convergence
line is denser than that in the convective updrafts, resulting in a
cold surface outflow toward the west and northwest. This cold outflow
air is the product of cool, saturated convective downdrafts behind the
leading edge (Zipser, 1977).

156
4.2 East-west Cross Section
The east-west cross section at 28 km north of TOGA, showing the
vertical distributions of storm-relative wind and vertical velocity is
presented in Fig. 7. The cross section passes through one of the convective
cells. Two reflectivity maxima are associated with two cells
with the main (old) cell on the west and the new cell to its east. On
the east side of the updrafts, the front-to-rear flow dominates at all
levels with maxima at lower and upper levels. In the lower troposphere,
the environmental high 6e air is lifted at the convergence
zone, feeding the convective updrafts at the leading edge (heavy dashed
line). The main updraft, with a maximum speed of 12 m s , is tilted
toward the west in the lower layer (0-4 km), and becomes almost erect
in the middle and upper layers (z > 4 km). A secondary updraft (new
cell) to the east of the main updraft is relatively weak with a maximum
speed of 6 m s . The updraft tilt keeps water loading and evaporative
TAMEX IOP2 OOH3L Y=28KM dBZ „ 10 M/SEC
'-I' ' ' -12
KM EAST OF TOGA
-2
0043
28
o
Hi
-30 -20
KM EAST OF TOGA
-10
Fig. .7. The east-west cross section at 28 km north of TOGA, showing
the (a) storm-relative wind with reflectivity contour superimposed,
and (b) vertical velocity. Contour intervals for Z
and w are 10 dBZ and 3ms , respectively, with negative
values dashed and positive values hatched. The heavy dashed
line represents the gust front at the leading edge.

157
•
cooling to a minimum from the falling rain. As the rain falls out the
updraft, it increases the density of the downdraft air below the tilted
updraft. The downdraft air tends to conserve its momentum, undercutting
the updraft air on the downshear side of the storm. As a result
of updraft-downdraft interaction in the lowest layers, the growth of
the new cell occurs ahead of the old cell and results in the storm
tilting into the shear instead of with it. The lifted air exits from
the updrafts in the middle and upper layers. It continues to move
westward, heading toward the trailing stratiform region of the squall
line as in the tropical squall line studied by Chong et al. (1987).
The rear-to-front cooler air enters from the western edge of the convective
region (left side of the figure) in the 2-4 km layer. Buoyancy
of the air is further decreased due to evaporative cooling and precipitation
loading in the region immediately west of the high reflectivity
core. As the descending air approaches the lowest layers, it spreads
out to form a diverging outflow in the boundary layer. Part of the
spreading air moves toward the east, colliding with the front-to-rear
low-level environmental inflow and forming a weak gust front in the
area slightly to the east of the 45 dBZ contour at x - -12 km. The
interaction between the front-to-rear environmental low-level inflow
and the rear-to-front midtropospheric flow is responsible for maintaining
the updraft-downdraft structure in the convective region. On the
other hand, the front-to-rear flow near the surface behind the leading
edge (i. e., x < -12 km) is quite similar to a shallow layer of cool,
saturated air near the surface reported in Zipser (1977). This flow is
the product of convective-scale saturated downdrafts (-24 < x < -19 km)
west of the main updraft. Upward motion dominates in the convective
region behind the gust front with downward motion in the region between
the main cell and the new cell and behind the main cell. Such downward
motion coincides with the relatively high reflectivity region, indicative
of the important contribution of precipitation loading to convective
downdrafts.
Vertical distributions of P * and T
f
along the same cross
section are shown in Fig. 8. Notice that high pressure forms on the
upshear side of the updraft with low pressure on the downshear side. A
mesohigh occurs in the boundary layer at x - -16 km behind the leading
edge. The mesolows underneath the main and secondary updrafts at x =
-15 km and x = -9 km, respectively, are hydrostatic in nature since the
updraft air, fed by the low-level environmental inflow, is warmer and
lighter than the surroundings. A deviation vertical pressure gradient
is directed upward from the lowest layer to 2 km and then reverses its
direction (downward) in the layers above. These deviation vertical
pressure gradients are needed in order to maintain the net vertical
acceleration within the updraft. The distribution of T ' is closely
related to the updraft-downdraft structure (Fig. 7b) and the reflectivity
pattern (Fig, 7a). In the updraft region, warming prevails due
to the release of latent heat by condensation. Conversely, cooling
dominates in the downdraft region. The cool area underneath the main
updraft with high reflectivities (-16 < x < -13 km) is attributed in
part to rain loading. This cool area extends downward to the lowest
layer behind the gust front, indicative of a cell's cold surface outflow.
. ' '' •
;
• .' • ' " " •• •' . '. , ' ' ' '• ' ' " .' •

158
KM
-20
EAST OF TOGA
-10
0043
28
X 10
o
Fig. 8.
-30 -20 -10
KM EAST OF TOGA
As in. Fig. 7 except for (a) deviation perturbation pressure,
P ' and (b) deviation virtual temperature, T ' Contour
Y
intervals are 10 Pa (0.1 mb) and 1 C for P J and T va
respectively.
5. CONCLUDING REMARKS
The kinematic, dynamic and thermodynaraic structures of a subtropical
squall line, which occurred on 17 May 1987 over the Taiwan
straits, were investigated using the TAMEX dual-Doppler data and the
thermodynamic retrieval method. Results show that the overall structural
features of a subtropical squall are quite similar to those of a
tropical squall line. On the front side of the leading edge, the
strong front-to-rear environmental flow dominates at all levels with
maxima at lower and upper levels and a minimum at middle levels. Conversely,
a shallow layer (1-3 km) of the rear-to-front flow enters
from the back side of the convective region. It transports cooler midtropospheric
air into the lower layer resulting in a sloping downdraft
behind the main cell. At the leading edge, high pressure forms on the
upshear side with low pressure on the downshear. In the lowest layers,
a mesohigh develops behind the gust front, due to water loading, with
low pressure to its east. A hydrostatic low-pressure center under the
convective updrafts is associated with the ascending warm environmental
air. The mesoscale pressure gradient accelerates the updrafts frontto-rear,
producing a westward tilt of the updrafts into the shear in
the layer below a LLJ. The retrieved temperature deviations seem consistent
with the updraft-downdraft structure. In the convective
updrafts, warming dominates due to the release of latent heat by condensation.
Conversely, cooling occurs in the convective downdrafts.
The near-saturated downdraft behind the leading edge is caused by precipitation
loading in the high reflectivity region.

159
6. ACKNOWLEDGMENTS
The authors wish to express their appreciation to those scientists,
technicians and staff members who participated in the TAMEX project.
Special thanks go to the National Science Council of Republic of
China and the National Science Foundation for supporting the field
experiment. This research has been supported by the Division of Atmospheric
Sciences, National Science Foundation, under Grant ATM-8609150.
7. REFERENCES
Barnes, G. M., and Sieckman, K., "The Environment of Fast- and Slowmoving
Tropical Mesoscale Convective Cloud Lines", Mon. Wea.
Rev., 112, 1782-1794 (1984).
Chen, G. T. , and Yu, C, C., "Study of Low-level Jet and Extreamely
Heavy Rainfall over Northern Taiwan in the Mei-Yu Season", Mon.
Wea. Rev., 116, 884-891 (1988).
Chong, M. , Amayenc, P., Scialom,G., and Testud, J., "A Tropical Squall
Line Observed during the COPT 81 Experiment in Vest Africa.
Part I: Kinematic Structure Inferred from Dual-Doppler Radar
Data", Mon. Wea. Rev., 115, 670-694 (1987).
Gal-Chen, T., "A Method for the Initialization of the Anelastic Equations:
Implications for Matching Models with Observations",
Mon. Wea. Rev., 106, 587-606 (1978).
LeMone, M. A., Barnes, G. M. , and Zipser, E. J., "Momentum Flux by
Lines of Cumulonimbus over the Tropical Oceans", J. Atmos.
Sci., *I» 1914-1932 (1984).
Lin, Y. J., Wang f T. C., and Lin, J. H., "Pressure and Temperature Perturbations
within a Squall-line Thunderstorm Derived from
SESAME Dual-Doppler Data", J. Atmos. Sci., 43, 2302-2327
(1986).
Rotunno, R., and Klemp, J. B., "The Influence of Shear Induced Pressure
Gradient on Thunderstorm Motion", Mon. Wea. Rev., 110, 136-151
(1982).
Roux, F. , "The West African Squall Line Observed on 23 June 1981 during
COPT 81: Kinematics and Thermodynamics of the Convective
Region", J. Atmos. Sci., 45, 406-426 (1988).
Zipser, E. J., "Mesoscale and Convective-scale Downdrafts as Distinct
Components of Squall-line Circulation", Mon. Wea, Rev., 115,
1668-1589 (1977).

160
Radar Observation of Precipitation Systems in Taiwan
PAY-LIAM LIN AND TAi-Cm CHEN WANG
INSTITUTE OF ATMOSPHERIC PHYSICS,
NATIONAL CENTRAL UNIVERSITY,
CHUNG-LI, CHINA 32054
ABSTRACT
The mesoscale data sets of two cases in TAMEX phase 1,1986
and in TAMEX ,1987 were analyzed to study the characteristics of
convective system during Mei-yu season. The synoptic conditions
and the pre-storm environment were examined. Based on digital
radar data, a detailed discussion of the reflectivity characteristics
such as horizontal pattern, vertical structure and movement is
given. In order to understand the kinematics of the convective
systems from conventional radar, the internal motion fields of
these cases were also calculated through the TREC technique. In
the TAMEX case study, a few very long-lived cells along the shear
line developed and maintained their three dimensional structures.
During its long life cycle, its 3-D airflow may evolved into supercell
like pattern with strong rotational characteristics.
1. INTRODUCTION
Flash floods are major meteorological diasters which occur around the globe.
They present a challenging problem for both scientific research and operational
forecasts because they involve complex dynamics and microphysical processes of
mesoscale connvective systems and their interaction with many other scales of
motion(Maddox et al. 1979). The heavy rainfall during typhoon and "Mei-Yu"

161
seasons sometimes brought great damages in the Taiwan area. On the other hand,
rainfall is also a very important water resource to the island. A raingauge network
consisting of 74 surface weather stations and 126 self- recording raingauges was
set up by Central Weather Bureau in the past few years. Five conventional radars
constitute the radar network(Fig.l). Two S-band radars have provided digital
data since 1985 summer. The other three radars have also provided digital data
since 1988. The first Doppler radar was set up in 1987 by CAA(Civil Aeronautics
Administration) at CKS( Chiang Kai-Shek) airport(Fig.2) in Taiwan area. This
radar has proved its ability to observe precipitaiton systems in northern Taiwan.
A field program of the Taiwan Area Mesoscale Experiment (TAMEX) was
operated from 1 May to 29 June, 1987 in the Taiwan area. Three Doppler radars
were positioned along the west coast of Taiwan, forming a triple Doppler network
(Fig.2). The primary scientific objective of TAMEX is to improve our understanding
of the mesoscale dynamics of the convective systems responsible for heavy precipitation.
In May and June 1986, a field experiment called " TAMEX phase I"
was held in Taiwan area as a testing field program for 1987 TAMEX.
The mesoscale data sets of two cases in TAMEX phase I, 1986 and TAMEX
,1987 were analyzed to study the characteristics of convective system during Mei-
Yu season. Based on the digital radar data, a detailed discussion of the reflectivity
characteristics such as horizontal pattern, vertical structure and movement was
illustrated. In order to understand the kinematics of the convective systems in
TAMEX phase I, the internal motion fields of thise case were also calculated
through the TREC technique. In the TAMEX case study, structural features and
internal motion of one long-lived and slow moving MCS rainband originated from
the Mei-Yu frontal unstable zone which occurred on 25 June 1987 in IOP-13 over
the Taiwan area were investigated .
2. RADAR ESTIMATE AND GUAGE MEASUREMENT
During the TAMEX phase I period, digital radar data with 0.5-hr interval and
rain gauge data with 10 minutes interval in the major events are available. From
May 20 to May 21 1986,a precipitation system was observed across the Taiwan
Strait. It later brought a rainfall accumulation of approxmately 200 mm in two
days in some major cities in the central and northern Taiwan. From 2300 May
20 to 0700 May 21, sixteen reflectivity distributions for every 30 minutes were

162
Fig. 2. A triple doppler network
of TAMEX showing the location of
CP-4, TOGA, and CAA radar.
The radar network
Fig. 4. Spatial distribution of G/R
ratio, (the relative position of optimizing
technique calculation domain
is marked in Fig. 3.). (after
Wang et al.,1988)
POCC "B 0030 b.;/!9ee
Oi-TSNCr. FK'.r

163
carefully analyzed. In Fig.3 the echo pattern of six different times are selected to
show the evolution of the precipitaion system.
In order to decide a proper radar data acqusition scheme for the different
types of rain, this case was selected by Wang et al.(1988) to show the variation of
G/R(Gauge/Radar) ratio. The spatial distributions of the G/R ratios within a 160
km x 200 km domain (Box I in Fig.3) are illustrated in Fig.4 at the selected time
0500. A well organized NE-SW oriented rainband had already passed Wu-chi and
Taidmng at this time (Fig.3). The heavy rain core was located east to Taichung.
The reflectivity was 30 dbz at Taichung and 20 dbz at Wuchi. Gauge measurements
were 10.2 mm/hr at Taichung and 0.6 mm/hr at Wuchi. An overestimation (G/R
4.2) was found near Taichung, and an underestimate (G/R 0.5) was found near
Wuchi. This sharp gradient of calibration factor could be the consequence of large
variation of real rainfall in both space and time.
The rainfall optimizing scheme suggested by Brandes(1974) had been applied
in this case. Within the domain of Box I, the G/R ratio was interpolated onto each
grid by the objective analysis ( Barnes 's scheme 1973 ). In the gauge data void
area , an average G/R ratio was assigned to the grid. After the G/R ratios were
interpolated, the radar estimate were multiplied by the gridded ratio to provide
the "corrected" rainfall estimate. We found that the objective anlysis is able to
bring up the radar estimate near heavy rain core and retain the spatial variation
of the sharp gradient of rainfall.
3. INTERNAL MOTION BY TREC
Convention radar is a powerful tool for providing three dimensional echo
structure with high resolution in space and time. However it can not supply
the kinematic parameter for study of the internal airflow structure. In order to
understand the kinematic properties of mesoscale precipitation systems, a pattern
recognization technique TREC (Tracking Radar Echo by Correlation) suggested
by Rinehart(1979) was adopted by Wang(1988) to calculate the internal motion
field from consecutive digital radar data set.
Since the data set of this TAMEX Phase I case was a so called CV map which
only contained the maximum reflectivity in the vertical column, so the motion field
we obtained can only be interpreted as the steering flow of convection at middle
to lower levels. The motion fields at four different times on May 21 were selected

164
for this presentation Fig. 5a(0000 to 0030) , Fig.5b (0130 to 0200),Fig.5c (0330 to
0400) and Fig.5d (0500 to 0530).
In Fig. 5a most vectors were pointing towards the NE, this indicates that
the SW flow dominanted the whole domain. However about 100km northwest to
Kaoshiting, vectors pointed to the north. About 120 km north of Kaoshiung wind
was very light. Between the fast moving southwesterly flow and the southerly flow
there was a confluent region. The location of this region is in agreement with the
echo development latter (Fig.3 ). When the prevailing southwesterly flow blew
against the terrain of the island, its speed would decrease near Taiwan. Also it
might be diverted to flow around the terrian. Part of the diverted flow would blow
towards the north. This would explain the southerly wind and the confluence.
In Fig 5c the southwesterly flow still behaved very similar to the field occured
two hours earlier, but there was a distinctive feature in the stratiform area. Some
wind vectors changed into a WSW or W or even WNW direction which caused
wind a shift line oriented NE-SW. This line would coincide with the the surface
front. This wind shift may be due to the fact that the depth -occupied by the
stratiform rain was different from the depth of the deep convection. Therefore the
steering wind was different. It could also be a frontal wind shift line.
4. SYNOPTIC CONDITION OF TAMEX IOP-13
By evening 24 June 1987 the surface high pressure area behind the surface
front over mainland China enhanced and moved southward, resulting in the southward
movement of the Mei-Yu front with a strong 850 mb shearline oriented in
the west-east direction over the Taiwan strait. At 850 rnb a short westerly trough
appeared in the Fu-Jen coastal area ,then moved into the northern part of the Taiwan
Strait and deepened. The northern part of Taiwan and the Taiwan Strait area
are located in the confluent airstream region, while a difSuence region was over
Taiwan area at 300 mb and 200 mb(Fig.6), This synoptic pattern was favorable
for convection to occur in the vicinity of the northern Taiwan area.
Fig.7 shows the selected satellite pictures. These pictures show the evolution
and movement of the convective cloud systems. During the early evening of June
24, strong convective cloud was observed over the coast of mainland China. At
2300 LST June 24 a group of cloud lined up in the east-west directionin association
with the east-west oriented 850 mb wind shear line was observed across the strait

165
Fig. 6. (a) The 850 mb map of
0800LST June 25 (b) The 300 mb
map of 0800LST June 25
Fig. 7. The IR satellite pictures at
15Z, 18Z, 21Z on June 24 and OOZ
03Z, 06Z on June 25, 1987.
Fig. 8. Wind hodograph of Panchaio
at 12Z (prefront) and 21Z (postfront)
on June 24, 1987.
Fig. 9. The reflectivity field from
Kaohsiung radar during 0000-1500
LST on June 25, 1987. first contour
is 15.5 dBZ, the contour interval
is 10 dBZ.

166
near the northern tip of Taiwan. Wind hodograph in Fig.8 illustrates wind change
with height at different time over Pan-Chiao station. The pre-frontal low level jet
and warm advection was very obvious. Behind the surface front, low level wind
shifted to the north- easterly and backed with height. This indicates cool and dry
air advection prevailed in the low level.
5. RADAR ANALYSIS
Fig. 9 shows the reflectivity field observed by the Kaoshuing radar from 0000
LST to 1500 LST June 25 with one hour interval. These Figs are the CV maps
showing the horizontal distribution of the maximum reflectivity in the vertical
column. Following the southward movement of the surface front, convection was
observed along the coast of mainland China during the evening hours of June 24,
the convective activity moved into the strait afterward. At 0000 LST June 25,
a line echo associated with the frontal system developed into a strong convective
rainband in the northern part of the Taiwan Strait and slowly moved southward.
At 0100 LST June 25, this line rainband transformed into a wave-like pattern(Fig.lO).
At the tip of the wave-shape echo, the radial velocity field showed
convergence centers and rotation signatures (lebelled by CON in Fig. 10). The
analysis of the vertical cross section indicated that this area was also in association
with stronger convection. In addition to the convergence zone, we also noted
that there was a divergence signature appearing in the rear of this MCS rainband.
We think it was the result from the rear downdraft motion aloft. The coupling
between southwesterly inflow updraft and this rear downdraft was probably responsible
for the organization of this strong convection.
The convective activity associated with the slowly moving rainbaad continued
as it approached the coast , Stratiform precipitations were noted ahead of and
behind of the system. From the analysis of the reflectivity field and radial velocity
field of TOGA radar during the period of 0700 to 1000 LST, a long-lived cell (>
30 dBz) with 40km long and 20km wide (cell L) was observed right in front of
the strong wind shear line (Pig. 11). Although this cell had changed its shape and
intensity, it did not dissipate completely over four hours. Due to the slow motion
of this cell, the accumulated rainfall was very high near Taidhung area. Taichung
reported nearly 175mm of rain by 2100 LST, and the TOG A radar site was flooded
and nearly ceased operation of the radar.

167
i: Fig.10. The CAA radar reflectivity
|ii| field at 0100 LST (a) 1 Km height
F (b) 3 Km height and the radial velocity
field at 0100 LST (c) 1 Km
height (d) 3 Km height. The CAA
radar site is located at (0,0) corrdi-
] nate.
(a)
-rrr-h
Fig. 12. Cross section perpendicular
to the shear line (marked with
A in Fig. 20) at 0802 LST on 25
June 1987 from TOGA Doppler radar
measurements, (a) reflectivity field
(b) radial wind. The position of
convergence area is marked with CON
the single letter indicates the wind
direction.
Fig. 11. Successive echo configration of the long-living
convection overlaid with radial winds for the period
» 0700-0900 LST at 1 km height taken from TOGA mea-
•• surement. The radar site locates at (-14, -42) coordi-
| nate. Graduated shading represents 10 dB intervals in
- reflectivity. Heavy solid line indicate the positive radial
wind while heavy dashed line note the negative radial
wind. Reflectivty contour are at 10 dB intervals with
SddBZ isopleth shown as a thin solid line, (a) 0700 LST
(b) 0802 LST (c) 0856 LST

168
In order to understand the internal circulation of this long-lived cell, the
single Doppler analysis at 0701, 0802 and 0856 LST were shown in Fig. 11. Prom
these three successive echo configurations with overlaid radial wind, we noted that
the configurations of the cell were almost steady and the isodops showed sharp
turning. This signature indicated that a clear wind shear line just located at the
back edge of reflectivity core(> 30 dBZ). From the configuration of isodops we
observed the strong southwesterly LLJ heading toward the shear line. To the rear
side of the shear line the westerly flow met the southwesterly one and created strong
convergence.This southwesterly low level jet was responsible to supply the moisture
to this prefrontal precipitation system as described in Browning and Pardoe(1973).
A cross-section pendicular to the shear line was illustrated in Fig. 12 ( the position
of the crossection was marked in Fig. 11) 30 dBZ contours were tilted toward the
southeast direction. From Fig. 12, at the lowest level 0.5km we can identify that the
shearline position was at x=-16 km, a strong convergence zone near this shearline
was found .Consequently a strong updraft will erect right above this region. From
5km to 15km height to front side of updraft, a strong wind component blowing
toward the southeast (northwest wind 16ni/sec) was observed, while only about
9m/sec north-west wind at this levle was observed on the rear side. Obviously, an
outflow region was located above updraft center. At 5 km height a convergence
was found near 24 km to the northwest of the strong convection, this convergence
resulted in a rear flank downdraft.
6. CONCLUSION
Based on the digital radar data, a detailed discussion of the reflectivity characteristics
such as horizontal pattern, vertical structure and movement was illustrated.
In order to understand the kinematics of the convective systems in TAMEX
phase I, the internal motion fields were also calculated through the TREC technique.
The TAMEX case study indicate that within the MCS rainband, a few very
long-lived cells along the shear line develop and maintain its three dimensional
structure. During its long life cycle, the 3-D airflow may evolved into supercell-like
pattern with strong rotational characteristics. From the configuration of isodops
we observed the strong southwest LLJ heading toward the shearline. To the rear
side of the shear line the westerly flow met the southwesterly and created strong

169
convergence. Consequently a strong updraft will erect right above this region.
From 5krn to 15km height to the front side of updraft, a strong wind component
blowing southeastward (northwest wind 16m/sec) was observed, while only about
9m/sec north-west wind at this level was observed on the rear side. Obviously, an
outflow region was located above updraft center.
The development and maintance of these long-lived convective precipitation
systems and its unique structure to interact with its larger scale environment will
be pursued in the future. The difference between mid-latitude supercell and this
kind of long-lived cell within Mei-Yu frontal zone will be a very interesting research
topic.
Acknowledgments. We wish to thank Miss Chin-Chin Yeh for her help in
the data processing of Doppler radar,CAA(Civil Aeronautics Administration) for
kindly supplying of radar data and Micro-Computer center of the Department of
Atmospheric Physics, NCU. for the use of computing facilities. This work was
supported by the National Science Council under Grant NSC 79- 0202-M008-02.
REFERENCES
Barnes, S.L., 1973: Mesoscale objective map analysis using weighted time-series
observations, NOSS Tech. Memo. ERL NSSL-62, Norman, Oklahoma, 60pp,
Brandes, E., 1974: Radar rainfall pattern optimizing technique. NOAA Tech.
Memo. ERL NSSL-67, Norman, Oklahoma, 16pp.
Browning, K. A., and C. W. Pardoe ,1973: Structure of low-level jet streams ahead
of midlatitude cold fronts. Quart. J. Roy. Meteor. Soc., 95, 619-638.
Maddox, R. A., C. F. Chappell,and L. R. Hoxit,1979: Synoptic andd meso-a scale
aspects of flash flood events. Bull Amer. Meteor. 5oc.,60,115-123.
Rinehart R.E., 1979: Internal storm motions from a single non-Doppler weather
radar. NCAR Technical Note, NCAR/TN-146+STR.
Wang, T-C. Chen, 1988: The radar analysis of two precipitation systems during
1986 Mei-Yu season. Papers Meteor. Res., 11, 63-94.
Wang, T-C. Chen, Long-Nan Chang and Pay-Liam Lin, 1988: Rainfall estimate
from digital radars in Taiwan area. Tropical Rainfall Measurements. P.471-
482, A. Deepak Publishing.

170
REMOTE SENSING OF ATMOSPHERIC COMPOSITIONS AND OPTICAL
CHARACTERISTICS
Lin Hai
Lu Daren
(Institute of Atmospheric Physics, Beijing)
Zhou Xiuji
(Academy of Meteorological Science, Beijing)
ABSTRACT
The research results on atmospheric optical
characteristics, compositions and visibility with the
lidar system and solar spectrophotometer developed in
recent years by the Institute of Atmsopheric Physics,
Chinese Academy of Science and the Academy of
Meteorological Science, State Meteorological
Administration are presented in this paper. The results
contain retrieval algorithms of active and passive
optical remote sensing and field observation of aerosol
size distribution, refractive indices, extinction
coefficients, meteorological visibility, ozone and water
vapor, Especially, we developed simultaneous inversion
of aerosol size distribution and complex refractive
indices with extinction-forward scattering method, lidar
remote sensing of slant visual range at airport, and
differential absorption lidar (DIAL) technique of
measuring atmospheric composition*
Based on lidar and spectrophotometer field
observations, this paper also analyses and compares
atmospheric optical characteristics in Beijing, Tibetan
Plateau and desert area of northwestern China, including
termporal and spatial variations of aerosol parameters
and thier variation mechanism, and presents validating
results of lidar measuring visibility with other mthods.
In addition, examples of lidar measuring ozone and water
vapor concentrations are given.
Finally, the future plan of optical remote sensing
of atmospheric trace gases and aerosol particles will be
introduced.
* Present affiliation: National Nature Science Foundation of China.

171
Atmospheric optical remote sensing includes passive remote
sensing by means of photometers and active remote sensing with the
lidar system. The laser atmospheric remote sensing has a history of
less than thirty years. It began in 1963, the fourth year after the
first laser came into being. The development of laser atmospheric
sensing, to a great degree, depends on the development of laser
technology. Before the mid 1970 f s, the laser radar was mainly based
on the principles of elastic scatteringCmolecular and aerosol), which
realized quantitative or semi-quantitative measurements of atmospheric
boundary aerosols, visibilities, stratospheric aerosols and clouds
etc.. Since then, thanks to the emergance of various tunable lasers
one after another, research on laser remote sensing of trace gases has
made great progress. Remote sensing of atmospheric water vapour
distribution, ozone and other trace gases has also obtained some
achievements. During this period, the differential absorption
lidar(DIAL) was mainly used. Looking forward to the 10-20 years from
now, along with the development of laser and photoelectron technology,
laser remote sensing from satellites and ground-based optical remote
sensing of local environmental compositions will gradually occupy an
important and special position.
Research on laser atmospheric remote sensing in China began
relatively early. In 1966,the Institute of Atmospheric Physics and
the Shanghai Institute of Optics and Fine Mechinery, Chinese Academy
of Science developed cooperatively the first 100-MW ruby laser
radar'- 1 -' in China. It was only four years after the first ruby lidar
used in atmospheric probing in the world and its performance was close
to the meteorological lidar developed by the Stanford Research
Institute in the USA at that time. After then, type II, III lidars
developed later came into use in succession in the probing of cloud
layer^, atmospheric optical parameters^|5^, visibility ^6" 8^,
aerosols I- 9 > 1 °} f smoke plumes dispersion^11^ and atmospheric diffusion
parameters'- 12^ etc. In 1980, the first tunable ruby lidar in China,
based on the differential absorption principle, was developed and
realized the probing of water vapour distribution in the lower
atmosphere^131
0

172
In the meantime, with the increasing concern about the impact of
atmospheric trace gases and aerosols on the environment, climate and
ecology in recent years, we have strengthened the research on the
probing of atmospheric trace gases and aerosols and developed
successfully some inversion methods, such as solar extinction-forward
scattering method of remote sensing of aerosol size distribution and
complex refraction indextl^-loj^ and some other combined inversion
methods. Through these methods, the ordinary photometer technology
can be used in a broader area.
1. The Laser Probing of Visibility
Laser radar is almost the only effective means of probing
visibility due to the fact that it can probe efficiently the spatial
distribution of atmospheric extinction coefficient* In the last 20
years, there were many studies in using laser tchnology to probe
visibility in both China and abroad^6»7,18-20]^ but all these were
still at the theoretical and experimental stages. Our research on the
laser probing of visibility has put significant step. A small version
of YAG lidar for visual range probing has been using preliminarily at
some airports in China.
1. Visibility mainly depends on the extinction coefficient of
the path, the brightness of observed objects and its background, and
the brightness of the sky. Lu Daren*-^ analyzed the relationship
between visual range and other atmospheric and background parameters,
derived some simplified relations for lidar-sensing visual range.
Zhao Yanzeng^J used a lidar to measure the atmospheric extinction
coefficient, and used a photometer to measure brightness of targets,
background and the sky, solving successfully the probing of slant
visibility. Her results coincide well with pilots* visual
observations.
2. Qiu Jinhuan^1°-f,
based on radiative transfer computation
results, presented an empirical formula with which the proportion of
downward average sky brightness to quasi-horizontal sky brightness can
be calculated. Since the target and background can be considered as a

Lambert reflector, there is no need to measure the brightness of
targets and background with a photometer. At the same time, through
the investigation of the properties of quasi-horizontal sky brightness
and the reflective properties of airport's cement runways and white
marks on the runway, he suggested the concept of "effective
reflectivity"* According to this concept, low visibilitiesdess than
2 kilometers) can be probed^.
3* In the numerical research based on the double scattering
lidar equation*- 21 J, the empirical correction of multiple scattering
was made to the extinction coeficient*- *, so as to raise futher the
probing precision of low visibility.
4. By means of IAG frequency doubling technology^, we adopted
a laser beam with a wavelength of 530 nm; it is closer to the
chromolight that is most sensitive to the human eyes(550nm), so it is
quite favourable for the improvement of probing precision of
visibility.
5. The validating experiments made at airports showed'- 22 *' that
the correlation between the laser probed and human eye observed
visibilities is as good as 0.93Cbased on 65 pairs of horizontal
visibility samples). For various ranges of visibility, the errors are
small enough to meet the requirements of aviation. When the
visibility is less than 500 meters, average error is 44 meters. As to
slant visibility, the comparing results of 21 pairs of samples are:
the average error of laser probed visibility and human eye observed
slant visual range is 10 percent; the maximum error is 33 percent*
Only two pairs of samples exceed the limitation of aviation for
probing precision of visibility(20 percent).
173
2. Multi-Parameter Remote Sensing of Aerosol Optical Properties
Remote sensing of atmospheric aerosols is better than in-situ
measurements in many situations. It can maintain the natural state of
atmospheric aerosols, and produce timely and promptly the spatial and
temporal variation of aerosols. At present, the passive remote
sensing of aerosols is mainly with the solar extinction

174
method (scanning frequency) and the forward scattering method (scanning
angle). Laser radar has the original advantage in aerosol probing,
By combining these methods, it is possible that aerosol concentration,
size distribution and complex refraction indices etc. be obtained
simultaneously:
1. Qiu Jinhuan, Zhou Xiuji and Zhao Yanzeng*-^] introduced a
sensitive function of volume scattering function to the real part and
imaginary part of the refraction index. They pointed out that the
volume scattering function has two relatively sensitive scattering
angle areas for both real and imaginary parts. They are in the
vicinity of 180° and 90° for the real part and 180° and 50° for the
imaginary part. The numerical experiments show that, through
selecting reasonable channels on the basis of sensitive analysis and
correlation analysis, it is feasible for a certain aerosol size
distribution to probe remotely the real and imaginary parts of the
refractive indices with the forward scattering method*
2. Because the aerosol extinction and forward scattering are
sensitive to different effective aerosol size ranges, Lu Daren, et
al*- 2 •* presented a combined extinction-forward scattering method of
probing aerosol size distribution and complex refractive indices.
They made numerical experiments in two algorithms-sectorized
retrieval, and simultaneous retrieval, and obtained satisfactory
results* Measurements with refitted photometer were also carried
3» Since 1975, we have carried out comprehensive aerosol
observations with solar photometers in succession in Beijing, Lhasa of
Tibet, and Taklamakan Desert, and obtained respectively the optical
depth (atmospheric turbidity), aerosol size distribution and complex
refractive indices. Here we use the wavelength of 550 nm as an
example to compare the results from different locations. In Beijing,
from October to early November, the sky is relatively clear and the
atmospheric turbidity is between Ov24-0.5 1 r^; in winter, the
observed atmospheric turbidity is up to 0.77*" •* and the Junge
parameter is between 2.7-3- 1 * because of heating and more serious
pollution. In the Taklamakan Desert, high atmospheric turbidity often

175
occurs and the average is 0.^1. During a dust storm, horizontal
visibility can be two orders of magnitude less than the normal and
atmospheric turbidity is up to 4.73^3. xhe actual value may be
higher because the severe dust storm may totally extinct the direct
sun light so that we were unable to make observation. The normal
Junge parameter in Taklamakan Desert is 2, basically belonging to
neutral extinction. During a dust storm, anomalous extinction may
happen and the Junge parameter is less than 2. There is a large
member of dust particles in the atmosphere. In Lhasa of Tibet,
because of its location(on plateau) and extremely less artificial
pollutants, the atmospheric turbidity is only half that of Beijing;
r7i
the maximum is 0.20 L

176
For sharp absorption lines, the two channels close to the lineCinside
and one outside the line) can be easily selected. These two channels
will be close enough so as their contributions of aerosol extinction
and molecular scattering can be considered as the same, and with
various methods we can obtain the contents of the composition with
relatively strong absorption. But when the contents of the
composition is low and the aerosol content is high, there will be
obvious errors. In this case, aerosol and molecular extinction
corrections have to be made. Using the multichannel method presented
rp7l
by Wang Pengju and Zhou Xiuji L^' J the correction can be made. They
measured successfully the total ozone content and total water vapor in
Lhasa. In June of 1986, they found through the ozone Chappuis band
retrieval that the daily mean of total ozone content is 306(m-cm~atm)
in Lhasa. The value coincides well with the valueCmonthly mean of the
year) calculated through linear interpolation by latitudes from the
data observed at Xianghe, Hebei province(40° N), with Dobson ozone
spectrophotometer in their observations. At the same time, they also
obtained the total moisture content of Lhasa through the band of water
vapor.
In 1985 Zhao Yanzeng et al. •* made up a fast-tuning element
with calcite-KDP crystal. With the element they equipped a ruby laser
and tuned to two wavelengths within vapour absorption lines, 694.380
nm and 69^.215 nm. This DIAL system can change the wavelength from
absorption peak to valley within 100-300 microseconed. The stability
of the wavelength was better than +0.002 nm and the width of the line
is 0.001 run . Therefore, the remote sensing of water vapour
distribution can be carried out with this DIAL system. The probing
range is 2-3 kilometers'- 1^;'.If we futher improve properties of the
instruments and equip them with automatic data processing system and
vehicles, it is possible to realize the remote sensing of threedimensional
water vapour field in a range of several tens of
kilometers.
The description above is an outline of our research on remote
sensing of atmospheric compositions and optical properties. More
detailed results are described in respective references.

In order to futher broaden applications of laser in atmospheric
probing, we have begun some theoretical research. For example, Huang
Runheng^29^ put forward the idea of using laser .scintillation effects
in wind sounding, Qiu Jinhuan and Lu Daren^0^ studied preliminarily
the Inversion algorithm of aerosol remote sensing from space-borne
lidar. It is expected that, in next ten years, space-borne and airborne
lldars, along with microwave active remote sensing, will improve
greatly the earth observation system.
177
References
[1]* Meteorological Lidar, "The Collecting Issue of the Institute of
Atmospheric Physics; Laser Application on Meteorological
Monitoring", China Science Press, No. 1, (1973).
[2]* Laser Cloud Probing, ibid.
[31* Lidar Application to Atmospheric Extinction Coefficient and
Visibility Eemote Sensing, ibid.
[4]* Lu Daren, Wei Chong, Lin Hai, "Vertical Distribution of the
Extinction Cooefficient of Lower Atmosphere explorated By Lidar",
Scientia Atmospherica Sinica, 3, 199-205(1977).
[53 Zhou Xiuji, Qiu Jinhuan, "Characteritistics of Atmospheric
Extinction«to-backscattering ratio in ruby lidar measurements",
Advances in Atmospheric Sciences, 1, No. 2, 179-287(1984).
[6]* Lu Daren, Wei Chong, Zhan Jianguo, "An Experiment of Visibility
measurement by Laser", Scientia Atmospherica SIniea, No.1,55-
61(1976).
[71* Zhao Yanzeng, Tao Lijiun and Hao Nanjun, "An Experiment of
Slant Visibility measurement by lidar 11 , Scientia Atmospherica
Sinica, 1980, 1, No. 2, 168-175(1980).
[8]* Qiu Jinhuan et al., "An Experimental Study of Airport Runway
Slant Visual Range Measurement by Lidar", Scientia Atmospherica
Sinica, 1988, J£, No.3, 292-300(1988).
[9]* Qiu Jinhuan et al., "Remote Sensing and Analysis of Aerosol
Optical Properties in Beijing", Acta Meteorologlca Sinica,
Mi No.1, 49-58(1988)>

178
[10]* Sun Jinhui, Qiu Jinhuan et al., "Distribution of Strateosphere
Aerosol Back- Scattering Ratio Measured by Lidar", Scientia
Atmospherica Sinica, J£, No.4, 431-436(1986).
[11]* San Jingqun, Yang Ming, "Lidar Observations of Smoke and Dust
accumulation Layer", Scientia Atmosphirica Sinica, No. 2,
156-158(1977).
[12]* Sun Jingqun et al., "Remote Probing of Atmospheric Diffusion's
Parameters by Lidar", Scientia Atmospherica Sinica, No.1, 36-
42(1977).
[13]* Zhao Yanzeng, Wu Shaoniin, "Ruby Lidar for Remote Sensing Water
Vapor in Atmosphere*, The collected Papers of Atmospheric
Probing, 33-37(1983).
[14]* Qiu Jinhuan, Wang Hongqi, Zhou Xiuji, Lu Daren, "Experimental
Study of Combined Remote Sensing of Atmospheric Aerosol size
Distribution by Solar Extinction and Forward Scattering Method",
Scientia Atmospherica SInioa, 1, No.1, 33-41(1983).
[15] Qiu Jinhuan, Zhou Xiuji, "Simultaneous Determination of Aerosol
Size Distribution and Refraction Index and Surface Albedo from
Radiance——Parti : Theory", Advances in Atmospheric Sciences,
1, No.2, 162-171(1986).
[16] Qiu Jinhuan et al., "Simultaneous Determination of Aerosol Size
Distribution and Refraction index and Surface Albedo from
Radiance PartII, Application", Advances in Atmospheric
Sciences, 1, No.2, 342-348(1986).
[17]* Xue Qingyu, Niu Jiangguo, Zhou Xiuji, Zhao Xuepeng, "Development
of a Solar and Skylight Spectrophotometef for Trace Gas and
Aerosol Studies", The Master Paper of Academy of Meteorological
Science, (1989).
[18]* Qiu Jinhuan, "Numerical Experiment on Slant Visibility and Its
Calculation Formula", Scientia Atmospherica Sinica, 12. t No.4,
404-411(1987).
[19] Hagard A., "Extinction and Visibility Measurements in the Lower
Atmosphere with UY-Y.AG Lidar", The 13th ILRC, NASA Conference
Publication, 14(1986).
[20] Balin Yu et al., "Lidar Measurements of Slant Visual Range", ibid*

[21]* Lu Daren, "Lidar Equation Taking Consideration of Double
Scattering and its Application for Low Visibility Measurements",
Acta Geophsica Sinica, £5, No.1, 1-9(1982).
[223* Qiu Jinhuan et al«, "Simultaneous Measurements of RVR and SVR
and Cloud Height by Lidar 11 , Scientia Atmospherica Sinica( special
issue), 330-340(1988)
[23]* Qiu Jinhuan, Zhou Xiuji, Zhao Yanzeng, "Theoretical Analysis of
Remote Sensing of Aerosol Refraction Index with the Forward
Scattering Method", Scientia Scientia Sinica B, No. 10, 958-970
(1984).
[24]* Lu Daren, Zhou Xiuji and Qiu Jinhuan, "Theory and Numerical
Simulation of Remote Sensing of Aerosol Size Distribution by
Combined Solar Extinction and Forward Scattering Method",
Scientia Sinica, No.12, 1516-1523(1981).
[25]* Lin Hai, Wei Chong, "Primary Measurements of Solar Visible
Radiation and Atmospheric Transmit tances in Beijing", Scientia
Atmospherica Sinca, No.2, 52-61(1976).
[26]* Wang Dongliang, Qiu Jinhuan, "The Study of Atmospheric Aerosol
Optical Properties in Taklamakan Desert in Spring", Scientia
Atmospherica Sinica, 1£, No.1, 75-81(1988).
[27]* Wang Pengju, Zhou Xiuji, "Measurements and Analysis of
Atmospheric Optical Properties over-Tibet an Plateau", Journal of
Academy of Meteorological Science, 1, No.1, 46-55(1988).
[28]* Zhao Yanzeng et al., "An electro-optically Fast-Tuning Ruby
Laser for Water Vapour Sounding in Atmosphere", Chinese Journal
of Lader, J2, No.4, 44-49(1985).
[29]* Huang Runheng, "A Scheme on the Measurement of Wind by the Laser
Scintillation Effects", Scientia Atmospherica Sinica, No.1, 44-
49(1977).
[30]* Qiu Jinhuan, Lu Daren, "A Study of Inversion Algorithm for
Determining Atmospheric Aerosol Profile from Simulated Spaceborne
Lidar Signals", Scientia Atmospherica SinicaC special
issue), 258-270(1988).
179
(*, in Chinese)

180
COMPARISION OF RADAR ESTIMATES AND SURFACE RAINFALL
DURING RAINSTORMS IN 1987-88
H. LAM and L.O. LI
Royal Observatory, Hong Kong
INTRODUCTION
This paper presents the comparison of radar estimated rainfall
with surface rainfall during major rainstorms in 1987-88. Correlation
and ratio of radar to gauge measurements were investigated and
relations between radar reflectivity and rate of rainfall in Hong Kong
were inferred.
The Royal Observatory (RO) operates a 10-cm digital radar system
since 1983. The radar has a beam width of 2 degrees, a peak power of
650 kW and the pulse length is 2 fts and is electronically calibrated
routinely. The radar completes a volume scan once every 6 minutes
during which time the antenna is elevated in 12 steps from angles of .5
to 34.7 degrees. The reflectivity data are digitized and converted to
rainfall rate using Z =200 Rl-6.
Five-minute rainfall amounts from 24 telemetering tipp ing-bucket
raingauges operated by the Observatory were used in this study. They
are distributed over an area of approximately 44 km X 60 km, all within
about 40 km of the radar site. (Figure 1)
METHOD OF ANALYSIS
Nineteen major rain cases during 1987-88 were chosen.
For each
case, the 5-minute rainfall amounts from available gauges were
extracted. The CAPPI data consist of rainfall rate with resolution of
2 km X 2 km in the horizontal and 1 km in the vertical. The 3 km
CAPPPI element closest to each gauge was extracted and interpolated in

time to derive the radar estimated rainfall for the same 5~minute
periods as the raingauge. The CAPPI element corresponging to each
gauge is highlighted also in Figure 1. Data pairs with gauge value
less than ,5 mm and radar value of less than .05 mm were excluded.
For each gauge site, the correlation coefficient (c) between
radar and gauge data, the sum of radar estimated rainfall (Ra) and the
sum of gauge rainfall (Rg) over the duration of the storm as well as
the ratio of radar to gauge total (Ra/Rg) were computed. The mean
(Ra/Rg) and standard deviation (s.d.) of Ra/Rg over all available
gauges for a rain case were then calculated. Assuming that extreme
values of Ra/Rg were likely to result from bad radar or gauge data,
Ra/Rg from a gauge site which was more than two s.d. from the mean was
discarded. The analysis was repeated in each rain case with 1 km CAPPI
data to study the differences in the radar estimates in the vertical.
181
RATIO OF RADAR TO RAINGAUGE TOTAL
Table 1 summaries the comparison of radar to gauge rainfall total
for the rain cases investigated in this study. The first case in which
hail was reported will be discussed separately because of large spatial
variation of Ra/Rg (a s.d. of .98 vs a mean of .97). The average of
Ra/Rg (3 km) over the remaining 18 cases is .48, indicating that the
radar estimated rainfall is only about half of that recorded by
raingauges. The average relative dispersion about the mean (s.d.
/mean) for the 18 cases is 34 % showing considerable in-storm spatial
variation in Ra/Rg.
In the 6 cases of tropical cyclones, Ra/Rg ranges from .12 to
.52, which are on the low side. Except in the last two cases involving
Typhoon Warren, the centre of the tropical cylcones were more than 200
km from Hong Kong. This suggests that the radar tends to underestimate
surface rainfall more seriously in the peripheral of tropical cyclones.
In the 9 trough cases, Ra/Rg ranges from *30 to .92 and tends to be
higher for troughs in the months of April and May but lower for those
in the months of June and July. The ratio Ra/Rg tends to be more
variable spatially for tropical cyclone cases (average relative
dispersion of 41%) than for trough cases (average relative dispersion

182
of 29 %). The Ra/Rg and the relative dispersion associated with cold
fronts are close to those of the average for the 18 cases. The radar
also tends to underestimate more seriously surface rainfall during the
months of June to August (average Ra/Rg of .39) than during April and
May (average Ra/Rg of .60).
In all cases except 4 (marked with * in Table 1), there is no
significant difference between Ra/Rg derived from 3 km and 1km radar
estimates, confirming that the use of 3 krn CAPPI in general is as good
as 1 km CAPPI in estimating rainfall amount. In the 4 exceptional
cases, the 3 km Ra/Rg are 120%, 71%, 34% and 60% of the 1 km Ra/Rg for
the trough of 870406, cold front of 870412, Tropical Storm Ruth and
Typhoon Betty respectively. During the passage of a cold front and
during light rain conditions under the peripheral of a tropical
cyclone, 3 km radar data can substantially underestimate surface
rainfall. It may be worthwhile to use lower CAPPIs in such conditions.
Ra/Rg can vary considerably within a relatively short period of
time. Examples are the trough situations of 870405-07 and 870727-30.
The synoptic conditions in the respective periods remain more or less
unchanged. However, Ra/Rg changes significantly within the periods and
it is considered more appropriate to stratify the rain episodes into
three and two rain cases respectively.
CORRELAION OF RADAR AND RAINGAUGE ESTIMATES
The mean correlation coefficient (c) over available gauges for
each rain case is presented in Table 2. The average of c over all rain
cases is .46 with a mean relative dispersion of 55% for 3 km while
those for 1 km are .58 and 3.7% respectively. This shows that 1 km
CAPPI data are better correlated with surface rainfall than 3 km in
general.
The correlation between gauge and radar rainfall improves with
decreasing time resolution, when the time interval for data analysis
increases from 5 to 60 minutes, the spatial variability of c decreases
and the c increases for all rain cases which cover a sufficiently long
period of time (Figure 2). This is expected because coarser time
resolution tends to reduce gauge and radar sampling differences. For

time interval of 60 minutes, the average of c is .80 with a mean
relative dispersion of 16 % for the nine rain cases shown in Figure 2.
Figure 3 shows the correlation of gauge data with the radar data
25, 20, 15, 1Q and 5 minutes before as well as 5 and 10 minutes after.
For almost half of the 19 rain cases, the gauge data are best
correlated with the radar data 5 minutes before. For the remaining
cases, the gauge data are best correlated with the radar data of the
same time. Similar results are obtained with 1 km CAPPI data.
183
THE RAIN CASE OF 15-18H,870405
On April 5 1987, an active trough of low pressure developed over
the South China coast bringing squally thunderstorms and heavy rain to
Hong Kong. Heavy showers affected the northwest part of the territory
at first and slowly spread across Hong Kong. Around 17H hail was
reported in Lai Chi Kok (roughly between R17 and R01) and at Sha Tin
(station SHA). The deluge continued into the next day.
The Ra/Rg for this rain case shows extreme variation spatially.
The Ra/Rg (3 km) for R17 is 3.5 while the mean Ra/Rg over remaining
gauges is only about .5. The time series of radar and gauge rainfall
for location R17 are shown in Figure 4a. The 3 km radar trace contains
a peak with little corresponding gauge catch around the time when hail
was reported. A similar, though weaker peak occur in the 1 km radar
and gauge trace. The peaks in the radar return are likely due to
presence of hail. This feature in the radar and gauge traces can be
useful for identifying hail. However, it does not necessarily occur
when hail is present. For in the case of Shatin, Ra/Rg is .58 and
there is reasonablly good agreement between radar and gauge traces
(Figure 4b). For this rain case, a larger time lag between radar and
gauge data is observed. When all gauges are considered, the gauge data
are best correlated with radar data 10 to 15 minutes before (Figure 3).
RELATION BETWEEN RADAR REFLECTIVITY AND SURFACE RATE OF RAINFALL
Possible sources of error which give rise to the disagreement
between radar and gauge measurements have been discussed in details by
others, see for example Wilson and Brandes 3], They can be grouped

184
under 1) errors in estimating radar reflectivity factor, 2) variations
in the Z-R relationship/ 3) gauge and radar sampling differences and 4)
errors in gauge measurement. More investigation will be required into
say the horizontal wind, vertical motion as well as the storm element
size, intensity, duration and motion of rain etc., if the results above
are to be interpreted with respect to the these sources of errors. If
it is assumed that variations in the Z-R relationship (i.e. the
variation in the rain drop size distribution) is the main factor
contributing to the differences between radar estimated and gauge
rainfall, data collected in this study may be used to infer appropriate
Z-R relations for the rain cases in the Hong Kong as reported below.
For each rain case, the 5-minute radar and gauge rainfall pairs
are first converted into rate of rainfall which are then averaged over
the duration of the storm and all available gauges. The mean radar
estimated rainfall intensity is then inverted to obtain the mean
reflectivity (Z) for the storm, using the Marshall- Palmer relation.
Linear regression equations can be obtained for the log Z and the log
of mean rate of rainfall from gauges(R). The results obtained for
various catagorization of the 18 rain cases are shown in Figure 5.
CONCLUSION
In this study, the correlation and ratio of radar and gauge
rainfall in 19 rain cases during 1987-88 were examined to gain better
insight into the performance of the digital radar in various types of
weather systems in Hong Kong.
The radar system is found to underestimate gauge rainfall by
about 50 % on average. It tends to underestimate more seriously in
months, of June to Augugst than in April to May. Rainfall estimated
from 3 km CAPPI can be significantly different from 1 km CAPPI during
passage of cold fronts and at the peripheral of tropical cylcones.
The average correlation coefficient over all rainstorms is .46.
The correlation between radar and surface rainfall improves with
decreasing time resolution. In general the 5-minute gauge rainfall are
best correlated with radar estimated rainfall for the same time period
or for the previous 5 minutes.

The digital radar at RO has been very useful for indicating the
occurrence of intense rain and is indispensible for the issuance of
landslip and flooding warnings during inclement weather. However, for
estimating quantitative rainfall and for producing objective rainfall
forecast2] r means of calibrating the radar with raingauges should be
used.
185
ACKNOWLEDGEMENT
The authors wish to thank Mr. F.C. Lam for supplying programs to
extract CAPPI data, Mr. Y.C. Yeung for his help in data processing and
Mr. L.K. Yau for supplying software to plot radar/gauge time traces.
REFERENCES
1. Austin, P. M. r "Relation between measured radar reflectivity and
Surface Rainfall", Mon. Weather Rev., Vol 115, 1053-1071 (1987).
2. Lam, C.Y.,"Digital Radar Data as an Aid in Nowcasting in Hong Kong",
Proc. Nowcasting-II Symposium, Norrkoping, Sweden, 3-7 Sept. 1984.
3. Wilson, J, W. and Brandes, E. A., "Radar Measurement of Rainfall - A
Summary", Bull. Amer. Meteor. Soc., Vol 6iQ, No. 9, 1048-1058 (1979).
4. Zawadzki, I., Desrochers, C., Torlaschi, E. and Bellon, A., "A
Radar-Raingauge Comparison", Preprints, 23rd Conf. on Radar
Meteorology, Snowmass, Amer. Meteor. Soc., 121-124 (1986).
Notes for the Tables: 1) I indicates that gauges with Ra/Rg outside
2 s.d. of the Ra/Rg.for the storm, have been retained. 2) * denotes
cases with significant difference between 1 km and 3 km radar estimated
rainfall. 3) Weather is the severest weather reported during the rain
case. TS denotes thunderstorms. 4) The maximum intensity (MI) is the
average of the 10 largest rate of rainfall detected by the radar over
the area of Hong Kong during the rain case. 5) The RO gauge total is
the sum of the gauge data in the series of available 5-minute
radar/gauge pairs for the Royal Observatory (RD1) during the rain
case. It is not identical with the rainfall recorded at the Royal
Observatory for the same time period because radar data is not always
available.

186
Table 1.
Summary of Ra/Rg for the 19 cases
—Ra/Rg (3 km) —
Time
Type(weather) mean s.d. gauges
87040515-0518 trough(TS,hail) .97! .98 19
87040518-0521 trough(TS)
.35 .15 19
87040600-0703 trough(TS)
.79* .12 20
87041206-1212 cold front(isol TS) .37* .12 21
87041212-1400 cold frbnt(isol TS) .59 .18 21
87050308-0312 cold front(TS) .64 .23 12
87051600-1700 trough(water spout) .92 .23 15
87052200-2300 trough(TS)
.49 .19 15
87061900-2000 T.S. Ruth(showers) .12* .06 11
87072200-2306 trough(TS)
.57 .19 14
87072712-2900 trough(TS)
.53 ,15 17
87072900-3003 trough(TS)
.30 .06 15
87081500-1600 T. Betty(isol TS) .23* .08 16
87081600-1714 after T. Betty(TS) .44 .11 14
87082000-2100 T. Cary(TS)
.33 .13 10
88052606-2621 trough(TS)
.62 .18 9
88062300-2500 trough(TS)
.46 .13 16
88071900-1906 T. Warren(TS) .36 .16 10
88071916-2008 T. Warren(TS) .52 .28 16
—Ra/Rg (1
mean s.d.
.80! .63
.42 .17
.66 .15
.52 .14
.56 .19
.73 .20
.87 .27
.58 .22
.35 .16
.54 .21
.48 .09
.33 .06
.38 .14
.44 .12
.43 .12
.56 .19
.43 .11
.47 .15
.52 .25
km) —
gauges
19
21
20
21
21
12
16
15
14
15
14
15
15
15
9
9
16
9
16
Table 2.
Mean correlation coefficient for the 19 .cases.
Time
87040515-0518
87040518-0521
87040600-0703
87041206-1212
87041212-1400
87050308-0312
87051600-1700
87052200-2300
87061900-2000
87072200-2306
87072712-2900
87072900-3003
87081500-1600
87081600-1714
87082000-2100
88052606-2621
88062300-2500
88071900-1906
88071916-2008
3 km CAPPI-—
mean %s.d . gauges
,24
.35
,49
.32
.19
108
63
31
75
137
17
19
18
19
20
.50 58 10
.51 39 15
,38
,48
,49
.50
.64
.52
.55
.54
.52
.42
71
48
69
54
17
48
40
54
35
50
13
11
13
16
14
15
12
9
8
15
.53 28 9
.66 20 •16
1 km CAPPI —
mean %s,d , gauges
,47
.58
.68
.43
.30
.57
.70
.64
.58
.66
.53
.72
.67
.58
.58
.48
.53
,57
.76
47
34
19
47
60
54
21
33
45
45
42
13
31
41
43
50
42
23
16
12
19
12
14
15
8
11
10
11
11
11
10
12
10
5
7
10
4
13
MI
(mm/h)
172
81
97
63
43
66
102
68
26
57
38
56
35
40
35
50
55
74
81
RO gauge
total (mm)
4.5
36.0
136.0
43.0
21.0
31.5
61.5
73.0
42.0
66.5
79.5
139.5
21.5
54.5
38.0
28.5
120,5
70.5
86.0
mean
,46
55
.58
37

187
FIGURE 1.
LOCATION OF RADAR AND TELEMETERING RAINGAUGES IN HONG KONG
30
is
z
K-A
27
^28
R2l||
IJ
^IF
SHA4
R31
RADAR
\r _\
^
V
X
J
n
35
Selected
s, $
Vr
£2^
clement
for
raingauge
FIGURE 2. CORRELATION COEFFICIENT BETWEEN 3 KM CAPPI AND GAUGE DATA
VS SAMPLING INTERVAL
MEAN CORR COEFF OVER AVAILABLE GAUGES
LEGEND
o 87040600-0703
+ 87041212-1400
* 87051600-1700
a 87072200-2306
* 87072712-2900
* 87072900-3003
. 87081500-1600
^ 88062300-2500
x 88071916-2008
SAMPLING INTERVAL (MINUTES)

188
FIGURE 3. CORRELATION COEFFICIENT OF GAUGE AND 3 KM CAPPI DATA VS THE
TIME THAT RADAR DATA LEAD GAUGE DATA
U.8
MEAN CORR COEFF OVER AVAILABLE GAUGES
0.6
0.4
0.2
0.0
-0.2
•0.4
^-^^^^^
^-*** ^ ^^^::::^!!^^::::::^^
^ ^^^^
J9&'
f
uo'S^^^^^^
^
X * 37040515-0518
d i57041212-1400 h * 37061900-2000
a 87040518-05121 e 87050308-03: L2
i 87072200-23( 36
b
f
j
t i37081500-1600 m 87081600-17; L4 n
u 38062300-2500 v 88071900-19( 36 w
Y
si^Cv
^r-»
1=^==^
i I I
i
-20 -10 0
10 20 30
TIME RADAR L& =yDS GAUGE DATA (MINUTES)
87040600-0703 c 87041206-1212
87051600-1700 g 87052200-2300
87072712-2900 k 87072900-3003
87082000-2100 o 88052606-2621
88071916-2208
FIGtJRE 4.
10
8
6
4
2
0
, ,L
\
mm
10
8 I GAUGE
6
4
2
[a) R17
I 1KM CAPPI
TIME SERIES OF GAUGE I5ATA AND RADAR ESTIMATES FOR THE PERIOD
15H-21H ON 5 APRIL 19 57
- , r^^
3KM CAPPI
^A-^xA__
i ijn
0
n[J]
5 i 18 21
I
^_A --
yiiljk.
mm
10
8
6
4
mm
2
10 0
8
6
4
mm
0 10
8
6
4
(b)SHA
~ 1KM CAPPI
: ...L
3 KM CAPPI
„, J\^~
I GAUGE
i
^^f\

iU
9
FIGURE 5. REGRESSION RELATION BETWEEN LN Z AND LN R
a) ALL CASES (exclud. hail case) c) CLASSIFICATION BY TYPE
-K~ all ^
CORR COEFF - 0.8(r> JpX< "^
1L*****~~
IU
9
In Z
— o— cold fronts ^f^'^
Z = 86R'»6 ^

190
DOPPLER WEATHER RADAR OBSERVATION
AND AVIATION WETHER SERVICE
Tseng Hsien-Yuan
Taipei Meteorological Center
Taiwan
ABSTRACT
The Doppler weather radar is able to detect
the echoes of the cloud, precipitation and wind
field. An experienced weather forecaster can
identify the propagation of gust fronts, downbursts
and low-level wind shears according to
Doppler radar. When air traffic controllers and
pilots get these information, they can guide the
aircraft to avoid the hazards ahead.
Since a Doppler weather radar was set in
the CKS International Airport in April, 1987,
we have collected a lot of Doppler radar data.
When the severe weather occurred, we can send
these information to the flight units by telephone
or TELEX immediately, so that they can
take necessary steps to keep the flight safety
in time.
1, Introduction
It has been over forty years since countries worldwide
set up the weather radars after World War Two. Due to the
development and the use of the weather radars in the past

were mainly focused on predicting fronts or thunderstorms,
this kind of radar is unable to satisfy the needs of the
users today, particularly, the needs for aviational weather
services.
Using the principles of Doppler effect to detect the
weather elements has been discovered and experimented by
scientists (Lhermitte, 1964, Fujita, 1963), but it is in
the recent ten years development have just been made for
replacing the conventional weather radars with the Dopplar
weather radars. For example: NOAA, FAA, NCAR, NASA, as
well as NEXRAD and TOWR ( Terminal Doppler Weather Radar )
are expected to take the place of the conventional radars
in the USA and abroad. ( Zcnic, 1982; Mueller ,1987) In
addition, European countries are also devoting themselves
to the development of Doppler weather radars; and some
commercial products therefore have appeared.
A doppler weather radar not only possess the ability of
the traditional radars to observe the clouds and the
intensity of rain's echo and show-them on PPI or RHI, but
also can calculate the average wind speed with its spectrum
width and show the result on the monitor, therefore, with
these data forecasters can indicate not only fronts,
squall-lines, thunderstorms, typhoons, but also low level
wind shears, downbursts, gust fronts, and the location of
low level wind shears, which are hazardous to flying. And
then the abovesaid data are forwarded to the control tower
or air traffic controllers, thus enabling them to take
necessary actions to guide the aircraft to avoid hazards.
Taiwan is located in sub-tropical area where the
weather change in a year is influenced not only by the
Eastern Asia continental weather system but also the air
mass from the Pacific Ocean. In order to more understand
and observe the singnificant weather in this .area and
provide the best aviational weather services, CKS( Chiang
Kai-Shek ) International Airport set up a Doppler radar in
the April of 1987 for the detection of hazards weather,
such as downbursts, turbulences, and low level wind shears
191

192
which are concerned about by the developed countries.
2. Hazardous Weather at the airport
According to the statistic of FAA, 23 severe plane
crashes happened in USA from 1970 to 1982, seven airplanes
were attacked by microburst, the others were attacked by
front wind shears, heavy rains, heavy thunderstorms, strong
gusts, and leeward side flow in rainshower. According to
the records of the events investigated.and those happened
in the rest of the world which have not been investigated
systematically, we know 30% of aircraft crashes were caused
by the hazardous weather while the aircrafts were landing
or taking off (Fox, 1988).
Actually,the so called hazardous weather at the airport
include more elements than those the FAA reported. As a
matter of fact, the elements of the hazardous weather are
usually interrelated, for example : the happening of
thunderstorms and the formation of heavy rains have something
to do with the approaching and passing cold fronts,
downbursts, gust fronts , low level wind shears, and
turbulences
caused by thunderstorms.
For the fear that the significant weather at the
airport endangers
flying, American government has invested
a.great deal of funds to undertake the projects of NIMROD,
SESAME, JAWS, CLAWS, and COHMEX. In those projects, Doppler
radars have been mainly employed to observe and watch the
weather so as to get more data about , the significant
weather at the airport, especially the microbursts
and low
level wind shears.These data can be used by the specialists
in research and will be a great help to them for more
understanding the weather system and improving the design
and application of Doppler radar in the
and at the airport.
aviational weather
According to the results of two studies, NIMROD and
JAWS, Fujita (1985) , downbursts are classified into the

193
following two kinds:
(a) Macroburst:
This is a large downburst with outburst wind extending
over 4 km(2.5 miles)in horizontal dimension. Damaging winds
,lasting 5 to 30 minutes, could be as high as 60 m/sec.
(b) Microburst:
This is a small downburst with outburst wind extending
only 4 km (2.5 miles ) or less, damaging winds as high as
75 m/sec.
The height of downburts usually remains about 1 km or
less. Flying over 1000 m can reduce or even avoid the
risks(Lin, 1984).
3. The non-meterologists' knowledge of the Doppler radar
data
Aviational weather services involve many units. These
services not only need the weather observators to provide
various weather data, but also data concerned have to be
reported to the users, aviation controllers, operational
dispatchers of the airlines, v and then the pilots in the
cockpit through ATS.
; -• Therefore, in order to secure flying safety, every
people related to aviational services should clearly
understand the weather at the airport and nearby after and
before flying, especially they should inquire about the
weather detailedly at-the schedualed. time of taking off
and landing.
In order to protect the aircraft and passengers, the
pilots should change the time of taking off or landing if
they are flying,in an area where the Doppler indicates the
appearance of significant weather or they are going to be
influenced'by the said hazardous weather.
Unfortunately, the display of significant weather on
the PPI of Doppler radar is easily neglected by the weather
forecasters and airlines personnel, because this kind of
weather usually lasts not more than 10 minutes. In order to

194
fly on schedule, take off and land on time for a high
reputation, the airlines occasionally tend to ignore flight
safety. Because respecting the weather information afforded
by the weather forecasters may sometimes delay the flights.
However, in order to cope with emergent conditions, it is
a must to follow the flight regulations and well utilize
the reports attained through Doppler.
4. Singnificant weather cases
CKS International Airport is located in Northeast
Taiwan. Many cold fronts pass over it in winter and spring,
and it is sometimes attacked or effected by typhoons in
summer and fall. Besides, there are local thunderstorms
happpening in the area of the said airport. In order to
have the signifcant weather in grasp, the airport setted
up a set of Doppler weather radar for providing the best
aviational weather services to the aircraft taking off or
landing at the airport.
In past two years,
the Doppler weather
radar at the said airport
participated in
the project of TAMEX
and gathered a lot of
data available to the
meterologists for their
reference. It also
observed many data
about the significant
hazardous weather.These
data were not only
sent to the flight
unit for their reference
but also recorded
Fig 1.Squall line of maximum horizontal
with tapes for further
reflectivity(dbz) in Doppler mode
analysis and study.
on May 17, 1987

195
Case 1: Squall line
Fig. 1 is the reflectivity
chart at
0122L May 17, 1987.
It showes that the
squall line formed at
the northern Taiwan
strait and passed through
the northwestern
Taiwan. Wide-spreading
thunderstorms/ rainshowers
with local heavy
rainfall were observed
at the northwestern
Taiwan when the squall Fig 2.Squall line of horizontal wind
line passing through.
field(m/s) in Doppler mode on May
Fig. 2 is the radial
wind field at the
17,1987
same time as fig.l . From this chart, we can find all
the wind blowing to the radar station at the quadrant
from northwest to southwest. The maximum wind area located
at the southwest, it reached to 20m/s ( severe turbulence).
Case 2: Typhoon
The tropical storm ALEX ( July 27, 1987 ) circulation
center can be easily defined when it approaching to the
northern Taiwan due to the circulation pattern is perfect.
The circulation center is located at the cyclonic vortex
center in the weak reflectivity area indicated by green
(see Fig.3). The circulation center can't be defined easily
in Fig.3-(4),(5), due to the circulation is destroyed
by the topography and the title vertical center structure
after the circulation center makes a landfall in the
northern Taiwan. We can also define the circulation center
by using the zero isodop and the core of the maximum plus

196
and minus value.
Fig. 4 is the
distribution of
horizontal radial
wind field(m/s).
According this
wind field distribution
, the
forecaster can
easily identify
the strong windshear
area and
then pass these
information to
the aircraft.
5. Conclusion
Doppler weather
radar's observation
possesses
an acute
ability to predict
the significant
weather such
as downburst
front gust, low
level wind shear,
and turbulence .
Adapting the traditional
surface
and high level
data, satellite
cloud data, and
NWP data to
Doppler weather
Fig 3.The distribution of max horizontal
reflectivity (dbz) in Doppler mode
on July 27, 1987. (1)0432L (2)0502L
(3)0533L (4)0602L (5)0632L (6)0932L
(7)1032L (8)1132L

197
radar's data
makes the weather
prediction in the
shortest time
more accurate.
After Doppler
radar was set up
in CKS two years
ago, the weather
forecasters not
only are able to
announce to the
tower controllers
the significant
weather for their
reference to take
necessary actions
for avoiding dangers
, but also
work at improving
Radar's soft and
hardware which
promotes the Radar's
automatic
warning anouncement
of nowcasting
and improves
the aviational
weather service.
Fig 4.The distribution of horizontal wind
field(m/s) at attitude
500M in Doppler
mode on July 27, 1987.(1)0432L (2)0502L
(3)0533L (4)0602L (5)0633L (6)0932L
(7)1032L
(8)1132L

198
REFERENCES
1. Donald Turnbull et. al.: The FAA Terminal Doppler
Weather Radar 24th Conf. on Radar Meteorology. Florida,
AMS 414-419, (1989) .
2. Fox, T.: Wind Shear the Most Dangerous Terminal
Event First Asia/Pacific Seminar on Wind Shear and
Weather Related Aeronautical Problems. Bangkok, Dec.
12-16, (1988).
3. Fujita, T.T.: Analytical Mesometeorology, A Review ,
Severe Local Storms, Meteor. 27, AMS. Boston, 77-125 ,
(1963).
4. _ . _— : The Downburst Microburst and macroburst.
The University of Chicago, SMRR Research Paper Number
210, 8, (1985).
5. Lin,Y.J.: Microbursts and Aviation Hazard: A Review,
Conf. on Aviation Weather and Flight Safety, Sep. 4-5,
1984. CAA Taipei, (1984).
6. Lhermitt, R, M.: Doppler Radar as Severe Storms Sensors
, Bull. Amer. Meteor.Soc., 456,587-596, (1964).
7. Mueller: Dynamics of A Thunderstorm Outflows. J.Atoms.
Scl., 44,1879-1898, (1987).
8. Tseng, H. Y., Lee, C.W,: A Primary Study on the CKS
International Airport Low Level Windshear and Turbulence.
Proc. First Regional Aviation Safety Workshop, Jan 29-30,
CAL Taipei, (1985) .
9. Tung, M,S., et. al.: CKS Doppler Observations of Wind
field within the Thunderstorm of TAMEX IOP-2, Proc.
TAMEX Workshop, June 22-30, 1989, Taipei 90-94,(1989).
LO. Wilson J., et. al.: Microburst Wind Structure and
Evaluation o.f Doppler Radar for Airport Windshear
Detection. J. Appl. Meteor.,.23..898-915, (1984).
LI. Zrnic, D. S.: Consideration for Observations of Storms.
NEXRAD Doppler Radar Symposium. 124-143, (1982).

799
Revival of the Tipping-bucket Raingauge
Sheng-I Hsu
Department of Geography
The Chinese University of Hong Kong
ABSTRACT
INTRODUCTION
A tipping-bucket raingauge was used to test the rateerror
on various rainfall intensities. It is found that
the rate-error could be as large as a 35% underestimation
of the actual rainfall amount at upper measurable limit
of the gauge. However, if a laboratory-based regression
equation was applied to normalize the rate-error on each
of 10-minute records, the error of underestimation could
be reduced to as low as 0.8%.
Rainfall measurement has been an important task in the
application of agriculture, hydrology, and socio-economic activities.
Rain measurement during intense rainstorms is particular vital to the
study of surface run-off, land-slide, and flooding problems.
Consequently, precise assessment of rainfall data is critical in
meteorological instrumentation.
Among numerous available raingauges, the tipping-bucket, on
account of its low manufacturing cost and its telemetric recording
capability, is currently the most widely used raingauge in a
meteorological station.
Its tipping character facilitates the
measurement of rainfall intensity in a short period of time.
There are
many factors affecting the accuracy of rainfall
measurement. These factors include the engineering of a gauge such as
adhesion, colour, diameter, and the exposing environment such as

200
evaporation, condensation, wind, and height of exposure. Research
papers dealing with these factors were summarized in a WMO report No,
343*'. Nevertheless, the rate-error inherent in a tipping-bucket
gauge is considered to be the most annoying problem requiring
attention.
TIPPING-BUCKET AND RATE-ERROR
A raingauge using the "tipping bucket" principle with a single
vessel was first attempted by Sir Christopher Wren in 1662. Later, a
"double-tipping bucket" was developed and was commonly used in North
2)
America and the rest of world '.
The tipping-bucket gauge consists of two balanced bucket-vessels
in unstable equilibrium. When a certain amount of water from the
collecting funnel has been added to one bucket, it tips suddenly, to
the other position.
bucket begins to fill.
The full bucket drains water while the other
Each reversal of the bucket produces a pulse
which operates a counter or recorder either mechanically or
0\
electrically . The total number of pulses accumulated will then be
used in computing the rainfall amount.
Each tip of the bucket takes approximately 0.2 second /.
Consequently, when the full bucket begins to tip with a designated
amount of water, rainfall is still running from the collecting funnel
into this bucket through the first half of its motion.
Therefore, the
amount of water collected for each tip of bucket tends to be more than
the designated amount.
The difference between the two is a function
of the rate of rainfall; the higher the rate of rainfall the bigger
the difference.
This difference, normally termed as the rate-error of
a tipping-bucket raingauge, is not serious in a light rain, but can
cause significant error in a heavy rainstorm.
In the past, great efforts had been devoted to the reduction of
the rate-error.
Major improvements include the following;
(1) To increase the bucket size so that the designated amount of
water per tip is increased.
When doing this, the rate-error can
be reduced proportionally:'-but not eliminated totally at the
expense of sacrificing the measurement sensitivity for small

201
rainfall,
(2) Donnelly ' introduced a digital raingauge recorder which is
composed of an electronic interface with a switchable mechanism
by employing three sets of tipping-buckets (rainfall resolution
of 0.013 mm, 0.064 mm and 0,254 mm), to measure the variation of
rainfall rate and to increase the gauge sensitivity.
step further toward the improvement of the
raingauge.
This is a
tipping-bucket
However, the rate-error still remains in the system
and the manufacturing cost is escalated.
6)
(3) Parkin and King ' designed and built 103 automatic recording
raingauges at low cost for use in a cloud-seeding experiment.
They minimized the rate-error by interposing a siphon device
between the collector outlet and the buckets, and claimed to have
an error of ±1% with rainfall rate up to 375 mm per hour at
resolution of 0.2 mm. The new design is by-far the most
acceptable means for the solution to the rate-error of a tippingbucket
raingauge.
Yet, calibration of the siphon rate for
removing the rate-error could be a lengthy process.
In order to rectify the rate-error, firstly, a laboratory
experiment is designed to develop a regression curve of the rate-error
on the number of tip counts.
Then, the raingauge was set up with an
electronic datalogger in the field to measure an expected forth-coming
rainstorm. Through this arrangement, the recorded rainfall amount
could be normalized in accordance with the experimental regression
equation, thus resulting in improved rainfall measurement.
LABORATORY EXPERIMENT
A tipping-bucket raingauge manufactured by Weather Measure (Model
6018A) was used in the laboratory for the rate-error test. The
rainfall resolution was set to 0.205 mm, or equivalent to 15.0 grams
of water per tip (with a 305 mm collector inlet).
Simulation of constant rainfall rate was essential to perform the
experiment.
In order to solve this problem, a mini water tower was
designed and built to maintain a constant flow rate.
As shown in
Figure 1, the water is supplied by a laboratory faucet with a macro-

202
control valve which permits ample amount of water to flow over the
tower. The spilling water is directed to a sink. At the bottom of
the tower, there is a pipe with a micro-adjustment valve which permits
a constant stream of water to flow into the collecting funnel of the
gauge. During each experiment at a fix flow rate, the total amount of
water tipped from the buckets is collected by a pan for the subsequent
computation of the mean flow rate. Depending upon the set flow rate,
each experiment runs from several minutes to several hours to make up
50 tips for observation.
A total of 80 experiments were conducted to simulate the rainfall
rate from as low as 1.5 mm per hour to an extreme of 685.8 mm per
hour. Among these experiments, 61 of them were observed to be within
the measurable limit of the instrument. The remaining 19 cases were
regarded as out of the limit, with a water stream so intense that the
draining bucket was unable to spill out the water completely before
the opposite bucket started to tip. It was observed that it required
at least 1.5 seconds to drain out the bucket water, and therefore an
upper measurable limit for this instrument is 400 counts per 10
minutes, which is equivalent to a rainfall intensity of 492 mm per
hour.
Figure 2 shows the rainfall rate and the rate-error for all 80
experiments. The data indicate a regressable change from low rainfall
up to about 492 mm per hour. Beyond this extreme, the relationship
breaks down. A regression equation within the measurable extreme (61
observations) is obtained as follow;
E = 1.266430 + 0.112730.R - 6.78523 x 10~ 5 .R 2
where E is the rate-error in percentage of underestimation, and R is
the rainfall rate represented by total number of tip counts per 10
minute.
Very seldomly will the world record of rainfall intensity exceed
the limit of 492 mm per hour. In 1956, 31.2 mm of rainfall recorded
in one minute at Unionvill, Maryland is considered to be far beyond
the measurable limit of this instrument. A rainfall of 205.7 mm
recorded in 20 minutes at Curtea de Arges, Romania is also beyond the
limit. While in 1907, a rainfall of 304.8 mm recorded in 42 minutes

203
at Holt, Missori is just barely under the limit '.
A rainfall of 50 mm per hour is regarded as an intense
rainstorm 4 '. However, much higher rainfall intensity may frequently
be experienced throughout the world. As a result, rate-error
rectification under intense rainfall becomes necessary if a tippingbucket
raingauge is used for rainfall measurement.
FIELD TEST
After calibration and the rate-error test of the instrument, a
tipping-bucket raingauge was installed in the weather station on the
campus of The Chinese University of Hong Kong. Alongside of it, a
standard gauge measuring 24-hour precipitation, and a siphon raingauge
were also installed in the station for comparative study of the
rainfall measurement.
On the 22nd of April 1989, thunderstorm brought in by a lowpressure
trough caused heavy rainfall. The total rainfall recorded by
the various gauges during the period from 9:00 am April 22 to 9:00 am
April 23 is as follow;
Gauge Type
Standard gauge
Tipping-bucket (0.205 mm/tip)
Siphon-gauge
Total Amount
110.00 mm
101.07 mm
110.00 mm
The number of tip counts in every 10-minute for the bucket gauge is
extracted and listed in Table 1. While the recording sheet of the
Siphon-gauge is shown in Figure 3.
The total amount of rainfall, (9;00 am April 22 to 9:00 am April
23, 1989) as registered both by the standard gauge and by the siphon
gauge is 110.00 mm, While the tipping-bucket registered 101.07 mm
(493 tips). If 110,0 mm is regarded as an accurate amount of rainfall
in the 24-hour period, then the rainfall measurement errors for the
bucket gauge may be computed as 8.1% underestimation of the actual
rainfall. It is apparent, that this large percentage of the rateerror
need to be eliminated in order to present a better record in the
rainfall measurement.

204
Table 1
Rainfall Records of a Tipping-bucket Gauge
(0.205 mm resolution)
NiiflberX hour
of tip \
17 18 19 20
21 22 23 24 1 2 3 4
5 6 7 8
10*
8 5 1 1
20
1 1
44 1
30
3
50 1 3 1
40
27 5
50
60
7
15
1 4
9 0
2
1
Total
1 0 28
115 7 1 12 2
3
* Ending time
RECTIFYING THE RATE-ERROR
The regression equation obtained from the laboratory experiment
is applied to rectify the rate-error for the tip-bucket. Table 2
shows the tip counts per 10-minute (column 3), the rate-error (column
4) computed from the regression equation, and the rectified rainfall
(column 5). Summation of every 10-minute rectified rainfall, from
9:00 am April 22 to 9:00 am April 23 1989, yields a total amount of
110.88 mm which is slightly over estimated by 0.8% when compared with
the actual amount 110,0 mm.
OTHER POSSIBLE ERRORS
Other possible errors of the bucket tipping process were also
observed in the laboratory. These errors are generally quite small
and variable, and consequently are difficult to quantify. For
example, the small amount of rain water remained in a bucket may vary
from time to time and is subject to the surface condition of the
bucket. This variable residual water generally causes a small
fluctuation of the measurement accuracy. It is also observed that the

205
Table 2
Rainfall Intensity and the Estimated Rate-error
(April 22 9:00 am - April 23 9:00 am, 1989)
Date
Time*
Tip Count
per 10 minute
Estimated
rate-error (%)
Rectified
rainfall (mm)
4/22
4/22
4/23
1700
1710
2000
2020
2030
2040
2050
2100
2110
2120
2130
2140
2150
2200
2210
2220
2230
0130
0140
0150
0200
0210
0230
0850
0900
(no rainfall
2
1
1
9
2
28
32
26
117
119
51
32
1973
2
12
145
1
117
1
recorded between 9:00
1.492
1.379
1.379
2.286
1.492
4.476
4,943
4.243
15.385
15.642
7.192
4.476
3.433
2.059
1.605
1.492
2.629
2,858
1.832
1.379
1.379
1.379
2.059
1.379
am to 17:00 pm)
0.42
0.21
0,21
1.89
0.42
6.00
6.88
5.56
27.68
28.21
11.21
6.85
4.03
1.46
0.62
0.42
2.52
2.95
1,04
0.21
0.21
0.21
1.46
0.21
Total 493 110.88
* Ending time
length of time for the final drop of rain water to drain out of the
bucket varies from 1.5 seconds to 31 seconds.
Since the water drop
weights 0.1 gram, any tip action occurs within 31 seconds, the
measurement accuracy may be affected by as much as 0.7% (0.1/15.0 =
0.7%). And any rainfall intensity exceeds the draining capability of
the bucket, say 1,5 seconds in this study, is considered to be out of
the instrument measurable limit as shown in Figure 2.
Combining these effects, the unpredictable minor errors for

206
various rainfall intensity should fall within 2% aside of rate-error.
Scattering of experimental data along the regression curve shown in
Figure 2, clearly reflects the marginal range of the understated
errors.
SUMMARY
On account of its low-cost, its minimal maintenance requirement,
and its telemetric recording capability, the tipping-bucket gauge has
been widely used in a meteorological station for the measurement of
rainfall amount and rainfall rate.
However, the inherent rate-error
associated with the bucket tipping motion may be as high as 35% as
observed in the laboratory experiment when measuring a very intense
rainstorm.
Removal of the rate-error is possible and practical.
The rateerror
of 8.1% before rectification is brought back to 0.8% after
rectification as demonstrated in this paper for the measurement of a
rainstorm in a 24-hour period.
Therefore, it is believed that the
rate-error removal technique recommended here ought to stimulate
reconsideration of the conventional tipping-bucket raingauge as a
reliable meteorological instrument.
REFERENCES
1) World Meteorological Organization, "Annotated Bibliography on
Precipitation Measurement Instruments", WMO-No. 343 (1973).
2) Chu, Ping-hai (ed.), "Dictionary of Meteorology", Shanghai
Publishing Co, (1985).
3) Middleton, W.E.'K., "Catalog of Meteorological Instruments in the
Museum of History and Technology", Smithsonian Institution Press
(1969).
4) Middleton, W.E.K. and Spilhaus, A.F.,"Meteorological Instruments",
University of Toronto Press (1953).
5) Donnelly, Denis P., "Digital Raingage Recorder", Journal of Applied
Meteorology 16, 205-207 (1977).
6) Parkin, D. A,, King, W...D* and Shaw, D.E., "An Automatic Recording
Raingage Network for a Cloud-seeding Experiment", Journal of Applied
Meteorology 21, 227-236 (1982).
7) Gedzelman, Stanley David, "The Science and Wonders of the
Atmosphere", John Wiley & Sons (1980).

207
valve
micro adjustment
valve
water faucet
1"
Y
/^
D
a
Figure 1
A mini water tower set-up for rainfall simulation
with steady intensity

208
c 400--
£
o
"5 350 H
— A
— B
-C
-D
A. World rainfall record in one minute
(Unionville, Maryland )
B . World rainfall record in 20 minutes
( Curtea de Arges, Romania )
C. Upper measurable limit of the
tipping-bucket'raingage.
Used in this study
D. World rainfall record in 42 minutes
(Holt, Missori )
E. A storm record in CUHK
(April 22,1989)
1550-
1500-
500-
450-
1906.5-
1845.0-
492.0-
430.5-
^ 300-
Q)
.0
369.0-
250- 307.5-
200-
150-
100-
50-
123.0-
61,5-
20 25 30
rate-error (%)
35 40 45
Figure 2
An experimental regression curve of rate-error
on the number of tip-count in 10 minutes

9 10 11 12
- 25 -
_ n A . _ _ _ _ —
1S M (5 f« 1 (I
________ 25
-._
_ _ „ . _ . . , . . - .. _ .
r:::::ii-:iriiiiz.; r:: :.:::::^-:z;.;...•.:-:::-.-: v
»
30
If 1
Hfl!'!l : t ii e i i i
l
/ : : • : ~ 25
4 J t J t 9
25
- 2 ffl|| |7 - - . . _. 2 ET —
- - -
b-z-r -- = :.:;.-::. :::•
::^^
: 15 _i.::i,::::::d~-^~z~ : -Yo~-:~-'---':~- r -~-'--
- •- - - -
._ _ . .
•-- • :
t j~i i i. i.TLni. fl.i T i i i i .i.,i.^t I i i
-
._..
--.__ziiini: = ii:iii t5 :ii
TfllTmtM 1
J
._:__
/
I i H M t 1 fi'i.'ri-f-H ll MA i u
1 M 1 1 M i I i. 1 1 M M M M u.i
Scale. 5 mm -•= 10 mm on chart Station.CbinesG. .University,. Total 9h to 9h
Recorder,.,
Check gauge
mm
mm
HtlHllD IN IHGUHD No. 57 I
Figure 3
Recording sheet of a siphon raingauge
(reduced by a factor of 0.71)

210
IMPACT OF HOURLY S-VISSR SATELLITE IMAGERY
ON OPERATIONAL FORECASTING IN HONG KONG
S.T. LAI, B.Y. LEE, and H.K. LAM
(Royal Observatory, Hong Kong)
1. INTRODUCTION
In 1977, Japan launched the first of a series of Geostatioary Meteorological
Satellite (GMS) into an orbit 35 800 kilometres over the equator
at longitude 140 deg E. Imageries in the visible (VS) and infrared
(IR) channels were available every three hours in analogue facsimile
form. The first GMS was replaced by GMS-2 in 1981 which was subsequently
substituted by GMS-3 in 1984. To cope with a change in the transmission
mode of the satellite from analogue facsimile form to
stretched-VISSR (Visible Infra-red Spin Scan Radiometer) digital data,
the original reception system at the Royal Observatory (Hong Kong) was
replaced and put into operation in November 1988.
Since then, satellite pictures became available every hour with additional
playback and enhancement capabilities. Considering the relatively
short time frame since the commissioning of the new GMS system in
Hong Kong, there is no doubt that its potential as a forecasting aid is
yet to be exploited to the full. However, there are already two immediate
visible advantages brought about by the hourly GMS imageries.
Firstly, the continuity of mesoscale weather systems, which have typical
time scales of hours and could easily go undetected on conventional
weather charts, can now be closely monitored. Section 2 describes such
a case during a rainy period in April 1989.
Secondly, the position and intensity of tropical cyclones can now be
monitored on a near real-time basis. Continuity and reliability have
also been improved by the usage of special enhancement techniques. As an
illustration, section 3 presents the life history of Typhoon Brenda
(16. - 21 May 1989).

211
In the final section, the paper concludes with some reflections and
comments with regards to adopting satellite analysis into an overall
forecast strategy.
2. MESOSCALE CLOUD SYSTEMS (3-9 APRIL 1989)
2.1 Synoptic Background
Cloudy and rainy weather persisted over the coastal areas of southern
China for about a week in the early part of April. In Hong Kong, significant
rain started on 3 April and got progressively heavier.
Rainfall over the 2-day period .of 7 - 8 April amounted to nearly 100
millimetres. The rain eased off on 9 April and the sky finally cleared
during the night. Altogether, a total of about 125 millimetres of
rainfall was recorded during the week.
Fig. 1 is a collection of the daily 500-hPa streamline charts as analysed
at 1200 UTC during the period 1 - 9 April 1989. The positions of
the surface and 850-hPa troughs are also shown for easy reference.
Initially, the surface trough stretched across central China with the
500-hPa jet located aloft over its eastern section. At the time, a
well-organized westerly trough could also be analysed over western China
at 500-hPa level. This trough subsequently broke down and the 500-hPa
flow became dominated by minor waves. While the continuity of these
waves was often hard to follow, the surface trough maintained a steady
southward movement and the 500-hPa jet also became more active south of
25 deg N.
The surface trough crossed the coast later on 4 April. It then degenerated
into a relatively weak feature over the northern part of the South
China Sea. The minor peak in precipitation in Hong Kong on 5 April (see
Fig. 5) can probably be attributed to its passage. However, an explanation
for the major rainy episode on 7 and 81 April is not so readily
forthcoming. Meanwhile, the 850-hPa trough remained quasi-stationary
over southern China. At 500-hPa level, there was no major trough and
the wind field only reflected some ill-defined short waves. What looked
suspicions, however, was the emergence of a semi-permanent jet core
over the Yunnan highlands since 3 April. This jet subsequently made two
eastward excursions - once to the coast of Guangxi on 5 April and once
to the north of Guangxi on 7 April. Those were times when cloud and
rain developments were particularly active over southwestern China.
Ironically, it was not until a major westerly trough moving out over
mainland China that clearance from the northwest occurred on the night
of 9 April.

212
Fig. 1. Daily 1.200 UTC • 500-hPa
streamline analysis for the period 1
- 9 April 1989 with the jet (winds
exceeding 25 m s"" 1 ) highlighted in
shaded areas and trough axes marked
in thin dashed lines. Positions of
surface trough (thick solid line)
and 850-hPa trough {thick dashed
line) are also shown alongside.
JOSE
HOE-

213
2.2 Observation From Satellite Imageries
A video tape of the satellite animation sequence for the period 3-9
April 1989 has been prepared for presentation. A day-by-day commentary
on the sequence of events is reproduced below.
3 April - With the surface trough moving into southern China, convection
grew more active over northern Guangdong. At this stage, the cloud
features already had strong wave-like signatures. As yet, most of the
disturbances passed to the north of Hong Kong.
4- April - Although the major synoptic feature (i.e. the surface
trough) that affected Hong Kong approached from the north, the source of
clouds and rain originated from the west. Towards the evening, small
cellular clouds with dimension of tens of kilometres formed over western
Guangdong. These developed into cloud masses of hundreds of kilometres
in dimension as they moved further east.
5 April - Isolated convective cells appeared in the morning along 22
deg N. Cloud development went into a brief lull for the latter half of
the day.
6 April - Cloud development over western Guangdong was rejuvenated in
the early hours and remained very active for the rest of the day. There
was again an evolutionary pattern in cloud size as individual systems
moved eastwards (sree Fig. 2) .
7 April - Wave-like disturbances continued to propagate eastwards
across the coastal areas. The origin of the wave trains seemed to have
receded even further west with cellular clouds forming in Guangxi as
well. Convection was also enhanced over western Guangdong and grew more
extensive as the disturbances passed over Hong Kong later in the day.
8 April - An extensive area of downstream convective development
moved over Hong Kong in the early morning (see Fig. 3) and brought the
heaviest downpour of the entire period. Convection then became more
subdued for the rest of the day and was mainly confined to isolated
cells over the coastal waters. Cellular cloud formation also became
less apparent although cloud masses continued to move in from the west.
9 April - Wave-like cloud features became less distinct. At the same
time, a clear-cut back edge appeared to the northwest of the cloud mass
(see Fig. 4.). The back edge advanced rapidly southeastwards and brought
dramatic clearance during the night.
In summary, cloud systems over southern China during this period were
characterised by strong wave-like pattern on a-'sub-synoptic scale. The
origin of the wave trains seemed to be located over southwestern China
and Hong Kong suffered from the downstream effect as the waves propagated
eastwards. At one stage, a well-defined evolutionary behaviour in
cloud size could also be observed.

I
Fig.
2. GMS-3 IR picture taken at
7:31 a.m. on 6 April 1989.
Fig. 3. GMS-3 IR picture taken at
4:32 a.m. on 8 April 1989.
Fig. 4. GMS-3 IR picture taken at
1:31 p.m. on 9 April 1989.
E 70 •
£60
= 50
£
c
o 30
>, 20
d 10
7,
X
NXXXX
y
/
y
x
vXXXN
' Jl sTI /
Jl , 0 1 . BJ] 1 , p
yt
5 6 7 8
APRIL
,J1'
7
Sri
&
o
6 jjj
5 &
A &>
« 3 |
- 2 -8
y 1 /
6
Fig. 5. Daily rainfall (shaded blocks) recorded at
Royal Observatory (Hong Kong) and daily occurrences
of wave passage (unshaded blocks) across 115 deg E.

215
To ascertain the activity of the wave disturbances, the longitude of 115
deg E was arbitrarily chosen as a reference line. Any significant cloud
system that crossed this line between the latitudes of 20 and 25 deg N
was taken to represent the passage of one wave train. The daily frequency
of wave passages is shown alongside the daily rainfall amount in
Fig. 5. Although the frequency of wave passage may not be quantitatively
accurate owing to subjective judgment, its trend nonetheless corresponds
fairly well with the trend of precipitation in Hong Kong.
2.3 Triggering Mechanisms
Considering the temporal and spatial scales of events, the cloud evolution
and behaviour during the period strongly suggests the propagation
of gravity wave across the coastal areas of southern China. As for the
triggering mechanisms behind the wave generation, effects due to terrain
are possible but rather unlikely. Mountain ranges of the Yunnan highlands
are mainly to the west of 105 deg E while most of the cloud development
took place east of 110 deg E. Convection can also be safely
ruled out as a trigger. Cloud formation over southwestern China had no
indication of convective activity as precursor. In fact, convection was
often initiated downstream and to the south of 22 deg N. Thermodynamic
instablity indices actually indicated slightly more stable conditions at
a time when mesoscale development was getting more active. The indices
also showed the area of maximum instability mainly confined to the
coastal waters which was consistent with the observed spatial extent of
convective activity.
This leaves shear instability as the major candidate. In a similar case
which occurred in the United States on 9 May 1979, Stobie et. al.
(1983)^ concluded that shear instability was an important mechanism in
the generation and maintainance of gravity waves. In the current case,
circumstantial evidences also strongly suggest the important role played
by shear instability. From Fig. 1, it can be seen that the enhanced
activity of the 500-hPa jet coincided with the period when generation of
wave trains was most rampant. As an attempt to quantify the effect of
the jet, vertical shears between 500-hPa and 850-hPa winds were computed
from the analysed wind field (over an area bounded by 20 and 30 deg N,
100 and 115 deg E) of the Royal Observatory Limited Area Model (ROLAM).
The results presented in Fig. 6 show a steady increase in vertical shear
towards 7 April followed by a dramatic drop between 7 and 9 April.
Unlike the 1979 case, the wave trains here remained active for a more
prolonged period and there was an apparent growth in wavelength as the
disturbances propagated eastwards. The latter could be an example of
upscale scattering of wave energy from the smaller scale Kelvin-Helmholtz
waves - a topic which received detailed theoretical treatment in
Chimonas and Grant (1984) 2 '.

216
110E 20 115E
30. 20" X J
4o*N'\ 'Y\ .,--'""\ / 1
. VA Pi Y"r\ ....... 2siT-
3 APRIL ] 7
*"' HONG KONG
115E
J 2b 105E
\ \
5 APRIL
110E
X"" x x
•

217
the west-northwest and swept across the islands of central Philippines
and southern Luzon on 17 May 1989. After entering the South China Sea
the following morning, it turned to a more northwestward course while
its forward speed was reduced by half. Brenda continued to intensify
during the process. It eventually attained typhoon strength on 19 May
with maximum sustained winds near its centre at the time estimated to be
about 120 km/h. It also accelerated slightly to a speed of about 18
km/h and at one stage seemed to be heading straight towards Hong Kong.
However, Brenda turned slightly to the west-northwest on 20 May, skirting
past at a distance of 130 km to the southwest. It finally made
landfall during the night over the coastal areas of western Guangdong
and dissipated rapidly over land in the morning.
A video tape on the life history of Typhoon Brenda has been prepared for
presentation using a special enhancement look-up table for tropical
cyclones. By observing the changes in cold tops and banding features
near its centre, forecasters can make reliable centre fixes and intensity
assessment. The observational aspect was covered in detail in the
video presentation. Discussion in this section will mainly concentrate
on the operational problems confronting the forecasters on the morning
of 19 May.
The provisional best track of Brenda along with forecasts and warnings
current at the time are shown in Fig. 7. JTWC (Joint Typhoon Warning
Centre, Guam) went for a basically northward track to eastern Guangdong.
ECMWF (European Centre for Medium-range Weather Forecasting) opted for a
gentle curve to the west-northwest in the general direction of western
Guangdong. Between the two extremes were the prognoses by ROLAM and the
analogue model. Both predicted a gentle curve to the north with landfall
position in the vicinity of Hong Kong. In terms of consistency,
ECMWF had been maintaining its scenario while JTWC had shifted from a
recurving scenario to a track that was further and further to the west.
Synoptically, there was a deep westerly trough moving out over eastern
China which provided some justification to the JTWC forecast. However,
ECMWF insisted that the trough would soon relax and a ridge would quickly
re-establish to the north of Brenda.
Therefore, it was a problem of assessing whether the westerly trough
would be able to pick up Brenda and the early part of 19 May was the
critical period if this should occur. Fig. 8 shows the visible GMS
picture received that morning. The cloud band extending to Japan was
associated with the westerly trough in question. It appeared to have
linked up already with the outer cloud bands of Brenda. However, subsequent
events as observed increasingly favoured those tracks which pointed
to the west of Hong Kong. Firstly, clouds previously clearing southeastern
China after the passage of the westerly trough had returned to
the region - an indication that the circulation of Brenda became more
dominant and the trough effect, if any, was on the decline. In fact,
when the picture shown in Fig. 8 was taken, the first of Brenda 1 s outer
rainbands had just reached Hong Kong, Secondly t Brenda intensified
further and an eye appeared for a short time. Good banding features
allowed a very reliable fix on Brenda's position on an hourly basis. It

218
Fig. 7. Provisional best track of Brenda (solid line) and forecast
positions (as received in the morning of 19 May) by ECMWF, JTWC, ROLAM
and the analogue model. 12-hourly positions of best tracks are designated
by ' ", and forecast positions are given by "x". Adjacent 4-digit
figures (DDZZ) give time of validity in day (DD) and hour (ZZ, in UTC).
Fig. 8. GMS-3 VS picture taken at 7:31 a.m. on 19 May 1989.

was found that instead of slowing down further, as might be expected if
recurvature was imminent, Brenda actually picked up speed to 18 km/h on
a steady northwestward course.
Post-analysis showed that ECMWF was able to capture the directional
trend but had pushed the storm too fast. Both ROLAM and the analogue
model performed fairly well in terms of distance from actual storm
positions. However, without the benefit of hindsight, it was the satellite
analysis which allowed forecasters to make timely operational
decisions on forecasts and warnings to be issued.
219
4. COMMENTS
With increasing sophistication and reliabilty, numerical models will no
doubt remain the main stay in the field of weather forecasting. Nevertheless,
satellite data (and for that matter radar data as well) must
still be regarded as a basic and important source of information which
enables forecasters to react to short term changes. Its primary role in
terms of nowcasting can hardly be replaced as yet, even with the trend
of regional models moving towards finer resolution.
Moreover, case studies presented in this paper reveal other dimensions
in the application of satellite data. Firstly, the visual impact of the
satellite imageries can impress upon forecasters the nature and characteristics
of events, particularly if the events are repetitive or conform
to a certain pattern. Armed with the knowledge of the probable
causative mechanisms, forecasters can apportion the appropriate extent
of emphasis on certain aspects of the analysis. An intelligent projection
can then be made using the critical information derived from the
analysis and the numerical products. Secondly, satellite imageries can
serve as a short term verification guide to the performance of various
prognoses. This is particularly useful in cases of conflicting scenarios
when forecasters are under pressure to make an operational choice.
With the new hourly CMS imageries, a more continuous and cohesive picture
can be assembled and this allows an optimization of the important
functions described above.
REFERENCES
1) Stobie, J. G., Einaudi, F. and Uccellini, L. W., "A case study of
gravity waves - convective storms interaction 9 Kay 1979." J.
Atmos. Sci. 40, 2804-2830 (1983).
2} Chimonas, G. and Grant, J. R., "Shear excitation of gravity waves,
2. Upscale scattering from the Kelvin-Helmholtz waves." J. Atmos. Sci.
41, 2278-2288 (1984).

220
PREDICTABILITY OF LOW FREQUENCY MODES
T.N. Krishnamurti, S. Moten, D. Oosterhof, and G. Daughenbaugh
Department of Meteorology, Florida State University
Tallahassee, Florida 32306
U.S.A.
ABSTRACT
In this paper we propose a procedure for the extended integration of
low frequency modes of the time scale of 30 to 50 days. A major
limitation of the extended integrations arises from a contamination of
low frequency modes as a result of energy exchanges from the higher
frequency modes.
In this study we show an example on the prediction of low frequency
mode to almost a month which is roughly 3 weeks beyond the
conventional predictability of these modes. This was accomplished
by filtering the higher frequency modes from the initial state. The
initial state included a time mean state and the low frequency mode.
The sea surface temperature anomalies on this time scale and its
annual cycle were also prescribed.
The specific experiment relates to the occurrence of a dry spell in the
monsoon region. The meridional passage of an anticyclonic
circulation anomaly over the lower troposphere and the eastward
negative velocity potential anomaly (implying convergence) over the
upper levels of the Indian monsoon, on time scales of 30. to 50 days,
are reasonably predicted. Suggestions for further experimentation on
the predictability of low frequency modes are proposed.

221
1. OBSERVATIONAL ASPECTS
Observations during the FGGE year and subsequent to that for the recent
8-year period have clearly shown the presence of low frequency motions on time
scales of roughly 30 to 50 days. There are several regional and global aspects of
these oscillations that have been emphasized in recent literature. Among these, we
shall be addressing the following four observation aspects of low frequency motions.
a) Meridionally propagating 30 to 50 day waves in the lower troposphere of the
monsoon region.
This is a family of trough-ridge
systems that can be seen as the
streamline-isotach charts of the time filtered motion field. The passage of a trough or
a ridge line over central India generally coincides with the occurrence of a wet or a
dry spell respectively. The meridional scale of this system is roughly 2000 to 3000
kilometers. The speed of meridional motion is roughly 1° latitude/day. During
certain years, the meridional motion and passage of these systems during the summer
monsoon season is quite regular while over other years the motion is somewhat
irregular. The reasons for this type of interannual behavior are not quite clear at the
present time.
b) Zonally propagating planetary scale divergent circulations on this time scale.
These seem to have a dominant scale of wave numbers 1 and 2. They
traverse the globe, from west to east, in roughly 30 to 50 days. The largest amplitude
is in the equatorial latitudes. The divergent circulations have a large meridional
extent, the east-west circulations can be seen even as far as 35°N and 30°S. These
broad divergent circulations appear to be related to equatorial and monsoonal heat
sources and sinks. The interannual variation of these eastward propagating waves has
also been studied and one finds that the propagations have been somewhat irregular
during several recent years.
c) Air-sea interactions.
We have recently examined the oceanic fluxes of sensible and latent heat on
this time scale. When fluxes are calculated using the so-called 'surface similarity
theory 1 , the basic variables are the SST, the surface wind, the temperature and
humidity at the top of a constant flux layer. The similarity fluxes are defined from

222
expressions that invoke the Monin-Obukhov length and a non-linear coupling of the
momentum, heat and moisture. Because of this non-linear coupling, one can diagnose
the relative importance of the low frequency variations of SST, surface wind, air
temperature and humidity and assess their role in the contribution to fluxes on this
time scale. A detailed diagnostic study was recently completed by Krishnammti et al
(1988). It was found that the latent heat flux on the time scale of 30 to 50 days can
be as large as 10 to 20 watts/m 2 , which was about 5 to 10% of the total flux over the
Indian and Pacific oceans. It was also noted that wind variations on this time scale
were very important contributors; next in line were the contributions from SST
variations on this time scale. The variations in air temperature and humidity were
relatively less important. Although the amplitude of SST anomaly was only of the
order of 0.8 to 1°C on this time scale, the SST coupled with wind variations of the
order 3 to 5 ms- 1 contributed to significant latent heat fluxes, i.e., » 10 to 20 watts/m 2 .
The sign of these low frequency fluxes are preserved for a couple of weeks, thus their
role can become significant.
d) Energetics in the frequency domain.
The maintenance of low frequency modes has been addressed via detailed
computations of energetics in the frequency domain using daily globally analyzed data
sets over many years.
energetics in the zonal wave number domain.
These studies are somewhat analogous to the estimates on
In the latter approach, one speaks of
kinetic energy exchanges from zonal flows to eddies of certain scales, and of waves to
waves via nonlinear interactions. The other important interactions in the wave
number domain are those from potential to kinetic energy for fixed zonal wave
numbers. In the frequency domain, analogous selection rules govern the exchanges of
energy. Here, one can visualize a breakdown among long term mean, low frequency
and high frequency motions.
In a frequency domain, the kinetic to kinetic energy
exchanges can occur among long term time mean flows and other frequencies, or
among triads of frequencies (analogous to the wave number domain). The potential to
kinetic energy exchanges are restricted to occur at the same frequencies.
The results' of these energetics calculations performed by Sheng (1986) show
that the kinetic energy of low frequency modes on the time scale of 30 to 50 days are
maintained by the following processes:

223
and
i) They receive a substantial amount of kinetic energy from high frequency
modes;
ii) They lose kinetic energy to the long term time mean flow;
iii) The annual cycle is a source of energy for other frequencies;
iv) They receive a smaller amount of energy from the potential energy on the
same frequencies.
The above observational findings were important for our design of
experiments to increase the predictability of low frequency modes.
In this context, we should mention some results on the predictability of low
frequency modes from long term integration of a global model. Dr. William Heckley,
of the ECMWF, examined the presence (or absence) of the monsoonal low frequency
modes from several ensembles of predicted data for the FGGE period. The zero day
ensemble of 365 days is a string of initialized FGGE data. This string, as to be
expected, contained meridionally propagating low frequency modes. However, as the
strings of the ensemble of 1, 2, 3, 4 and 5 day forecasts were examined, in this
context, it was noted that the low frequency modes were lost by about day 5. The
conclusion was drawn that the global model has a predictability, for the low frequency
modes, of about 4 days.
We feel strongly that this loss of-predictability for the low frequency modes
is largely due to the errors the model makes in the prediction of the higher frequency
motions. These, in turn, contaminate the low frequency modes by the transfer of
errors in these energy exchanges.
2. Modelling Aspects
The aforementioned observations and ideas were useful in the design of a
class of long term integration experiments in order to extend the predictability of low
frequency modes.
Specifically, we designed the following experiments:
a) A control experiment with a comprehensive global spectral model
(Krishnamurti et aL, 1989). The model was initialized using nonlinear normal mode
initialization using global data for July 31 1979 (1200 UTC).
This experiment
included an annual cycle of SST. A 270 day integration was carried out starting from
that day.

224
b) A low frequency mode experiment where the initial state was obtained as
follows:
i) A time mean state was obtained for all variables at all vertical levels using
the data sets for a 120 day period preceding the initial date (i.e., July 31
1979 12 UTC).
ii) A low frequency mode for the initial date: This was based on the data sets
for the same preceding 120 day period.
iii) SST anomalies on the time scale of 30 to 50 days: These were updated
during the course of integration. This data is described in Krishnamurti et
al. (1988). In order to enhance the response for the model resolution, the
anomalies were multiplied by a factor of two.
iv) A long term averaged annual cycle of SST that varies from month to
month. (The same fields were also used in the control experiment.)
The premise here being that if the high frequency modes are filtered out from
the initial state, the contamination for the energy transfers as errors grow could be
reduced, thus we might extend the predictability of low frequency modes.
retention of the time mean state and the SST anomalies was to provide some of the
other important energy sources for these low frequency modes.
The results of these experiments were quite successful to about 30 days, after
which the energy exchanges grew quite large as high frequency motions evolved. The
main results of these experiments were as follows:
a) The control experiment failed to predict meridional motion of low
frequency modes. The anomaly experiment was quite successful The meridionally
propagating low frequency modes of the lower troposphere over the Asian monsoon
region were very accurately predicted for the first 30 days. The occurrence of a dry
spell over central India in the middle of August was well predicted. The amplitude of
the low frequency mode was somewhat underpredicted; however, the phase errors
were very small.
b) The eastward propagating planetary scale divergent wave at 200 mb was
very reasonably predicted to almost 25 days in the anomaly experiment The phase
speed was very close to that based on observations. The control experiment failed to
show an eastward motion of the divergent wave.
c) The energetics of the control revealed a large transfer of kinetic energy
from the low frequency modes to the higher frequencies. That must be an important
The

225
factor for the rapid collapse of predictability in the control experiment.
In the
anomaly experiment (where we had included a time mean state, a low frequency mode
and the SST anomalies on the time scale of 30-50 days), the kinetic energy exchange
was consistent with the observational results. This experiment was extended to 270
days. The energetics in the frequency domain were calculated for the 270 days of
data and also from day 31 to day 270. As stated earlier, the predictability of the low
frequency modes was good to about 30 days in this experiment. For the entire 270
day period, the kinetic energy exchange from the high to the low frequencies was
positive, but greater than that for the period from day 31 to day 270. This implies that
the contributions from the first 30 days, for which the predictability was large, was
consistent with the observational estimates. As the errors grow, the contribution of
the energy exchange slowly reversed signs and the low frequency modes were
contaminated.
In conclusion, the present approach simply delays the rate of contamination
of low frequency modes, thus extending their predictability to almost one month.
Eventually, higher frequency motions do form and the rate of contamination increases
rapidly.
Further work is needed to assess the relative importance of the details of the
time mean, the definition of the low frequency mode and the details of the SST
anomalies in the increase of predictability of low frequency modes. Further
experimentation is needed to predict the irregular behavior of low frequency modes,
as was noted from the observations from some recent years. If such experimentation
is successful in predicting the irregular behavior, then there is a need to diagnose the
importance of the aforementioned input parameters.
This approach may hold much promise for predicting the dry and wet spells
that are related to the passage of low frequency modes on the intraseasonal time
scales.
Detailed results of this paper will appear in the Journal of Meteorology and
Atmospheric Physics (1990) under the same title.

226
A CLOUD WAVE THEORY AND ITS APPLICATION TO THE
30-50 DAY OSCILLATION IN THE EQUATORIAL ATMOSPHERE
Qiu-shi Chen
Department of Geophysics
Peking University
Beijing, China
and
Kuo-Nan Liou
Department of Meteorology/CARSS
University of Utah
Salt Lake City, Utah 84112, U.S.A.
1. Introduction and the Basic Idea for the Cloud Wave
The 30-50 day oscillation has recently been investigated by Murakami and
Nakazawa (1985), Lau and Chan (1985), and Knutson et al. (1986) using the
outgoing longwave radiation (OLR) dataset.
From the analysis of the observed
OLR, the most dominant patterns consist of an east-west oriented dipole that
propagates eastward at a speed of about 4-5 m/s over the equatorial Indian and
western Pacific Oceans.
The half wavelength of the OLR patterns is about
6000-7000 km. The strongest OLR pattern is primarily developed in the Indian
and western Pacific Oceans, while the wind pattern appears to have a global
influence.
The variations of the u-wind at 850 mb and 150 mb levels at Truk
(152°E, 7°N) are out of phase on both levels.
The 30-50 day oscillation can
be clearly observed in the u-wind at this station (Madden, 1986).
What is the physical implication of the observed OLR patterns in the
tropics
The smaller OLR values imply the presence of high tropospheric

227
clouds.
These clouds trap the IR radiation emitted from the warm surface and
reradiate it at the cloud top with a much colder temperature in the upper
troposphere.
Thus the transient patterns of the OLR are directly related to
the transient patterns of the clouds and their associated radiative heating
field.
In the recent theoretical studies of the 30-50 day oscillation presented
by Lau and Peng (1987) and Chang and Lira (1988), the direct consequence of
these OLR patterns, which is related to the cloud field and cloud radiative
heating, was not considered.
the convective condensation.
Rather, these patterns were used as indices of
Lau and Peng (1987) proposed a wave-CISK
hypothesis to produce a condensation heating that is associated with the lowlevel
convergence circulation.
In their model, they generated a Kelvin wave
that propagates eastward at a speed of about 20 m/s if the vertical profile
has a maximum in the upper troposphere, and a speed of about 10 m/s if the
heating has a maximum in the lower troposphere. However, most budget studies
(Reed and Recker, 1971; Yanai et al., 1973; Kuo and Anthes, 1984) show that
the vertical profile of latent heat for deep convection has a maximum in the
upper half of the troposphere.
In this case, the wave speed simulated by Lau
and Peng (1987) is faster than the observed value.
From the observed clouds studies during winter monsoons, Webster and
Stephen (1980) found that the most common species over the equatorial South
China Sea and Indonesian region are thick (optically black) middle and upper
tropospheric extended clouds with the top in the vicinity of the 200 mb level.
The cover area of the extended upper and middle cloud decks is perhaps an
order of magnitude larger than the convective region itself.
Based on the
cloud and atmospheric state from the MONEX data, they computed the radiative

228
heating indicating a substantial net heating at the base of the cloud and
cooling at the top. It was argued that the radiative effects cannot be
ignored in the total diabatic heating fields in such tropical extended cloud
systems. We consider that the transient OLR patterns associated with the 30-
50 day oscillation are direct consequences of the observed radiation field. A
theory that is developed to explain the observed transient OLR patterns must
ultimately be consistent with these patterns in the radiation field.
For
these reasons, we shall undertake an investigation of the 30-50 day oscillation
from the perspective of clouds and cloud-radiative heating.
In order to extract the pure effect of clouds on radiative heating, we
may separate the total radiative heating into two parts, i.e., Q = Q + Q -
Q + (Q - Q ), where Q and Q represent radiative heating for a clear sky
and cloud field, respectively. In a clear sky, we have Q - 0. We assume
that a large-scale cloud has been formed in the equatorial region, as shown
in Fig. 1.
The horizontal structure of the cloud-radiative heating can be
seen from the observed composited OLR patterns associated with the large-scale
cloud or super cloud clusters, as shown in Fig. 2 (Nakazawa, 1988).
As for
the vertical structure of the large scale clouds, three possible different
distributions are shown in Fig. 3(a)-(c).
The cloud radiation heating is
computed by using Ou and Liou's (1988) radiation scheme.
The vertical
profiles of computed IR cloud radiative heating are shown in Fig. 3(d)-(f).
The mean vertical profiles of the temperature and humidity within the cloud
cluster In'the.West Pacific tropical region studied by Gray (1975) are used in
these computations.
The computed OLR at the top of the atmosphere for three
vertical cloud distributions shown in Fig. 3(a)-(c) have the same value
(127,4 W/m 2 ) and the difference from its value of clear sky is about -120

229
W/m 2 .
It can be seen that the result of the OLR is the same if there is the
same value of temperature on the cloud top, no matter if the vertical
distributions of the cloud cover are quite different or whether or not
convection exists.
From Fig. 2 it is shown that, along the equator, positive OLR regions are
associated with clear sky, and they are located on both sides of the negative
OLR region, which has a zonal width of about 6000-7000 km.
Thus a large-scale
horizontal temperature gradient must be developed between the large-scale
cloud field and clear regions due to differential radiative heating.
If most
large scale clouds have vertical structure as shown in Fig. 3(b) or 3(c),
based on the hydrostatic equation, the vertical and horizontal configurations
of the differential radiative heating will result in a large-scale geopotential
perturbation (Fig. 4) on the isobaric surface in the middle level of the
cloud indicated by NN' in Fig. 1.
The quasi-geos trophic motion is a good
approximation for the planetary scale motion in the equatorial atmosphere.
The quasi-geostrophic motion in the equatorial atmosphere was first investigated
by Matsuno (1966) as shown in Fig. 4.
Owing to the quasi-geostrophic
motion, a zonal westerly current in the middle level of the cloud must be
developed during which high geopotential perturbations are produced. Fig. 4
can be used as a schematic diagram for the geopotential height and wind field
on the isobaric surface in the middle level of the cloud.
This zonal current
would generate the movement of large-scale clouds eastward in the equatorial
region.
The large-scale moving clouds are referred to in this paper as the
cloud wave.

230
If
N
Fig. I,
A schematic diagram for the large-scale clouds in the tropics.
SON
CLR COMPOSITED 1980
15N
EQ
15S
SOS
»OE
SOW
Fig, 2. Composited transient OLR fields. Contour interval is
5 W/m^ and the negative regions are stippled. The minimum
value is -50 W/m 2 (after Nakazawa, 1988).

231
Fig. 3.
The vertical distribution
of the cloud cover and IR
cloud-radiative heating.
1000 0 20 40 60 60 100
Cloud Fraction (%)
6-4-20 24 6
Qc(°C/day)
Equator
Fig. A.
Pressure and velocity distribution for longwave
approximation (V -' 0) (after Matsuno, 1966).

232
2. A Quasi-Diabatic Geostrophic Model for the Tropical Planetary Scale Low
Frequency Motion
In order to illustrate the dynamic essentials of the low-frequency
planetary-scale motion in the tropical atmosphere, it is necessary to use
simplified dynamic and thermodynamic equations.
It is well known that the
basic structure of the extratropical atmosphere may be largely described by
the quasi-geostrophic theory.
However, such a theory has not emerged for the
tropical atmosphere.
Two simplified approaches have been made to understand
the physical processes that dominate the tropical atmosphere.
The first
approach was originally proposed by Charney (1963) by using the comparative
scale analysis.
The second approach followed the classical tidal theory.
Matsuno (1966) and Longuet-Higgins (1968) were among the first to realize the
importance of the tidal equations.
In the present paper, we shall demonstrate
that these two approaches can be unified based on the scale analysis for lowfrequency
motion in the tropical atmosphere.
For planetary scale motion, the characteristic values of the physical
quantities are as follows:
v « U • vj_, U .a* 10 m/s Characteristic horizontal velocity
(x f y) - L(xi»yi)» L » 10' m Characteristic horizontal length
z' - H zi, H » 10^ m Characteristic height
f -.fpf^-, f o » 10"-* s"^- Characteristic Coriolis parameter
The nondiniensional quantities v^, x^, y]_, and
fj_ are on the order of unity.
In the latitude region, f 0 has a value of 2Q sin^ between 3-4° and is
equivalent to the characteristic value of f in the equatorial region between

233
equatorial region, the meridional width of the disturbance must be larger
than 1800 km.
Based on the preceding values, the governing equations for the
specific humidity q, liquid water content 2 C , temperature T, wind field,
continuity, and hydrostatic equilibrium for a time scale on the order of
about 30 days may be written in the forms
q _ „ £S + (c
an ,
_ V * c - „ jf - ^ (P o UQ y H- (C - E c - P)
• [L (C - E C - E r ) -f Q c + Q rQ ]
P
+ F + F T - 0 , (3)
h v
-20 sinaS k x V - V + F^ + F V - 0 , (4)
h v
(5)
where C denotes condensation, E c and E r evaporation due to cloud droplets and
raindrops respectively, P precipitation, w c the bulk terminal velocity for
liquid water, and F^ and F^ the horizontal and vertical fluxes due to
diffusion of water vapor, liquid water, and temperature respectively, where

234
X can be q, £ c, v, or T. The other notations are conventional. If convection
occurs, the term (C - E c) may be replaced by a parameterized expression for
cumulus convection (Kuo, 1974; Anthes, 1977).
Parameterizations of condensation,
evaporation, and precipitation processes follow those developed by
Sundqvist (1978).
Equations (3) -(6) represent a set of simplified balance
equations for diabatic heating and geostrophic winds.
It has been found by
Gill (1980, 1982) that the geostrophic wind for the planetary wave In the
tropical atmosphere is as good an approximation as in the midlatitudes.
The
present analysis illustrates that the geostrophic balance is only a part of
the total balance between the diabatic heating and atmospheric motion as a
whole.
This total balance is referred to as a diabatic geostrophic balance.
The dynamic and thermodynamic equations presented above are not explicitly
dependent on time.
However, the time variation in the velocity and
temperature fields may be realized through the time variation in the diabatic
heating.
Equations (l)-(6) form the quasl-diabatic geostrophic model in the
tropical atmosphere.
The tropical atmospheric motion described by this model
is referred to as the quasi -diabatic geostrophic motion.
The method that has been used in solving the quasi -geostrophic model in
the midlatitudes may be utilized to solve the present model.
This method
involves the computation of the advection terras first.
Subsequently, the
balanced equation is numerically solved.
If the large-scale clouds have
already been generated in the atmosphere, except the horizontal advection, the
high frequency variation in the clouds may be neglected.
In this case, Eq.
(12) may be simplified as follows:
i

235
Using a stable leap-frog scheme, Eq. (7) may be solved by the following:
where i and j denote the indices for the grid points used in the model. After
the cloud liquid water content has been changed to ^n+ -'-, the cloud- radiative
heating may be subsequently computed and is denoted by Q n+1 .
Thus the
balanced equations denoted in Eqs. (3) -(6) can be solved under the condition
that Q n+1 is known.
By using the ^-plane, "longwave" (in the zonal direction), "Rayleigh
friction", and Newtonian cooling approximations for a clear sky, Eqs. (3)-
(4) , may be rewritten in the forms
- o> T - ~- Q* +1 - aT , (11)
P
where a denotes the linear damping coefficient and F the hydrostatic instability
parameter.
If Q n+1 is known, Eqs. (3. 17) - (3 .19) can be solved by an
expansion of the parabolic cylinder function, which is similar to that used by
Gill (1980).
It follows that v n+1 , w n+1 , T n+1 , and $ n+l can be computed. The
objective of solving Eqs. (9)-(10) is to obtain a new diabatic geostrophic
balanced state after the cloud- radiative heating is changed due to the
advection of large-scale clouds.

236
On the basis of the preceding discussion, there are two computational
processes in each time step.
The first is to obtain the cloud variation
caused by advection and the corresponding cloud-radiative heating.
The second
is to derive a new balanced state due to the variation in cloud-radiative
heating.
In general, the convective and stratiform condensations are included
in the balanced equation.
The solution of the balanced equations in the
second step may not be straightforward.
However, the diabatic geostrophic
balance can also be obtained from the adjustment process.
3. Results and Conclusion
From the numerical results, we find that if large-scale cloud-radiative
heating occurs in the equatorial region, the diabatic-geostrophic balanced
motion has a number of basic features.
These include a Walker and an anti-
Walker circulation with different characteristic horizontal scales developed
to the east and west, as well as two Hadley cells developed to the north and
south of the equatorial heating region.
An area of upper tropospheric
divergence is generated over the heating region, while an area convergence is
located at a distance of about 90-100° longitude to the east of the heating
region.
Anticyclone dipoles occur in the upper tropospheric stream function
field on both sides of the equator.
Their centers are located close to the
west of the equatorial heating center.
A high pressure region and zonal
westerlies are developed in the upper troposphere over the equatorial heating
center.
A broad easterly and a relatively narrow westerly are produced in the
lower troposphere to the east and west of the diabatic heating, respectively.
The zonal current in the upper troposphere over the center of the cloudradiative
heating region is about 4 m/s.
Thus Eq. (7) may be rewritten in

237
the form
d&
BS.
—£ „ _ n °-
3.t ax
This implies that the large-scale clouds move eastward with a zonal speed of
about 4 m/s.
In each computational step, two associated processes are
involved.
One is the movement of the large-scale clouds caused by the
equatorial zonal current and the cloud-radiative heating associated with the
moving clouds.
The other is concerned with the new diabatic-geostrophic
balance caused by the cloud-radiative heating that produces the equatorial
zonal current.
The physical essential of the 30-50 day oscillation is a
result of the interactions between the preceding two processes.
If no other
process is considered, the large-scale clouds and quasi-balanced atmospheric
circulation caused by the cloud radiative heating will produce eastward
motions with a speed of about 4 m/s.
If the half wavelength of the largescale
clouds is about 7000 km, the period of the oscillation is 2ir/kc - 40
days.
We propose that this is the basic mechanism for the 30-50 day oscillation
in the equatorial atmosphere.
We shall refer to the preceding behavior of large-scale clouds as the
cloud wave.
The movement and development of the cloud wave are described by
the quasi-diabatic geostrophic model.
In the abow discussion, the cloud wave
has a speed of about 4 m/s.
This wave can be expressed not only in the cloud
field, but has its components in the temperature and wind fields, which can
influence the entire globe.
The upper tropospheric divergence and low-level
convergence occur in the maximum cloud cover region of the cloud wave.
In
addition, Walker and anti-Walker circulation are developed to the east and

238
west of the maximum cloud cover region.
The zonal winds are out of phase at
the upper and lower levels.
On the basis of the preceding discussion, it is our view that the 30-50
day oscillation in the equatorial atmosphere is not produced by the Kelvin
wave, but is primarily caused by the cloud wave.
In the present quasidiabatic
geostrophic model, the Kelvin wave is filtered out.
In the present
quasi-dialictic geotrophic model, the Kelvin wave is filtered out.
This is
similar to the case where the gravity wave is removed from the quasi-geostrophic
model in the midlatitudes.
From the traditional viewpoint of the dynamics of the midlatitudes, the
motion systems play the dominant role, whereas the formation of clouds is
subject to the motion systems.
However, from the standpoint of the cloud wave
for the tropical low frequency motion, the role of cloud systems and atmospheric
motions is reversed.
The cloud system plays the leading role, while
the motion systems are secondary and are determined by the cloud systems
through the cloud-radiative heating.
References
Anthes, R. A., 1977: A cumulus parameterization scheme utilizing a onedimensional
cloud model. Hon. Wea. Rev., 105, 270-286.
Chang, C. P., and H. Lim, 1988: Kelvin wave-CISK: A possible mechanism for
the 30-50 day oscillations. J. 'Atznos. Sci., 45, , 1709-1720.
Charney, J, G., 1963: A note on large scale motions in the tropics, J.
Atmos. Sci., 20, 607-609.
Gill, A. E. , 1980: Some simple solutions for heat induced tropical circulation.
Quart. J. Roy. Meteor. Soc., 106, 447-462.
Gray, W. M., E. Ruprecht, and R. Phelps, 1975: Relative humidity in tropical
weather systems. Mon. Wea. Rev. 103, 685-690.

239
Knutson, T. R. , K. M. Weickmann and J. E. Kutsbach, 1986: Global-scale
intraseasonal oscillations of outgoing longwave radiation and 250 mb
zonal wind during Northern Hemisphere summer. Hon. Wea. Rev., 114, 605-
623.
Kuo, H.-L., 1974: Further studies of the parameterization of the influence of
cumulus convection on large-scale flow. J. Atmos, Sci., 31, 1232-1240.
Kuo, Y. H., and R. A. Anthes, 1984: Mesoscale budgets of heat and moisture in
a convective system over the central United States. Mon. Wea. Rev., 112,
1482-1497.
Lau, K. M. , and P. H. Chan, 1985: Aspects of the 40-50 day oscillation during
the northern winter as inferred from outgoing longwave radiation. Mon.
Wea. Rev., 113, 1889-1909.
Lau, K. M. , and L. Peng, 1987: Origin of low-frequency (intraseasonal)
oscillations in the tropical atmosphere. Part I. The basic theory.
J. Atmos. Sci., 44, 950-972.
Longuet-Higgins, M. S., 1968: The eigenfunctions of Laplace's tidal equations
over a sphere. Philosph. Trans. Roy. Soc. London, Ser. A, 262, 511-607.
Madden, R. A., 1986: Seasonal variations of the 40-50 day oscillation in the
tropics. J. Atmos. Sci., 43, 3138-3158.
Matsuno, T., 1966: Quasi-geostrophic motions in the equatorial area. J.
Meteor. Soc. Japan, 44, 25-43.
Murakami, T. , and T. Nakazawa, 1985: Tropical 45-day oscillation during the
1979 Northern Hemisphere summer. J. Atmos. Sci., 42, 1107-1122.
Nakazawa, T., 1988: Tropical super clusters within intraseasonal variations
over western Pacific. J. Meteor. Soc. Japan, 66, 823-839.
Ou, S. C. , and K. N. Liou, 1988: Development of radiation and cloud parameterization
programs for AFGL global model. Final Report AFGL-TR-88-
0018,
Reed, R. J., and E. E. Recker, 1971: Structure and properties of synopticscale
wave disturbances in the equatorial western Pacific. J. Atmos.
Sci., 28, 1117-1133.
Sundqvist, H., 1978: A parameterization scheme for non-convective condensation
including prediction of cloud water content. Quart J. £07. Meteor.
Soc., 104, 677-690.
Webster, P. J., and G. L. Stephen, 1980: Tropical upper-tropospheric extended
clouds: Inference from winter MONEX. J. Atmos. Sci. 37, 1521-1541.
Yanai, M. , S. Esbenson and J. Chu, 1973: Determination of bulk properties of
tropical cloud clusters from large-scale heat and moisture budgets. J.
Atmos. Sci., 30, 611-627.

240
NORMAL MODES OF CLIMATOLOGICAL MEAN FLOW AND THEIR
ROLES IN ATMOSPHERIC GENERAL CIRCULATION
Zhang Zuojun
Institute of Atmospheric Physics, Academia Sinica
ABSTRACT
The loss of orthogonality between unstable normal modes is
general for any kind of eigen-analysis, in particular for an observed
climatological mean flow this is found to be very significant for the
development of perturbations. A small perturbation can have a very
large projection onto the most unstable normal mode. The adjoint eigen
mode is most efficient at exciting the normal mode. The "gain" on
projection is described by the projectibility. In general, growth rate
and frequency information should be augmented with the projectibility
and eigen vectors should be augmented by the corresponding adjoint
eigen vectors.
For the 300 hPa January climatological mean flow, the maximum
projectibility is found to be 7.8 and the adjoint mode corresponding
to the most unstable normal mode has large amplitude over the
subtropical Indian Ocean and southeast Asia. The adjoint mode when
used as initial perturbation yields an energy increase of a factor of
50 within 10 days even when a damping is added to make the system
stable. Both the initial value and forcing problems show that the
linear barotropic vorticity equation gives important ideas on the
atmospheric low-frequency variability and the role of the tropics.
Initial studies suggest that growth rates and projectibility
together give important information on the likely accuracy of mediumrange
weather forecasts.

241
A Study On The Radiative Balance Simulated
By a General Circulation Model
Jough~Tai Wang
Institute of Atmospheric Physics
National Central University
Chung-Li, Taiwan 32054
ABSTRACT
Simulated data from a general circulation model are used
to construct the earth-atmosphere's radiative balance. The balance
are investigated in a global mean sense and compared with
observed climatology and with recent results from the Earth Radiation
Balance Experiment.
The cloud-radiative-effects are also estimated through the
partition of the budget into a cloudy and clear skies cases. Seasonal
and regional characteristics of the radiative balance are also
constructed and discussed.
1. INTRODUCTION
The usage of a general circulation model (GCM) evolves as one of the powerful
tools to enhance our understanding of the climate system and interactions among
its components. The GCM can potentially be used to study the atmosphereocean
interaction or to estimate the future climatic change due to various processes
related to man's activities. For example, the effects of doubling C0% or the change
of surface types on future climate.
Development of the GCM has progressed from the pioneering calculations of
Phillips (1956), Smagorinsky (1963) and Arakawa et al. (1969) with simple formulation,
to nowdays with highly complicated physical processes and parameterizations
(Washington and Parkinson, 1986), Among various processes, the radiative
transfer parameterization is quite critical in the GCM simulation. Stephens (1984)
presented a review of the various methods used to compute the atmospheric radiation
embedded in numerical models of atmospheric .'circulation. Pitcher et al

242
(1983) pointed out that simulated climate changed significantly when different
radiation schemes were implemented in the model. Ramanathan et al. (1983)
investigated the response of a GCM to refinements in radiative processes. They
found that a GCM with improved radiation/cloud model was able to reproduce
many observed features. However, the detailed structure of the radiative balance
within a model was not constructed.
This study attempts to reconstruct the radiation budget envisioned from a
GCM and compares it with the observed climatology, especially results from the
Earth Radiation Balance Experimen (ERBE). The characteristics in global mean
sense, in different regions, and the role of cloud in the radiative balance of a climate
system will be identified.
2. DATA AND METHODOLOGY
The data used in this study is from the Oregon State University (OSU) twolevel
atmospheric general circulation model. Complete description of the model
can be found in Ghan et al. (1982). The model grids are 4° latitude by 5°
longitude, and with a time step of 10 minutes. A history tape with the primary
dependent variables and all the working variables is written every 6h. From an
integrations over 39 consecutive months, composites of four Januarys and three
Julys are the basic data set used in the present analysis (the same as in Wang et
al., 1984).
In general, the processes related to the radiation and heat balance in the
model include the longwave and shortwave fluxes, along with the surface latent
and sensible heat fluxes. Their formulation within the model will be discussed
here briefly.
The net upward flux of longwave radiation at any level R% is estimated as
RZ = RZ T ~~ RZ I where RZ t> RZ I is the upward, downward fluxes at level of
consideration (2), respectively. The upward and downward fluxes estimation are
based on the conventional physical concept. The absorber amounts are assumed
to be constant above certain levels. The absorbers of longwave radiation in the
troposphere considerred with this mdoel are water vapor and carbon dioxide. Some
approximations related to the simplification of the transmission function are also
introduced.
Following Joseph (1970), a simplification was made in dividing the solar flux
into a part that is scattered but not absorbed, and a part that is absorbed but
not scattered.

243
The net downward solar radiation at the model surface level and levels 1 and
3 are given by
£4 = 5; + 5: (i)
ASi = S a 0 - S 2 < (2)
AS 3 = SJ-S 4
a
.(3)
Superscripts a and a stand for the absorbed and scattered part of the shortwave
fluxes. The cloud effects are included in the cloudy atmosphere. Both the
detailed calculations for shortwave and longwave can be found in Ghan et al.
(1982).
The surface latent and sensible heat fluxes are estimated in the model as
qd (4)
Tt) (5)
p4 is the surface air density, CD the surface drag coefficient, V{ the effective surface
wind speed. q g and #4 are the ground and surface air water vapor mixing ratio
respectively. These estimates are the so-called bulk-aerodynamic method. These
processes are used in the computation of the ground temperature for nonwater
surface.
The longwave, shortwave and surface fluxes are the processes needed to be
identified and calculated for this study to construct the radiative balance within
a climate model.
3. RESULTS AND DISCUSSIONS
The monthly mean radiation budget is constructed for 4 Januarys and three
Julys. The composites of January and July are simply the average of the monthly
mean data involved. Annual mean is estimated from the average of the January
and July monthly statistics. The results will be presented in a global mean sense
and in terms of different regions.
3.1 Global structure
The simulated global annual mean of the earth-atmosphere radiation structure
is presented in Fig. 1. All the quantities are divided by the shortwave flux that

244
reach the top of model atmosphere to show its percentage contribution in the total
budget. The corresponding observed climatology from NRC (1975) is shown in
Fig. 2. From Fig. 1 for the shortwave radiation (SW) part, the planetary albedo
simulated by the model is 37.1%, the absorption within the atmosphere is 13%,
and the surface absorption is 49.9%. In contrast to the observed climatology, the
simulated planetary albedo is stronger than the observed (37% vs 30%), while
the atmospheric absorption is consequently reduced (13% vs 19%). The simulated
surface absorption is quite comparable to the observed (49.9% vs 51%). Simulated
surface latent and sensible heat fluxes contribute 30% of the heat transfer from
surface to the atmosphere, which are the same as those from observed climatology.
For the longwave radiation (LW), the net upward outgoing LW from earth's
surface is 20.8%, quite close to the observed estimate (21%). The simulated net
longwave radiative cooling (sum over the radiative cooling in the atmosphere to
the space and the absorption of longwave fluxes from surface to the atmosphere)
in the model is 41.4%, and the observed value is 49%.
The very recent ERBE estimate for the global radiative balance from Ramanathan
et al. (1989) are reconstructed and shown in Fig. 3. The ERBE results
differ with the NRC (1975) estimate slightly. The ERBE data shows a smaller
earth absorption of SW (49.4% vs 51%), with a larger atmospheric absorption
of SW (19.9% vs 19%). For the LW part, the ERBE estimate smaller net upward
outgoing LW (18.4% vs 21%), while the surface fluxes and planetary albedo
estimation are larger (31% vs 30%), (30.7% vs 30%), respectively.
The main discrepancy in the model simulation is related to the overestimation
of the planetary albedo and the underestimation of the atmospheric absorption in
the SW. For the net longwave radiative cooling, it is underestimated by about 7%,
which can clearly be attributed to the discrepancy in the shortwave simulation.
This net atmospheric longwave cooling, in a global mean sense, has to be balanced
by the atmospheric absorption of shortwave radiation and the heat fluxes from
the earth's surface to the atmosphere. Fig. 2 identifies that value from observed
climatology is 49% vs 19%+30% for NRC (1975) and 50.9% vs 19.9%+31% from
ERBE, an exact balance. While the simulated balance is 41.4% vs 13%-f 30%.
There is a small imbalance. This leaves the atmosphere with a small fraction of
heating contribution. The sampling problem may be the cause of this imbalance.
If more months can be included in the estimation of the annual mean, the balance
should be achieved.
As summarized in Ramanathan et al. (1989), clouds in general are shown
to trap longwave radiation and reduce the emission of longwave radiation to the

245
space. This effect is the green-house effect of clouds and is termed as the longwave
cloud forcing; clouds are also shown to reflect more shortwave radiation compared
to clear sky. This enhancement of reflection is referred to as the shortwave cloud
forcing. The net effect of cloud to earth-atmosphere system is the sum of longwave/shortwave
cloud forcing and is referred to as cloud radiative forcing. In
view of recent estimate of the cloud-radiative-effect as pointed out Ramanathan
et al. (1989), the global annual mean budget is further decomposed to identify
the cloud-radiative-effect in the model atmosphere. Fig. 4 indicates some budget
terms in the clear and cloudy atmosphere. Planetary albedo of 59.6% (16.6%)
in cloudy (clear) atmosphere is clearly shown, along with the surface shortwave
absorption of 28.3% (69.6%). The difference for the cloud/clear atmosphere for
the surface shortwave absorption is 41.3%, which is quite large. The atmospheric
absorption of shortwave radiation in cloud (clear) atmosphere is 12.1% (13.8%).
The outgoing surface longwave radiation in clear atmosphere is larger than the
cloudy atmosphere by 15.8% (the difference of 28.3% and 12.5%). So the cloudradiative-effect
to the earth-atmosphere system is mainly in the large reflection of
the shortwave fluxes and the modification of the outgoing longwave fluxes. The
change in absorption of shortwave is relatively small. Prom Ramanathan et al.
(1989), their estimate of the net cloud-radiative-effect is a cooling effect, on the
order of -14 to -21 w/ra 2 . In our estimate as illustrated in Figs. 1 and 4, this effect
realized by the present model is around .-35 w/ra 2 (10.9% of the total incoming
solar radiation). Although the quantity is larger that the observed estimate, the
simulated model cloud kept the consistent physical reasoning as observed.
Seasonal characteristics for the global mean is presented in Fig. 5. The
planetary albedo is larger in January than July. The atmospheric absorption of
shortwave radiation does not experience much seasonal variation.
3.2 Regional characteristics
The simulated global structure of the radiation budget indicated overestimation
of the planetary albedo, which may be due to the model's overestimation of
the cloudiness. The simulated atmospheric shortwave absorption and longwave
cooling were consequently reduced. Bear this characteristics in mind, the simulated
regional radiative balance will be presented here.
The whole globe is divided into five regions, those are 30*JV-30°S, 30°^-
60° JV, 30 0 5-60°5, 60°JV"-90°JV, 60*5-90*5. The radiative structure for the region
3Q 0 JV-3Q°S in January and July is presented in Fig. 6. The seasonal variation

246
is not clear in the tropics. This was understood by the general climatology of
that regions (Critchfield, 1983). The surface budget is in an equilibrium. The
structure also indicates the tropical atmosphere absorbed more energy than it
radiated out through the longwave part. This excess heat has to be transported
poleward dynamically to balance the heat loss there.
Figure 7 shows some quantities in the tropics for January and July with
respect to clear and cloudy atmosphere. The characteristics in tropics for the
clear and cloudy atmosphere are quite similar as those shown for the global mean.
This resemblance is a simple reflection that the tropical regions contribute a great
extent of the global mean. The tropical atmosphere has about 10% heat excess
(around 35 w/m 2 ) to be transported to the higher latitudes.
For higher latitudes regions, the results are not shown here. Their characteristics
are that net atmospheric cooling exists in January, in contrast to the heating
in July. This winter cooling is balanced to some extent by the heat transported
from lower latitudes.
4. SUMMARY
This study used the simulated data from OSU GCM to investigate the earthatmosphere's
radiation budget. The physical processes related to this study are
the shortwave (longwave) fluxes in the earth's surface and within the atmosphere,
along with the surface latent and sensible heat fluxes. These processes are identified
and used to construct the budget , also are compared with the observed
climatology. In annual mean sense, the simulated budget is reasonably good. For
the shortwave part, the simulated planetary albedo is overestimated by about 7%,
the atmospheric absorption is underestimated by 6%.
The surface absorption of shortwave, surface latent and sensible fluxes and
surface outgoing longwave radiation are quite close to the observed. The simulated
net longwave atmospheric cooling is short of 7%, compared with the observed
value. Seasonal variation of the atmospheric absorption of shortwave fluxes is
quite small. Clouds are found to affect the surface budget greatly, however, their
influence in the atmospheric absorption in annual mean sense is small.
Regional characteristics indicates that the tropical atmosphere has about 10%
heat excess to be transported to the higher latitudes. Seasonal reversal is obvious
in the higher latitudes for some processes, but not the atmospheric absorption of
the shortwave fluxes. In polar region, the surface fluxes are found to be quite small
in the summer hemisphere. For the winter hemisphere, the fluxes in that regions

247
are downward.
REFERENCES
Arakawa, A., A. Katayama, and Y. Mintz, 1969: Numerical simulation of the
general circulation of the atmosphere. In Proceedings of the WHO
/IUGG Symposium on Numerical Weather Prediction, Japan Meteorological
Agency, Tokyo, pp. IV-7 to IV-8-12.
Critchfield, H.J., 1983: General Climatology, Fourth Edition, Prentice-Hall
Inc., New Jersey, U.S.A., 453 pp.
Ghan, S. J., J. W. Lingaas, M. E. Schlesinger, R. L. Mobley and W. L. Gates,
1982: A documentation of the OSU 2-level atmospheric general circulation
model. Rep. No. 35, Climatic Res. Inst., Oregon State
University, Corvallis, 395 pp.
Joseph, J.H., 1970: On the calculation of solar radiation fluxes in the troposphere.
Solar Energy, 13, 251-261.
NRC (National Research Council), 1975: Understanding Climatic Change:
A Program for Action. U.S. Committee for the GARP, National
Academy of Sciences, Washington, D.C., 239 pp.
Phillips, N.A., 1956: The general circulation of the atmosphere: A numerical
experiment. Quart. J. Roy. Meteorol Soc., 82, 123-164.
Pitcher, E.J., R.C. Malone, V. Ramanathan, M.L. Blackmon, K. Puri, and W.
Bourke, 1983: January and July simulations with a spectral general
circulation model. /. Atmos. Sci., 40, 580-604.
Ramanathan, V., E. J. Pitcher, R. C. Malone and M. L. Blackmon, 1983:
The response of a spectral general circulation model to refinements in
radiative processes. J. Atmos. Sci., 40, 605-630.
Ramanathan, V., R.D. Cess, E.F. Harrison, P. Barkstrom, E. Ahmad, and
D.Hartmaim, 1989: Cloud-radiative forcing and climate: Results from
the Earth Radiation Budget Experiment. Science, 243, 57-63.
Rasmusson, E.M., and T.H. Carpenter, 1982: Variations in the tropical sea surface
temperature and surface wind fields associated with the southern
oscillation/El Nino. Mon. Wea. Rev., 110, 354-384.
Smagorinsky, J., 1963: General circulation experiments with the primitive
equations. 1. The basic experiments. Mon. Wea. Rev., 91, 98-164.
Stephens, G.L., 1984: The parameterization of radiation for numerical weather
prediction and climate models. Mon. Wea. Rev.,112, 826-867.

248
Wang, J.-T., J.-W. Kim and W.L. Gates, 1984: The balance of kinetic and
total energy simulated by the OSU two-level atmospheric general circulation
model. Mon. Wta. Rev., 112, 873-892.
Washington, W.M., and C.L. Parkinson, 1986: An introduction to threedimensional
climate modeling. University Science Books, CA, USA,
422 pp.
Annual mean (simulated)
-71
Fig. 1. The simulated global annual mean of earth-atmosphere radiation
structure. * stands for the planetary albedo, 2 and 3 stand for the
atmospheric and surface absorption of the shortwave radiation, respectively.
4 is the net outgoing longwave fluxes in the atmosphere, 5
the sum of the surface latent and sensible heat fluxes, c the outgoing
longwave radiation from surface. All the quantities are in unit of %.
The sum of all the shortwave radiation is 100%.
Annual mean (osbserved)
-'I
49 TV
30 7 1 21 /
Pig. 2. As in Pig. 1, except for the observed climatology (reestimate from data
in NEC (1975)).

249
data from ERBE
31 / 18.4
Fig. 3. As in Fig. 1, except for the observed data from ERBE (reconsrtucted
from Ramanathan et al., 1989).
(a) Cloudy atmosphere
(b) Clear atmosphere
7_
'"'I
Fig. 4. Partial budget quantities simulated by the model for clear and cloudy
atmosphere. The meaning of the superscripts are the same as in Fig.
1.
"'1
(a) Global (January)
(b) Global (July)
«•'••'r S
•
31.7 7 1 22.9
Fig. 5. The same as in Fig. 1, except for the January and July.

250
(a) 30°N-30°S (January)
(b) 30°7V-30°5 (July)
'I
3
U
24 71 19,3 7*
7
Fig. 6. The same as in Fig. 1, except for regions of 30°JV-30°5 and also in
January and July.
(a) January (cloudy)
(b) January (clear)
„'/
' I
(c) July (cloudy)
(d) July (clear)
4 ***
TV
Fig. 7. Same as in Fig, 4 except for regions of 30°A'*-30 0 S.

251
INTRODUCTION TO HEIHE BASIN FEILD EXPERIMENT (HEIFE)
— ATMOSPHERE-LAND SURFACE INTERACTIONS PROGRAM
Lin Hai
(National Natural Science Foundation of China)
ABSTRACT
In 1987, the National Natural Science Foundation of
China approved a major program "Atmosphere-Land Surface
Interactions of Heine Basin Field Experiment in Weastern
China (HEIFE)" proposed by the Lanzhou Institute of
Plateau Atmospheric Physics, Academia Sinica. The HEIFE
is one of the pilot land surface process experiments for
the World Climate Research Program (WCRP). It is an
interdisciplinary project combining basic theoretical
study with applied research. The HEIFE program started
in 1988 and lasts for at least 5 years. A pilot field
observation already implemented in the period from
August to September 1988.
The general objectives of the Heihe Basin Feild
Experiment are to investigate the physical processes of
the air-land exchanges of water, heat and momentum in
desert and agricultural irrigation regions of the midlatitudes,
test and improve the parameterization scheme
of such fluxes in GCM for the grid square of arid and
semi-arid regions of the mid-latitudes. It will also
collect the necessary ground-based data for developing
and validating the methods to convert satellite-observed
radiances to land-surface natures and climatological
variables. Because of the special geographical
environment of the experimental area, the investigations
of the concentration and particle size distribution of
atmospheric dust over desert, its dispersion and
transport and its effects on the radiative transfer will
be included in this program.
In this paper, the scientific problems, subjects,
observational plan and prospective results will be
introduced briefly.

252
An Important element of the earth-climate system is the
interaction between the atmosphere and the surface involving exchanges
of momentum, water, heat and trace gases. These are processes which
depend to a large extent on the nature of the surface. In order to
have a global perspective on land-atmosphere interactions, it is
necessary to understand water and heat budgets in areas that are
representative of different types of land surface over the globe. The
World Climate Research Program(WCEP) proposed two surface experimental
regions, including Southwestern France and Great Plains in the United
States. They are called the First ISLSCP Field Experiment (FIFE) and
Hydrological Atmospheric Pilot Experiment(RAPEX) respectively. The
Heine Basin Field Experiment(HEIFE) in western China headed by Prof.
Gao Youxi of the Lanzhou Institute of Plateau Atmospheric Physics,
Chinese Academy of Science, was approved by the National Natural
Science Foundation of China (NSFC) in 198 as one of major projects of
NSFC during the Seventh Five-Year Plan. As so much as Western China
is one of the areas in the world where orography and surface
characteristics are extremely complicated and consist of a variety of
surface features including glaciers, permafrost, desert, wilderness,
forests, meadows and cultivated land, the HEIFE differs from the other
two experiments (FIFE and HAPEX) and will become the third atmosphereland-surface
processes experiments for the WCRP*
I. Objectives
(1) To investigate the physical processes of the air-land surface
exchanges of water, heat and momentum, test and improve the
parameterization scheme of such fluxes in the General Circulation
Model (GCM) for the grid square of arid and semi-arid regions of the
mid-latitudes *
(2) To collect the necessary ground-truth data for developing and
validating the methods to convert satellite-observed radiances to land
surface netures (such as utilization of the land surface, state of

crop growth, and desert) and climatological variables.
(3) To investigate the concentration and size distribution of dust
over desert and its effect on climate.
(4) To study the water requirement by main crops and the
techniques of water-saving irrigation in Heihe district and to
estimate the water resource, its utilization ratio and the
agricultural potential in Hexi Corridor, Gansu Province in western
China.
253
II. Experimental Area
Heihe river basin is located in the arid region in the midlatitude
and one of the most important irrigated agricultural region
in northwest China with great agricultural potential. The annual
rainfall of the region is 100—150 mm.
The experimental area (99°30 f —101 o OO f E, 38°40 f — 39°40 f N) consists
of a rectangle with an area of 70km*90km and a triangle with an
average elevation of 1500m above sea level as shown in Fig. 1. South
to the experimental area is the Qilian mountain, the northern edge of
the Qinghai-Xizang (Tibet) Plateau. The Heihe river originates from
the glacier over the Qilian mountain. The northern boundary of the
experimental area stretches into the desert. The Heihe river flows
through the experimental area. North of the river, about half of the
experimental area is desert, some of which is Gobi. The south area of
the river is mostly flat irrigated cropland with Gobi in it. The main
crops in the area are wheat and maize. The farmland is bare in
winter.
Within the area, there are three standard meteorological stations
of the State Meteorological Administration (SMA), four hydrological
stations and an experimental station of the Lanzhou Institute of
Desert (LID) of the Academia Sinica, located in Linze County.
Five sites (Fig. 2) have been selected as basic experimental
stations. They are located in Zhangye city(cropfield with wheat),
Linze county (cropfield with maize), Pingchuan desert experimental

254
IOCS
I02S
3SN
Fig. 1
A general view of the experimental site.
LXOOi
The comprehensive observational station of the
Lanzhou Institute of Glacier and cryopedology,
Acadamia, Sinica

Rainfall station
Well for measuring water-table
Floating evaporation station
Meteorological station
Experimental station
*HydrograpMc station
Th« cxparlmantal station of the
^.tnzhou InstltuCe of Dosert,
AC»daml« Slnl'ca
Fig. 2 A schematic view of the experimental area, a

256
staion of LID (farmland in the forest belt between desert and oasis ),
Huayin, southwest of Linze county (Gobi) and North of Pingchuan
(desert)*
The central station is located at Linze, where the automatic data
collecting and transmitting system will be set up to remotely control
the operations of measurements at 5 basic experimental stations and to
record the observational data 'from them.
In addition, there are 1*1 rainfall stations and a mobile station
for multi-parameter observation.
III.
Scientific Problems
The main scientific problems for realizing the objectives
mentioned above are as follows:
(1) Development of measuring techniques and computing methodology
for the determination of the heat and evaporation fluxes over the
nonuniform ground surface in the arid area*
(2) Study on the vertical structure of the surface layer and
planetary boundary layer over various land surfaces (desert, Gobi,
irrigated farmland, etc.)* and the advection effectes on the vertical
transport of energy in the PBL over the whole experimental area.
(3) Study on the influences of surface and sub-surface conditions
(such as vegetative processes and soil moisture) on the surface water
and energy budget.
(4) Development of the methodology for spatial integration and
parameterization of net raidiation, heat flux, evaporation and
rainfall for the area with various types of land surface.
(5) The numerical experiment simulating the interaction between
climate and land surface.
(6) The characteristics of radiative physics and its variation for
various land surfaces, and the statistical study on the comparision
between the ground-truth and satellite-observed data (NOAA, GMS).
(7) The concentration and size distributions of atmospheric dust

over desert area and its possible effect on climate*
(8) The water requirement by main crops and the techniques of
water-saving irrigation*
257
IV* Research Subjects
The HEIFE consists of six research subjects:
Subject one. Observational study of the turbulent fluxes in the
surface layer and the structure of the planetary boundary layer headed
by Prof. Hu Yinqiao of the Lanzhou Institute of Plateau Atmospheric
Physics.
Subject two. Observational study of the surface radiation budget
and the radiation properties of the ground surface in the experimental
region headed by Prof. Ji Guoliang of the Lanzhou Institute of
Plateau Atmospheric Physics*
Subject three. Observational study of evaporation and water
balance in the Heihe area headed by Siner Engineer Chen Manxiang of
the Gansu Provincial Hydrological Station.
Subject four. Data collection, reorganization and preliminary
analysis of the whole experiment headed by Prof. Guo Youxi of the
Lanzhou Institute of Plateau Atmospheric Physics.
Subject five* Numerical simulation of the boundary layer headed
by Prof. Chen Linsheng of the Department of Atmospheric Sciences,
Lanzhou University.
Subject six. Study on the water requirement and techniques of
water-saving irrigation for main crops in Heihe district headed by
Prof. Li Shouqian of the Gansu Academy of Agriculture Science.
V. Experimental Time Periods
Formal Observing Periofl (FOP) The formal observing period will

258
start from October 1990 and end in the same month (October) of 1992.
Intensive Observing Periods CIQP) The formal observing period will
include 4 intensive observing periods, scheduled to take place in four
different seasons in 1991, each lasting 15 days. The scientific
objectives for the four lOPs are as follows:
(1) In winter (December or January), observational study on the
behaviour of the energy exchange between air and land surface, i.e.
whether the land surface is a heat sink or a heat source; and the
physical mechanisms of the energy exchange between air and land
surface when ground surface is a heat sink.
(2) In spring (April), two major scientific problems will be
studied: Air-mass transformation Atmospheric dust formation
(3) In summer (July), the main scientific objective is to observe
and study the structure of the PEL and the advection effect in the
PEL.
(4) In autumn (October), the seasonal transition of surface energy
budget*
Pilot Observing Period (]pOP) One pilot field observation was
conducted in August and September 1988 before the FOP. The aims of
the POP are to test the automatic data collecting and transmitting
system, and to carry out the field observational plan in order to
explore the scientific results and learn the logistical lessons.
VI* Measurements
(1) Measurements during JFOp Radiation — The radiation
measurements can be divided into two parts. Some are long term
measurements (lasting for two years) concerned with estimating the
radiative energy budget of the ground surface. Additional special
measurements of atmospheric and surface radiation properties are
directed towards providing corrections for satellite data. For this
reason, the measurements of the spectral solar and infrared radiations
at a well-chosen set of discrete wavelength bands should be made

discontinously in addition to the measurements of surface radiation
budget. Measurements of surface radiation budget include direct solar
radiation with MS-52, sky diffusive radiation with MS-42, global
radiation and surface albedo with MR-21, upward and downward long wave
radiation with Eppley Precision Infrared Radiometer, net radiation
with CN-11.
Measurement of wind velocity, temperature and humidity gradients
in the surface layer — Temperature, humidity and wind velocity on a
20m meteorological tower at six levels(0.5, 1.0, 2.0, 4.0, 8.0 and
20.0m), temperature and wind velocity at four levelsdO, 20, 40, 80cm)
in the crop canopy during the growth period of the crops.
Measurement of soil temperature, moisture and heat flux — Soil
temperature at the ground surface and the depths of 0, 5, 10, 15, 20,
40, 80, 120, 160 cm under the surface; soil heat flux at the depths of
2, 10, 40, 80cm under the surface with soil heat flux plate, CN-81;
soil moisture at the depths of 10, 20, 40, 80cm under the surface with
Neutron moisture meter.
Measurement of evaporation and evapor transpiration with Lysimeter
(only at Zhangye, Linze and Pingchuan stations).
Atmospheric aerosols and turbidity — Measurements of the
concentration of atmospheric dust:measurements of the particle size
distributions of atmospheric dust by using PM730 optical particle
counter in various seasons; measurements of the atmospheric turbidity
with sunphotometer; the measurement of spectral solar radiation with
Heissatat for studying the optical features of atmospheric dust and
other material.
(2) Measurements during the IQPs Measurements of turbulent fluxes
with eddy correclation method — U, V, W, T with sonic anemometer; q
with Lyman-o* hygrometer or infrared hygrometer.
Observation of the structure of the PEL and the IBL (Inner
Boundary Layer ) with tethered balloon.
259
VII. Data Collection and Processing

260
The measurements of surface radiation energy budget, wind,
temperature, and humidity gradients in the sureface layer and the soil
heat flux and temperature will be conducted 24 times each day* Each
measurement will be conducted 20 minutes around each hour and will
sample 2 times/min.
The measurements and data collection mentioned at all basic
stations are carried out automatically with dual-control f i.e., the
centralized-control by the Data Collection and Transfer System (DCTS)
at the central station, and independent-control at each station. In
this way, the observed data are recorded on tape and printed at each
station. They are also transferred to the central station
simultaneously and then recorded into the magnetic tape after they
have been preprocessed by the microcomputer system.
Direct measurements of various fluxes (i.e., spectral radiation,
momentum, heat and water vapour) and aerosols will also be conducted
30 minutes around each hour.
The data will be processed in the Lanzhou Institute of Plateau
Atmospheric Physics. The meteorological, hydrographical and various
field observed data will be recorded according to a fixed format on
tapes or diskettes for the purpose of supplying to the participating
units.

261
THE IMPACT OF URBANIZATION ON CLIMATE
IN HONG KONG AND ITS IMPLICATIONS FOR
HUMAN ENERGY EXCHANGES
by
William J. Kyle
Department of Geography & Geology
University of Hong Kong
Hong Kong
ABSTRACT
The impact of urbanization on long-term climatic changes
as evidenced by trends in temperature and wind speed
measured at the Royal Observatory, Hong Kong is examined.
Evidence shows a long-term warming trend concurrent with
increasing urbanization. Further investigation
attributes this trend to both horizontal expansion in
the built-up areas around the Observatory as well as
vertical expansion in more recent years. A definite
declining trend in post-war surface wind speed is also
evident and is linked to this latter phase of
urbanization. These two trends are examined within the
context of the commonly hypothesized causal factors for
the development of an urban canopy layer heat island.
The implications for human energy exchanges of this
warming trend are examined and an evaluation of the
likely effect on human thermal comfort is presented.
INTRODUCTION
It has been stated that the most obvious climatic manifestation
of urbanization is the trend towards higher temperatures, thi's trend
depending on the prevailing synoptic conditions (Landsberg, 6)). Urban
climates are really differentiated topoclimates and as such depend on
variation of radiative fluxes and turbulent exchanges, the contrasts of
which are greatest in clear, calm conditions and least in cloudy, windy
weather.

262
Secular trends in urban temperatures have been widely
investigated. In the absence of measurements prior to the establishment
of an urban area the clearest evidence is provided by confirming
a rising trend in local temperatures independent of any secular trend
in regional comate. A number of studies may be cited where this is
demonstrated.
In Japan, Fukui 3) documented changes, shown in Table 1, for
TABLE 1. Temperature Changes in Japanese Cities 1936-1965
Rapid growth
Slow growth
City
Temp, rise (°C/yr)
City
Temp, rise
(°C/yr)
Kyoto
0.032
Hikore
0.020
Osaka
0.029
Nemuro
0.005
Tokyo
0.032
Tyoshi
0.011
after Fukui 3).
three rapidly expanding cities and three smaller cities without much
urban growth. During the period 1936-65 the former experienced a
temperature increase of about 0.03 C per year whereas in the latter
the increase was only about 0.01°C per year. Immediately after World
War II when Tokyo was largely destroyed but thereafter rapidly reconstructed
even more dramatic changes are cited. From 1947-1963; the
period of reconstruction, daily maxima increased at a rate of 0.036 C
per year while the minima rose at an even faster 0.047 C per year.
A similar trend of rapidly rising minima and more slowly rising
maxima is cited by Detwiller 2). During the 78 year interval (1891-
1968) a rising trend of 0.011°C per year and 0.019°C per year was
reported for the daily maxima and minima respectively.
In Hong Kong, Peterson 8) presented data, shown in Table 2,
indicating that no significant trend was detectable in temperatures
measured at Macau Observatory during the period from 1902-80. On the
contrary, from 1884-1980, increases significant at the 0.1% level of

263
TABLE 2. Longterm Temperatures Trends ( C) at the Royal
Observatory, Hong Kong and Macau Observatory
Royal Observatory, Hong Kong
Macau
Observatory
Date
Mean
Max.
Mean
Mean
Min.
Date
Mean
Max.
Mean
Min.
1884-1893
1894-1903
24.4
24.8
21.8
22.1
19.9
20.2
1904-1913
24.7
22.2
20.1
1902-1910
24.7
19.6
1914-1923
24.9
22.3
20.3
1911-1920
24.9
19.6
1924-1933
1934-1939 ^
1947-1950 )
24.9
25.3
22.3
22.4
20.3
20.3
1921-1930
1931-1940
1941-1950
26.0
25.8
25.9
19.5
19.8
18.0
1951-1960
25.6
22.6
20.3
1951-1960
25.8
19.7
1961-1970
25.9
23.0
20.7
1961-1970
25.3
20,0
1971-1980
25.9
22.9
20.8
1971-1980
24.8
20.0
after Peterson 8).
about 1.5 C in mean maximum and 0.75 C in mean minimum temperature
have occurred at the Royal Observatory.
This study investigates these trends in more detail using 10
year running mean data from the Royal Observatory. These records
extend from 1884 although there is an unfortunate gap from 1940-46 as
a consequence of World War II. The data period represents one during
which major urbanization has occurred. It is possible, because of the
detailed and reliable map records available in Hong Kong, to trace the
secular development of this process with a high degree of reliability
thus providing a means of assessing its impact on urban climate. Such
an assessment, in turn provides insights into the ways in which urbanization
affects radiative fluxes and turbulent exchanges leading to
the development of the urban heat island in Hong Kong. Finally, since
urban Hong Kong is home to a substantial number of people, the
implications of this warming trend for human thermal comfort need to
be evaluated.

264
LONGTERM TEMPERATURE TRENDS AT ROYAL OBSERVATORY
Figure 1 shows the secular trend in temperatures measured at
the Royal Observatory using 10 year running means to reduce year to
year fluctuations. As in the studies already cited there has been a
CO
CO
UJ
u
CD
LU
o
27
22
21
20
*/*
" ;»»****
t^ff"^
_- -——-
•fc|tf+T*
^*tH^
..—%M.
V*"*"*
—
... *•••* ******
•mf
+
JH
MEAN
_
—%
>

265
TABLE 3.
Trends in 10 Year Running Mean Temperatures (°C/yr)
at the Royal Observatory, Hong Kong (1893-1988)
Period
Mean Maximum
Mean
Mean Minimum
1893-1947
0.0082
0.0081
0.0102
1947-1969
0.0455
0.0495
0.0541
1969-1988
-0.0254
-0.0059
0.0131
1893-1988
0.0156
0.0110
0.0085
is confirmed as in other studies. After the war and up to the end of
the sixties a period of very rapid increase is evident with increases
of 0.0455, 0.0495, and 0.054l°C year again consistent with previous
studies. From the early seventies to the present a third trend appears
with mean maxima declining by -0.0254 C per year while mean minima
continue to rise although at a slower rate of 0.0131 C per year.
Consequently mean temperature appears to have stabilized though
showing slight decline of -0.0059 C per year.
All of these trends are significant at the 0.1% level and appear
to be independent of any secular trend in regional climate. Consequently,
they need to be explained in terms of the urbanization that has
occurred in the respective periods and how this has influenced the
processes controlling local microclimate.
TRENDS IN URBANIZATION
Good quality maps of the area around the Royal Observatory are
available on a regular basis for this century. Similarly, aerial
photographs of the area are also available, though less commonly in
the early years. If it is assumed that the urbanization process can be
expressed in terms of changes in the percentage of built-up area and
gross building volume in the vicinity of the Royal Observatory then it
is possible by a combination of map and aerial photographic analysis
to provide reliable estimates of the changes in both of these parameters
for most of the period for which temperature data are available
(Kwok, 4)).

266
Using these data sources It has been possible to produce graphs
of the trends in built-up area and estimated building volume in an
area defined by a circle of 1 km diameter centred on the Royal
Observatory measurement site for the period 1904-1987. It has been
assumed that the changes that have occurred in such an area will have
a dominant impact on the radiative and turbulent exchanges which in
turn control local temperatures.
Figure 2 shows the change in built-up area expressed as a percentage
of the total area so defined. It is clear that over the
period there has been a steady, nearly linear (r = 0.9778) increase in
the percentage of built-up area at an average rate of just under 0.5
3U '
AK .
40
in -
5
| C .
i n .
5 .
• *
*
*
0 -
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990
FIGURE 2.
percent per year.
YEAR
.
* «
• »• *
Change in Built-up Area around Royal Observatory
(1904-87).
In 1904, just under 5% was occupied by buildings.
This had risen to slightly over 20% by the end of the World War II.
The increase continued at a higher rate in the early post-war period,
interrupted only by a period in the mid 1950 f s when there was a widespread
demolition of pre-war buildings.
New construction continued
until the 1970's by which time most of the available land had been
occupied so that the present level of about 40% has not changed
greatly over the past 20 years.
m

267
The change in estimated building volume (Figure 3) shows much
more dramatically the developments that have occurred. Prior to the
war increase In building volume was closely correlated with the
I £. ~
1 n -
x EST. BLOG. VOLUME
8 .
* BEST FIT LINES
6 .
4.
2 _
x. • •
n .
x.
*
/
. * • •*•V x" " . • x* * X
.-x*x*^'
X
*
1900 1910 1920 1930 1940 19SO 1960 1970 1980 1990
YEAR
.X-
FIGURE 3.
Change in Estimated Building Volume around
Royal Observatory (1904-87)
expansion of built-up area since buildings were predominantly of the
same height throughout the period. From the end of the war and
particularly from the mid 1950 f s onwards there was a market increase
in vertical development resulting in a rapid increase in building
volume. This reached a maximum in the early seventies when most of
the buildings in the area had been redeveloped to the statutory
height limit imposed by government as a consequence of the proximity
of the area to the airport. In terms of urbanization, therefore,
three distinct phases can be identified:
(1) a slow, steady increase in the built-up area with little vertical
development up to about 1950;
(2) a more rapid increase in built-up area concurrent with a rapid
expansion in building height from the mid 1950's to early 1970's;

268
(3) a nearly static phase where built-up area and building height have
reached a maximum from early 1970's to the present.
INTER-RELATIONSHIPS BETWEEN TEMPERATURE AND URBANIZATION TRENDS
What is readily apparent from these two secular trends is the
coincidence in the break points in each, namely around 1950 and around
1970. This statistical association is readily confirmed by regression
analysis which yields correlation coefficients of 0.9329, 0.9407, and
0.8345 between percent built-up area and mean maximum, mean and mean
minimum temperatues respectively. The corresponding values for
estimated building volume are 0.9298, 0.9741, and 0.9029. Such
evidence suggests a strong association between the secular trend of
urbanization and temperature. Since it is the processes of energy
partition and redistribution that influence climate, and thus temperature,
within the urban canopy layer explanation of this association
has to rely on exhibiting links between urbanization processes and
these controls on climate.
Various causal factors have been proposed for the existence of
the canopy layer heat island, all of which tend to make the urban area
warmer (Table 4). However, the relative role and importance of each
within the urban canopy layer varies with city structure and function,
time and season, and regional climate (Oke, 7)), In summer, factors 3,
4 and 1 y and to a lesser extent 5 and 6, are likely to dominate and
combine to make the urban canopy a store of sensible heat by day'thus
raising maximum temperatures. In contrast, at night, factors 2 and 7
act to prevent the rapid dissipation of this store and hence keep
urban minimum temperatures higher. The role of factor 5 is clearly
dependent on city size and function as well as regional climate. In
areas such as mid latitudes in winter the magnitude of this factor can
potentially exceed that of the natural net available energy and can
thus dominate as an urban climate control. The importance of factor 1,
similarly, depends on the type of activities in the city and the degree
and nature of urban air pollution.

269
TABLE 4.
Commonly Hypothesized Causal Factors of
the Urban Canopy Layer Heat Island
Factor
Characteristics
1 Increased longwave counter radiation due to absorption of
outgoing longwave radiation and re-emission by the
polluted urban atmosphere.
2 Decreased net longwave radiation loss from urban canyons
due to reduction in their sky view factor.
3 Greater shortwave radiation absorption due to the
combined effects of canyon geometry and reflectivity.
4 Greater daytime heat storage due to the thermal
properties of urban materials, and night-time release
of this stored energy.
5 Addition of anthropogenic heat from buildings, people,
industrial and commercial activity, and vehicles.
6 Decreased evaporation du£ to removal of vegetation
from and surface waterproofing of the city.
7 Decreased loss of sensible heat due to the reducation
of wind speed in the canopy layer.
after Oke 7)
In Hong Kong in the absence of detailed flux measurements we
can only speculate on the relative importance of each of these
factors. During the daytime 3, 4, 5, 6, and 7, are all likely to be
significant, and at night 2, 5, and 7 probably are the most important
although the relative role of each is unknown. The associations
evident in this study do, however, allow us to attempt an assessment
of the importance of each and thus point towards further study.
In the prewar period the dominant urbanization process was a
steady expansion of built-up area around the Royal Observatory.
Factors 3 and 4 and to a lesser extent 5 and 6 are thus likely to
increase daytime heat storage in the urban canopy and thus exert
dominant influence on daytime maxima. The consequent release of the
excess stored heat similarly would increase nighttime minima. The

270
greater rate of increasing minima in such situations is consequent
upon the varying proportion of the excess store due to urbanization
relative to the daytime and night-time radiation balances, the impact
on the former being less than on the latter. This is consistent with
the view that a major cause of the urban heat island is the change in
radiation balance and not ejected anthropogenic heat.
In the early postwar period this increase in built-up area
continued but a new trend of increasing building height also developed.
Consequently, additional factors may be considered to come into the
relationship. In particular, the reduction in windspeed in the canopy
layer (factor 7) due to the increasing building height would tend to
reduce loss of sensible heat by turbulent exchange both day and night.
This decrease in windspeed near the surface, particularly in the
1960 f s, is clearly evident in Figure 4 where the surface mean wind
3 •
4C -
* * ' * ,
*
•
3 5 -
3.
* .
* * • •
• • *
*
t.
1955 I960 1965 1970 1975 1980 1985 19
YEAR
FIGURE 4,
10 Year Running Mean Postwar Annual Surface Wind
Speed at Royal Observatory, Hong Kong.
speed decreased from about 4.5 to around 3 metres per second from
1960 to 1970. This trend has been reported by Chen 1) who clearly
links it to the erection of tall buildings in the vicinity. Indeed,
the anemometer at the Royal Observatory was raised 6,8 m in 1959 in

271
an effort to minimize the impact of this building activity. In
contrast the wind measured at the 900 hPa level in the free atmosphere
did not change appreciably during the same period (Peterson, 8))
indicating this feature as the likely source of the reduction in
surface windspeed.
In addition, this increase in building height would result in
the presence of warmer buildings occupying a larger solid angle
relative to the cooler sky. Such a reduction in sky view factor would
reduce the loss of heat by longwave radiation from the surface (factor
2) thereby tending to contribute to elevated night-time minimum
temperatures. Finally, during this period factor 5 probably assumed
increasing importance with greater building volume, increasing
population, more human activity and traffic all contributing to the
anthropogenic heat added to the urban canopy. Again, the trend of
more rapidly increasing minima than maxima during this period is
consistent with this scenario.
Lastly, in the 1970 T s and 80 f s the increase both in built-up
area and building volume has nearly ceased as the area has become
almost totally built-up to the maximum permitted by zoning regulations.
However, the increase in population density, traffic volume and the
use of air conditioning has meant that factor 5 probably has assumed a
more dominant role than heretofore. While a continuing increase in
night-time minimum temperatures is consistent with this effect it is
more difficult to explain the steady decline in daytime maxima during
this phase. The explanation may, however, be linked to the fact that
the Royal Observatory is still surrounded in its immediate area by
actively transpiring vegetation. Additional available energy may
enhance this transpiration thus causing a cooling effect not only offsetting
the tendency to further increase in daytime maxima but even
reversing it. In effect the postulation is that factor 6 may not
apply in this situation.
While all that has been postulated so far must be considered
as derived from observed associations it appears to be internally

272
consistent with accepted theory. Two conclusions follow: first, a
detailed study of the fluxes is needed to establish the relative
importance of the factors controlling urban climate in Hong Kong and
second, that an assessment is required of the impact of such changes
on the living environment of the population. In connection with the
latter, some initial investigations have already been made.
IMPLICATIONS FOR HUMAN THERMAL COMFORT
Assessment of human thermal stress in Hong Kong has been made
on the basis of an energy balance equation which specifies the energy
exchange processes between a body and its environment (Kyle, 5)).
This assessment found that while cold stress conditions (body losing
more energy than it gains) prevail in the winter these can be readily
overcome by those wearing light to moderate clothing and engaged in
moderate work. Clearly, the implication of rising temperatures is to
decrease the rate of heat loss and so reduce cold stress conditions in
the winter months. As such it may be considered environmentally
beneficial.
Conversely, in the summer, the study found that heat stress
conditions (body gaining more energy than it loses) regularly prevail,
so that in the hottest months, with customary clothing levels,
undesirable heat stress is incurred by those in moderately heavy work
and thermal discomfort is experienced even though no physical activity
is being undertaken. In this case the implication of rising
temperatures is to exacerbate this situation and must be viewed as a
detrimental environmental effect.
CONCLUSIONS
A strong statistical association has been demonstrated between
long-term secular trends in temperature and urbanization in the area
around the Royal Observatory, Hong Kong which has permitted postulations
concerning how urbanization processes have influenced measured
temperature through changes in radiative and turbulent exchanges in the
vicinity to be advanced. These appear to provide a plausible
explanation for the observed phenomena and provide a basis for further

273
detailed study of changes in the relevant fluxes. The implications
of such inter-connections for human energy exchanges and thus thermal
stress have also been examined. These suggest both beneficial and
detrimental aspects which require further investigation to assess
which is dominant.
REFERENCES
Chen, T.Y., "Comparison of Surface Winds in Hong Kong 1 ', Tech.
Note No. 41. Royal Observatory, Kong Kong, 43 p, (1975).
Detwiller, J., "Evolution Seculaire du Climat de Paris
(Influence de 1'Urbanisme)", Mem. Meteorol. Natl., Paris No.
52, 83 p, (1970).
Fukui, E., "The recent rise in temperature in Japan", Japan
Progr. Climato1., Tokyo Univ. of Education, Tokyo, 46-65, (1970).
Kwok, W.Y., "The Impact of Urbanization on Microclimate as
Evidenced by Longterm Temperature and Wind Speed Trends at the
Royal Observatory, Hong Kong", B.A, Dissertation, Dept. of
Geography & Geology, Univ. of Hong Kong, 106 p, (1982).
Kyle, W.J., "Human thermal stress in Hong Kong", Ann. G.G.A.S,
Univ. of Hong Kong, 10, 1-9, (1982).
Landsberg, H.E. "The Urban Climate", Academic Press, New York,
275 p, (1981).
Oke, T.R., "Inadvertent modification of the city atmosphere
and the propects for planned urban climates", in Proceedings of
Symposium on Meteorology Related to Urban, Regional and Landuse
Planning, Asheville, North Carolina, Tech. Note No. 444,
World Meteorological Organization, Geneva, 151-175, (1976).
Peterson, P,, "Extreme Temperatures in Hong Kong". Te'ch.
Note (Local) No. 22 (revised), Royal Observatory Hong Kong,
34 p, (1981).

274
RECENT APPLICATIONS OF THE PENN STATE/NCAR MESOSCALE MODEL TO
SYNOPTIC, MESOSCALE AND CLIMATE STUDIES
Richard A. Anthes
University Corporation for Atmospheric Research
ABSTRACT
This paper summarizes recent studies of a variety of atmospheric phenomena in different
parts of the world using the Penn State/NCAR mesoscale model. These phenomena include
explosive cyclogenesis over the North Pacific and North Atlantic oceans, cyclogenesis over Europe
and associated ozone transport during the ALPEX experiment, heavy fainfall and flash flood events
over Pennsylvania and China, "Plateau" and "Southwest" vortices over China, severe storms over
the United States, mesoscale convective complexes, elevated mixed layers and "lids," an
Australian Southerly Buster, low-level damming of cold air to the east of the U.S. Appalachian
mountains in winter, urban heat island effects, and regional acid deposition. I also review
Observing System Simulation experiments (OSSEs), several sensitivity studies, the nesting of the
mesoscale model in a global climate model for regional climate studies, and some recent real-time
forecasting studies conducted by the Pennsylvania State University.
An important result of these and earlier studies is that a general mesoscale model with
realistic treatment of surface conditions and physical processes and initialized with good largescale
conditions is capable of simulating and predicting a large variety of synoptic and mesoscale
phenomena in different parts of the world. The model simulations also provide high-resolution,
dynamically consistent data sets which are useful in understanding the physical behavior of
complex mesoscale systems.
1.0 INTRODUCTION
It has been slightly more than a decade since Anthes and Warner (1978) first described
"the development of a general, predictive, hydrostatic meteorological model," which was "suitable
for a wide variety of problems, ranging from the synoptic scale to the small end of the mesoscale."
In this paper, Anthes and Warner hypothesized that many mesoscale phenomena would have some
predictability, even if the model's initial conditions were based only on synoptic-scale
observations: "Thus, we hypothesize that in many synoptic situations, if local forcing is modeled
correctly, the details of the initial conditions are not particularly important." (Anthes and
Warner, 1978, p. 1046).
Since the publication of the first description of the Penn State mesoscale model, there
have been many changes and improvements to the model by students and scientists from The
Pennsylvania State University, NCAR, and many other institutions around the world. The model
has evolved into a 4th-generation model, now designated the Penn State-NCAR mesoscale model
version 4, or MM4, This paper describes recent (over approximately the past two years) results
from a number of studies using MM4, which is being used to study and understand many different
phenomena in several countries. Most of these studies use a version of MM4 very similar to those
described by Anthes et a], (1987).
2.0 DESCRIPTIVE AND DIAGNOSTIC STUDIES
2,1 Explosive Cyclogenesis Over The North Atlantic And North Pacific
Anthes eiak (1983) used the MM4 to study the development of the famous Queen
Elizabeth II (QE-H) storm (Gyakum, 1983) which developed over the North Atlantic. Kuo and Reed
(1988) used the MM4 to study the development of a phenomenal storm that developed over the
eastern North Pacific in 1981. Both of these studies showed the importance of accurate initial
conditions, adequate horizontal resolution, and the release of latent heat in the storms'
development.

0: O O
o S
E
S 1
- I
• £ o
° £} £
4- o oo
a;-:
3 Q§ S
O er
c o f 1
o
g o
*-* £• ^*
Od 0

276
upward motion (Fig. Ib) of air with high equivalent potential temperature. The model predicted a
48-h rainfall maximum of 213 mm over the Sichuan Basin, in good agreement with observations.
Diagnostic analysis of the model results showed that the formation of the SW vortex was
induced by the strong variations in terrain elevation. As the southwesterly monsoon current
impinged upon the mesoscale Yun-Gui Plateau, which extends from the southeastern corner of the
main Tibetan Plateau, the low-level flow was blocked. The air aloft descended into the Sichuan
Basin to the lee of the Plateau, creating cyclonic relative vorticity through vertical stretching.
Differential surface friction and diabatic effects were not responsible for the initiation of
the vortex. However, both latent heating and surface fluxes of sensible and latent heat were
essential to the full development of the SW vortex.
Li ejLaL (1987) simulated another case of extreme rainfall over China. During the period
1200 UTC 19 June to 1200 UTC 20 June 1982, heavy rain fell over the Yangtze Valley, with more
than 300 mm falling near the city of Wuhan. The numerical simulations of this case showed that
the interactions of the latent heat release and the low-and upper-level jets were critical to the
maintenance of the slowly moving convective rainfall systems.
2.4 Simulation Of The Johnstown, Pennsylvania Flood Of 1977 And
Other Mesoscale Convective Systems.
Many successful simulations have been obtained with versions of MM4 that use horizontal
grids of 50-100 km and a relatively simple parameterization of cumulus convective effects.
However, the simulation of many mesoscale convective phenomena requires higher horizontal
resolution and inclusion of more complete physical processes such as downdrafts and
microphysical processes. Da-Lin Zhang and his colleagues at The Pennsylvania State University
developed a nested-grid version of MM4 for study of some of these complex moist convective
events. The modified version of MM4 features a two-way interactive nested grid, (Zhang et al..
1986) with a fine mesh of 25 km and a coarse mesh of 75 km; a version of the Fritsch/Chappell
convective parameterization scheme on the fine mesh; and a Kuo-type convective parameterization
on the coarse mesh. Variations of this model have been used to study the Johnstown flood of 19-20
July 1977 (Zhang and Fritsch, 1986), a mesoscale convective complex over Oklahoma on 7-8 July
1982 (Zhang and Fritsch, 1988c), and a squall line that developed over Oklahoma during 10-11
June 1985 (Zhang, filaL, 1989).
In all of these simulations the model, initialized with conventional observations,
reproduced well many (though not all) observed meso-beta scale features of the convective
systems, supporting the hypothesis that under certain synoptic-scale conditions, mesoscale
convective systems can be predicted using observations from a synoptic-scale network if
appropriate model resolutions and complete physical parameterizations are used.
2,5 Simulations Of The Effect Of "Lids" And Soil Moisture On The
Development Of Mesoscale Convective Systems
The importance of elevated mixed layers and the associated capping inversions ("lids") in
the initial suppression, then initiation of severe convective storms is well known (Carlson eiaL
1983). Idealized simulations by Lanicci elaL (1987) and real-data simulations by Lakhtakia and
Warner (1987) showed that numerical models can accurately simulate the development of these
"lids" and that horizontal variations in soil moisture are often critical to the formation of the lid.
Fig. 2 shows a west-east vertical cross section of an elevated mixed layer and lid from a 9-h model
simulation of the development of the environment of the severe convective storms of 9-10 May
1979. The model simulated well the dry line between 101 and 102 W longitude, the elevated
mixed layer, and the lid. Lakhtakia and Warner's studies showed that differential surface heating
at the edge of the lid was the most important single factor in initiating undemmning of moist
boundary layer air and the subsequent development of convective precipitation in this case.

277
2100 GMT 9 MAY 5 104 J03 IO2 101 100 99 98
LONGITUDE («W)
2.6 An Australian "Southerly Buster"
In the late spring and summer, the southeast coast of New South Wales in Australia often
experiences an intense squall with an abrupt wind change into the south. The squall, which is
associated with a cold front, is sometimes accompanied by severe weather, and can cause loss of
life and property. The event is referred to as the "Southerly Buster."
Although the "Southerly Buster" refers to the narrow (approximately 50 km) squall and
wind shift ahead of the cool southerly current, Howells and Kuo (1988) were able to simulate many
features of the rnesoscale environment of a "Southerly Buster" that occurred on 1 December 1982
using MM4 with 40-km horizontal resolution. The simulated southerly flow was a relatively
shallow and narrow current of cold air. Model sensitivity experiments indicated that orography
and land-sea contrasts played major roles in the production of the observed strong gradients of
temperature and wind in the lower troposphere.
Analysis of the model simulations indicates that the following physical processes
contribute to the development of the environment of the Southerly Busier:
(i) channelling of cool maritime air around the southeastern extent of the mountain
ranges,
(ii) blocking of the flow on the western side of the mountains,
(iii) adiabatic downslope heating of the prefrontal westerly air flow,
(iv) sensible heating of the prefrontal continental air mass,
(v) reduced fractional coupling of the warm prefrontal westerlies over the cooler ocean,
and
(vi) greater friction over land which retards the progress of the front over land compared
to over ocean.
2.7 Appalachian Cold-Air Damming And Coastal Frontogenesis
In the winter and spring when anticyclones are centered over New England, a ridge of high
pressure accompanied by a shallow layer of cold air often extends southward to the east of the
Appalachian Mountains. This persistent cold air is sometimes associated with freezing rain as
warm, moist air from the southwest overruns the stable air mass. A quasi-stationary coastal front
typically separates the cold air over land from the milder air over the oceans. Because of its
shallow nature, the phenomenon has been difficult to predict with operational numerical models.

278
Stauffer and Warner (1987) used the MM4 to simulate the development and maintenance of
an episode of cold-air damming that occurred during 13-14 January 1980, and was associated with
a moderate ice storm during this period. Fig. 3 shows a west-east vertical cross section across the
Appalachians and the coast at 12 h of the simulation. The section shows the shallow dome of cold
air trapped between the Appalachians and the coast. Low-level east and northeast How
maintaining this cold air is overrun by southwest winds aloft. A strong baroclinic zone inland
from the coast depicts the coastal front The simulation also showed realistic vertical profiles of
temperature and wind, including a mountain-parallel jet which is commonly observed along
mountain ridges during episodes of cold air damming,
a
600
CRW BKW
ROA RMT
HAT
CD
Fig. 3: West-east
isentropic cross
section and winds
(m/s) from 12~h
simulation
verifying at 0000
UTC 14 January
1980. One full
barb represents 5
m/s. The caret
denotes the
location of the
coast. (Stauffer
and Warner,
1987).
HTS
DAN
2.8 Urban Heat Island Effects
EWN
Although the MM4 is a hydrostatic model, it has been used at horizontal resolutions as
high as a few km under conditions when nonhydrostatic effects are assumed to be small. In
simulations of the flow over St. Louis, Missouri, Seaman eliL (1989) used a two-way interactive
nested grid version of MM4 with 7.5 km resolution on the outer grid and 2.5 km resolution on the
inner mesh. Realistic ihree-dimensionaliy variable initial and lateral boundary conditions were
specified from observations in order that numerical simulations could be used for quantitative
evaluation of urban effects on the atmosphere. After it was demonstrated that the model could
accurately simulate the observed heat-island effects in a control simulation, the importance of a
number of processes in the urban PBL were investigated. These PBL effects were isolated in
different simulations by using the present surface parameters associated with St. Louis as well as
parameters appropriate to the pre-urban environment and hypothetical parameters associated
with a possible future enhanced urban environment.
2,9 Use Of MM4 To Drive Regional Acid Deposition Model
It is well-known that the three-dimensional synoptic and mesoscale variations in wind
flow, humidity, temperature, clouds and precipitation are important in regional-scale transport,
transformation, and deposition of chemical species in the atmosphere. However, because
atmospheric chemistry has only a small effect on the atmospheric dynamics over periods as short
as a few days, it is possible to use output from a meteorological model as input into a model that
calculates the evolution of chemical trace species. This strategy has been followed in the
development of a three-dimensional Eulerian regional acid deposition model to calculate episodic
chemical concentrations and dry and wet deposition of acidic material over North America (Chang
1987).

279
It is interesting to note that an initial impetus to develop the MM4 system in the mid
1970s was to produce an atmospheric and chemical model to study regional air pollution. It is
also worth noting that a version of the MM4-RADM system, the European Acid Deposition Model
(EURAD) has been adapted for the use in Europe (Ebel el a].. 1989). Hass et al.
(I989)successfully simulated the transport and depositing of radioactive material in Europe
following the Chernobyl accident using a version of EURAD.
3.0 SENSITIVITY STUDIES AND OBSERVING-SYSTEMS SIMULATION
A number of recent studies have examined the sensitivity of regional-scale, limited area
models to variations in initial conditions, lateral boundary conditions (LBC), and treatment of
physical processes. Anthes et al. (1989) analyzed results from 72-h simulations and forecasts for
12 cases to understand the contribution to model error or uncertainty introduced by these factors.
The simulations and forecasts were verified for both the 12 individual cases and for the ensemble
average of the 12 cases, using several objective measures of skill. The differences in these skill
scores were tested for their statistical significance.
The results showed that the use of observed LBC exerts a strong control on the growth of
errors over a domain size of 3600 x 4800 km. The errors showed little growth beyond about 36 h,
so that the 72-h simulations were nearly as accurate as the 36-h simulations. On these time and
space scales, the quality of the LBC were more important than any other factor tested in the
temporal evolution of the model errors. The results showed that the large-scale atmospheric
structure has a major effect on the evolution of small-scale features in the model.
Anthes e_t ah. (1989) introduced the concept of climatological use and verification of
regional models. The climatological skill of the control version of MM4 was quite good. The skill
scores of the ensemble average simulations were considerably better than the average scores of the
individual simulations. The model has small bias errors and the horizontal structure of the
model-simulated atmosphere is similar to the observed structure for scales of motion resolved by
the upper-air observational network over the United States.
In a related study, Warner et al. (1989) performed a series of observing-systems
simulation experiments (OSSEs) with MM4 in order to determine the effect of horizontal and
vertical data resolution, data location, and measurement error on mesoscale forecast accuracy. In
agreement with previous results, the error generally decreased with increasing forecast time.
Major contributors to the observed error reduction included nonlinear effects, geostrophic
adjustment and surface forcing, in addition to the use of identical LBC.
Although the studies by Anthes et aJL (1989) and Warner gt al. (1989) showed that small,
random errors in the initial conditions had relatively minor effect on the simulations, one should
not conclude that models are insensitive to changes in the quality or density of initial data. When
the initial data contribute information that improves the initial analysis in a horizontally and
vertically coherent manner, they can have a major positive effect on the forecast. For example,
Douglas and Warner (1987) showed that incorporating 110 temperature and humidity soundings
derived from the visible and infrared spin-scan radiometer sounder (VAS) into the MM4 produced
large changes in the analysis over the North Pacific Ocean, and that these changes resulted in a
positive impact on the forecast of cyclogenesis over the North Pacific.
A challenge of model initialization is to incorporate the divergence and vertical motion
fields into the model's initial conditions in a way that balances these fields with the diabatic and
other forcing in the model initial data. When precipitation exists at the initial time, the adiabatic
cooling associated with upward motion in the precipitation system is nearly balanced by the
diabatic heating associated with latent heat release. This balance suggests that observed
precipitation rates could be used in a dynamical initialization procedure to generate realistic
vertical motion and related divergence fields in the numerical model. Wang and Warner (1988)
tested a scheme to use observed precipitation rates in the initialization of MM4, the use of the

280
observed precipitation rates focuses the development of precipitation
results in a superior short-range forecast.
early in the model and
Y.-H. Kuo and his colleagues have used MM4 to investigate ways of using wind and
temperature soundings derived from a network of profilers. Kuo eiaJU (1987a) conducted a series
of OSSEs which tested the sensitivity of MM4 forecasts to characteristic measurement errors
associated with a number of hypothetical profiler networks. These results, which used a static
initialization technique, indicated that profiler wind observations would have a positive impact on
short-range numerical weather prediction. They also showed that the use of temperatures derived
from the profiler winds (Kuo filal., 1987b) gave superior results to those derived from the
profiler radiometric measurements; however the derived temperatures were less accurate than
radiosonde temperature measurements.
In an extension of earlier work, Kuo and Guo (1989) tested a dynamic initialization
procedure for continuous assimilation of observations from a network of profilers. The results of
this study indicate:
(i) Dynamic initialization by nudging can successfully assimilate wind profiler
observations to improve the initial analysis compared to a static initialization scheme.
(ii)
(iii)
The improved initial analysis results in a superior forecast.
Assimilation of the wind field is superior to assimilation of the temperature field;
however best results are obtained when both wind and temperature observations are
assimilated.
4.0 REAL-TIME FORECASTS USING MM4
A stringent test of any modeling system is its use in real time. Recently, Warner and
Seaman (1989) have adapted a version of MM4 for running in real time at the Pennsylvania State
University for use in research, instructional, and public service applications. A two-way
interacting nested grid system is used with a fine mesh of 30 km and a coarse mesh of 90 km.
The quasi-operational version of MM4 has been tested on a number of cases since it's
inception in April 1989. The results indicate that MM4 has the potential to contribute to the
improvement of real-time forecasts of rnesoscale phenomena. Work is now underway at Penn State
to make the operational version of the model relocatable and to run the model in real time on a
large number of cases.
5.0 REGIONAL CLIMATE SIMULATION
Realistic simulation of regional climates probably requires a horizontal resolution of
approximately 50 km x 50 km, which represents an increase of a factor of 10 x 10 over that of
present climate models. Such an increase would require an increase in computational power of
approximately ten thousand over present capability. Until such computational power becomes
available, one approach to modeling regional climates over certain areas of interest is to embed a
high-resolution limited-area model in a GCM over the specific region of interest. The GCM is run
first and the output used to provide LBC to the regional model. In this way the large-scale global
climate forcing can be supplied to the regional model, which can in principle add regional details
such as those associated with high-resolution variations in terrain or land-surface
characteristics.
The nesting of a limited-area model (MM4) into a GCM (the NCAR Community Climate
Model-CCM) for the study of the regional climate of the western United States has been
undertaken by Dickinson ejtaL. (19.89). " In initial tests of the coupled CCM-MM4 system,
Dickinson gtal.. (1989) performed 20 days of January simulations to examine five precipitation
events selected from three Januaries of the CCM simulation. Dickinson et al (1989) show that the

281
climatology of the MM4 precipitation for these five cases corresponds much more closely to the
observed average January precipitation over the western U.S. than does the CCM precipitation
climatology, and that the individual storms simulated by MM4 are much more realistic than those
in the CCM.
Using the same CCM-MM4 system, Giorgi and Bates (1989) simulated the period 1-31
January 1979, in which 9 Pacific storms moved across the western U.S. The MM4 captured most
regional features of the orographic forcing of precipitation. Compared to station data,
precipitation amounts tended to be overpredicted. Daily precipitation threat scores for various
precipitation thresholds varied between 0.315 and 0.385. However, the threat scores for the
monthly precipitation, more indicative of the model's ability to simulate regional climatological
precipitation, were higher, ranging from greater than 0.8 for light precipitation to 0.5 for
moderate to heavy precipitation. Snow depths predicted by the model also showed realistic
regional features.
In the first known application of a mesoscale model to paleoclimate studies, F. Giorgi, S.
Hostedtler, and L. Benson (personal communication) have simulated the effect of two large
prehistoric lakes, Lake Bonneville and Lake Lahoutan, which existed in the western U.S. during
18,000 B.P., on the wintertime precipitation of the region. The hypothesis was that the enhanced
precipitation resulting from the large lakes would increase runoff from the winter snows and help
maintain the high lake levels during this period. Fig. 4 shows the differences in the January 1979
precipitation simulated by MM4 between model runs with and without the enhanced lakes.
Maximum differences of more than 2.5 cm exist in two locations in the vicinity of the lakes,
suggesting that the lake-effect snowfalls could have contributed significantly to the precipitation
of the region during this time.
SIMULATED> JANUARY 1979 PRECIPITATION
DIFFERENCES LAKES-CONTROL
CONTOUR INTERVAL OF 1.0 cm
Fig. 4: January
1979
precipitation
differences
between
simulation
containing Lake
Bonneville and
Lake Lahoutan as
they were 15K BP
and the present.
Contour interval
is 1.0 cm. (Giorgi
1989)
6.0 SUMMARY
These results of many studies using MM4 show that a general mesoscale model with
realistic treatment of surface conditions and physical processes and initialized with good largescale
conditions is capable of simulating and predicting a large variety of synoptic and mesoscale
phenomena in different parts of the world. In a significant number of cases, realistic mesoscale
features develop during the simulation even though only large-scale data are used to initialize the
model. The model simulations also provide high-resolution dynamically consistent data sets
which are useful in understanding the physical behavior of complex mesoscale systems.
In addition, the MM4 has been coupled with the NCAR Community Climate Model (CCM) to
study the regional climate of the western United States. Five-and thirty-day simulations of the

282
coupled model system show that the higher resolution of MM4 produces significantly more
realistic cyclonic storms and regional details of precipitation than does the coarse-resolution of
the CCM. These preliminary results indicate that coupled model systems may be very useful in
simulating past, present, and future climates over regions of interest.
' • ' ' ' ' ' - . . ' . ' ' ' . • ' '
Acknowledgments
I acknowledge with thanks the contributions of materials from Larry Benson, Julius
Chang, Bob Dickinson, Matthew Einson, Mike Fritsch, Filippo Giorgi, Ying-Hwa Kuo, Yu-Fang Li,
Simon Low-Nam, Nelson Seaman, Tom Warner, Da-Lin Zhang. Susan Stilwell prepared the final
version of the manuscript.
REFERENCES
Anthes, R.A., 1977: A cumulus parameterization scheme utilizing a one-dimensional cloud model,
Mon. Wea. Rev., 105, 270-286.
Anthes, R.A. and T.T. Warner, 1978: Development of hydrodynamic models suitable for air
pollution and other mesometeorological studies, Mon. Wea. Rev., 106, 1045-1078.
Anthes, R.A., Y.-H. Kuo and J.R. Gyakum, 1983: Numerical simulations of a case of explosive
marine cyclogeaesis. Mon. Wea. Rev., Ill, 1174-1188.
Anthes, R.A., E.-Y. Hsie and Y.-H. Kuo, 1987: Description of the Penn State/NCAR Mesoscale
Model Version 4 (MM4). NCAR Technical Note, NCAR/TN-282 + STR, 66 pp.
Anthes, R.A., Y.-H. Kuo, E.-Y. Hsie, S. Low-Nam and T.W. Bettge, 1989: Estimation of skill and
uncertainty in regional numerical models. Q. J. Royal Meteor. Soc., 115, 763-806.
Arakawa, A. and W.H. Schubert, 1974: Interaction of a cumulus cloud ensemble with the largescale
environment. Parti. /. Atmos. Sci., 31. 674-701.
Carlson, T.N., S.G. Benjamin, S.G., G.S. Forbes and Y.-F. Li, 1983: Elevated mixed layers in the
regional severe storm environment - Conceptual model and case studies. Mon. Wea. Rev. t
111, 1453-1473.
Chang, J.S., R.A. Brost, I.S.A. Isaksen, S. Madronich, P. Middleton, W.R. Stockwell and CJ. Walcek,
1987: A three-dimensional Eulerian acid deposition model: physical concepts and
formulation. J. Geophys. Res.,92, 14,681-14,700.
Dickinson, R.E., R.M. Errico, F. Giorgi and G.T. Bates, 1989: A regional climate model for the
western United States. Climatic Change, (in press),
Douglas, S.G. and T.T. Warner, 1987: Utilization of VAS satellite data in the initialization of an
oceanic cyclogenesis simulation. Mon. Wea. Rev., 115, 2996-3012.
Ebel, A., H. Hass, HJ. Jakobs, M. Laube, M. Memmesheimer, H. Geiss, J.S. Chang, W.R. Stockwell,
CJ. Walcek and Y.-H. Kuo, 1989: Simulation of air pollution transport, transformation
and deposition over Europe using a regional chemical transport model. Proceedings,
International Conference on Atmospheric Chemistry, Beijing, 3'II May 1989.
Giorgi, F. and G.T. Bates, 1989: On the ciimatological skill of a regional model over complex
terrain. Mon. Wea. Rev., (in press).
Giorgi, F., S. Hostetler and L. Benson, 1989: (private correspondence).
Gyakum, J.R., 1983: On the evolution of the QE II Storm L Synoptic aspects. Mon. Wea. Rev., Ill,
1137-1155.
Hass,'H M M, Memmesheimer, H, Geiss, HJ. Jakobs, M. Laube and A, Ebel, 1989: Simulation of the
Chernobyl radioactive cloud over Europe using the EURAD model. Atmospheric
BnvirQmnejal (in press).
Howells, P.A,C. and Y.-H. Kuo, 1988: A numerical simulation of the mesoscale environment of a
Southerly Buster event. Mon. Wea. Rev., 116, 1771-1788.
Hsie, E.-Y., R.A. Anthes and D. Keyser, 1984: Numerical simulation of frontogenesis in a moist
atmosphere. J. Atmos. Sci., 41, 2581-2594.
Jakobs, HJ. and Y.-H. Kuo, 1988: Numerical simulation of a cyclogenesis event during ALPEX.
Annales Geophysicae, Special Edition, European Geophysical Society, XHIth General
Assembly, Bologna, Italy, 117-118.
Kuo,. Y.-HU-L.-Cheng; and R.A. Anthes, 1986: Mesoscale analysis of the Sichuan flood catastrophe,
11-15 July 1981. Mon. Wea. Rev., 114, 1984-2003.
Kuo, Y.-H., E.G, Donall and M.A. Shapiro, 1987a: Feasibility of short-range numerical weather
prediction using observations from a network of profilers. Mon, Wea, Rev., US, 2402-
-. •. ',2427. •; "• ••• • .•'. • •. • : •:•• .• • : ' • .

283
Kuo, Y.-H., D.O. Gill and L. Cheng, 1987b: Retrieving temperature and geopotential fields from a
network of wind profiler observations. Man. Wea. Rev., 115, 3146-3165.
Kuo, Y.-H., L. Cheng and J.-W. Bao, 1988: Numerical simulation of the 1981 Sichuan flood
catastrophe. Part I: Evolution of a mesoscale southwest vortex. Mon. Wea. Rev., 116,
2481-2504.
Kuo, Y.-H. and R.J. Reed, 1988: Numerical simulation of an explosively deepening cyclone in the
Eastern Pacific. Mon. Wea. Rev., 116, 2081-2105.
Kuo, Y.-H. and Y.-R. Quo, 1989: Dynamic initialization using observations from a hypothetical
network of profilers. Mon. Wea. Rev., 117, 1975-1998.
Kuo, Y.-H. and S. Low-Nam, 1989: Prediction of nine explosive cyclones over the Western Atlantic
with a regional model. Mon. Wea. Rev., (in press).
Lakhtakia, M.N. and T.T. Warner, 1987: A real-data numerical study of the development of
precipitation along the edge of an elevated mixed layer. Mon. Wea. Rev., 115, 156-168.
Lanicci, J.M., T.N. Carlson and T.T. Warner, 1987: Sensitivity of Great Plains severe-storm
environment to soil moisture distribution. Mon. Wea. Rev., 115, 2660-2673.
Li, Y.-F., A, Huang and K. Gao, 1987: The role of convective heating in heavy rain system during
Mei-rains season. Scienlia Sinlca (Series B), XXX, 182497.
Seaman, N.L., F.L. Ludwig, E.G. Donall, T.T. Warner and C.M. Bhumralkar, 1989: Numerical studies
of urban planetary boundary-layer structure under realistic synoptic conditions.
/. Appl. Meteor.. 28, 760-781.
Stauffer, D.R. and T.T. Warner, 1987: A numerical study of Appalachian cold-air damming and
coastal frontogenesis. Mon, Wea. Rev., 115, 799-821.
Wang, W, and T.T. Warner, 1988: Use of four-dimensional data assimilation by Newtonian
relaxation and latent-heat forcing to improve a mesoscale-model precipitation forecast: A
case study. Mon. Wea. Rev., 116, 2593-2613.
Warner, T.T., L.E. Key and A.M. Lario, 1989: Sensitivity of mesoscale-model forecast skill to some
initial-data characteristics; data density, data position, analysis procedure and
measurement error. Mon. Wea. Rev., 117, 1281-1310.
Warner, T.T. and N.L. Seaman, 1989: A real-time mesoscale numerical weather prediction system
for research, teaching and public service at Penn State University. Bull. Amer. Meteor.
Soc., (submitted).
Zhang, D,-L. and R.A. Anthes, 1982: A high-resolution model of the planetary-boundary-layer
sensitivity tests and comparisons with SESAME-79 data. /. Appl. Meteor., 21, 1594-
1609.
Zhang, D.-L., H.R. Chang, N.L. Seaman, T.T. Warner and J. M. Fritsch, 1986: A two-way interactive
nesting procedure with variable terrain resolution, Mon. Wea. Rev, 114, 1330-1339.
Zhang, D.-L. and J.M. Fritsch, 1986; Numerical simulations of the meso-^ scale structure and
evolution of the 1977 Johnstown flood. Part I: Model description and verification.
J. Atmos. Sci.. 43, 1913-1943.
Zhang, D.-L., 1989: The effect of parameterized ice inicrophyses on the simulation of vortex
circulation with a mesoscale hydrostatic model. Tellus, 41 A, 132-147.
Zhang, D.-L. and J.M. Fritsch, 1988a: Numerical sensitivity experiments of varying model physics
on the structure, evolution and dynamics of two mesoscale convective systems. /. Atmos.
Sci.,45, 261-293.
Zhang, D.-L. and J.M. Fritsch, 1988b: Numerical simulation of the meso-p scale structure and
evolution of the 1977 Johnstown flood. Part III: Internal gravity waves and the squall
line. /. Atmos. Sci. t 45, 1252-1268.
Zhang, D.-L. and J.M. Fritsch, 1988c; A numerical investigation of a convectiveiy generated,
inertially stable, extratropical warm-core mesovortex over land. Part I: Structure and
evolution. /. Atmos. Sex., 116, 2660-2687.
Zhang, D.-L., E.-Y. Hsie and M.W. Moncrieff, 1988: A comparison of explicit and implicit
predictions of convective and stratiform precipitating weather systems with a meso-j3-
scale numerical model. Q. J. Royal MeteoroL Soc. f 114, 31-60.
Zhang, D.-L., K. Gao and D.B. Parsons, 1989: Numerical simulation of an intense squall line during
10-11 June 1985 PRE-STORM. Part I: Model verification, Mon. Wea. Rev.,117, 960-994.

284
Baroclinic Instability of Modified Eady Waves
by
Wen-yih Sun
Department of Earth and Atmospheric Sciences
Purdue University
W.Lafayette, IN 47907
ABSTRACT
Baroclinic instability in an inviscid fluid with parabolic potential
temperature profiles is investigated. In addition to the long wave
disturbances (Mode I), shortwave disturbances (Mode II) can also develop
in the lower atmosphere, where the stratification is weaker. The growth
rate of the short waves increases with increasing stratification aloft.
Numerical simulations obtained from a nonlinear mesoscale model also
confirm that a short wave can develop into a surface front within a few
days. Those short waves may correspond to the medium-scale disturbances
observed over the AMTEX (Air Mass Transformation Experiment) region.
1: Introduction
One of important phenomena observed during the AMTEX was the medium-scale
disturbances over the East China Sea in winter 1]. The length scale of those
disturbances was 1000-2000 km in the east-west direction. They became active in a
moist lower troposphere with a less stable stratification, and were not associated with an
upper tropospheric trough. Conventional baroclinic instability and symmetric
instability have been applied to study those disturbances. Because a constant
Richardson number was used in the entire domain, the results obtained by Gambo 2»3]
and Tokioka4»5] fail to explain some important characteristics of those waves.
Observations indicated that the stratification in the lower atmosphere was much less
than in the upper atmosphere during the AMTEX. Here, the Eady model

285
has been investigated by Eady 6] and many others. Instability of short waves has also
been studied by Staley and Gall 7] by using a four -level model Blumen 8] used a
two-layer Eady model to study instability of short waves due to the jump of
stratification at the interface. Instability in a nongeostrophic system for a fluid, which
includes a smooth transition layer between two layers of different stratification, has also
been studied by Nakamura 9], Their results confirm that the short waves become
unstable if the stratification in the lower atmosphere is weak.
Here, a constant wind shear is assumed in the u-component. The vertical (potential)
temperature profile is described by a second order polynomial function, which provides
a weakly stable stratified atmosphere near the surface and a very stable layer aloft This
may resemble the climatic environment over the TAMEX region during winter.
2.2: Basic Equations For A Modified Eady Model
With the pseudo-height z=Hg/K (l-(p/p 0 ) K ) 1 0] being used as a vertical coordinate,
the basic equations for an inviscid, compressible atmosphere are identical to William's
model * 1] with Boussineq approximation. The initial basic (potential) temperature may
be represented by:
6 =az 2 +bz+0 0 +(30/3y)y. (2.1)
where a and b are constants, and will be discussed later. We are limited to stable
stratification in the whole atmosphere where the Richardson number is greater than one
to avoid symmetric instability 12]. The basic wind is assumed:
U
and the variation of the initial 0 in the y direction is
36 -f0o9U -f0oV
- = -- = . - = constant (2.3)
3y g 3z gH
In a two dimensional flow, we can have a single nondimensional equation for
perturbation pressure variables
* in a quasi-geostrophic system:
(i_ + r^_)( 3 > + !_ (I !i' )).!*! ^H» =0, (2.4)
3 1* 3 x* 3 x* 2 3 z* p 3 z* 3 x* d z*
where the Buger number p = (gH2/9 0 f2L2) (dG/dz) = N% 2 /f 2 1 - RiR 0 2 ,
N 2 =(g/ 0 0 ) (d 0/dz) and Ri , R o are Richardson number and Rossby number,
respectively. The boundary conditions of (2.4) are
1 3 3 3 * 3 6*
w* = ~- [(JL + z*^) ^~)~ ] «0, at z* = 0,L (2.5)
3t* 3x* 3z* 3x*

286
Eq (2.4) corresponds to the conservation of quasi-geostrophic potential vorticity q
which is defined as q=Q+q' with
d 1 dp' , dQ d. N
q'= C + f —(
r ) and = - — ( — o — ) (2.6)
dz N^dz dy dz N 2 dz
In (2.32) and (2.34), the solution is assumed in the form of
* = ® (z*) exp (i a (x*-ct*) ), (2.7)
we obtain a second order differential equation for
d 2 dp/dz* dO dp/dz* 1
-_1 + (_1 pa 2 ) 0=0, (2.8)
dz* 2 p dz* p (-c+z*)
while boundary conditions are:
(-c+z*) = at z* = 0, 1. (2.9)
dz*
Eqs (2.8) and (2.9) can be solved by the reduction and iteration method 1^].
3: Eigenvalues Of The Modified Eady Problem
The constants of Eq. (2.1) are: a=n*del, where n=l,2,3»4,5 for cases 2a to 2e, and
n=0 for an Eady problem (indicated by E); del=0.05 K km" 2 ; b=2.0 K kmfl; 8 0 =288 K,
and d6/dy = -10"^ K nr 1 . Besides, we also include a case (indicated as Es) with a=0,
and b=3.9 K tor 1 . The length scale L = 1000 km. The wave numbers a tested range
from 0 to 5 (according to wavelength X £1256.6 km). The observational horizontal
length of the medium-scale disturbances over AMTEX is about 1000 — 2000 km,
which is within the range of this study. The basic potential temperature profiles and the
corresponding Burger numbers are shown in Figs. 1 and 2, respectively.
The phase speed c r and growth rate oxj obtained for various cases are shown in
Figs. 3 and 4. In Mode I region, c r and cxcj decrease with increasing stratification.
Fig. 3 also shows that c r for the unstable waves is no longer a constant 0.5. It decreases
slowly with increasing wave number in the Mode I region, and decreases more quickly
in the Mode n region. The transition wave numbers between Mode I and n also
decrease from case 2a to 2e. Our c r are similar to those obtained by Blumen^] and
Nakamura^l, except no neutral wave exists in our results, which is consistent to the
critical layer instability ^»^1.
The maximum growth rate and the most unstable wavelength are also comparable

287
with the classical Eady waves, although c r gradually decreases with the increase of
stratification aloft. Fig. 4 reveals the short wave cutoff for classical Eady waves, but
short waves become unstable when the vertical variation of stratification is included (i.e,
3Q/3y * 0), as expected. However, ocq of the short waves increases with increasing
stratification aloft.
4: Structure Of The Baroclinic Disturbances
Figs. 5 - 6 show the pressure and temperature perturbations (* and 0*) at OC] =
0.5*7C (i.e, wavelength A, =4000 km) for case 2a, which are very close to the
P8TENTIRL TEMPERRTURE
BURGER NUMBER
/// .' /
.« .« .6 .7 .1 1.1 1.2 1.5 1.4 l.S l.t
1. The basic potential temperature profiles 2. Burger number (3 in conjunction with
in this study.
R£_fiT;VE. PHfiSE SPEED
temperature profiles shown in Fig. 1.
GR8HTH RfiTElMCl)
W-
D .5 1.0 l.S
3. Variations of the real part of eigenvalue, 4. Variations of the growjh rate,
c r ,with wave number.
aq, with wave number.

288
conventional Eady waves, except that the perturbations are slightly weaker near the top
5. Pressure perturbation * for 2a with 6. Temperature perturbation 8* for 2a
wavelength A,=4000 km, the height
with A.=4000 km.
of the domain is 8500 m.
7. Pressure perturbation 4* for 2e 8. Temperature perturbation 6* for 2e
with X=4000 km.
with X=4000 km.'

289
Figs. 7-8 (for case 2e) show that a stronger stratification aloft reduces the amplitude of
0* in the upper layer considerably. This also reduces the height of the steering level so
that c r of 2e is smaller than that of 2a, as discussed in Fig. 3. Overall, the fundamental
7'i'"r r in i 111 i IITTI r-f i i ri/ ri'i i;i i n rrrn TrrvT'i
9. * for 2a with X=1600 km, the height 10. 0* for 2a with X=1600 km.
of the domain is 8500 m.
11. d>* for 2e with X=1600 km. 12. 8* for 2e with X=1600 km.

290
perturbations in Mode I remains similar to the original Eady waves.
4> * and 0* for a 2 =1.25*rc (i.e., X, =1600 km) of Mode n are presented in Figs.
9-10 for 2a, and in Fig. 11-12 for 2e. The perturbations are more confined in the lower
atmosphere, especially for case 2e. The temperature slightly tilts eastward in the lower
layer, atop a transition layer, where it tilts westward drastically. The tilt of 9* is very
small in the middle and upper atmosphere. The structures of * and 0* for 2e are
similar to the short waves discussed by Nakainura ^1 Bretherton 15,16] argues that the
presence of the new unstable modes can be caused by the existence of a gradient of the
basic state potential vorticity (i.e, 3Q/3y ^ 0) in the interior of the flow, according to
Eq. (2.6).
The height of the disturbance may be measured by the Rossby depth of penetration,
h « c r /(V/H). h2a ~3 km for 2a, and h2e«2.5 km for 2e. Those heights are
comparable to the height of the transition level of 0 * in Figs. 10 and 12. The average
a0/3z=2.15 K km' 1 at z « 3 km for 2a; and 90/3z=2.63 K km" 1 at z « 2.5 km for 2e.
The effective stratification of 2a is smaller than that of case 2e, but the growth rate of 2e
is larger, as shown in Fig. 4. Therefore, the short waves are sensitive to the
stratification not only just in the lower atmosphere but also in the upper layer.
Comparing perturbations fields of 2a (Figs. 9-10) and 2e (Figs. 11-12), we can see that
decrease of I h. The change of w'0' across h is much smaller
in 2a than 2e, because motion in 2e is more confined in the lower layer and is able to

291
generate more positive kinetic energy in the lower layer. Hence, the total net buoyancy
production term per unit volume in 2e is 1.18xlO" 9 , which is larger than the value of
0.804x10-9 in 2a.
We may define the effective Burger number as
g (2h* )2 d6
fieff = — (4.2)
dz
where h* is the height of the maximum of the horizontal average w'9 1 . With X =1600
km, we obtain that Beff (2a) = 0.064 for 2a and Beff(2e) =0.047 for 2e, which may
explain a larger growth rate of the short wave for case 2e. For the long wave (X =4000
km and 2h*=6.68 km) for case 2e, we obtain Beff = 0.035, which is still less than that
of the short wave. This may suggest that Beff may be better than the average
stratification in order to determine

292
5: Summary And Remarks
Two different types of disturbances can be generated by baroclinicity with parabolic
temperature profiles in the vertical direction. The waves are unstable in all the test
wave numbers. The long wave disturbances are quite similar to the classical Eady
problem, which has a larger growth rate and propagates faster than short waves. The
short wave disturbances are mainly confined to the lower atmosphere, where the
stratification is weaker. The growth rate of these waves increases with increased
stratification aloft. Although the long waves are more unstable than the short waves,
according to linear stability in a quasigeostrophic system, the short waves may become
dominant through nonlinear interactions, and/or enhancement by diabatic heating in the
lower atmosphere, which remain to be investigated. The short waves generated here
may be associated with the medium-scale disturbances observed over the AMTEX
region, or the surface front in the lower atmosphere with a weak stratification.
Acknowledgments
Contributions of Ms. R. L. Kao and Mr. A. Yildirim on this work are appreciated.
Part of this work was supported by NSF under grants ATM-8313418 and
ATMS-8611729.
References:
1] Nitta, T., and M., Nanbu and M. Yoshizaki, 1973: Wave disturbances over the
China Continent and the Eastern China Sea in February 1968, J. Meteor. Soc.
2] Gambo, K., 1970a: The characteristic feature of medium-scale disturbances in the
atmosphere L I Meteor. Soc. Japan, 48.173-184.
3] Gambo, K., 1970b: The characteristic feature of medium-scale disturbances in the
atmosphere II. J. Meteor. Soc. Japan, 48, 315-330.
4] Tokioka,T., 1970: Non-geostrophic and non-hydrostatic stability of a baroclinic
fluid. I Meteor. Soc. Japan. 48.503-520.
5] Tokioka, T., 1971: Supplement to non-geostrophic and non-hydrostatic instability of
a baroclinic fluid and medium-scale disturbances on the fronts. J. Meteor. Soc.
Japan v 49.129-132.

293
6] Eady, E. T., 1949: Long waves and cyclone waves. TellusJ,33-52.
7] Staley, D. O., and R. L. Gall, 1977: On the wavelength of maximum baroclinic
instability. J. Atmos. ScL 34.1679-1688.
8] Blumen, W., 1979: On short-wave baroclinic instability. J. Atmos. ScL
26,1925-1933.
9] Nakamura, N., 1988: Scale selection of baroclinic instability-effects of stratification
and nongeostrophy. J. Atmos. ScL45.3253-3267.
10] Hoskins, B. J., 1971: Atmospheric frontogenesis: some solutions. Quart. J. Roy.
Meteor. Soc.. 97.139-153.
11] Williams, R. T., 1967: Atmospheric frontogenesis: A numerical experiment. I
Atmos. ScL. 24. 627-641.
12] Stone, P. EL, 1966: On non-geostrophic baroclinic stability. J. Atmos. ScL 23.
38-52.
13] Charney, J. G., and M. E. Stern, 1962: On the stability of internal baroclinic jets in
a rotating atmosphere., J. Atmos. Sci., 19,159-163.
14] Kuo, H. L., 1978: A two-layer study of the combined barotropic and baroclinic
instability in the tropics. J. Atmos. Sci.. 35. 1840-1860.
15] Bretherton, F. P., 1966a: Critical layer instability in baroclinic flows. Quart. J.
Rov. Meteor. Soc.. 92.. 335-345.
16] Bretherton, F. P., 1966b: Baroclinic instability and the short wavelength cut-off in
terms of potential vorticity. Quart. J. Roy. Meteor. Soc., 92, 335-345.
17] Sun, W. Y., 1976: Linear stability of penetrative convection. J. Atmos. Sci., 33.
1911-1920.
18] Sun, W. Y., 1989: Baroclinic Instability and Surface Waves (submitted for
publication).
19] Sun, W. Y., and W. R. Hsu, 1988: Numerical study of cold air outbreak over the
warm ocean. I Atmos Sci.,45., 1205-1227.
20] Kao, R. L., 1987: Baroclinic instability and frontogenesis. M.S. Thesis, Dept. of
Earth and Atmospheric Sciences, Purdue University, W.Laf. IN., 109pp.

294
INFLUENCES OF OROGRAPHY ON FLOW IN BOUNDARY LAYER
Wu Rongsheng
Department of Atmospheric Sciences,
Nanj ing University,
Nanjing, 210008,
China.
I . INTRODUCTION
The classic Ekman flow is the result of the balance of three
forces, Coriolis, friction and pressure gradient forces. However, when
Rossby number is not sufficiently small then inertial force will be
important for atmospheric motion. Furthermore, the orographic effect
plays an important rote in the tower atmosphere, especilly for the
boundary flow. In the paper, the geostrophic momentum approximation is
used to study the influence of orography when the inertial force is
taken into consideration.
With a -coordinate, the analytical solution of the flow in the
boundary layer is obtained with the geostrophic momentum approximation.
The vertical motion at the top of the boundary layer is studied in
detail. Numerical examples are given to show the features of the flow
over an ellipsoid mountain under the assumption of jet-type geostrophjc
wind aloft. It is pointed out that the relative position of the jet
stream and the ridge of the mountain as well as the vorticity
distribution of the jet is important for Ekman suction.
Ii. GOVERNING EQUATION OF BOUNDARY LAYER IN 0 -COORDINATE
or-coordinate is used in this study. Let a be
a= — , (1)
P*
where P 5 (*> y/ t) is the pressure at surface. The temporal average is
defined as usual, i. e.,
5= -/;< )dt . . (2)
At
The deviation from the average is expressed as ( )" with the
property,
r"3"= o v o)
The pressure at the surface can be expressed
Pj

295
But, for wind and geopotential height field, the average with p (x, y)
as the weighting function is used. The definition of weighted average is
("""% , (5)
h
where the tilde symbol denotes the weighted average and ( )' expresses
the deviation from the weighted average. Thus, V and $ can be
expressed as^,
Y = V + "V' , = ^ + 4»' . (6)
The charactistics of weighted average can be found in book of Cebeci
and Smith (1974). Some usefull relations are given as follows. For
example, according to the definition, we have
p $ V = (IT + p $ "XV* + V') = j£\T+ p 5 "Y + p 5 W . (7)
Averaging eq. (7), it reduces to
p V = p V + p V . (8)
With the difinition of weighted average, the following result is
obtained^
pjT'r 0 . (9)
Similarly, we have
P 5 7 4>'- 0 . (10)
In terms of these relations, the contiunity equation may be reduced as
follows after averaging over time
•- i '+ T-p'V-+ -— ( a) = 0 . (11)
$
at
c)

296
(13)
The next step is to determine the stress F^, F Assuaing that a
parcel of air loving froa a to cr+l, keeps its aoaentua constant
within the distant of I, then the deviation of aoeentua u', say, is
- u( a
3 a
where I is a non~di»ensional quantity with the randoi character. Now,
for convenience, it can be taken as an average value, as in the case of
aixing lenghth . If
u'—v'— a a ' , (15)
where a
is a paraaeter, introduced for hoaogeneity in diaension, then
...... ,

297
dary layer in z-coordinate to a -coordinate. Finally, the soaentui
equation are written as
Let
^ y
^
fr = i— + — 22-L ,
3 9x p 5 7x
y
P * Py
- f!T= +— . (19)
where if , v^ are the geostrophic wind coaponents in cr-coordinate. If
the following relations
>
(x, y) = L(x', y') ,
L
t = t' . (20)
and the stretching variables
1-a ^ K
are defined, then the non-dimensional moientua equation are reduced to
_ > /v ^ ^9 ^ ^ ^ 1 9*«"
R ( + u + v )ti. - v + v fl = ,
at ^x ay 3 ^ 2 a n
> ^ ^X C> ^ /^ /^ 1 P V
4 t 2>x ay ^ J
where R is Rossby number. Since the boundary layer is so thin, the
contribution of the rate of change of u and v with respect to a can
be neglected. In this case, eq. (22) can be reduce as
1 P K *s t 'V
—_. + a u + b v = c ,
2 Pn '
2 3 if *
a
X-HOV
(.23)
- 1

298
where
„
= 1 - R
a = - R 2>x
c _ v 4
9 x
= .- R - R (24)
This is saiae as that derived by Wu and Blunen. The tower boundary
condition is
TJ - 0 ; ^ 7= 0 .
(25)
The upper boundary condition is
v = V
(26)
since at the top of the boundary layer, a = a 9 , K-*0 then rj -*-oo. v^f
V T are the wind coiponents at the top of boundary layer. As a matter
of fact, they can be calculated by neglecting the friction ter»s in eq.
C23> as
D = a b-
ay
(28)
The solution to eq. (23) under the boundary condition (25), (26) are
r^ ~f
-1 -ft
u = u -u e cos£- c / D e r sinp
(29)
^ •»$ -2 ,5
v = V T -V e conp- c^D e r sinp . (30)
Foraally, they are the sane as the result obatined by Wu and filuaen
(1982). However the leaning of TJ is different since the orographic
effect has been included. To facilitate comparison with the
results obtained in p-coordinate, the solutions will be returned to the
p~coordinate t Since in the problem, T» , *¥* are assumed constant in
^ *^ 9 *
0 direction. That means
equal to zero. Thus from eq.(2)>
we have
V
= yy/
The quantities in (29), except J* , can be directly replaced by those in
p-coordinate, while 0 is
(32)

299
As R -*0> 0=1, U T = u , V T ~ u , then the solutions iaay be reduced to
the classical Ekman flow as
^ ^ ^ -vj -- ,*
u = u - u e cos q - v e'sinn,
5 3 *
T = / vl- v;e cos q -IT e'sintj (33)
The detail discussion of the solution aay be found in Wu (1989).
EL VERTICAL MOTION AT THE TOP OF 'BOUNDARY LAYER
Chen (1979) pointed out that for large scale motion, the
continuity equation in a coordinate can be simplified as follows
according to scale analysis
_C3f * _/r
3
da 5
Using i\ instead of a, then nondimensionai equation reads
V p V + - (p a>= 0 . (34)
^ —

300
— -~ 1, thus the above relation becomes
IV
Sf ^
a = - » T . (41)
Then the vertical velocity at the top fo the boundary layer in 2
coordinate reads
< 42)
III order to appeciate the orographic contribution, the influence of
geostrophic momentum can be neglected temporarily. In this case
CZ* *^f
D = f, C f = Yj, t z -- -7y y^ . (43)
And the vertical velocity is expresed as follows
1 = -av 4 .Vlnp -* - (K • VAVlnp-) + - L - (44)
T
4 2 2 ^
These three tens on the right Jiand side of eq. (44) denote three
different physical processes* First, the orographic effect, ~ay*vlnp".
If the surface is flat then vp^ =0, and this ter» vanishes. Second,
frictional effect, £ . This means that in a tow pressure
2 3
system, £ > Q, and this term makes positive contribution. The otherwise
is true in a high pressure system. Since the flow is always deflect
to the low pressure side when the infuence of friction is taken into
consideration. Thus it results in convergence and vertical motion at
the top of the boundary layer in a region of low pressure as the
continuity principle requires. Similar argumemts apply to high pressure
systems. This was discussed by Charney and Eliassen in 1949. The third,
. _
is the joint effect of orography and friction, (K * VAvlnp ).
2 *
If the surface is flat or the motion is fric-tionless, then .lftp^= 0
or K= 0 respectively. In this case, the third term vanishes. It is
non-zero only if both effects coexist. The physical explanation of
this term is that under the influence of friction, even if the
geostrophic wind is parallel to the configuration of orography, the
ageostrophic component will force the air upward or downward. These
three factors can be visualised in the following schematic
illustrations.

301
Cb)
CO
Fig. la shows the orographie forcing. Fig.ib shows that even if the
geotrophic wind is parallele to the configvration of the orography, the
ageostrophic comporent due to friction will cause upward or downward
motion over terrain. Fig.lc shows the vertical motion due to friction.
3V. NUMERICAL RESULTS
For large scale motion, R is small and is neglegible. Thus
vertical velocity at the top of boundary layer may be reduced to
the
I = -aV 5»
Assume the distribution of p
as
2 3
p = 200(3 + 4A - 2A ) , A 1
where a, b denote the major and minor axes of an ellipse respectively.
The surface pressure at the peak of the orography is 600 hPa. The
geos trophic wind is assumed as
where u f , 6^ are parameters. Y 0 denotes the position of the axis of
the jet. In the numerical experiment, a is taken 0.0841.
I . Y = Oj jet is coincident with major axis.

302
The vertical velocity at the top of the boundary
figure 2.
layer is shown in
Fig,2 Vertical velocity at the top of the boundary layer over the
mountain; a) Total amount ofthree factors, b) two factors without
the joint effect.
The distribution of upward and downward motion is symmetrical. Figure
2b shows the distribution of vertical velocity when the joint effect
is negligabte. In this case the positive area moves nouthward and
negative area southward. This can be explained by the non-uniform
distribution of geostrophic vorticity.
II. Y = -0.3; jet is located at southern slope of mountain.
The distribution of vertical velocity is shown in Figure 3. It is
apparent that the strength and location of upward and downward centers
are different.
Fig. 3 Same as Figure 2 with jet over the southern slope of the moutain.
R number is taken into consideration, the velocity is
approximately expressed as
Take
V-Y 2B 3 3 D y 2B f
^ ; : --KMif -•-
u.= L5e , v = 0.
The orography is assumed to be
p = 50C7A - 3A 4- U) ", A

2
y
-f -5
* b*
The numerical result is to some extent similar to those of Urge scale
motion. The illustrations are oaitted for brevity
303
V. CONCLUSION
The veritcal motion at the top of boundary layer over Mountains
is affected by three factors, friction convergence, orographic forcing,
and joint effect of both friction and orographic forcing. Numerical
results show that these three factors are of the same order of
importance. The position of the geostrophic jet stream relative to the
major axis of an ellipsoid orography is also important in determining
the distribution of vertical motion.
REFERENCES
Atpert, P. amd I. Neuman. (1986), Horizontal components of the
frictional flow in the a-coordinate system for mesometeorological flow
model, Sixth conferice on Nnmerical feather Pridiction, Amer. Met.
Soc., 310-312.
Cebeci, T., and A. M. 0. Smith (1974), Analysis of turbulent boundary
layer,, Academic Press. 404.
Charney, J. G., and A. Eliassen (1949), A unmerical method for
predicting the perturbation of the midlatitude westerlies, Tellus, 1
30-54.
Chen Qiushi (1979), The simplified equations and the phsical processes
of the mountain effects on large scale motion in middle latitudes, Acta
meteorological Sinea, 37, 88*102.
Godev, N. (1983), Contribution of mutual effect of orographic and
nthermal inhomogeneous and surface friction to the generation of
Mediteranean cyclones, WMQ Short- and medium range weather predictions
research publication series, 3, 75-119.
Haltiner, G. Jl and R. T. Williams. , Numerical prediction and dynamic
meteorology, 2nd ed. John Wiley * Sons, 477(1980).
Wu, R, and W. Blumen, An analysis of Ekman boundary layer dynamic
incorporating the geostrophic momentum approximation, J. A. S. 3J,
1774-1782(1982).
Wu, R. The influences of orography upon the flow within Ekman boundary
layer under the assumtion of geostrophic momentum, Adv. Atm, Sci. l f
1-7(1985).
Wu, R. Effects or-ography and air flow in Ekman boundary layer, Acta.
Met. Scinlca, 47. 137-146(1989).

304
THE MIGEOPHYSICS OF A MEI-YU CASE:
DATA ANALYSIS
Chung-Ming Liu and K. Kenneth Lo
Department of Atmospheric Sciences
National Taiwan University, Taipei, China
ABSTRACT
The microphysical processes in the melting layer and
the warm-rain region in the stratiform precipitation area
of a MCS (Mesoscale Convective System) observed during
TAMEX is studied through analyzing the microphysical data
collected by the PMS (-Particle Measuring Systems, Inc.)
2D-C imaging cloud probe and the 2D-P precipitation probe
onboard the NOAA WP-3 aircraft. The observed MCS consists
of a multi-cell convective line system oriented
approximately north-south, and a widespread stratiform
precipitation area. Two box flight tracks are in the
stratiform area. Studies of the aircraft data collected
along these two tracks provide us with an opportunity to
identify the general features in the melting layer and the
warm-rain region.
The 0°C isothermal layer is approximately below the
5km level and is about a .few hundred meters deep. The
melting layer extends from the 0°C level toward the 2-3°C
level (about 4.5km in altitude) vertically, and is
characterized by a sharp decrease in particle number
density by an order of magnitude caused by^ the significant
increase of particle fallspeed as melting snow turns into
wet droplets. Meanwhile, a steady increase of the
percentage of large particles immediately above the
melting layer followed by a sudden decrease are noticed in
both probes' data. These phenomena stress that snowflake
aggregation generates large particles, in particular in
the 0 C isothermal layer where wet snowflakes have much
higher chance to collect other particles. Before reaching
the bottom of the melting layer, either the
collision-breakup or the spontaneous breakup process
causes a swift disappearance of these large, wet
particles, since the breakup efficiency increases with
particle size. Studies of the particle spectra above and
in the melting layer clearly reveal the physical processes
of aggregation, melting and breakup.
The warm-rain region below 4.5km interfaces both the
melting layer and the precipitation at the surface.
Analyses of the vertical profiles of raindrop median
diameter and droplet percentage in different size ranges
suggest that the collision-coalescence process is dominant
in the warm-rain region to generate larger drops at the

305
expense of smaller droplets and to increase the median
diameter by a factor of 2-3 from the bottom of the melting
layer downward to the ground level. Still the competition
between the coalescence and the breakup processes through
binary interactions eventually leads to an equilibrium
state. The analyses of the raindrops spectra reveal the
existence of three peaks below 1.5km as proposed by other
theoretical and observational studies for equilibrium
raindrop spectra.
1. INTRODUCTION
Using the data collected by the PMS (Particle
Measuring Systems, Inc.) 2D-C imaging cloud probe and the
2D-P precipitation probe onboard the NOAA WP-3 aircraft,
we have analyzed the microphysical processes in the
melting layer and the warm-rain region in the stratiform
precipitation area of an open-ocean mesoscale convective
system (MCS) which occurred on June 16-17 1987 just off
the southeast coast of Taiwan during the Taiwan Area
Mesoscale Experiment (TAMEX).
The observed MCS consists of a multi-cell convective
line system oriented approximately north-south and a
widespread stratiform precipitation area. Two box flight
tracks (marked by 1 and 2, respectively, in Fig. 1) are in
the stratiform area. In both tracks the aircraft descends
spirally from 5.5km toward. 0.35km in 19min (descending
speed about 4.5m-s~ 1 ). Studies of the aircraft data
collected along tracks 1 and 2 provide us with an
opportunity to identify the general features in the
melting layer and the warm-rain region.
The 0°C isothermal layer is approximately below the
5km level and is about a few hundred meters deep (Willis
and Heymsfield, 1989). The melting layer extends from the
0°C level toward the 2-3°C level (about 4.5km altitude)
(Stewart et al., 1984, Willis and Heymsfield, 1989). The
warm-rain region below 4.5km interfaces both the melting
layer and the precipitation at the surface.
The analyses schemes for the 2D-C and 2D-P imaging
probe data are those described by Heymsfield and Parrish
(1978), Heymsfield and Baumgardner (1985) and Parrish and
Heymsfield (1987).
2. THE MELTING LAYER
Analyses of the in-situ measurement data show that
from the top to the bottom of the melting layer, there are
(1) a sharp decrease of the LWC (liquid water content,
measured by the Johnson-Williams device) from 0.6 to

306
O.Qg-m" 3 , which is probably due to an increase of the
efficiency of the particles' collision-coalescence process
as melting snowflakes turn wet; (2) an obvious shift of
the horizontal wind (Fig. 2); (3) a transition toward
mesoscale descent motion below the melting layer. These
findings support the idea that the pressure perturbation
induced by the cooling of melting snow causes a decoupling
of dynamics above and below the melting layer (Johnson and
Young, 1983)
Analyses of the probe-imaging data show that the
significant increase of particle terminal velocity from
melting snow to liquid droplet results in a sharp decrease
of the particle number .density by an order of magnitude
(Fig. 3). Lo and Liu (1989) have applied a theoretical
approach to quantify such relationship.
Meanwhile, a steady increase in the percentage of
large particles (diameter 2500-5000jLim for 2D-C and
2000-lOOOOjLtm for 2D-P) immediately above the melting
layer followed by a sudden decrease are noticed in both
probes' data (Figs. 4 and 5), These phenomena stress that
snowflake aggregation generates large particles, in
particular in the 0°C isothermal layer where wet
snowflakes have a much higher chance to collect other
particles. Willis and Heymsfield (1989) have claimed that
the radar reflectivity maximum (bright band) is due to
these relatively few, very large aggregates that survive
to warmer temperatures. Before reaching the bottom of the
melting layer, either the collision-breakup or the
spontaneous breakup process causes a rapid disappearance
of these large, wet particles, since the breakup
efficiency increases with particle size.
Evolution of the size spectra (Fig. 6) also reveals
the physical processes of aggregation, melting and
breakup. From -1.7°C (Fig. 6a) to 0.2°C (Fig. 6b),
snowflakes aggregate to form a few particles larger than
0.7cm diameter at the expense of smaller particles and
therefore the total particle concentration. From 0.2°C
(Fig. 6b) to 1.4°C (Fig. 6c), both the melting and breakup
processes help to eliminate larger particles, while the
increase in fallspeed caused by the transition from ice to
water phase results in a significant decrease in total
particle concentration.
3 r THE WARM-RAIN REGION
In this section, our focus is on analyzing the
evolution of spectra along the vertical direction below
the melting layer. The liquid water content (LWC)
measured by the Johnson-Williams device shows a constant

307
low value of 0.4g-irT . The vertical profiles of
temperature and dew-point temperature indicate a saturated
environment. Vertical shear of the wind exists (Fig. 2)
in the descending flow. Hence, mixing of raindrops
through either localized turbulence or horizontal
advection cannot be ignored. However, since the aircraft
has taken only 19min to descend from 5.5km to 0.35km, the
effect of horizontal advection is limited to microscale.
The profile of the raindrop median diamter (Fig* 7)
shows a considerable variation of raindrop size
distribution vertically caused by turbulence and
advection. After averaging every 200m datasets, we
observe a general increaseing trend of the median diameter
by almost a factor of 2-3 from the bottom of the melting
layer (4.5km) downward to about 0.35km level. Both the
2D-C and 2D-P probes observe a similar increasing trend in
both the transitional and stratiform regions. This
phenomenon suggests that the collision-coalescence process
must be dominant in generating larger droplets and hence
to increase the median diameter.
The evolution of droplets in different size ranges
(Figs. 4 and 5) along the vertical direction confirms the
efficient generation of medium (diameter 800~2500Mm for
2D-C and 800-2000/im for 2D-P) and large (diameter
2500-5000/im for 2D-C and 2000-10000/im for 2D-P) particles
at the expense of smaller (diameter 300-800Mm for 2D-C and
400-800Mm for 2DH) and smallest (diameter ISO-SOO^m for
2D-C) particles through the collision-coalescence process.
Still, it is believed that the competition between the
coalescence and breakup processes will eventually lead to
an equilibrium state. Lo and Liu (1989) have studied this
subject through an analytic approach.
Recent theoretical and observational studies on the
equilibrium state of warm-rain process propose that there
are three peaks in the equilibrium raindrop spectrum (List
et al., 1987; Willis and Tattelman, 1989 etc.).The
increasing trend of the medium particles (Figs. 4 and 5)
below the melting layer does support the appearance of the
third peak near 2000/im. In Fig.8, a few raindrop spectra
are plotted, which show that peaks near 0.03, 0.08 and
0.18cm as proposed by List et al. (1987) can be observed
below 1.5km.
4 . SUMMARY AND CONCLUSION
In the melting layer, the significant increase of
particle terminal velocity from melting snow to liquid
droplet results in a sharp decrease of particle number
density by an order of magnitude (Fig. 3). Meanwhile,

308
snowflakes aggregation generates larger particles above
and in the 0 C isothermal layer. Before reaching the
bottom of the melting layer, either the collision-breakup
or the spontaneous breakup process causes a swift
disappearance of these large, wet particles, since the
breakup efficiency increases with particle size.
In the warm-rain region, the collision-coalescence
process is dominant in generating larger particles. Such a
phenomenon is distinguishable in the vertical profile of
median diameter (Fig. 7) after averaging every 200m data
to smooth out the effect of localized turbulence and
horizontal advection. Meanwhile, a few raindrop spectra
below 1.5km (Fig. 8) stress the existence of three peaks
as those proposed by List et al. (1987), etc. for an
equilibrium spectrum maintained through the binary
interactions of coalescence and breakup.
ACKNOWLEDGEMENT
All the analyses done in this paper are executed on
the Micro Vax-II computer located in the Dept. of
Atmospheric Sciences, National Taiwan University through
the support of National Science Council Grant
NSC77-0202-M002-22. The assistance of Miss S. C. Lee is
appreciated.
REFERENCE
Heymsfield, A. J. , and D. Baumgardner,1985:
workshop on processing 2D probe dat&.
Meteor. Soc., 66, 437-440.
Summary of a
Bull. Amer.
, and J. L. Parrish, 1978: A computational technique
for increasing the effective sampling volume of the
PMS two-dimensional particle size spectrometer. J.
Appl. Meteor.,17, 1566-1572,
Johnson, R. H., and G. C. Young, 1983: Heat and moisture
budgets of tropical mesoscale anvil clouds. J.
Atmos. Sci. p . 40, 2138-2147.
List, R., N. R. Donaldson and R. E. Stewart, 1987:
Temporal evolution of drop spectra to collisional
equilibrium in steady and pulsating rain* J. Atmos.,.
Sci., 44, 362-372.
Lo, K. K., and C. M. Liu, 1989: The Microphysics of a
Mei-Yu case: Theoretical study. International
Conference on East Asia and Western Pacific:
Meteorology and Climate. July 6-8, 1989. Hong Kong.

309
Parrish, J. ^ L. , and A. J. Heymsfield, 1987: An
interactive system for processing and editing
PMS 2-dimensional imaging probe data using a
touch screen terminal. NCAR Tech. Note.
Stewart, R. E., J. D. Marwitz, J. C. Pace, and R. E.
Carbone, 1984: Characteristics through the melting
layer of stratiform clouds. J. Atmos.jSci. r 41,
3127-3133.
Willis, P. T.^ and A. J. Heymsfield, 1989: Structure of
the melting layer in mesoscale convective system:
Stratiform precipitation. (will appear in J. Atmos.
Sci. )
-, and P. Tattelman, 1989: Drop-size distributions
associated with intense rainfall. J. Atmos. Sci. r
28, 3-15.
6.
x o
Iv
2 1
TRflCK
Fig. 1: The flight track of WP-3D
aircraft on June 16, 1987 14-20 UTC
during TAMEX lOP-lOa. Two box sounding
tracks are marked by track- 1 and 2,
respectively.
Fig. .2: The vertical profile of wind
field observed along tracks 1 and 2.

310
(a) TRP.CK - 2 (1652 0 - 1711 0) (a) PR08E C TRflCK - 2 (1652 0 - 1711 OJ
PR0BE C PR0BE P PR0BE OP
(ISO - SOOO U«l (100 - 10000 UM1 USQ - IOGOO UM)
SMRLLEST SMRLLER MEDIUK LflRGE
( ISO- 300 UH1 ( 300- 800 UMJ t 800- 2SOO l .W 12SOO- SOOO U
o
UJ
o f o o o S o S o S S o
PflRTICLE NUMBER DENSITY (#/L)
(b) TRflCK - 1 (HIS 0 - H34 01
PR08E C PR08E P PR08E OP
USD - SOOO UM HOO - 10C300 UH1 (ISC - 10000 UHI
8 ° 5 g ° S 8 0 S
PERCENTflGE
PR08E C TRRCK - 1 (HIS 0 - H31 03
SMRLLEST SMflLLER MEDIUM LflRGE
( ISO- 300 UM1 I 300- BOO UM> ( BOO- 2500 UHJ {2500- SOOO UM)
r~*
O
£ ^
s;
•5 1-1
o
S ~
z:
o
UJ X
S | S g S | 3 S S S S S
PRRTICLE NUMBER DENSITY
PERCENTflGE
Fig. 3: The profile of particle number
density (I/L) measured by probe 2D-C,
2D-P and C4-P, along track (a) 2 and (b)
1. Dot points are every 10s samples. The
solid line is every 200m averaged data.
Fig, 4: The profile of the percentage
of particles in different size range
measured by 20-C along track (a) 2 and
(b) i. Dot points are every 10s
samples. The solid line is every 200m
averaged data.

31
1987 6/16 165310 T= -1.7 C Z = 5.3' h
(a.) PR0BE P TRflCK - 2 C1652 0 - 1711 0)
SMflLLER MEDIUM LflRGE-
( «00- BOO UHI t 80C- 2000 UHJ 12000- 1000OWI
PR0BE
C
X o
(a)
0.0 0.2 0.1 0.6 0.8 1.0
DlftMETER (CM)
1987 S/16 ISS1SO T= 0.2 p I - 1.71 KM
PERCENTRGE
PRBBE
hrm.
C
(b) PR0BE P TRflCK - i I HIS 0 - 1*34 0)
SMRLLER MEDIUM LflRGE
(100-800 (BOO- 2000 t2CCO- 10000UM!
UM) UHJ
0.0 0.2 0.1 0.6 0.8 1.0
(b)
DIflMElER (CM)
1987 6/16 165530 fs 1.1 C 2 = 1.55 KM
.£-01
,c-oi
.e ns0
.e-oo
.C-M
b-n '•'•'• I
PRB8E P '
• 1 -O.JJ •
; * »•»
PERCENTflGE
.1 DSfl
.£•00
.C-01
f f.M
Nu_
^^"^T 1 !
PRBBt C*f '
rg:U
0.0 0.2 0.< 0.6 0.8 1.0
(c) DlflMETER (CM)
Fig. 5:
probe.
Same as Fig. 4, except for 2D-P
Fig. 6t The particle spectra obtained
at (a) -1.7°C, (b) o.2°C/ and {c) 1.4°C
levels along track 2. The spectra is
empirically fitted to an exponential
function N e" A (solid curve) with the
correlation coefficient being
written along the index "2 W , and.^a
three-parameter function N d° e~
(dashed curve) with the Sorrelation
index being written along the index W 3".

312
1987 6/16 17QSSO U 0.9 C 2 = 1.12 KM
PR0BE C»P
(a) ~ 2 a652 I7n OJ
° -
PR08E C PR08E P PR3BE OP
(iso - sooa UNI (too - loaoa UNI nsa - 10000 uti
X
o
(a
O.J 0,2 0.3
DifiHETER
(CM)
1987 6/16 $70710 T= 22.9 C 2 = 0.97 «H
PR08E C'P
MEDIflN DlflMETER
(UM)
(b)
TRflCK -
\ (1415 0 - 1*3* OJ
PR08E C PR0BE P PR08E OP
tlSO - 5000 UHI f«0 - lOQOQ UW USD - 10000 U«J
OJ 0.2 0.3
OlfiHETER (CM
l.E+02
I.E'QJ
x
X
Z l.E+00
O
1987 6/16 m03Q Ts 2S. I C 2 = 0.38 KM
PR88C OP
2 -0.94 -
3 0.91 •
§ 1.E-Q2
1 I-E-03
I I 11 .S. i § §
2 I.£-01
MEDIflN D1RMETER CUM)
0.0 OJ 0.2 0.3 0.1
(c) DIflMETER CCM)
Fig. 7: Same as Fig. 3 f except for the
median diameter (pin).
Fig.' 8: same as Fig. 6, but at (a)
2o!9°C, (b) 22.9°^ and «c> 25.1°C
levels.

ON DYNAMICAL STUDIES OF ORO GRAPHIC ALLY INDUCED
MESOSCALE PHENOMENA
313
SIU-SHUNG HONG CHUNG-YING Hu FU-SHAN WENG
Department of Atmospheric Physics
National Central University
Chungli
ABSTRACT
We show in this article that (l)heavy precipitation on the
upwind side of a mountain is possibly induced by the upstream
retardation of airflow upon approaching the terrain rise; (2)the
severe surface gust occurring over Hengchun Peninsula, the socalled
'fall wind', is a manifestation of multiple-reflection of gravity
waves between the critical level and the ground.
1. Introduction.
Taiwan is an island with complex terrain. The island is over 380 km in the
north-south direction, and near 140 km in the east- west direction. The Central
Mountain Ranges, which run approximately in the middle of the island and divide
the island into the east and west parts, rise up to an elevation of over 3000 m
within 50 km in most parts. In addition, the island is surround by oceans, with
the Pacific Ocean in the east, the South China Sea in the south, the Taiwan Strait
in the west, and the East China Sea in the north. The Kuroshio Purrent lies next
to the island in the east.
Weather systems passing over Taiwan are subject to significant modifications,
with a characteristic length scale of hundreds of kilometers, which is of mesoscale
in nature. With these complicated , unique backgrounds, the orographically
induced phenomena also bear special characteristics.
In this paper, we will focus on two main subjects, namely, the upstream heavy
rainfall and the fall wind phenomena. Both phenomena will be introduced and a
theory for each will be provided.

314
2. Heavy Rainfall Upstream of a Mountain.
2.1 Introduction. That precipitation systems intensify during the process of
approaching the Taiwan Island is quite a common feature and in many events lead
to heavy rain and even flash flood. The satellite pictures of one of the typical
examples is shown in Figure 1.
This event occurred in the period of May 30 through May 31, 1983, also in the
so-called Mei-Yu season. It caused flash flood in the Chungli-Hsinchu area. The
pictures clearly show that the convective systems intensified when they approached
the land, and diminished upon moving over the land.
This research studies the possible mechanisms which may be responsible for
the enhancement of convection on the upstream side of a mountain. The proposed
mechanism is that upstream retardation of airflow should generate uplifting of air
before reaching the mountain slope. This uplifting will further induce persistent
convection on or before reaching the slope when certain conditions are met. The
conceptual diagram shown in Fig. 2 demonstrates the idea.
The answer to what condition (or conditions) could be responsible is not easy
to deduce. The effect of upstream retardation is inversely proportional to the
Froude Number, and is therefore eminent in a stable atmosphere; but convection
would be suppressed by the vertical stability. An atmosphere with conditional
instability favors convection to occur and be sustained; but upstream influence
would be minimum. Then, is there some combination that might constitute the
condition we are looking for Finally, we found that in an atmosphere with a
stable lower part and at the same time having a conditionally unstable situation
just above the lifting condensation level, upstream heavy precipitation is indeed
possible.
Since the problem involves changing phases of the water substances, in addition
to being non-linear, we employed numerical dynamic model simulation for
our purposes. The model will be briefly described in the next sub-section, and the
results will be presented in sub-section 2.3.
2.2 The model. The dynamic model is basically a two-dimensional, primitive
equation one, in flux form and

315
lateral boundaries. The model includes dry convective adjustment scheme as proposed
by Klemp and Lilly 4 } . The moist treatment is done by a diagnostic cloud
model with Kessler's cloud physics^ . For the boundary layer, when needed, the
parameterization of Zhang and Anthes 5^ is adopted.
The model had been tested extensively. It worked very well in simulating,
mountain waves (linear or non-linear), land-sea breeze circulation, as well as fall
(downslope) wind storm, etc.
2.3 Results and discussions. In order to investigate how stability affects the
result, we divide the basic atmosphere into three layers. Two of the combinations
are shown in Figures 3 and 4. The stable layer is labeled with 'S', whereas the
conditionally unstable layer is labeled with 'IT. For example, Fig. 3 is the result
for an atmosphere with stable upper layer, stable middle layer, and conditionally
unstable lower layer (abbr. as SSU). The mean wind is 4-10 m/s. Thus the left
hand side of the mountain is the upstream side. The mountain bears the Gaussian
profile with maximum height of 1000 m, and a half width of about 30 km. For
this case, the convection is strong and has a maximum located near the top of
the mountain as expected. Fig. 4 illustrates the result for SUS atmosphere. Now
there is a second maximum in rain-water content (denoted as QR) occurring at
the upstream foothill. The main maximum occurs over the top of the mountain
though. At this point, we should mention that the base of the conditionally
unstable layer lies at 3 km above the mean sea level. The shape of the mountain is
an important factor but still is not able to change the rainfall pattern (figure not
shown). It is also interesting to note that the contrast in surface roughness between
the land and the sea is not important. Actually the effects of the boundary layer
axe not important in this problem.
In order to demonstrate further that the condition as shown in Fig. 4 is realistic,
we choose the case which occurred on 31 May 1983 as mentioned previously for
a particular numerical simulation. Fig. 5 shows the total rainfall amount during
the day. After comparing with Taiwan's topography as shown in Fig. 6, one can
easily observe that the maximum rainfall occurred in front of the terrain rise, i.e.,
on the upstream side. The two-dimensional elevation of Taiwan's northern part is
shown in Fig. 7. Fig. 8 shows the sounding taken at Makung, Special at tension
should be paid to the inversion layer at about 1 km.
The sounding data are used as the initial condition. For the dry case, reverse
flow in the lower layer of the upstream side of the mountain is clearly generated
(figure not shown). When the moisture is put into the model, the results are
shown in Fig. 9. Now the upstream convective cells are indeed generated. The

316
accumulated rainfall amount on the ground after five hours integration is shown
in Fig. 7, the upper solid curve. A comparison with the observations shows very
good agreement.
2.4 Conclusion. We have demonstrated above that upstream heavy rainfall is
quite possibly induced by uplifting due to upsteam retardation of airflow. The most
favorable conditions for upstream intensification or induction of strong convection
appear to be: (1) weak surface wind; (2) the atmosphere has a stable lower layer
and conditionally unstable layer aloft;(3) high mountain with large aspect ratio
especially on the windward side.
3. The Fall Wind over Hengchun Peninsula.
3.1 Introduction. The so-called 'fall wind' is a peculiar phenomenon which occurs
every year from October till April of the next year over the western coastal
area of the Hengchun Penisula (the southern tip of Taiwan). The phenomenon is
characterized by a strong, gusty wind storm. The surface gust can easily reach 25
m/s from calm condition in a couple of hours. One of the typical cases occurred
during the period from November 15 till 17, 1986. In the following we will use it
for evaluating the proposed theory.
The surface gusts recorded at Hengchun station were surveyed for the last ten
years. According to the record, we define the fall wind event to be the one having
maximum surface gust of over 15 m/s. Table 1 shows the averaged number of
events for each month from 1981 to 1986. One can see that the October-December
is the pre-dominant period. There was at least one event every two days.
Table 1. Number of cases of fall wind event
at Hengchun in 1981-1986 period
Month
No. of
cases
JAN
78
FEB
68
MAR
45
APR
.21
MAY
18
JUN
6
JUL
1
AUG
2
SEP
5
OCT. NQV
57 92
DEC
93
Averaged
No. of
cases/yr
"13.0
11.3
7.5
3.5
3.0
1,0
0.1
0.3
0.8
9,5 1.5.3
15,5
For most cases there appeared no distinct mechanism which can cause the
surface gusts (weather maps not shown here), except that only a few cases were
due to typhoon but occurred very rarely. However, when we examine the upper

317
air sounding, over Panchiao (although a little too far from Hengchun Peninsula,
yet its sounding is the most reliable and is typical for regions not seriously affected
by the high mountains) on 17 November 1986 for example, we can see that there
existed a critical level with small Richardson number at about 2.5 km from the
mean sea level (Fig. 10). This is actualy a quite general situation over Taiwan in
the September-April period.
Therefore, multiple reflection between the ground and the critical level must
play an important role in the fall wind event 2 !, and the height at which the critical
level is located should be one of the determinating factors. Again, since the
problem is non-linear, we use a numerical model for the purpose of dynamical
study.
3.2 The model and results. The model used here was designed for smaller
scales than the previous one. It is capable of switching to the quasi-hydrostatic
approximation. It is a primitive equation model in terrain-following coordinate.
For the background atmosphere, Panchiao's upper-air sounding data were
examined (Fig. 10), and analytical functions for 0 and U were written as functions
of height to simulate the sounding at OOZ 17 November 1986. The reason for
doing so is that in the experiments that follow the height of the critical level is
the sole controling factor, all other conditions are exactly the same throughout
the atmosphere. The results are shown as dash-dot curves in Fig. 10. In the
Figure the real sounding data are represented by solid curves. The dashed and
the dash-double-dot curves represent the Richardson number calculated from the
real sounding and the analytical functions, respectively.
According to the linear analysis, the analytical solution without considering
the critical level is
w zz + M 2 w = 0 (1)
where
7172 J7
M 2 = / 2 -fc 2 = ^-^-A; 2 (2)
For the Hengchun Peninsula, L x » 36 km, k = 1.745 x W~ 4 m~ l . If U=
-10 m/s, P > k 2 , so that M 2 > 0, wave is the solution with vertical wavelength
L z « 4.3 km. This value is the key in determining why the surface wind is strongest
when the critical level is situated at about 2.5 km.
Numerical experiments are conducted for exactly the same conditions except
that the location of the critical level varies. Table 2 summarizes the cases which
were executed. To be concise, only the results of cases 03 and 00 are to be shown.

318
Figures 11 and 12 demonstrate the resulting surface wind speed and the Reynolds
stresses for these two cases. One can see that when the critical level is located at
2 ~ 2.5 km, the surface wind speed is enhanced very much because the Reynolds
stress grows consistently showing coherent interference through multiple reflections.
Table 2. The tested cases
Case
03
02
01
00
-1
-2
-3
Zo
(km)
4 0
3,3
3 0
2 47
2.2
2 . 0
• 5
Finally, numerical simulations are conducted for the Nov. 17 case, using the
corresponding Panchiao sounding data as the initial condition. The results are
shown in Fig. 13. The fact that strong surface wind is being set up together with
the consistently growing Reynolds stress is clearly shown.
3.3 Concluding Remarks* The Hengchun fall wind is the result of multiple
reflection of the mountain wave between the critical level and the ground. The
favorable conditions are: (l)The critical level is located at about 2 to 3 km with
local Richardson number smaller than or near 1; (2) Easterly-wise low level wind
with sufficient strength; (3)Stable lower atmosphere.
Acknowledgment. Researches related to this article were supported by various
NSC grants, including NSC73-Q202-MOQ8-09, NSC77-0202- M008-14, NSC78-0202-
M008-18. Special thanks are extended to Ms. C.-H. Shiax> for assistance in various
stages during the preparation of this paper.
REFERENCES
R., Frequency filter for time integrations. Mon. Wea. Rev., 100, 487-490
(1972),
^Breeding, R.J., A nonlinear investigation of critical levels for internal atmosppheric
gravity waves. J. Fluid Mech., 50, 545-563 (1971).
^Kessler E., On the distribution and continuity of water substance in atmospheric
circulations. Met. Monogr., No. 32, 1-84 (1969).
^Klemp J.B. and Lilly D.K., Numerical simulation of hydrostatic mountain waves.
J. Atmos. ScL, 35, 78-107 (1978).
5 1 Zhang D. and Anthes R.A., A high resolution model of the planetary boundary
layer - sesitivity tests and comparison with SESAME - 79 data. J. Appl
Met., 21, 1594-1609 (1982).

319
Fig.l Satellite imagery for May 30 -31
1983. Time sequences from left to
right are respectively 1229Z, 1620Z,
2121Z, 0006Z, 0306Z, and 0606Z.
Fig.2 Conceptual diagram for the
proposed mechanism for upstream
heavy precipitation.
DEGREE K
-ISO • -50
X-AXIS ' KM )
150 -50 50
X-AX!S ( KM 1
W ( CM/S )
RH ( 7. )
-150 -50 50
X-AXIS I KM )
Fig.3 The results for moist SSU atmosphere.
a)#, b)vertical velocity w, c)perturbed horizontal velocity,
d)rain-water content, e)cloud-water content, f)relative humidity.
The regions enclosed by dashed curves are negative values.

320
DEGREE K
IT ( M/S )
QC (G/KG)
-so so
X-AXIS ( KM )
-ISO "50
X-AXIS
Fig.4 Same as in Fig. 3, except for moist SUS atmosphere.
Fig,5 Observed accumulated rainfall
amount on 31 May 1983.
Fig. 6 Taiwan's topography. The
contour lines shown are 100 m,
200 m, 2000 m and 3500 m.

321
-ISO -50
X-AXIS C KM
Fig,7 The two dimensional
enveloped orography of the
northern Taiwan and the
accumulated rainfall amount
obtained from simulation
(solid curve), and from observation
(denoted by a, b, ... ., etc.)
/ I983/OS/31/OOZ
STATION MAKUNG
Fig.8 Sounding over Makung at OOZ, 31
May 1983.
DEGREE K IT ( M/S )
-ISO -50
X-AXIS ( KM )
W ( CM/S )
OR (G/KG)
-130 -50 50
X-AXIS ( KM )
-ISO -50 50
X-AXIS C KM )
Fig. 9 The simulated results for May 31 after
3 hrs integration, the moist case.

322
270 290 310 330 350 370
POTJEMP CKj. .
-40 -32 -24 -16 -8 0 8 16 24 32 40
CROSS HIND CM/SJ
Fig.10 Comparisons between the
analytical functions and the
real sounding over Panchiao,
at OOZ 17 Nov. 1986. See text.
Fig. 11 Surface wind speed and Reynolds
stress calculated from experiment 03
after 4 hours integration.
T-rr-1 ri "Till i i f > i i i i i i i i
10 20 30 40 50 60 70 80 $0 JOO UO 120
WEST EAST CKMJ
i i vTrr i ri 1 n 1 i i i i 'i '." rrr i
10 20 30 40 50 SO 7Q 80 90 !00 110 120
WEST EAST CKM)
1 2 3 4
Rf-TNOLDS STRESS C-0.1)
Fig. 12 Same as Fig. 11 ,• except for
case 00.
Fig, 13 Simulated results for the case
of 17 Nov. 1986.

323
Numerical Simulations of topographical effects on airflow and precipitation
Su-Tzai Soong, Mukut Mathur 1 and Wei-Kuo Tao 2
University of California, Davis, CA 95616
* National Meteorological Center, Washington, DC 20233
2 Goddard Space Flight Center, Greenbelt, MD 20771
ABSTRACT
A primitive equation model was used to simulate the topographic and diurnal
heating/cooling effects on the airflow pattern and precipitation during
undisturbed conditions over Taiwan. The model uses the s coordinate and
includes the soil and boundary layer physics. The simulation without
heating/cooling reached a quasi-steady state in 12 hours. The mountain blocking
effect is clearly evident and there is almost no cross mountain flow except in the
southern most part of Taiwan, where the mountain range is lower. The
simulation with diurnal heating/cooling generated a distinct cyclonic circulation
over southeast Taiwan where the heating in the afternoon was a maximum. This
airflow pattern created an area of low level convergence and precipitation
extending from the west side of the mountain in southern Taiwan, across the
central mountain range, and to the northeast coast of Taiwan
1. INTRODUCTION
One of the major objectives of Taiwan Area Mesoscale Experiment (TAMEX) is to
study the effect of topography on airflow and precipitation during undisturbed conditions.
During the period of June 19-21, 1987, no frontal passage was anticipated and the
Intensive Observing Period (IOP) No. 11 of TAMEX was carried out specifically for this
purpose. Isolated thunderstorms were predicted for all three days, especially in the late
afternoons due to sea breeze effect. The NOAA P-3 aircraft also flew a topographic
mission around the Taiwan island on June 20, 1987. No deep clouds were observed by
satellite at 0800 LST on that morning. By 2000 LST, deep clouds from the southwest to
the northeast of the island were observed by satellite. Precipitation was reported over the
northeast part of Taiwan, where Taipei recorded 27 mm and I-Lan 19 mm (Wu and Chen,
1987). Precipitation was also reported over the Central Mountain Ranges.

324
A primitive equation model was used in this study to simulate the topographic and
diurnal heating/cooling effects on the airflow pattern and precipitation. The simulation
started at 0500 LSI on June 20, 1987 and ended at 0500 LST on June 21, 1987,
completing a 24-hour cycle. Section 2 gives a brief description of the model Section 3
describes the large-scale and initial conditions. The results and discussions are in section
4 and the conclusions and future work in section 5.
2. THE MODEL
The model framework is based on the Quasi-Lagrangian Nested Grid Model
(QNGM) developed by Mathur (1983). The quasi-Lagrangian scheme used in this model
evaluates the horizontal advections by tracing back the original position of a grid point
using a nine point interpolation scheme. This scheme is second order in nature but it also
considers the acceleration on the path during advection to increase the computation
accuracy (Mathur, 1983).
The model uses the 0 coordinate in the vertical For this study, 22 vertical layers
are used with an interval of Aa = 0.05 from a= 0 to 0.95. A better resolution is used in
the lowest 3 layers where Aa= 0.01 for the lowest layer and ACT = 0.02 for the next two
layers. There are 31 by 31 horizontal grid points and the horizontal grid spacing is 20
km. The model domain and the topography are shown in Fig. 1.
r / / / / /• / / / / // >>>>>>>> sv '/ / s */ '/ '/ V-'/.
/ /• //////////// /'X + * , ,\, t
//////////// /''/ /
////////////// ,
* /'
xl
•//•/// y A.
, -1 / //////.
. / i / ////// , t V f / /;///// A
//;/// ///////.
_/V/ /////////A
' //.//,//•
Ff^. J T/ze wotfe/ domain and the
topography of Taiwan. The contour
interval zs 500 m.
Fi^. 2 r/ze .srea^y state airflow in the
lowest layer after 12 hours simulation.
The maximwnwind vector is 95 mis.

325
The physical processes of the original model include the sensible and latent heat
fluxes from the ocean surface, dry convective adjustment, large-scale condensation and a
Kuo-type cumulus parameterization. Since one of the objectives of this study is to
simulate the diurnal variation of airflow and precipitation, we made extensive modification
of the boundary layer processes. The current version of the model uses a soil layer to
determine the soil surface temperature. The physical parameters included in the
computation are the albedo, the downward long wave radiation from the atmosphere and
outgoing long wave radiation from the earth's surface, the sensible heat flux and the latent
heat flux. It is assumed that the top 40 cm of the soil layer will be affected by these
physical processes. The lack of a coupling hydrological model at this moment forced us
to choose an arbitrary Bowen ratio of 1 to partition the amount of sensible and latent heat
fluxes.
Vertical turbulent fluxes are computed using turbulent exchange coefficients based
on a K-theory. The values of the turbulent exchange coefficients are evaluated based on
the Richardson number using the empirical formulas developed by Blackadar (1979). The
first layer is treated as the surface layer and the fluxes are calculated by assuming that the
lowest layer satisfies the Monin-Obukhov similarity theory (Lattau, 1979).
The Long wave radiative cooling plays an important role in the development of the
nocturnal boundary layer. The computation of radiative processes is usually very
complicated. In consideration of computational economy, the evaluation of the cooling
rate due to long wave radiative flux divergence employs the approximations of Andre, et
al. (1978).
3. THE LARGE-SCALE AND INITIAL CONDITIONS
Since one of the objectives of this work is to study the topographic effect on airflow
over a meso-p scale mountain, it is necessary to use a representative large-scale condition
not contaminated by the topography as the initial condition. The prevailing wind on June
20,1987 over Taiwan area was from either south or southwest The observed wind field
showed an around mountain flow pattern and it had undoubtedly been affected by the
island topography and might not represent the large-scale condition. After inspecting all
the rawinsonde data for 0800 LST, June 20, 1987 in conjunction with the surface and
upper air maps at that time, we decided to use the 0500 LST sounding at Panchiao to
initialize the model. This selection was based on the wind direction at Panchiao that
matched most closely to the geostrophic wind direction on all levels below 500 mb. The
sounding at 0500 LST instead of 0800 LST was selected because the 0800 LST sounding

326
had already been affected by solar heating and would not have been suitable to study the
diurnal heating/cooling effect on airflow and precipitation over a 24 hour period.
4. RESULTS AND DISCUSSION
The simulation was carried out in two parts. In the first part, we studied only the
topographic effect on the airflow without surface and radiative heating or precipitation. In
the second part of the simulation, we studied the diurnal heating effects on airflow,
temperature and precipitation.
4.1 The topographic effect on airflow
In this part of the simulation, the initial condition consisted of a horizontally
uniform wind field. The observed temperature and mixing ratio profile at Panchiao at
0500 LST on June 20, 1987 were combined into one single variable of virtual potential
temperature and it is used as the only thermodynamic variable. This value was applied at
the center of the domain and the virtual potential temperature at other grid points were
computed to satisfy the thermal wind relationship. A 12-hour integration was made and
the wind and temperature fields reached a quasi-steady state. During the last 3 hours of
the simulation, the maximum change in wind speed in the lowest model layer (~ 50 m) is
only 0.6 m s , in comparison, the 12 hour accumulated maximum change in wind speed
is 10.7 ms . Fig. 2 shows the airflow pattern in the lowest model layer around Taiwan
at the end of the 12 hour simulation. The mountain blocking effect is clearly evident and
there is almost no cross mountain flow except at the southern most part of Taiwan.
Comparing these results with the surface weather map of 0800 LST of June 20 (p. 297,
Cunning, 1988), several well defined features are realistically simulated. One is the
strong southwesterly wind to the southeast of Taiwan, where the simulated wind speed is
close to 10 ms~*> which matches the observed wind speed at Lanyu. The other is the
weaker southerly wind over western Taiwan, where the simulated and observed wind
speed are both about 5 ms . Over the northeast of Taiwan area, there is a line of
convergence.
The topographical effect on the wind and potential temperature change can be seen
more clearly in a east-west cross section over southern Taiwan in Figs. 3 and 4. The
isentropes bend upward, indicating a cooling, over the windward side and bend
downward, indicating a warming, over the lee side of the mountain. In response to the
temperature change, there is a pressure increase over the wind ward side of 2 mb and a
pressure decrease of 3.5 mb over the lee side of the mountain.

327
Fig. 3 The east-west cross section of the
steady state airflow at grid 11 from south.
The maximum wind vector is 15 mis.
4.2 The diurnal heating/cooling effect
p. 4 Sam£ as Fi 3 butfor potentiai
temperature subtracted by 300K. The
comour intefyal is 2K
In this part of the simulation, the results of part one were used as the initial
condition after separating the virtual potential temperature into the potential temperature
and the mixing ratio. This was done by assuming that there was no horizontal gradient in
relative humidity on a pressure surface and the relative humidity profile at Panchiao was
adopted over the entire domain of the simulation. Sunrise was set at 0600 LST and sunset
was set at 1800 LST. The intensity of solar radiation was prescribed by a sinusoidal
curve with a maximum intensity of 950 w m~^ at noon.
From 0500 LST to 0800 LST, the simulated change of temperature at the lowest
model level was less than IK. This is due to a combination of cooling before sunrise and
a weak wanning after sunrise. There is a rapid warming between 0800-1100 LST and the
temperature difference between 1100 LST and 0500 LST (Fig. 5) shows that the
maximum warming occurs at the east and southeast coast of Taiwan with a maximum of
6.7 K near Taitung. There is also significant warming over the northwest coast The
warming over the mountain top areas are quite small and the warming over the northeast
and southwest are less than 4K. The warming continued during the next three hours and
at 1400 LST the temperature at Taitung was 9.4K warmer than at 0500 LST (Fig. 6).
Another location with large warming is near I-Lan where the temperature at 1400 LST is
7.4K warmer than the early morning. Over the north and west coastal area, the prevailing
wind keeps the warming in the range of 3-5 K. The temperature changed little between

328
I
.011
I 1...J/1 l\l l-L-l L I L I 1 /I 1 1 I I I I 1 I t i l I
Fig. 5 The potential temperature Fig. 6 Same as Fig. 5 but for the period
difference between 1100 and 0500 LST, between 1400 LST and 0500 LST, June
June 20, in the lowest layer. The contour 20.
interval is 2K.
, VsNiV
\N -
, \ \ \ \
r y\ \ \
hv
Y
StQJM' ' '. /'»-..:* .-' J ' r.A-.,.-.-H
Fz^. 8 T/ie change of wind flow between
Fig. 7 Same as Fig. 5 but for the period 1400 LST and 0500 LST, June 20, in the
between 2000 LST and 0500 LST, June lowest layer. The maximum wind vector
20, is.72 mis.
1400 and 1700 LST and a cooling trend started after 1700 LST. By 2000 LST (Fig. 7),
the temperature at Taitung is still 6.5K warmer than at 0500 LST but significant cooling
occurred over the southwest coastal area where the temperature at that time is L5K cooler
than the early morning. This cooling can be explained by a combination of the advection
of cooler maritime air from the ocean by the prevailing wind and the radiative cooling. By

329
0500 LST the next morning, the temperature in the entire north, west and south coast area
has dropped to 2-3K cooler than it started 24 hours ago.
Sea breeze started at 1100 LST and lasted until 2000 LST. Fig. 8 shows the
airflow pattern associated with the sea breeze at 1400 LST. A strong sea breeze occurred
over the entire west coast and the central part of the east coast of Taiwan. The sea-land
circulation on the north and south coastal areas is greatly skewed by the terrain and the
mean wind. Moderate precipitation (Fig. 9) occurred at this time over the western slope of
the mountain in the south and over the eastern slope of the mountain in the north-central
areas. During the next 6 hours, the sea breeze pattern was modified by the Coriolis force
and both the intensity and the area coverage of precipitation have increased. Fig. 10
shows the sea breeze pattern at 2000 LST, 2 hours after sunset. A distinct cyclonic
circulation pattern was generated over southeast Taiwan where the heating was a
maximum. This airflow pattern created an area of low level convergence and precipitation
extending from the west side of the mountain in southern Taiwan, across the central
mountain range, and to the northeast coast of Taiwan (Fig. 11) This precipitation pattern
matches the satellite cloud distribution at the same time (Fig. 12).
Land breeze first started at 2300 LST over the southwestern Taiwan near
Kaohsiung and gradually spread over other parts of the coastal areas during the next 6
hours. The wind speed of the land breeze is about half as large as the sea breeze. As the
land breeze started, the precipitation intensity weakened gradually.
i r t v i i i r 1
H
.00
.827 ''
Fig. 9 The 3-hour accumulated
precipitation ending at 1400 LST, June
20. The contour is 05 cm.
Fig. 10 Same as Fig. 8 except for the
period between 2000 LST and 0500 LST,
June 20. The maximum wind vector is
6.3 mis.

330
«'..
Fig. 11 Same as Fig. 9 except for 2000 Fig. 12 The IR satellite picture at 2000
LSI, June 20. The contour interval is LST, June 20.
0.5 cm.
5. CONCLUSIONS AND FUTURE WORK
A numerical model was used to study the topographic and diurnal heating/cooling
effects on airflow and precipitation. The topography of Taiwan creates a strong blocking
effect during undisturbed conditions. No airflow is able to cross the Central Mountain
Range except over the area of lower mountains in southern Taiwan. During the period of
prevailing southwesterly flow, the blocking effect produces a strong southwesterly flow
to the southeast of Taiwan and a weak southerly flow over the west coastal area. The sea
breeze is strong over the entire west coast and over the central part of the east coast but
relatively weak over the southern and northern part of Taiwan. The daytime temperature
increase ranges 3-5K over the southwest, 5-7K over the north and 7-9K over the
southeast of Taiwan. Precipitation due to the sea breeze started first over the west of t
mountain in southern Taiwan and over the east of the mountain in north-central Taiwan in
early afternoon. The precipitation gradually increased in intensity and in area of coverage
throughout the afternoon, and by 2000 LST, it covered the area from southern Taiwan
across the central mountain range to the northeast coast
The numerically simulated airflow pattern and precipitation area matched, in
general, the surface wind observation and satellite cloud coverage. The assumption of the
Bowen ratio of 1 may not be exact and may need to be improved in the future. Fairly
simple relationships between latent heat flux and soil moisture have been used in the past

331
(Walker and Rowntree, 1977). More complex couplings, involving vegetational effects,
have recently been attempted (Sellers, et aL, 1986). We feel that for a simulation of one
or two days, as in this study, the simple treatment of the soil moisture as proposed by
Walker and Rowntree (1977) should provide reasonable results without using excessive
computer time. Another future improvement of the model is in the treatment of the
precipitation processes. Lin, et al. (1983) first proposed to separate the hydrometeors into
cloud, rain, ice, snow and graupel. This microphysical representation has been used in
several cloud models (see McCumber, at al., 1987 as an example). Previously, we have
experimented with several kinds of microphysical processes in cloud models (Soong and
Chen, 1984, Tao and Soong, 1986, Tao, et al., 1987). The one including the ice phase
produced the large anvil structure, similar to those observed in deep clouds. For a study
of precipitation in a mesoscale model, a more detailed microphysical process gives at least
two advantages: (1) it provides cloud cover, which affects the amount of radiation
reaching the ground and the soil surface temperature; and (2) it lets the precipitating
particles drift in the air and fall in areas outside the areas of cloud formation.
6. ACKNOWLEDGMENT
The authors thank Mr. J.-C. Jang for writing the code of boundary layer processes.
The work of Soong is supported by NSF under Grant ATM-84-19811. The work of Tao
is supported by the NASA Headquarters Mesoscale Processes Program under Contract
2926-SS-202.
REFERENCES
Andre, J. C., G. de Moore, P. Lacarrere, G. Therry and R. du Vachat, 1978: Modeling
the 24-hour evolution of the mean and turbulent structures of the planetary boundary
layer. /. Atmos. ScL> 35, 1861-1883.
Blackadar, A. K., 1979: High-resolution models of the planetary boundary layer.
Advances in Env. Sci. & Eng., 1, 51-85.
Cunning, J. B., 1988: Taiwan Area Mesoscale Experiment: Daily operations summary,
NCAR Technical Note, NCAR/TN-305+STR, 361 pp.
Lettau, H. H., 1979: Wind and temperature profile prediction for diabatic surface layers
including inversion cases. Boundary-Layer Meteor.* 17,443-464.
Lin, Y.-L., R. D. Farley and H. D. Orville, 1983: Bulk parameterization of the snow field
in a cloud model. /. Climate Appl Meteor., 22, 1065-1092.

332
Mathiar, M. B., 1983: A quasi-Lagrangian regional model designed for operational
weather prediction. Man. Wea. Rev., Ill, 2087-2099.
McCumber, M. C, W.-K. Tao, J. Simpson, R. Penc and S.-T. Soong: Comparison of
ice-phase microphysical parameterization schemes in tropical fast-moving convective line
simulations. Preprint Volume, 17th Conference on Hurricanes and Tropical Meteorology.
April 7-10, 1987. Miami, Ha, Amer. Meteor. Soc., Boston, Mass. 117-120.
Seller, P. J., Y. Mintz, Y. C. Sud and A. Dalcher, 1986: A simple biosphere (SiB) for
use within general circulation models. /. Atmos. ScL, 43, 503-531.
Soong, S.-T., and S.-C. Chen, 1984: The effect of wind shear and ice phase on the
structure of a tropical cloud cluster. Postprints, 15th Conference on Hurricanes and
Tropical Meteorology, January 9-13, Miami, Florida, 181 -182.
Tao, W.-K., and S.-T. Soong, 1986: The study of the response of deep tropical clouds to
mesoscale processes: Three-dimensional numerical experiments. /. Atmos. ScL, 43,
2653-2676.
Tao, W.-K., J. Simpson and S.-T. Soong, 1987: Statistical properties of a cloud
ensemble: A numerical study. J. Atmos. ScL, 44, 3175-3187.
Walker, J. M. and P. R. Rowntree, 1977: The effect of soil moisture on circulation and
rainfall in a tropical model. Quart. J. R. Meteor. Soc., 103,29-46.
Wu, T.-Y.and G. T.-L Chen, 1987: Taiwan Area Mesoscale Experiment, National
Science Council, Science and Technology of Disaster Prevention Program, Technical
Report 76-19, 133pp.

Study on the Frontal Cyclone System in Southern China and the Vicinity
of Taiwan Area during Late - Winter and Early-Spring
Huo-Ming Jiang
Institute of Atmospheric Physics
National Central University
Chung-Li, Taiwan 32054
ABSTRACT
The structure of the upper level and stationary surface
frontal zones associated with a cyclone developing over southern
China on 3-5 February 1979, is examined and discussed. The
dynamics of the transverse circulation which plays important
role in the frontogenesis process is also discussed .
From the cross-section analysis of the potential vorticity
structure, it was shown that the upper sub-tropical front was
caused by the tropopause folding of high potential voricity
stratosphere into the troposphere, but the stationary surface
frontogenesis was caused by the sub-geostrophic flow whicg was
induced by the Tibetan Plateau.
1. INTRODUCTION
According to the three-cell meridional circulation model
(Palmen,1951), the hemispheric airis divided into three principal air
masses : tropical air mass, mid-latitude air mass and polar air mass. The
polar fronts are considered as the boundary between polar and mid-latitude
air masses. The contrast between these two air masses on the poleward edge
of the Ferrel cell is relatively distinct in the low and middle troposphere
and less distinct higher up. Owing to the low-level divergence in the
boundary region between the Hadley and Ferrel cells, there is no marked
air-mass boundary in the lower troposphere. A contrast between tropical
and mid-latitude air is found in the upper troposphere where there is
meridional convergence between these circulation cells.
Fig.l (after Palmen and Newton, 1969) shows the typical structure of
the northern hemisphere. The scheme is most applicable to the atmospheric
structure during the cold season when all features are more pronounced
than during the warm season. The subtropical front is indicated as a
sloping upper-tropospheric layer between the tropical and the mid-latitude
air masses.
Cyclogenesis frequently happened in the East China Sea during latewinter
and early-spring (Fig.2). According to the results of Hanson and
333

334
Long (1985), the maximum occured to the northeast of Taiwan, while the
maximum frequency in surface frontofenesis was in southern China and
north Taiwan (Yeh and Chen, 1984).
From the air-mass concept discussed above, the subtropical fronts
are upper fronts. However, surface fronts often developed in southern
China which is a subtropical region, and cyclogenesis happened in the
frontal zone. At present, I would like to describe a primary case study
showing the interesting process of surface frontogenesis and cyclogenesis
over there. The data were from the level III-B data of the First GARP (
Global Atmospheric Research Program ) Global Experiment (FGGE).
In the following section, viz. section 2, the synoptic situation and
horizontal analysis will be presented. Vertical cross-section will be
discussed in section 3. In section 4 I will discuss the dynamics of the
transverse circulation. Finally the conclusion will be given in section 5.
2. SYNOPTIC SITUATION
The synoptic situation at OOOOGMT 04 and 05 February 1979, as
analysed by the Japan Meteorological Agency, are replasented by the
surface isobars and fronts shown in Fig.3 and Fig.4 respectively. A
stationary surface front was located over the East China Sea and southern
China for a few days. Eventually, a low pressure center with cyclonic flow
appeared in the frontal zone.
The frontal zone were defined according to the structures of the
horizontal temperature gradient, static thermal stability and potential
voticity. Fig.5 shows the horizontal temperature gradient at 1000 Hpa
surface at OOOOGMT 04 February 1979. A maximum region was found
over the coast of southern China where the frontal zone was according to
the subjective analysis in Fig.3.
The potential vorticity can be written in isobaric coordinate as
P = -g(fk+Vx\Ove (1)
where kis a unit vertical vector and V is the three-dimensional gradient
operator in (x,y,p) space. Due to the strong stability and the cyclonic shear
vorticity, the potential vorticity should be stronger in the frontal zone than
in the non-frontal region. Fig. 6 shows the potential vorticity at 950 Hpa at
OOOOGMT 04 February 1979. We can see that the values were more than
0.5 P V units in the surface frontal zone (* * 1 PV unit = 10-6 m2/ S e C /kg).
A high-pressure system developed near the ground over China. The
surface winds over southern China were northeasterlies (Fig.7). The cold
anticyclone was very shallow and there was strong zonal ( westerly ) flow
at 700 Hpa (Fig.8). The backing of the wind with height and large vertical
shear implys strong cold advection below 3 km.

Fig.9 indicates the ageostrophic components of horizontal wind. Due
to the influences of the Tibetan Plateau, the wind tended to flow around the
high mountain near the earth's surface and was decelerated in the upstream
region. Hence, the real wind was much smaller than the geostrophic wind
over southern China. In other words, there was extremely strong subgeostrophic
flow in this region for several days (through 2 to 5 february
1979). This highly sub-geostrophic area was consistent with the stationary
frontal zone. It suggested that the topographic effect of the Tibetan Plateau
offered the strong ageostrophic motion in southern China and in term the
ageostrophic (sub-geostrophic) motion induced the frontogenesis there.
3. VERTICAL STRUCTURE OF FRONTAL ZONE
Cross-section analysis is a useful tool for studying the vertical
structure of frontal systems. Fig.10 (a)-(d) present cross-sections of
temperature gradient for 00, 06, 12, 18GMT 04 February 1979,
respectively. The cross sections were taken along 116.25°E. The frontal
surface tilted northward. It contained two maximum domains, one was
associated with the surface stationary front and the other was associated
with the upper subtropical front.
Fig. 11 (a)-(d) show cross-sections of vertical vorticity for the same
period. In the warm sector and the cold region, there was strong negative
vorticity, but positive wvorticity was found in the frontal zone. The axis
of positive vorticity was parallel to that of maximum temperature gradient
but was located a little bit northward.
The conservative property of potential vorticity under conditions of
adiabatic frictionless motion was used as the basis to illustrate the concept
of tropopause folding and downward intrusion of stratospheric air into the
troposphere (Reed and Danielse,1959). The narrow tonque of high
potential voticity extended downward from the stratosphere into the uper
portion of the sub-tropical frontal zone as shown in Fig.12. This result was
consistent with the diagnostic vertial motion calculated for the frontal case
by Shapiro (1970) which showed a strong narrow band of sinking motion
extending downward from the stratosphere into the upper frontal zone. In
the surface stationary frontal zone, potential voticity was increasing with
time because the ageostrophic motion offered strong shear vorticity and
then created a stable layer near the ground.
The transverse circulation is important in frontogenesis process
(Eliassen, 1962). Fig.13 shows the transverse circulatin along 116.25°E for
04 February 1979. The thermally direct circulation was in the warm
sector, and a downslope flow was found along the frontal surface. After 12
hours, the circulation decayed and upward motion was found in the whole
domain. Fig. 14 shows the secondary circulation along 123.75°E for 04
335

336
February 1979. It clearly indicates that a cyclone developed to the
northeast of Taiwan, i.e. in East China Sea.The downward motion
intensified in the cold region during the cyclogenesis. As shown by Shapiro
(1981), frontogenesis can be expressed as a purly ageostrophic process
which evolves much faster than the quasi-geostrophic process. The
dynamics of ageostrophic transverse circulation will be discussed in the
next section.
4. DYNAMICS OF THE TRANSVERSE CIRCULATION
The charecteristics of frontal structure depends upon the transverse
circulation in the vicinity of frontal zone. So, the generation of vorticity in
the y-z plane will be discussed.
The x-direction component of vorticity and the vorticity equation
can be expressed as
( w) = - ( v w ). ( + + + - ( + )
at ay az a y az a z p 2 Va z a y a y az v }
where p is the density, p is the pressure, f is the Coriolis parameter and
(u,v,w) are the wind components in (x,y,z) respectively.
Eq.(2) and (3) can be transformed into the p-coordinate as
8v pg 303 i 1 \/y 30 . do) \J\Af

337
£ * ,-
(6)
dp dy
where u g is the geostrophic wind in x-direction. The third term and the
fourth term can be combined as
where u ag is the ageostrophic wind in the x-direction. R ag is called the
ageostrophic residue because it represents the departures of the x-velocity
and temperature fields from geostrophic balance (Orlanski and Ross,
1977).
Jiang and Kau (1987), Jiang and Jeng (1987) estimated the
magnitudes of each term using case studies and numerical model
simulations, respectively. They found that the first and the second terms
were negligibly small. The generation of horizontal vorticity depends
primarily upon the ageostrophic residue. Hence, knowing the behavior of
Rag we can easily predict the rate of generation of cross-stream vorticity
along the front.
5. CONCLUSION
In late-winter and early-spring, the Topographic effect of the
Tibetan Plateau induced a strong sub-geostrophic motion near the ground
in southern China. The ageostrophic residue is the main effect of the
transverse circulation which is important in the frontogenesis process.
Therefore, stationary surface frontal zones often existed in southern China
in late-winter and early-spring.
According to the three-cell circulation concept and the conservative
property of potential vorticity, it appears to exist a semi-permanent upper
sub-tropical front from which the high potential vorticity stratospheric air
folds into the troposphere.
It is well known that the formation mechanisms of the upper subtropical
front and the stationary surface front over southern China are
quite different. However, we do not quite understand the coupling effects
of these two frontal systems until now. More research about this is needed,
REFERENCES:
Eliassen A., "On the Vertical Circulation in the Frontal Zone". Geofys.
Publik., 24, 147-160 (1962).

338
Jiang H. M. and Jeng H. N., "Numerical Experiments of the Transverse
Circulation in the Vicinity of the Frontal Zone". Tech. Report,
pp77, (1987) (in Chinese).
Jiang H. M. and Kau C. L.," Diagnostic Analysis of the Stationary Frontal
Zone in Eastern Asia". Tech. Report, pp74, (1987) (in Chinese).
Hanson, H. P. and Long B., " Climatology of Cyclogenesis over the East
China Sea ". Mon. Wea. Rev., 112, 697-707 (1985).
Orlanski, I. and Ross B. B. , " The Circulation Associated with a Cold
Front. Part I: Dry Case ". J. Atmos. ScL, 34,1619-1633 (1977).
Palmen, E. , f< The Role of Atmospheric Disturbances in the General
Circulation ". Quart. J. Roy. Meteorol. Soc.,77,337-354 (1951).
Palmen, E. and Newton C. W., " Atmospheric Circulation System ft .
pp603, (1969),
Read, R. J. and Danielsen E. F., " Fronts in the Vicinity of the Tropopause
". Arch. Meteor. Geophys. Bioklim., AIL 1-17,(1959).
Shapiro M.A., n On the Applicability of the Geostrophic Approximation to
Upper-level Frontal-scale Motions ". J. Atmos. Sci., 27, 408-420,
(1970).
Shapiro M. A.," Frontogenesis and Geostrophically Forced Secondary
Circulation in the Vicinity of Jet Stream-Frontal Zone System ". J.
Atmos. Sci., 38, 954-973, (1981).
Yeh F. W. C. and Chen G. T. J., "Some Aspects of Front Climatology
over Southern Part of China and the Adjacent Oceans in Winter Half
Year ".Qt J. Meteor., The Weather Central, Weather Wing, CAP,
10.1, 9-20, (1984)(in Chinese),
no m 120
US 120
Fig.l The principal air masses, tropopauses and
fronts, and jet streams in relation to the
features of the low-level wind systems,
(After Palmen and Newton, 1969 )
Fig.2 Cyclogenesis occurrence during February
and March through 1978 to 1987.

339
100 110 120 130
Fig.3 The synoptic situation over eastern Asia at
OOOOGMT 04 February 1979, as analysed
by the Japan Meteorological Agency.
Fig. 4 The synoptic situation over eastern Asia at
OOOOGMT 05 February 1979, as analysed
by the Japan Meteorological Agency.
1979/2/04/002 1000 MB TEMP-GRAO-CC/IOOKM] 1979/2/04/QOZ 950 MB PV C*PVU3
Fig.5 The horizontal temperature gradient at 1000
Hpa at OOOOGMT 04 February 1979. The
contour interval is 0.5 C/lOOkm.
Fig.6 The potential vorticity at 950 Hpa at
OOOOGMT 04 February 1979. The contour
interval is 0.5 PV unit.
1979/2/04/002 1000 MB CU,.V3 1979/2/04/OOZ 700 MB CU.J/}
Fig.7 Hie horizontal wind field at 1000 Hpa at
OOOOGMT 04 February 1979. The length
scale of the vectors is such that the longest
vector corresponds to the magnitude of 10
m/sec.
Fig.8 The horizontal wind field at 700 Hpa at
OOOOGMT 04 February 1979, The length
scale of the vectors is such that the longest
vector corresponds to the magnitude of 10
m/sec.

TEMP.GRAD.(C/100KM3
1979/2/04/OOZ 1000 MB tUAG.VAG)
20 25 30 35 40 (N) 2Q 25 30 35 40 (N)
j 1979/2/4/122 TEMP.GRAD.CC/100KM3 f1979/2/04/18Z TEHP.SRAD.[C/10QKH)
Fig.9 The ageostrophic wind component field at
1000 Hpa at OOOOGMT 04 February 1979.
The contour is the magnitude of the
ageostrophic wind and its interval is 5
m/see.
Fig.10 Cross-section analysis of temperature
gradient for 04 February 1979. The cross
sections were taken along 116,25°E. The
contour interval is 0.5 C/lOOkm. (a) at
OOGMT; (b) at 06 GMT; (c) at 12 GMT;
(d)at 18GMT.

•= 1979/2/04/002 VORTICITY KE-S/SEC 1979/2/04/06Z VORTICm »E-S/SEC ^ 1979/2/04/DOZ
1979/2/04/06Z
20 25 30 35 40 (N) 20 2S 30 35 40 (H)
30 35 40 (N) 20 25 30 35 40 IV)
"5 1979/2/4/12Z . VORTICITY «E-5/SEC 3 1979/2/04/18Z VORTICITY «E-5/SEC
•^ 1979/2/4/12Z
PV (*PVU) -5 1979/2/04/J8Z PV C*PVU3
.Fig.ll Gross-section analysis of vertical vorticity
for 04 February 1979. The cross sections
were taken along 116.25°E. Solid contours
are positive; dashed contours are negative
values. The contour interval is l.Oxl0-5/sec.
(a) at OOGMT; (b) at 06 GMT; (c) at 12
GMT; (d)at 18GMT.
Fig. 12 Cross-section analysis of potential vortitity
for 04 February 1979. The cross sections
were taken along 116.25°E. The contour
interval is 0.5 PV unit, (a) at OOGMT; (b)
at 06 GMT; (c) at 12 GMT; (d)ai 18GMT.

TKANS.C1R. 9 1979/2/04/OE2 •; J979/2/05/OOZ TRANS.CIR. •£ 1979/2/Q5/06Z TRANS.CIR,
se fO
20 25 30 35 40 W)
20 25 30 35 40 (N) 2Q 25 30 35 40 {H»
20 2S
3° 35 40 !N>
3.1979/2/4/1^ TRftNS.ClR. 1979/2/04 /18Z TRftNS.CSR.
•21979/2/05/122 TRANS.CIfU -5 1979/2/05/182 TRANS-C1R-
20 25 30 35 40 (HS
Fig.13 Transverse circulation near the frontal
zone for 04 February 1979. The cross
sections were taken along 116.25°E, (a) at
OOGMT; (b) at 06 GMT; (c) at 12 GMT;
(d)at 18GMT.
20 25 30 35 40 («>
Fig. 14 Transverse circulation near the frontal
zone for 05 February 1979. The cross
sections were taken along 123,75°E. (a) at
OOGMT; (b) at 06 GMT; (c) at 12 GMT;
(d)at 18GMT.

343
The Microphysics of a Mei-Yu Case:
Theory
K. Kenneth Lo and Chung-Ming Liu
Dept. of Atmospheric Sciences
National Taiwan University
Taipei, China
ABSTRACT
Observational results from airborne observations on June 16-
17, 1987 over the Pacific Ocean east of Taiwan show, 'among other
things, the following results:
(1) Within the melting layer, the particle number density
decreases significantly from the top to the bottom of the melting
layer. If the particle spectra are assumed to satisfy
N 0 {exp(-AD) ) , where No is the intercept and X the slope, N 0
decreases significantly from top to bottom while the slope
decreases much more slowly. The median diameter increases
slightly from the top to the bottom of the melting layer.
(2) In the warm rain region, the particle number density of
raindrops initially decreases with the decrease of height and
then comes to a roughly constant value. The particle median
diameter initially increases with the decrease of height and
eventually also cornes to a rather constant value.
This paper will explain theoretically the observed results.
Melting Layer:
Within the melting layer, since both the particle number
density of raindrops and the intercept decrease significantly
from the top to the bottom, while the slope changes only
slightly, it is postulated that the decrease of the first two
parameters is mainly due to the increase in terminal fallspeed
when snow crystals melt to become raindrops. Other processes,
such as condensation and binary interactions are comparatively
not that important.
It is obvious that the precipitation irate should be the same
above and below the melting layer, otherwise there would be
convergence or divergence of precipitation particles somewhere.
Then, using the conservation of precipitation rate, we can
formulate an equation describing the change of size distribution
slope before and after melting.
The results show that this simple calculation can explain
the sharp decrease in slope. Despite the fact that there are a
great variety of processes going on within the melting layer,

344
however, the sharp decrease in particle number density with the
decrease of height within the melting layer can be attributed
mainly to the increase in particle terminal fallspeed during
melting. Even without taking account of binary interactions,
this simple calculation can duplicate the significant change of
the intercept.
Warm-rain region:
In the warm-rain region, it is observed that the particle
number density initially increases with the decrease of height
while the median diameter initially increases with the decrease
of height. It is postulated that these are due to collisional
coalescence being more effective than breakup. The change in the
particle size distribution, f(x,h,t) (x being the mass of a
raindrop, h the height and t the time), due to collisional
coalescence and breakup can be described by using the stochastic
collection equation for binary interactions.
Expressing f(x,h,t)=N 0 (exp(-XD), there are two dependent
variables, i.e. N t , and \. Thus, two equations are needed in
order to obtain the solutions. The two most easily-derived
equations are the first moment and the second moment equations of
the stochastic collection equation.
The results show that the comparatively slower decrease of
particle number density and increase of median diameter can be
explained by coalescence being more effective than breakup. The
theory shows that the raindrop size distribution will eventually
achieve an equilibrium state. This is consistent with previous
findings.

345
1. Introduction
Observational results obtained from airborne observations on
June 16-17, 1987 (Liu and Lo, 1989} show, among other things, the
following results:
(1) Within the melting layer, the total number of particles
in a unit volume decreases significantly from the top to the
bottom of the melting layer. If the particle spectra are assumed
to satisfy the analytic function of Notexp{-AD)), where No is the
intercept and ,X the slope, No decreases significantly from top to
bottom. The slope first increases and then decreases. If we
consider only the values of the slope above and below the melting
layer, the change is slight compared with the enormous change of
the intercept. The median diameter increases slightly from the
top to the bottom of the melting layer.
(2) In the warm rain region, the total number of raindrops
in a unit volume decreases with the decrease of height. The
particle median diameter increases quite significantly with the
decrease of height.
This paper attempts to explain theoretically the observed
results.
2. Melting Layer
Within the melting layer, since both the total number of
raindrops in a unit volume and the intercept decrease
significantly from the top to the bottom, while the slope changes
only slightly, it is postulated that the decrease of the first
two parameters is mainly due to the increase in terminal
fallspeed uhen snow crystals melt to become raindrops. Other
processes, such as condensation and binary interactions are
comparatively not that important. The results will show that the
postulates are correct.
It is obvious that the precipitation rate should be the same
above and below the melting layer, otherwise there would be
convergence or divergence of precipitation particles somewhere.
Then, using the conservation of precipitation rate, we can
formulate

346
00
" XD 3 b
D 3 aD b dD = [ N. e XD pD q rnD n dD = R ri\
0 - ~ •'o 1 Ic v
* '
wh.ere, Ni = Intercept of snow-size distribution,
Nw = Intercept of raindrop size distribution,
X - Slope of size distribution, taken to be the same for
snow crystals and raindrops,
Vi z Terminal fallspeed of snow crystals,
Vw = Terminal fallspeed of raindrops,
M = Mass of snow crystals,
D = Particle diameter.
The mass-diameter and fallspeed-diameter relationships of
snow crystals are taken from Locatelli and Hobbs (1974). Two
formulation of raindrop fallspeed are used, namely that from
Rogers (1979) and Liu (1986). The former raindrop fallspeed
formulation is simple enough to allow a full analytic solution
while the latter formulation requires numerical solution.
Comparing the results with Liu and Lo (1989) shows that this
simple calculation can explain the sharp decrease in slope (Table
1). Certainly, it is understandable that the results will not
match exactly for the many assumptions made in this simple
calculation. For the C-probe, the results from using the
terminal fallspeed of Rogers (1979) and that of Liu (1986) are
not that different. The P-probe results show that the terminal
fallspeed of Liu (1986) is better than that of Rogers (1979).
3. Warm-rain region
3.1 Formulations
In tb-s warm-rain region, it is observed that the particle
number density increases with the decrease of height while the
median diameter increases with the decrease of height. It is
postulated that these are due to collisional coalescence being
more effective than collisional breakup. The change in the
particle size distribution, f(x,h,t) (x being the mass of a
raindrop, at height h and time t), due to collisional coalescence
and breakup can be described by using the stochastic collection
equation for binary interactions (Scott, 1968, Drake, 1972,
Passarelli, 1978, Srivastava, 1978, Lo, 1983):
w
,t) = I J f(x-x',h,t)£(x',h,t)K(x~x f ,x')q(x-x',x')dx'
0
- f(x,h,t) f f(x f ,h,t)K(x,x')q(x,x')dx'
o
00 TO (2)
+ .3 I J S(x|s',x f }£(x',h,t)£(x",h,t)K(x' ,x")U-

347
where
V(x) = terminal fallspeed of raindrop with mass x
K(x',x") = collisional kernel between raindrops of mass x 1
mass x"
and
q(x',x") = probability of coalescence when two raindrops of
mass x j and x" collide
S{x|x',x") = function of the number of fragments with mass
between x and x+dx when raindrops of mass x 1
collide and break up
and x"
The first term on the right hand side is the production of
raindrops of mass x due to coalescence. The second term is the
depletion of raindrops of mass x due to coalescence. The third
term is the production of raindrops of mass x due to breakup and
the fourth term is the depletion of raindrops of mass x due to
breakup.
In order to simplify the derivations, the vertical updraft
is taken to be a constant. No and A are assumed to change only
with height and not with time. Other formulations are:
raindrop size distribution: f(x,h)dx = N Q (h) e" D> iD
raindrop mass: x = a D
fallspeed: V = a D
Collfsional Kernel: K(x',x") = \ (D'+D") 2 E|V(D")-V(D J )|
breakup probability: qtx'/.x") - e
2 — A x
fragment number probability: SCxjx'rX 11 )' = {x'+x M )A e
Expressing f(x,h,t)=No{exp(-XD), there are two dependent
variables, i.e. No and A- Thus, two equations are needed in
order to obtain the solutions. The first moment and the second
moment equations of the stochastic collection equation are
derived. Physically, the first moment equation is to keep track
of the total particle mass in a unit volume against height and
the second moment is to keep track of the radar reflectivity
factor against height. The results from this set of two
simultaneous equations give the No and A values at different

348
heights. The median diameter and the particle number density can
then be calculated from:
The resulting equations are:
Median Diameter, Do=ln2/X
Particle Number Density, Na=No/A
(5) Cife+Zl Si
V 2' b+a dh
- b±ii f^ ------ (4)
dh x dh
I - f x 3 y 3 (x*y) J (x b -y b ) e'^^'d* dy
1 J o 4 o
f°° f X 3 ,2 , b b. - . ,
I a x '(x+y) (x -y ) e dx dy
J o J o
T = f f x 3 ( X+ y) J (y b -x b ) e- Cx+y> J dx dy
o o
f 00 f x b
a ', 'v^ / b ^ -' ,
I 4 - = I I x (x-J-y) (x -y ) e dx dy
o o
I B = f J* V (x t y) a (y b -x b ) .-*"*" dx dy

349
3.2 Results
with
Figs. 1
height.
and 2 show the change of particle number density
The number density initially decreases with the
decrease of height. Eventually, a constant value of number
density is reached and the number density does not change with
height. Physically this means that initially coalescence is more
effective than breakup and so there is a net depletion of
raindrops and so the number density decreases. However, the
efficiency of breakup increases with the increase of raindrop
sizes. Eventually,
coalescence is reached.
an equilibrium between breakup and
In this mathematical formulation, the breakup probability
and fragment number probability increase with the size of
raindrops. This means that coalescence will be more effective
when there are more larger raindrops. In this calculation,
'coalescence is initially more effective than breakup, producing
more larger raindrops. The increase in the number of large
raindrops will in turn enhance the breakup efficiency. Therefore
although the breakup probability and fragment number probability
are initially prescribed so as to give coalescence an advantage,
the coa1esoeriee process itself acts as a negative feedback.
Eventually, an -equilibrium is reached between coalescence and
breakup.
The values of c arid A used for track 1 are 20 and 0.01 while
the values of c and A for track 2 are 4 and 0.15 respectively.
Physically a smaller c implies two particles are more likely to
coalesce when they collide. (If c=0, two particles will always
coalesce when they col1ide.) A smaller A implies that there are
fewer fragments when two particles collide. In order to model
the continual decrease of particle number density of track 1,
large** c and smaller A values are used. These mean that two
raindrops are more likely to break up when they collide but the
collision will produce fewer fragments. The two processes will
mean that it will take a longer time to reach equilibrium.
Pigs. 3 and 4 show the change of median diameter with
height. The median diameter initially increases with the
decrease of height. But then it reaches a constant value. The
explanation for this phenomenon is similar to the explanation for
the change of particle number denisty. Initially coalescence is
more effective than breakup, resulting in the depletion of small
raindrops and the production of large raindrops. Therefore the
median diameter increases. But eventually the coalescence and
breakup effects balance one another and the median diameter
arrives at a constant value. The result is similar . to the
observations and the equilibrium between coalescence and breakup
agrees with the findings of Srivastava, 1978; List et.-al.:, 1987;
arid /a wad a It i and de Agost. inho Antonio, 1988.
4. Conclusions:
Despite the fact that there are a great variety of processes

350
going on within the melting layer, however, the sharp decrease in
particle number density with the decrease of height within the
melting layer can be attributed mainly to the increase in
particle terminal fallspeed during melting. Even without taking
into account of binary interactions, the simple calculation can
duplicate the significant change of intercept.
In the warm-rain region, the comparatively slower decrease
in particle number density and increase in median diameter can be
explained by coalescence being more effective than collisional
breakup. The theory shows that the raindrop size distribution
will eventually achieve an equilibrium state. This is consistent
with the findings of Srivastava (1978). Observational results
for the lower troposphere also show no indications of obvious
increase or decrease of number density or median diameter.
5. Acknowledgment
This research was supported by National Science Council
under Grant NSC77-0202-M002-22. Miss S.C. Lee is being thanked
for assistance in plotting.
References:
Drake, R.L., 1972: The scalar transport equation of coalescence
theory: Moments and kernels, J rV Atmos.. Sci . , 29 f 537-547.
List, R., N.R. Donaldson and R,E. Stewart, 1987: Temporal
evolution of drop spectra to collisional equilibrium in
steady and pulsating rain. J. Atmos. Sci • . t 4,4 , 362-372.
Liu, C.M., 1986: Effects of the raindrop terminal velocity on
raindrop growth. Pap Me tea. Res..._, 9, 79-104.
and K.K. Lo, 1989: The Microphysics of a Mei-Yu Case:
Observations. International Conference on East AsjLa and
Western Pacific Meteorology and Climate, 6-8 July 1989, Hong
Kong.
Lo, K»K., 1983: Growth Processes of Snow. Sc.D. Thesis.
Massachusetts Institute of Technology. 193 pp.
Locatelli, J,D. and P.V, Hobbs, 1974: Fall speeds and masses of
precipitation particles. JN_ Geophy. Res..,., 7.9, 2185-2197,
Passarelli, R,E. f Jr., 1978: An approximate analytical model of
the vapor deposition and aggregation growth of snowflakes.
J. Atmos. Sci . , 2.5 ,'. 54-65.
Rogers, R»R..., 1979: A Siiort Course in Cloud Physics* 2nd Ed.
Pergamon Press. 235 pp*
Scott, W.T., 1968: Analytic studies of cloud droplet coalescence
X* J, >;tmos. Sci.. 25, 54-65.

351
Srivastava, E.G., 1978: Parameterization of
distributions. J. Atmos. Sci . , 35, 108-117.
raindrop
size
Zawadzki, I. and M. de Agostinho Antonio, 1988: Equilibrium
raindrop size distributions in tropical rain. J. Atmos. Sci,
45., 3452-3459.
Melting:
water:
u a 8000 r = 4000 D
N = 0.49 cm" probe p:
17.6+4.25 = 10.925 cm
(Ice
phase)
108.1D°'
1.0Q5xlO~ 7 D 1>4 5.37xlO~ 4
Crystal Fallspeed and Mass
V=115.6D
Intercept (Raindrop)
9.8xlO~ 4
147.1D°*
177.36D 0
2. 939x10" 3 D 1 " 9
2.939xlO~ 3 D 1 ' 9
7.88x!0" 4
7.84x!0" 4
probe c:
. 21.6+13.4 „ 17 , ^ -t
as ~ s i/.o cm
Crystal Fallspeed and Mass
L15.6D 0 ' 16 Mai.834x10"'D 1 ' 4 .
177.36D
108.1D°'
ID°" 2T 2. 939x10" 3 D 1 " 9
1.005xlo"" 3 D 1 ' 4
N 4 =1.74 cm (Ice
phase)
Intercept (Raindrop)
l,0989x!0" 2
6.629x10" 3
6.17xlo" 3
6.137xlo" 3
Table 1: Tables of the intercept of particle spectra before
and after melting. Different crystal fallspeeds and
masses are quoted from Locatelli and Hobbs
Units used are c.g.s. units.

352
TRFICK - I (HIS 0 - H31 0J
PR0BE C PR0BH P PR08E C + P
O
LU
x:
o ^
PRRTICLE NUMBER DENSITY ( XL)
Fig.
1: The profile of particle number density as computed
for track 1. (cf. Liu and Lo (198^)
fRRCK - 2 11652 0 - 1711 QJ
PR8BE C PR0BE P PR08E C + P
to
o
V
o o
CM
UJ x:
r
nq iinn| II|IM
.
r
nnm-rnwrrrm*
r"
o
o
t
I
U
f
UJ
U4 UJ
lAJ(l
o.
8 S £
; 8 s E
S'e
ij UJ Ul Ml W
PRRTICLE NUMBER DENSITY (./L)
Fig. 2: Similar to Fig. 1 but for track 2.

353
TRRCK - 1 (1115 Q - 1131 QJ
PR0§E C PR08E P PR08E C + p
to
o
\
,
,
x:
i —
0 o
UJ ^
in
'
o
o
3
LI
S 3
5 »
UJ UJ U
m * r> f\
O C '
u uJ U.
T) «•
XI
UJ
MEDinN DIRMETER (UM)
Fig. 3:
The profile of median diameter as computed for
track 1. (cf. Liu and Lo (1989))
TRRCK - 2 11652 0 - 1711 0)
PR0BE C PR086 P PROBE C + P
o
LU

354
MONTHLY AND SEASONAL FORECASTS AND
TROPICAL OCEAN - ATMOSPHERE INTERACTIONS
Chao Jiping , Ji Zhengang and Wang Xiaoxi
(Nanioal Research Center For Marine
Environemental Forecast, Beijing,China)
ABSTRACT
The experimental forecasts results by use of the
air - sea /land coupled anomaly model, which
is developed ten years ago have been represented
in this paper. It is also shown that this anomaly
model is also effective to study the
teleconnection effect of the sea surface
temperature anomalies (SSTA) in the tropical
ocean upon the changes of the global atmospheric
circulation . In order to develop this model
into one of including the Equator, the study on
the air - sea interaction in the tropical ocean
has been undertaken based on the same idea of
the model mentioned above. It is pointed out that
a kind of slow wave, which is caused by the air -
sea interaction exists. This slow wave may be
regarded as a monthly and seasonal predictor, as
a consequence, a new global anomaly model
including the tropical ocean of long - range
forecast is being developed .
1. MONTHLY AND SEASONAL NUMERICAC FORECASTS BY USING THE
ANOMALY OCEAN /LAND - ATMOSPHERE COUPLING FILTERED
MODEL
Ten years ago, a numerical model was proposed for
the monthly and seasonal forecasts, which is an atmosphere
- ocean/land coupled model (Chao et al, 1977, 1982). The
model was different from the usual general circulation
model (GCM) in that the physical qualities such as
geopotential height (GPH) , sea surface temperature
(SST) were separated into climatological means and
anomalies. The climatological means were obtained ^or
diagnosed from the observations as known fields putting
into model, so only the anomalies are needed to be
predicted. Such a model has been called anomalous
model. Moreover, based on the analysis of linear system
of the model, the transient Rossby waves have been

355
filtered from model because it can assumed as » high -
frequency noise" in comparing with the low frequency air -
sea/land interaction waves which are just what we are to
forecast in the time scale of one month or more.
According to above two basic ideas first suggested by
Chao et al., an anomalous atmosphere - ocean/land coupled
filtered model (AFM) has been developed. Recently, eight
examples of monthly forecasts during the winter months in
1976 - 1977 and 1982 - 1983 El Nino events have been
performed. Their correlation coefficients between
observations and predictions as well persistences of
monthly prediction are listed in Table 1. On the average,
the prediction of both surface temperature and height
fields are superior to the persistences. The predictions
(shown in chart (a)) of the anomalous surface temperature
and SOOhPa height fields for March 1977 are illustrated in
Fig. 1 and Fig. 2 resperctively. For the sake of making
comparison, the observations (shown in chart(b)) are also
given in the figures.
In the meanwhile, the seasonal forecasts for the
winter months 1976 - 1977 and 1982 - 1983 El Nino events
also completed. The results show in Table 2(figures
omitting) . As shown in the Table 2, it can be seen that
the seasonal forecasts in the sense of average are
promising, that means the AFM to have the potential
ability in predicting the seasonal anomalies of
large - scale circulations.
2. TELECONNECTIONS OF THE SST IN THE TROPICAL OCEAN WITH
THE 500 hPa GEOPOTENTIAL HEIGHT IN THE NORTHERN
HEMISPHERE
In recent years, many authors are interested in the
influences of the tropical SST anomalies upon the global
atmospheric circulation. Some examples have been
illustrated in this section in order to show AFM having
the capability in studying this kind of teleconnection.
The correlation between SST in the Indian Ocean (83*E
- 91 ° E, 5 ° N - 15*N) and 500 hPa GPH anomalies in the
Northern Hemisphere is. given in Fig.3. It can be seen in
Fig. 3, that there exists positive correlation in the
regions of lower and higher latitudes, while negative
correlation mainly emerges in middle latitudes* The
teleconnections also studied by using AFM based on the
fact mentioned above. Fig.4 describes the teleconnection
map of 500 hPa GPH in the Northern Hemisphere with the
heating source in the Indian Ocean (83*E - 91°E, 5 N° -15*
N) . Some interesting finding from Fig. 4 is that the
anomalous perturbations of 500 hPa GPH in the Northern
Hemisphere are spiral wave structure which is independent
with the positions of heating sources.

356
In another experiment, the heating sources is located
at the eastern tropical Pacific Ocean(5*N, 130°W) where is
so-called " key region lf by Pan and Oort(1983). The
result of simulation illustrated that the spiral structure
anomalous perturbations are covered the whole Northern
Hemisphere at 500 hPa . And it is also shown that the
atmospheric circulation of the Northern Hemisphere is
characterized by anomalous wavetrains across the Pacific
and Northern America continent, which is somewhat similar
to PNA pattern. In Fig.5. the anomalous wavetrains are
denoted by duble thick line, and the spiral shaped
perturbation are linked by single thick line.
On the other hand, we set three heating sources in the
Eastern Tropical Pacific Ocean (2 N°- 10°N , 125*W - 135°W
, (3 a c)), Western Tropical Pacific Ocean (2°N - 10*N , 125°W
- 155°W / ( ~1.2°c)) , and the Indian Ocean (5*N - 15°N ,
83° E - 91°E ,(l.8°c)) , respectively, which are referred
to the patterns of anomalous SST fields in January 1973.
The response of 500 hPa GPH to these three heating sources
are shown in Fig. 6 . The observational 500 hPa anomalous
GPH for January, 1973 is shown in Fig. 7 . It shows that
the main features of Fig. 6 is closed to Fig. 7 .
3. TROPICAL UNSTABLE ATMOSPHERE INTERACTION
To apply above model to the tropics, we should firstly
discuss the tropical atmosphere - ocean interaction. Some
authors (e.g. Philander et al, 1984) constructed models to
investigate air - sea interactions using so called
"slavery atmosphere", i.e., the atmosphere model is
steady, only the ocean model is time - dependent. Ji
(1987) argued that this kind of coupling is actually to
parameterilze the forcing of wind stress in an ocean
model. Here we will set up an analytical coupled model, in
which the motion equations of the atmosphere and ocean
are all time - dependent, and explore coupling mechanism
of atmosphere - ocean system (AOS).
The basic equations of the atmosphere and ocean
are similar the ones derived by Anderson and Gill(1975),
which are shallow water equations and inertial gravity
waves are filtered. Chao and Ji (1985, 1986) used these
equations to discussed the tropical oceanic waves. The
coupling between the atmosphere and ocean are similar to
Philander et al 's (1984), i.e., changes in the depth of
the thermocline affects the SST which, in turn, heats (or
cools) the atmosphere. Atmospheric winds ar^ assumed to
act as a body force in the ocean. Parabolic cylinder
functions are used to solve the equations. It is a simple
truncated filtered model. The detailed descriptions of
this model are given inseparate papers (Ji and Chao, 1989
Chao and Zhang, 1988).

357
The dispersion relationship is shown in Fig. 8a, as
coupling coefficients a=10 /s and b=5 10 /s which are the
same as the ones taken by Philander et al (1984).
Comparint Fig.Sa with the result of Matsuno (1966), it can
be seen that the asymmetric component of AOS (m, n=l,
which are asymmetric about equator) are stable. The slower
wave numbers. The faster one is not changed much, which
shows that coupling interactions of AOS do not influence
the high frequency wave very much. In symmetric component
of AOS, there is also a eastward low frequency wave . It
is worthy to be mentioned that the inertial gravity wave
and Kelvin wave are all filtered in the basic equations.
Hence the eastward slow wave shown in Fig.Sa should be the
results of coupling atmosphere - ocean interactions.
Rasmusson (1985) analysed several parameters of AOS during
1982/83 ENSO, he found that their anomalies extended and
propagated eastward consistantly. Though we can not affirm
that the eastward wave mentioned here is just the physical
causes of anomalies moving eastward uncovered by
Rasmusson, however, we believe that this kind of eastward
wave produced by air - sea interactions has certain
influences upon the propagation of anomalies during ENSO.
Ji (1987) investigated the physical mechanisms of eastward
propagation and declared that it is the comprehensive
result of the earth rotation and coupling interactions
between the atmosphere and ocean.
Another striking feature in coupling AOS is the
instability. Fig. 7b shows the e - fold time is about one
month or so. Instable length is 800 - 11000 km, the most
instable wave length is around 4000 km . Fig. 9 shows the
relationship between parameter a X b and instable wave
lengths. Instable area is hatched. From Fig.9 it can be
seen that AOS is easier to be instable at smaller or
larger wave length. As a X b increases, the range of
instable wave lengths decreases . And as a X b reaches a
critical value, AOS returns to be stable. It illustrates
that only coupling coefficients a and b vary in certain
range, AOS could be instable. And this kind of instability
is the result of internal coupling interactions of the
AOS. Ji and Chao (1989) also discussed the possible
relation of this wave ENSO events.
The structure of instable waves are shown in Figs. lOa
and lOb (wave length L=5000 km). Fig. 9a is the
perturbational dynamical height of the model ocean and
Fig. 9b is the corresponding perturbational height of the
atmosphere. The black arrows indicate the convergence and
divergence of the atmosphere and ocean. Comparing Fig. lOa
and lOb , it can be seen that the convergence (divergence)
of oceanic currents corresponds to the convergence
(divergence) of wind field. So the AOS can be organized to
form positive feedback process which , in turn, leads
perturbation fields of AOS to develop furthermore.

358
4. CONCLUSION
The forecasting experiment mentioned above has shown
that the anomaly model has forecasting capability of the
monthly and seasonal time scale. It is feasible to use
this long - range forecasting model in the operational
services, because only the general GTS data were used in
the experiment.
It is also shown the capability of this model in
studying the teleconnection. On the other hand, it is
revealed that the tropical ocean SST is very significant
to the changes of the atmospheric circulation of the mid
and high latitudes. Therefore, it is necessary to take the
effect of the teleconnection of the tropical ocean into
accoount in a .good forecatsing model .
On this account, we start from the study on the
characteristics of the air - sea interaction of the
tropical ocean to develop a new global anomaly model
including the tropical ocean of the long - range
forecast.

359
REFERENCE
Chao Jiping and Ji Zhengang, " On the influences of large
- scale inhomogeneity of sea temperature upon the
oceanic waves in the tropical regions - Part I:
Linear theoretical analysis", Advances in
Atmospheric Sciences, Vol. 2.,295 - 306 (1985).
Chao Jiping et al," On the physical basis of a model of
long - range numerical weather forecasting", Scientia
Sinica, 20, 377 - 390 (1977).
Chao Jiping et al, "a theory and method of long - range
numerical weather forecasts", J. Meteo. Soc. Japan, 60,
282 - 291 (1982).
Chao Jiping and Zhang Renhe, "The air - sea interaction
waves in the tropics and their instabilities ", ACTA.
Meteor. Sinica, 2, 275 - 287 (1988).
Gill, A. E.,"some simple solutions for heat - reduced
tropical circulation model", Q. J. R. Meteor. Soc.,106,
447 - 462 (1980).
Ji'Zhengang, "Some researches on the large -scale dynamics
of the tropical atmosphere and ocean", Ph.D.
diss. Institute of Atmospheric Physics, Academia
Sinica (1987).
Ji Zhengang and Chao Jiping, 11 On the influences of large -
scale inhomogeneity of sea temperature upon the oceanic
waves in the tropical regions - Part II: Linear
numerical experiment", Advances in Atmospheric
Sciences, Vol.3, 238 - 244 (1986).
Ji Zhengang and Chao Jiping, "Teleconnections of the sea
surface temperature in the Indian Ocean with the sea
surface temperature in the eastern equatorial Pacific,
and with the 500 mb geopotential height fields in the
Northern Hemisphere", Advances in Atmospheric Sciences,
Vol.4, 343 - 348 (1987).
Matsuno, T. , "Quasi - geotrophic motions in the equatorial
area", J.Meteor.Soc. Jan,,44, 25 - 42 (1966).
Pan Yihang and A. H. Oert, "Global climate variation
connected with sea surface temperature anomalies in the
Eastern Equatorial Pacific Ocean for the 1985 - 1973
period", Mon. Wea. Rev.Ill, 1244 -2358 (1984).
Philander, S. G. H. et al, "Unstable air - sea
interactions in the tropics", J. Atmos. Sci., 41, 604 >
613 (1984).

360
TABLE 1. COMPARISONS OF THE CORRELATION BETWEEN
OBSERVATIONS AND PREDICTIONS AS WELL AS PERSISTENCES
OF MONTHLY PREDICTION
CASES
Nov. -Dec. , 1976
Dec., 1976
-Jan., 1977
Jan. -Feb. ,1977
Feb. -Mar. ,1977
j Nov. -Dec. ,1982
Dec. ,1982
-Jan. ,1983
Jan. -Feb., 1983
Feb. -Mar. ,1983
AVERAGE
stants for the j
A
T'
0.42
0.41
0.56
0.59
0.60
0.48
0.44
0.62
0.52
I
B
0.39
0.19
0.25
0.31
0.47
0.56
0.51
0.56
0.41
predict:ion ar
700 1: iPa
A
0.28
0.48
0.61
0.27
0.26
0.27
0.17
0.51
0.35
B
-0. 15
0 . 38
0.47
-0.14
0.20
0.35
0.61
0.44
0.27
500 1iPa
id B is3 the I>ersis1:ences
A
0.31
0.58
0.70
0.32
0.34
0.23
0.40
0.47
0.42
B
-0.10
0.01
0.53
-0.12
0.42
0.12
0.58
0.22
0.21
1
A
300 hPa
fl "
0.27
0.64
-0.17
0.28
0.66 0.42
0.38 0.15
0.16 0.15
0.33 0.38
0.21 0.36
0.53 0.46
0.40 0.25
TABLE 2. COMPARISONS OF THE CORRELATION BETWEEN OBSERVATIONS
AND SEASIONAL PREDICTIONS OR PERSISTEN
CASES
Nov. ,1976
-Feb., 1977
Dec. ,1976
-Mar. ,1983
Nov., 1982
-Feb., 1983
Dec. ,1982
-Mar. ,1983
A
0.46
0.36
0.15
0.45
*}["
i
B
0.39
0.02
0.15
0.44
700 * iPa
A
0.58
0.16
-0.03
0.36
B
-0.17
-0.10
0.18
0.31
500
A
0.74
0.29
-0.09
0.40
hPa
B
-0.13
-0.05
0.06
0.39
300
A
0.76
0.34
-0.09
0.48
hPa
B
-0.03
0.07
-0.27
0.41
AVERAGE 0.36| 0.25| 0.28|0.06| 0.34] 0.07J 0.38] 0.05

361
Pig. 1. Anomalous fields of the earth's surface
temperature, March 1977. Chart (a) 10 the Monthly
prediction, (b) is the observation.
Fig. 2. As in Fig.l except for 500 hPa GPH

362
Fig. 3. Map of the correlation coefficients when SSTA in Fig. 4, The atmospheric response at 500 hPa to the
the region of 83'E - 91*E, 1S*N - 0*- lags the 500 hPa GJH heating sourc© in the region of 83*E - 91*E, 15'N - 5*N.
: ,wrt « of fl3"£ - 91*
2*N and 135*W - 125

363
Fig. a. Ac the coupling coefficient* of ACS a*b - 5 X 10
/* , a) the dipersion relationship between the real part
of wave frequency and wave length, b) the amplifying (or
decaying } rate*, solid (dashed)lines represent the wave
frequencies a* »,n-0,2 (»,n»l).
Fig. 7. The observational 500 hPa anomalous height field
for January, 1973.
. .
ight of the model ocean and atmosphere, respectively.

364
THE TELECONNECTION OF EQUATORIAL SST IN TAIWAN AREA
Cho-Teng Liu
Institute of Oceanography
National Taiwan University
Taipei, China
ABSTRACT
The oceanographic phenomena (like eddies) usually have much
larger time scale and much smaller spatial scale compared to the
atmospheric phenomena. This is not necessarily the case for the scales
of circulation of coupled atmosphere and ocean. Before an El Nino and
Southern Oscillation (ENSO) event, there is a gradual build up of warm
surface water in the western tropical Pacific along with the
abnormally high trade wind. The rain rate, river run-off and the sea
level in Taiwan are also abnormally high during this event. Since part
of the western boundary current Kuroshio, comes from the warm surface
water in the western tropical Pacific, both the SST and the sea level
around Taiwan will also increase when the Kuroshio carries this warm
water to the Taiwan area. The squared correlations are 231 between the
hindcasted anomaly of equatorial annual mean SST (HSST) and the
observed SST anomaly, 11% between the HSST and the KeeLung sea level,
40% between the HSST and the Naha sea level, and 21% between the HSST
and the KaoPing river run-off. The varying degrees of correlation may
be interpreted as following: since Naha is on the warm side of the
western boundary current Kuroshio, both the anomalous high HSST and
the Naha sea level represent the thickening of warm surface layer, or
the increase of heat content over the tropical and subtropical western
Pacific. The sea level over KeeLung which is on the cool side of
Kuroshio, is more related to the sea level of East China Sea and
therefore shows the lowest correlation with the HSST. The medium
correlations for both the KaoPing run-off and the observed equatorial
SST may be the result of high intrinsic variabilities of localized
wind pattern and ocean current.

365
A THREE-DIMENSIONAL MODEL OF THE CHINA SEAS AND
ASPECTS OF TYPHOON SURGE PREDICTIONS
S. K. Liu 1 , P. Wu 2 ,T. Y. Wu 3 , and S. T.Wang 3
ABSTRACT
This paper describes the numerical modeling work of the China Seas.
The area covered by the model is shown in Fig. 1. Within the large model
which has a I/I by 1/1 degree grid network, there are two levels of
submodels nested within the modeling domain. The regional submodels have
spatial resolutions between 1/8 degree and 1/24 degree. These submodels
obtain their open boundary conditions from the China Sea model.
majority of these regional models are three-dimensional, baroclinic, and have
five layers. For several coastal and esturaine systems, there are
two-dimensional models with variable grid network. Their open boundary
points are connected to the regional models.
The primary purpose of the modeling work is for ocean engineering and
coastal planing.
The
One recent application of the model is to predict coastal
storm surge due to typhoon and astronomical tides. Due to the nonlinear
dynamics associated with the coastal currents, the abnormal water level due
to the com bined action of typhoon and tides cannot be computed separate ly.
This is particularly true in coastal areas with strong tidal currents and
overtides induced by bottom dissipation or geometry.
The paper also discusses several operational aspects associated with the
forecast and areas for improvements.
1 RAND Corp. USA, and Consultant, Science & Tech. Advisory Group.
2 Science and Technology Advisory Group, the Executive Yuan.
3 Central Weather Bureau.

366
3-*-j-] dag grid
% ^Showing every other grid)
Tidal amplitudes in meters
f /
:
^
Tidal phases in degrees GMT
Fig. 1 - Oceanic areas covered by models of various resolutions

367
HYDRODYNAMIC MODELING
The hydrodynamic model used for the study is formulated according to
the equations of motion for water, continuity, state, the balance of heat, salt,
pollutant, and turbulent energy densities, on a three-dimensional finite
difference grid. The governing equations of the model are as follows:
dr
da. , frr
dz
5s

368
oBE
(6)
fl(uP) , fl(vP)
... 0(wP
#y
= o (7)
Using the standard finite difference notation on a horizontal C-grid
central in space and time, the model equations are:
'
Mean sea level
-*-. •
I VI I
i i
_ .+ _ +
ati,j,n (8)
at i+l/2,j,k,E (9)
where £ represents free surface, h, i, j, k, n denote layer thickness x, y, z

369
grid, and time level respectively.
To be brief, here we only present the computation of free surface by
continuity equation and the momentum equation in the x direction.
Equations for y-momentum, salinity (s), temperature (T), SGS energy (e),
pollutant (P) are similar.
In the present model, the horizontal grid conforms to the earth's
ellipsodial coordinates and the arbitrary vertical grid spacing approximates
the bottom topography of the modeled area. The results are subsequently
transformed into the Universal Mercator projection for graphical
representation.
The vertical momentum, mass, heat, and turbulent energy
closures are computed implicitly. In the horizontal direction, the exchange
coefficients are computed in two parts as a function of the local vorticity
gradient and the local grid dimension.
The second part represents the
contribution from the homogeneous subgrid scale turbulence above the
spatial Nyquist frequency, which is computed according to the turbulence
spectrum theory (i.e.,proportional to the four-third power of the gridsize).
The model detail can be found in Liu (1983, 1985, 1988) and Liu and
Leendertse (1978, 1987).
The-five-layer China Sea model covers the area from 110°E to 129°E,
and from 19°N to 45°N with a 1/4 x 1/4 degree grid resolution (Fig. 1). For
the Southern China Sea area, a 1/8 x 1/8 degree model covers up to 29°N.
The areas surounding Taiwan are covered by a high resolution model of 1/24
by 1/24 degree resolution.
To establish the initial and boundary data bank for these models, a
substantial effort has been undertaken since the summer of 1980. These data
include depth, salinity, monthly temperature, and tidal prediction
coefficients at the models open boundaries.
made for the China Sea model.
After adjustments have been
It is then used for prescribing the open
boundary conditions for the nested models. In the process, interpolations
were made in the frequency domain by the cross-spectral analysis and
transfer function technique. These statistical methods were applied because
field data usually contain errors.
After the initial adjustment phase, the
model gives reasonably good results in making astronomical tidal
predictions The model has also been used for generalting co-tidal charts for

370
the major tidal components over the China Sea region (Liu, 1983, 1988) like
the one shown in Fig. 1 for the Mg component. Some tidal prediction results
shown in Fig. 2 were made independently by the Naval Hydrographic Office.
In many occasions numerical tidal models underestimate tidal level for two
reasons. First, models are driven at the open boundaries by fewer tidal
components than there are in nature. Secondly, some meteorological
disturbances existed during the verification period which were not included
in the simulation. It is also quite costly to collect tide data for model
verification purposes. Many bottom pressure recorders were lost due to
fishing activities.
TYPHOON SURGE PREDICTIONS
Numerical models described in this paper are developed primarily for
ocean engineering and coastal planning purposes. One such application is to
make typhoon surge predictions. The oceanic areas covered by the model are
generally larger than the local Rossby radius of deformation which is a
function of latitude and depth. For the storm surge computation, we use a
parametric typhoon model basied on gradient wind equation at this time.
NWP model will eventually be used for this purpose. The parametric model
contains marine boundary layer computations which includes air-sea
temperature differencs. Correction for sea surface wind speed/direction from
the computed theoretical gradient wind were based on data collected mainly
from extratropical cyclones over the North Sea and Bering Sea, (Liu and
Leendertse, 1987). An additional veering angle of 5 degrees was added based
on the theoretical considerations for applications in the tropics. Additional
studies are needed in this respect. Orographic effects on typhoon due to
Taiwan's central mountain range are corrected according to data analyses
made by the Central Weather Bureau (Wang, 1980), Further research is
also required in this area, however.
Fig. 3 shows comparison between the computed and the observed water
level at two locations during the passage of typhoon Trix (August 7, 1960).
The maximum effects of typhoon over the normal tide level is approximately

Model verification with only astronomical tides. 1-6, December 1986
T~
KATER LEVEL AT TANSUI RIVER ENTRANCE
Obs/
Obs.
Hwa Li an
Tansui river entrance (obs.)
Tansui coast (comp.)
Taichung
TRIX
Obs.
Pon Hu
Fig. 3 astronomical tides and
storm-induced surge heights.
Obs.
Suo
Keelung
Comparisons between predicted and the observed water levels at five
commercial harbors using the numerical model.

372
one meter. Most of the water level gauges are located within a harbor or on
bridge pier in a river system. In the first case, the computed water levels are
usually higher than the harbor water level due to damping effects of the
harbor mouth. In the second case, the computed surge level is usually lower
than those observed in a river due to the effects of flood stage. These factors
have to be considered during the model adjustment stage.
Fig. 4 shows the distribution of water level induced by a typical typhoon
toward 300°T at 5.1 m/s with a central pressure of 965 mb. The maximum
positive surge (157 cm) is located near the Tansui River entrance while the
maximum negative surge (-70 cm) is located near Keelung. With the model,
we have made 1300 simulations using typhoons of various strengths and
directions passing through Taiwan. For each coastal area, surge heights are
being analyzed for engineering design and planning purposes. For example,
near Taipei, severe surge heights are ranked and tabulated on top of Fig. 5.
The lower diagram in Fig. 5 shows the storm track index map with
probability roses for typhoon's movements within each 1° by 1° square
according to past data.
In the south-western part of Taiwan, the elevation of many coastal ares
are very low. During the high tide conditions, even a small typhoon passing
through the Phillipines may cause floodings near the SW coast. The writers
believe that those areas are located off deep waters of relatively low latitude.
The large Rossby radius of deformation (700-900 km) when combined with
the coast-trapped waves made the problem worse.
In our prediction operation, improvements are needed in four major areas,
namely, better typhoon track forecast (e.g., see Tsay 1982), higher coastal
resolution, more verification/adjusment, and a better wind field model.
ACKNOWLEDGMENTS
The execution of this study would not have been possible without the
help of many colleagues. They are: Drs. S. Y. Chen, L. H. Chang, -L. H.
Fong,C. Y'Kwqh, C. K, Li, T, C. Li, F. C. Liu, W, M. Liu, from the Central
Weather Bureau; Messrs. Y. C. Hwang, W, Lin, N. H. Ma, from the Science

Rank
DESIGN TYPHOON SURGE HEIGHT NEAR TAN SU1 COAST
Surge Height in cm
1
*^K
1 ^
*«•-»
^f
v:-.,
^VM -" «
*
•5
- 4 Water level ia cm from normal tide
i" l '»' = "»s l " j"»
iV'"'*V "*'
E
, ^
"'"«'» i
Storm track index map for lyphoon surge forecast. Probability roses within
each 1 deg x 1 deg grid are derived by CWB data.

374
and Technology Advisory Group, the Executive Yuan; Drs. C. G. Tu and G.
N. Yao, from the Naval Hydrographic Office; Drs. C. S. Chen, W. S.
Chuang, C. Y. Li, N. K. Liang, C. T. Liu, C. Y. Tsay, J. Wang, from the
National Taiwan University; Dr. L. W. Tang, National Cheng Rung
University; Dr. F. Yin, National Culture Univ. ; Dr. EL W. Li, National
College of Oceanography; Drs S. C. Chow, W. G. Chen, H. S. Sou, Ministry
of Transportation, and Mr. Y. H. Wu from the Chung San Institute of
Technology.
REFERENCES
Liu, S. K. (1983, 1988): A Three-dimensional Model of the China Seas,
Edition I, 1983, Edition II, 1988, Pub. by the Science and Technology
Advisory Group, the Executive Yuan.
Liu, S. K. (1985-1988): A Typhoon-surge prediction Model of Taiwan and
SE China Sea. VoL 1-4, Central Weather Bureau.
Liu, S. K. and J. J. Leendertse (1978): Multidimensional Numerical
Modeling of Esturies and Coastal Seas, in ADVANCES IN
HYDROSCIENCESII, Academic Press, New York.
Liu, S. K. and J. J. Leendertse (1987): Modeling the Alaskan Continental
Shelf Waters, RAND-3567-NQAA/RC.
Liu, S. K. and J. J. Leendertse (1987): A Three-dimensional Model of the
Gulf of Alaska, COASTAL ENGINEERING, VoL 21, American Society of
Civil Engineers, New York.
Tsay, C. Y. (1982): Numerical prediction of typhoon tracks, in Methods of
Typhoon Forecasting in Taiwan, ROC, Sept. 1982
Wang, S.T. (1980): Prediction of the behavior and strength of typhoons in
Taiwan and its vincinity, T. R. 18, NSC-67M0202-05(01).

INTERAiWAL VARIABILITIES OF THE WESTERN TROPICAL
PACIFIC OCEAN AND LOW-FREQUENCY RESPONSE
OF THE SUBTROPICAL HIGH OVER THE NORTHWEST PACIFIC OCEAN
375
Pu Shuzhen
Yu Huiling
First Institute of Oceanography, SOA, Qingdao, China
ABSTRACT
Computation and analysis of the interannnal variabilitiesof the area,
strength,and westward extention of the snbtropical High over the Northwest
Pacific Ocean (NWPQ) are made to show that the variabilities, with a phase
lag, are correlated to the El Nino events. In order to explain the
correlation between the two, data obtained by the Sino-Acierican bilateral
TOGA Program are used to find scientific evidences for the interannnal
variabilities of the thermal structure and circulation in the upper Western
Tropical Pacific Ocean (WTPQ) and to discuss their effects on the subtropical
High over the NWPO, It is found that the WTPO features daring the 1986/1987
El
Nino event are obviously different from those before or after the event.
The main variabilities in the WTPO during the El Nino event, which lay
affect the snbtropical High over the NWPO, are as follows: (1) east-west
spreading along the equator of the warmer water in the waning pool; (2)
decrease of the northward transportation of the warmer water from the
area (south of 18° N, west of 130° E) where the Inroshio originates, The
former could anomalously heat the atmosphere over the Western Equatorial
Pacific Ocean(WEPO),and the latter could decrease the heat content in the
upper water under the subtropical High-over theJWPO, The both will drive
the
and
Hadley circulation in the related longitudes running faster than normal
strenghen the downward motion in the upper air and the divergence in the
lower level of the subtropical High over the NWPQ*Therefore, the subtropical
High over theJfWPO becomes stronger and larger than normal.

376
INTRODUCTION
In 1%6,
Bjerknes, using bathythermograph sounding data in the longitude
sector 140° —150° W (Sep., Oct., Dec. 1955 and Dec. 1957), studied the abundance
of warm water in the Central Pacific Ocean during the 1957/1958 El Nino event,
and suggested that the wanning could make the Hadley circulation run faster than
normal in the affected longitude sector and transport absolute angular momentum
to the subtropical jet stream at a faster rate than normal. The continued
poleward and downward flux of absolute angular momentum in the belt of the
surface westerlies could then be assumed to maintain stronger than normal in the
middle latitudes of the longitude sector. He also pointed out that the
subtropical High over the Eastern Pacific Ocean was farther south in the
1957/1958 winter than normal . Since that time, the correlation between
tropical ocean and global atmosphere has attracted more attention of
meteorologists and oceanographers. But those papers did not work out the
quantitative correlation by means of statistic theory and did not touch the
mechanism of the relations. The authors in the present article intend to discuss
the frequency responses among the indices of the subtropical High and El Nino,
and suggests an assumption for explaining the interannual variabilities of the
subtropical High over the NWPQ. The air-sea Interaction reasoning for the
Interannual variabilities of the subtropical High is based on evidences for
showing interannual variabilities of thermal structure and currents in the upper
WTPQ. It is assumed that the WTPO can affect the subtropical High over the IWO
more directly and essentially than the Eastern Tropical Pacific Ocean (ETPO) ,
and the correlation between the subtropical High over the N$PO and the ETPO Sea
Surface Temperature(SST) Is only appearance of the things*
LOW-FREQIBICY VARIABILITIES OF THE SUBTROPICAL HIGH OVER THE NWPO
In order tho describe the low-frequency variabilities of the subtropical High
over the N^PO, three indices, I.e. area, strength and westward extentlon of it,
are defined and the time series of the three indices are obtained (At a l month).
The area Index means the number of the grid points'(5°- lat.xiO 0 long.) enclosed
by the 588 isohypse on the 500 inb chart over the NWPO (110° E—180 0 . north of

10° N), the strength index is the sum of the differences of the geopotential
heights at the grid points from 5880 geopotential meters, and the westward
extention index is the longitude of the crossing point where the 588 isohypse
meets with the west end of the subtropical High axis. 1 " 1 ' 1 .
Fig.l
377
is the tirae series of the area index of the subtropical High, which is
the results after the 12-month moving averages for filtering the seasonal
variability from the monthly means. Fig.2 and Fig.3 are the corresponding time
series of the strength and westward extention worked out by the same
processing
as Fig.l.The obvious interannual variabilities can be seen from Figs, 1,2, and 3.
The authors have chosen the monthly means of the cold tongue Sea Surface
Temperature Anomalies (SSTA) in the ETPO as an El Nino index, and obtained its
r«yn
time series " . The period around the peak spectral value of the time series of
the
the
El Nino index is estimated about 3.5-4 years 1 " 31 * The corresponding period of
time series about the area'of the subtropical High is 0.023 month"" 1 , and the
other two periods corresponding to the strength of the subtropical High and its
westward extention are respectively 0.024 month' 1 and 0.023 month" 1 , which are
equivalent to about 3.6 years. It is shown that the four time series have the
similar period or frequency for their components with the greatest amplitude.
In addition, the computed coherence functions
of the four time series have
values identically close to 1 at the frequency of the peak spectral value
(0.023-0.024 month" 1 ) between each pair of the four time series. In other words,
they are almost perfectly coherent. However coherence among them at other
frequences is
less or much less than that about 0.023roonth" \ It is shown by
computing the frequency response functions that the variabilities of the westward
extention,
strength, and area lag from about 3 to 4 months behind the El Nino
index, in the given order mentioned above. The lag of 3-4 months might correspond
to the time which is taken for the Hadley circulation affected by the
above-normal equatorial heating to bring the absolute angular momentum to the
subtropical
High belt or the middle latitude westerly belt and to strengthen the
subtropical High in the longitudes under consideration.
r-1
The definitions are identical with those by the long-range
forecasting division of State Meteorological Administration, and the
data are kindly 'offered'by' the division

378
THE INTERANNUAL VARIABILITIES OF THE THERMAL STRUCTURE IN THE UPPER WTPO
After understanding the variabilities about the subtropical High over the
NWPO described in the above section, we are going to put forward a new question:
with more than one million kilometres away from the each other why the NwPO
subtropical High can correlate to the ETPO SST . In other words, what is the
mechanism for the remote correlation
It is common knowledge that El Nino phenomena are not a problem in isolation,
and related hot only the ETPO SST wanning but also a series of changings in the
air-sea coupling system. When an El Nino event occurs, the response of the
WTPO is apparent, with different features from the ETPO response .
Fig .4 is the equatorial sections of the sea temperature obtained during
Cruises 1-5 of the Sino-American TOGA program. We can see from Fig.4 that a large
quantity of warm water piled up at the warming pool in Jan. and Feb, 1986, the
months before the 1986/1987 El Nino, whth water warmer than 29° C spanning about
30° longitudes (from 145° E to 175° E) and more than 100 meters thick at the
deepest area( see Fig. 4(a)). The warm water in the warming pool spreaded further
westward and the thermocline became shallower with 29° C isotherm near 60-70
meters deep in Nov*~Dec. 1986, the months after the El Nino event had occurred.
The warmer water in the warming pool kept going much shallower, with the
shallowest mixing depth at 20-30 meters, the isotherms higher than 20° C rising,
and the water warmer than 29° C spreading further eastward, in Oct, 1987(the
3-rd Cruise), It is the time when the El Nino event was coming to its and. The
equatorial warm water seems to have started building up again and the thickness
of the top mixing layer had reached the 50m depth in May and June 1988. The
thickness for the piled warmer water was getting more than Cruise 3, with the
28° C isotherm below the 100m depth and the 25° C reaching the 150m depth,
showing some La Nina features* All those facts prove that variabilities of the
warming pool in the WTPO are important and significant for the heat energy to
re-distribute in the upper equatorial area during an El Nino event. Some El Nino
features in the WTPO are as follows' (!) the variance of the WTPO SST is the
minimum on the whole section, while the vaiance of sea temperature near the
thermocline is the maximum; (2) the cold water upwelling strengthens at the
thermoclines of the WEPO because the upper warm water spreads east-west; (3) the

warm water of the upper WEPO west of 141.5° E would keep higher than 29° C, a
period after the El Nino event had ended, and the therroodynamic effects
presumably make the Walker Circulation shift hack to its normal condition or a
La Nina condition; (4) one year after the El Nino had ended, the depth of the
warming pool had not reached the depth before the El Nino event,i.e. it will, not
lie built up within one year, which probably explain why the shortest interval
between any two successive El Nino events seems to be at least two years.
379
INTERANNUAL VARIABILITIES OF THE CIRCULATION IN THE UPPER OTPO
Not only the interannual variabilities of the thermal structure of the upper
JvTPQ but also the variabilities of its circulation can affect the subtropical
High over the NWPO, because the current is a causative factor for transporting
heat energy in the ocean. Actually, their interannaul variabilities have
attracted attention of oceanographers E4 " .
Fig.5 is the plots of the acoustic Doppler current profiling (ADCP) at
165° E from the 2nd cruise to the 4th cruise. In Dec. 1986, i.e. the early
stage of the El Nino event, eastward flow can be seen everywhere in the whole
section, showing above-normal heat transpotation to the east part of the ocean.
The Equatorial Under Current (EUC) surfaced and strengthened. Both the North
Equatorial Countercurrent (NECC) and the South Equatorial Countercurrent (SECC)
became stronger than normal. In Oct. 1987, when the El Nino event was coming to
its end, the NECC and the SECC weakened, the RJC almost disappeared, the South
Equatorial Current (SEC) re-eastablished but did not reach its normal speed, and
the shear among the currents was weaker than normal. In May 1988, the months
after the El Nino event had ended, the EUC and the SEC accelerated, both the SECC
and the NECC shrinked. It was the time when the warming pool was rebuilding.
Fig.6 is the ADCP plots along 18° 20*N section obtained in the same cruises
as in Fig.5. We can see from Fig.6(a) that northward current prevails except a
few patches of very weak southward currents. There might be two cores of the
northward flow. One is located between 127° £—128° E, with a highest speed of
about 40 cm/sec, and the other, i.e. the Kuroshio current near the Phillippine
coast, for some reasons,the maximum speed of the later is beyond the range of
fig.6(a).
In September 1987 (near the end of this El Nino event), the northward

380
current between 127° E—128° E became very slow
with the highest speed less than
20 on/sec although some data about the Kuroshio Current near the Philippine
coast was missed, while the southward flow was expanded and more obvious. The
deflection of the current directions (see Fig.6(b)) shows somethihg like an eddy
centered at 127° E. In April 1988, the northward flow (the Kuroshio current)
became stronger again, with a highest speed more than 80 cm/sec near the
Phillippine coast, and a much weaker southward current can be seen to the east of
124.5° E(see Fig. 6(c)). Therefore, the western boundary current at 18.3° N
section appears interannual variabilities related to the El Nino event.
Recently, Toole et.al have utilized the inverse method to work out the
variations of the transportation in a box of 8° N~18° 20'N, west of 130° E
between Cruise 3 and Cruise 4, based on the CTD data. Their result shows an
northward
transportation across the 18° 20*N section was 12.3xl0 9 kg/sec in the
top layer above a e -26.25 in the 3rd cruise and 30.6Xl0 9 kg/sec in the 4th
cruise
t61 *)
. Toole also estimates the annual mean transport of 15.7X10 kg/sec
crossing the section 18° 20*N for water warmer than 12° C potential temperature
(closely corresponding to the top layer in Reference 6) ''. Therefore, the result
estimated from the CTD data also show the interannual variabilities of the
northward transportation on the 18° 20*N section, similar to that proved by the
ADCP measurements.
We will assume, based on the analysis discussed above, that the northward
transportation of the heat energy from the area equatorward of 18° 20 *N and west
of 130° E significantly decreased in the period from the mature stage to about
the end of an El Nino event. The decrease of the heat transportation would result
in less heat content of the upper NWPO north of 18° "20% even further to the
middle latitudes where the kuroshio flows, with a certain time-lag or phase-lag.
Therefore, the upward heat flux from the NWPO to the atmosphere in the
subtropical High might ..be less than the normal conditions, in a period after an
El Nino event had started* It is the reason why the subtropical High over the
NWQ can be getting stronger and larger than normal after El Nino events have
started.
CONCLUSION
The above discussion based on the data related to the subtropical High over

the NWPO and the data obtained from the Sino-American bilateral TOGA program
seems to allow to draw the following conclusions'
(1) the strength, area and westward extension of the subtropical High over
the NWPO show interannual variabilities obviously related to El Nino events in
the equatorial ocean, and with a time-lag about 3~4 months;
(2) During the 1986/1987 El Nino event, warmer water in the wanning pool in
the WTPO spreads more east-westward, making the Hadley Circulation over the WTPO
run faster. The speeded Hadley Circulation then brings absolute angular momentum
to the poleward area from the equator at a higher rate than normal and then
downward to the surface north of the equatorial easterlies. The effects are
reflected by the larger and stronger subtropical High over the NWPO;
(3) In an El Nino event ( at least in the last stage), the heat transported
by the western boundary current to the subtropical NWPO obviously decreases,
which then reduces the heat content of the upper NWPO where the boundary current
flows, the changing tends to strenthen both the downward motion and flow
divergence in the lower atmosphere of the subtropical High over the NWPO. This
could be the additional mechanism for the strengthening of the subtropical High
over the WTPO, makes it occuping more area, and extending further westward
during an El Nino event than normal.
381
AdNOwIEDGEMENTS
The authors wish to take this chance to thank both SOA and NOAA which are
supporting the China and American bilateral TOGA program. We are grateful to Drs.
B.Taft 1 , M.Mcphaden 1 , L.Mangum 1 , J.Tooie 2 , KX.Miilard 2 and E.Firing 3 and other
U.S. col leagues, and Z.Wang 4 , Ou 4 , B.Li 4 and other Chinese colleagues of
for their scientific contribution to the program.
1. Pacific Marine Environmental Laboratory, Seattle, Washington U.S.A.
2. Woods Hole Oceanographlc Institution, Woods Hole, Massachusetle, U.S.A.
3. JIMAR, University of Hawaii, Honolulu, U.S.A.
4. First Institute of Oceanography, Qingdao, China.
ours
REFERENCE
(1). Bjerknes, J., A Possible Response of the Atmospheric Hadley Circulation to
Equatorial Anomalies of Ocean Temperature, Tellus, 18(1966),820-829.
(2). Pu Shuzhen and Yu Bulling, Characteristics of Frequency Response Between

382
El Nino and Southern Oscillation, Marine Science Bulletine., Vol.6, N'o.2, June
1987, 1-5 (in Chinese),
(3). Pu Shuzhen and Xu Hongcla, Responses of the WTPO Temperature to El N'ino
Events, ACTA OCEAXOLOGICA SINICA, Vol.9, N'o.2, March 1987, 262-266 (in Chinese):
(4). Firing, E., R. Lukas, J.Sadler and K. Wyrtki: Equatoial Undercurrent.
Disappears during 1982/1983 El Nino, Science, 9 December 1983, Vol. 222
1121-1123.
(5) Toole, J.M., E.zou and R.C.Millard, On the Circulation of the Upper Waters
in the Western Equatorial Pacific Ocean, Deep Sea Research, Vol. 35, !Vo.9,
1988, 1451-1482.
(6) Toole ,J.M.,R.C.Millard, Z.Wang and S.Pu, Observations of the Pacific N'orth
Equatorial Current Bifurcations at the Philippine Coast ( being submitted).
Fig.l. The time series of the area index of the subtropical High on the 500ml)
chart ( the top).
Fig.2. The time series of the strength index of the subtropical High on the
500mb chart (the middle).
Fig,3. The time series of the westward extention of the, subtropical High on
the same chart as in Fig.l. (the bottum).

383
UO V E ISO' 160 170' 18CT
0
15U" IT 170'E
100 -
2 00-:
300
400 -
500
141TE 150' 160'' 140 J E ISO* 160' H.tJ'E
U
\J
100
200
300
400
500
Fig,4.
(a) The sea temperature section along the equator in the 12th Jan.—10th
Feb. 1986, Cruise 1. of the China-US TOGA program; (b) the same as
Fig.4(a). exception in the 30th Nov.-the 11th Dec. 1986. Cruise 2; (c) the
same section in the 3rd-the 16th Oct. 1987, Cruise 3; (d) the same
section in the 5th-the 20th May 1988, Cruise.4J (e) the same section in
the 28th Oct.-the 13th Nov.1988, Cruise 5.

384
Fig.5. The ADCP measurements (see the text) along 165° £ section .The shaded area
Is for the westward flows and with negative values and the isotaches are
in units of cm/sec ( primarily calibrated and plotted by the group of E.
Firing), (a)—the 2nd cruise; (b)— the 3rd cruise; (c)--the 4th cruise.
(m)
124E 12BE 121C 1JT4C|J«
Fig;6. The ADCP measurements along 18° 20*N.The shaded area is for the southward
flows and with negative values in units of era/sec, (primarily
calibrated and plotted by E. Firing), (a)—the 2nd cruise; (b) the 3rd
cruise; (c)~the 4th cruise.

385
Large Scale Air-Sea Interaction in the Western Pacific
Region
Author: Joseph C. K. Huang, formerly Program Manager in the
Office of Climate Research, NOAA, U.S.A. and U.S.
Executive Secretary for U.S.-PRC Bilateral Program
on Air-Sea Interaction Studies in the Western
Pacific
Affiliation: National Ocean Service
National Oceanic and Atmospheric Administration
1825 Connecticut Avenue, N. W.
Washington, D.C. 20235
Abstract
Recent changes in climate, especially the severe droughts
over the last few year together with the related environmental
phenomena, such as the greenhouse effect, the ozone depression,
etc., have awaken mankind to pay specific attention to
understand the physical processes involved in the climate
system and to make increasing efforts to predict the climate
conditions for the future. There are several national and
international climate research programs planned and/or in
progress. Notably, one is the Tropical Ocean and Global
Atmospheric programs (TOGA), a large scale, air-sea interaction
program under the World Climate Research Program sponsored
by WMO, ICSU, and IOC. The TOGA program, started on
January 1, 1985 will last for at least 10 years. It has made
substantial scientific progress in the understanding of
important physical processes and modeling of the coupled
ocean-atmosphere system. However, there are still many gaps
of knowledge regarding processes within the warm pool region
in the Western Pacific Ocean and the role these processes may
have in modeling the ocean atmosphere system. In order to
gain greater insights into the physical nature of the air-sea
interactions in this region, the TOGA Coupled Ocean Atmosphere
Response Experiment (TOGA COARE) is designed. Major physical
processes related to air-sea interactions over the warm pool
water in the Western Pacific will be intensely studied and
climate modeling activities will be conducted in parallel with
the planning and implementing of the field experiment. the
success of TOGA COARE will hopefully improve the modeling of
climatic events.

386
I. Introduction
Climate is one of the basic factors in human life that
strongly affects the social and economic fabric of societies.
Recent changes in climate, especially the severe droughts over
most of the world in the last few years, have made us pay
specific attention to the earth environment and made us want
to increase our understanding of the causes of climate change.
Accurate predictions on climatic conditions are most valuable
for making intelligent decisions about how to overcome the
caused hardship and to make the best mitigatory adjustments.
Several international organizations such as the World
Meteorological Organization (WMO), the International Council
of Scientific Union (ICSU) and the Intergovernmental
Oceanographic Commission (IOC) are jointly encouraging the
industrialized nations to monitor, measure, and study the
earth environment as a single interactive system. Many
ongoing and planned programs such as the scientific research
program under the International Geosphere-Biosphere program
(IGBP), notably the Climate and Global Change Program (C&GC),
and the World Climate Research Program (WCRP), notably the
Tropic Ocean and Global Atmosphere Program (TOGA) and the
World Ocean Circulation Experiment (WOCE), are designed to
better understand and predict for the natural earth
environment on climate time scales.
The earth receives its energy from the sun. However, the
most obvious and dominating factor affecting climate changes
is the energy exchange between the oceans and the atmosphere.
Energy fluxes of momentum, heat, radiation, and water at the
air-sea interface drive the ocean circulation and ice
formation, and determine the coupling between the atmosphere
and the oceans. The ocean, mostly due to its large heat
capacity and slow motions, redistributes the energy, hence
plays a very important role in climatic changes. On climatic
time scales, i.e., greater than the weather prediction skill
of about two weeks, slowly varying conditions at the earth's
surface, especially the sea surface temperature, force the
climate system to respond accordingly and yield predictive
skill of a different type. Instead of being able to predict
the temperature and precipitation for a given day, we might be
able to predict the warmth or dryness of a given month, a
season to a few years in advance. At time scales from weeks
to years, only the upper layers of the tropic ocean need be
included because the atmosphere and the tropic oceans are
almost directly coupled. At a time scale of one hundred
years, the intermediate layers of the global ocean become
important. Similarly, deep water processes and ice processes
together with land processes become important as we move to
still longer time scales. As the time scale of interest

387
expands, more and more interactions in the various components
of the earth system become important. There is a hierarchy of
time scales leading to a hierarchy of complexities. It is a
strong challenge for earth scientists to understand the
various complexities of major physical processes that
contribute to the climate changes in the proper corresponding
time scales. How can we build these processes in climate
models to predict future climate changes One key step is to
understand the air-sea interactive processes that determine
the flux fields on the time scales of interest.
II.
The Western Pacific Ocean
The Pacific Ocean is our largest ocean basin. Observational
data and modeling results point to the Pacific Ocean as the
dominant basin in terms of driving the high-frequency side of
climate changes, i.e., in the TOGA El Nino/Southern
Oscillation (ENSO) signal domain. In addition to the huge
area and strong boundary current, the Kuroshio, there are many
physical features in the Tropic Pacific, especially in the
Western Pacific that are extremely important to climate
changes. These features are:
o
o
o
o
o
o
o
The huge warmest water pool in the world ocean;
The deepest thermoline in the tropical oceans;
A transient zone between the monsoon system and
the trade winds;
A region of strong atmospheric convections closely
correlated with a warm water pool;
A region of high variations of the Intertropic
Convergence Zone and the Southern Pacific Convergence
Zone;
The rendezvous area of the South Equatorial Current
and the North Equatorial Current and the Origin of the
North Equatorial Counter-Current, the South Equatorial
Counter-Current and Equatorial Under Current; and,
The source region of forcing, for equatorial large
scale waves, oceanic large scale, low frequency waves
along the equatorial wave guide.
III. The TOGA Program
The TOGA program is a major component of the
WCRP aimed

388
specifically at the modeling and predicting of climate
phenomena on time scales of months to years. It is a large
scale air-sea interaction climate program. Underlying the TOGA
program is the premise that the dynamic time scale of the
tropical oceans are far more rapid than that at higher
latitudes. Thus, disturbances emanating from the Western
Pacific Ocean may propagate to the eastern basin on the time
scale of weeks compared to years for corresponding propagation
at higher latitudes. The significance of the higher frequency
dynamic time scales in the tropical ocean is that they are
similar to those of highly energetic atmospheric modes. This
similarity allows the formation of coupled modes between the
ocean and the atmosphere in the tropics.
The specific goals and scientific objectives of the TOGA
program are (WCRP, 1985):
( i) To gain a description of the tropical oceans and the
global atmosphere as a time dependent system in order
to determine the extent to which the system is
predictable on time scales of months to years and to
understand the mechanisms and processes underlying its
predictability;
(ii) To study the feasibility of modeling the coupled
ocean-atmosphere system for the purpose of
predicting its variations on time scales of months to
years; and,
(iii)To provide the scientific background for
designing an observing and data transmission
systems or operational prediction, if this
capability is demonstrated by coupled oceanatmosphere
models.
The TOGA program began on January 1, 1985 and is planned
to last for at least ten years. From the TOGA large scale
monitoring and process studies, there has been significant
progress in identifying many of the important physical
processes driving the coupled ocean atmosphere system. Areas
where there appear to be substantial progress and
understanding include:
o
On ENSO time scales, the Pacific Ocean is
predominant. Oceanic disturbances appear to extend
from the western Pacific warm pool along the
equator to the east. The disturbed sea surface
temperature of the Pacific Ocean so alters the
large scale heating of the tropical atmosphere that
the basic circulation of the tropical atmosphere on
the largest scale is perturbed.

389
On ENSO time scales, the perturbed atmospheric
circulation drives changes in the sea surface
temperature in the Atlantic and the Indian Oceans.
While these changes are locally very significant,
they generally follow the Pacific variability and
are smaller in magnitude.
Two families of theories (instability and quasicyclic
equilibrium) of variability on the ENSO time
scale have emerged, both depend on the character of
the coupled system.
Several climate models have been developed and
model-based climate forecasts are being made.
These models are routinely initialized by data from
the TOGA monitoring program. Despite the
substantial progress gained through the TOGA
program, there is still considerable doubt
regarding which elementary physical processes
maintain the mean and transient state of the warm
pool regions of the western Pacific Ocean.
Processes and thought which need further study are:
- Atmospheric response in models has been shown to be
extremely sensitive to SST variations, especially when
the SST is warm. However, model assessments of ocean
structure almost universally predict temperatures which
are far too warm although the reason for these
discrepancies is not understood. Presumably, the
discrepancies are associated with poor assessments of
heat, momentum and water fluxes between the ocean and
atmosphere;
- The heat balance of the warm pool region of the
western Pacific Ocean is only known approximately with
discrepancies of the order of 80 win . The relative
involvement of the slowly evolving atmospheric flow
(such as the trade wind regime) or the higher
frequency, more episodic, equatorial events (such as
westerly bursts and organized convective phenomena) are
not understood;
- Theories of ENSO require the instability of, or
the oscillation about, the basic state of the
coupled system. Yet, the maintenance of that
basic state and why the temperature of the warm
pool lies between 28° and 30° C with a broad,
low gradient character is not well understood.

390
Even during the warm phase of the ENSO cycle, the
warm pool region varies by less than 1° C in the
western Pacific Ocean with major magnitude changes
occurring in the central and Eastern Pacific Ocean,
although these relatively small changes involves
critical influences in the atmosphere;
- The large scale time-mean atmospheric divergence
is collocated with sea surface temperatures in excess
of about 28° C together with the maximum time-mean
precipitation and centers of convective action (as
indicated by the variance of outgoing long wave
radiation). However, the reasons for these
collocations are not well known;
- Contribution to the long-term average divergence by
the shorter time scale variations (e.g., variations
on the 40-60 day and diurnal time scales) to the longterm
mean divergence is not understood although there
is considerable modeling and observational evidence
suggesting that there is interaction and organization
across a broad frequency band; and,
- There is substantial evidence that transitions in the
Pacific Ocean are accompanied, or even produced by
propagating eguatorially trapped planetary ocean modes.
However, it is not known whether these waves are part
of a long-term sequence of reflected waves, result from
inherent instability of the coupled system or are
forced by episodic high-frequency atmospheric forcing.
Indeed, it is not known whether all three processes may
be important at different times or are part of the same
mechanism.
It -is with these issues in mind that the TOGA Coupled Ocean-
Atmosphere Response Experiment has been proposed and
developed.
The TOGA COARE Program
The TOGA Coupled Ocean-Atmosphere Response Experiment (TOGA
COARE) is a regional process study designed to observe the
ocean-atmosphere structure of the warm pool in order to
understand both basin scale dynamics of the ocean and global
interannual climate variabilities* The scientific goals of
TOGA COARE (TOGA COARE Science Plan, 1989) are to describe and
understand:

I. The principal processes responsible for the coupling
of the ocean and the atmosphere in the western Pacific
warm pool system;
II. The principal atmospheric processes that organize
convection in the warm pool region;
III. The oceanic response to combined buoyancy and wind
stress forcing the western Pacific warm pool region;
and,
IV. The multiple scale interactions that extend the oceanic
and atmospheric influence of the western Pacific warm
pool system to other regions and vice versa.
The implementation of TOGA COARE is designed along three
components:
(A)
(A) An oceanographic component;
(B) An interface component; and,
(C) An atmospheric component.
The TOGA COARE Atmospheric Component
The primary goals of the atmospheric component of TOGA (SOTS
are:
(i)
To define the character of the large scale,
seasonally varying atmospheric circulation of the
western Pacific Ocean region, including the
detailed surface wind field and its relationship to
planetary scale phenomena;
(ii) To study the structure of the synoptic and mesocale
components of the large scale, slowly varying
atmospheric circulation in the warm pool. In
particular, determine the morphology of the most
convective stage of the 40-60 day mode and its subcomponents
;
(iii)To study the transition of the tropical planetary
boundary layer from the descending regions of the
eastern Pacific Ocean to the convective western
Pacific Ocean region and identify when the domain
shifts from easterly to westerly during, for
example, westerly wind bursts;
391

392
(iv) To investigate the morphology of the episodic
westerly burst phenomena of the western Pacific Ocean
region including their source whether in situ or
remote;
(v)
To study the vertical heating distributions and life
cycles associated with synoptic events and mesoscale
convective cluster components and their relationship
to the large scale flow of the tropics;
(vi) To describe the relationship of the phenomena
described in (i)-(v) to the heat moisture and
momentum fluxes at the ocean-atmosphere interface in
an attempt to relate the atmospheric phenomena of the
western Pacific Ocean region to transitions in the
dynamics and thermodynamics of the upper ocean
structure; and,
(vii)To determine the manner in which the motions
created by episodic and mean heating in the warm
pool regions transmit their signal to other
regions of the tropics and globe. Such
communications appear to occur at all phases of
the ENSO cycle.
(B) The TOGA COARE Oceanographic Component
The primary aims of the TOGA COARE oceanographic
component are:
(i) To estimate the magnitude of the vertical
mixing between the upper ocean and the cooler, more
saline, deeper ocean, and if necessary, find improved
means of parameterizing this mixing term;
(ii) To understand better the role of salinity
variations in the warm pool region and its
possible contribution to the ENSO mechanism;
(iii)To examine the role of western boundary dynamics
in the heat balance of the warm pool; and,
(iv) To gain a better understanding of the impact of
episodic atmospheric events such as the westerly wind
bursts on the surface fluxes and, in conjunction with
their impact on the heat budget of the upper ocean,

determine their influence on SST variations in the
western Pacific Ocean. In particular, map the spatial
and temporal variation of the barrier layer and
determine the processes by which it is modified and
maintained.
393
(C)
(i)
The TOGA COARE Ocean-Atmosphere Flux Component
The goals of the ocean-atmosphere flux component of
TOGA COARE are:
To quantify more adequately various empirical formulae
used to estimate net surface heat flux. Present
estimates of a real mean heat fluxes in the western
Pacific range from o to 100 Wm" 2 (Wyrtki, 1965, Weare
et al, 1981, Reed, 1985, Hsiung, 1985 among others);
(ii) To understand the impact of the full range of wind
structure, from the ambient trade wind regime through
the episodic westerly bursts, on the ocean-atmosphere
fluxes of heat moisture, radiation, and momentum; and,
(iii)To determine the fresh water flux between the
ocean and atmosphere.
Figure 1 is the proposed timeline of TOGA COARE which
establishes a timetable for implementing planned activities
in the elements of the program. The intensive observation
period (IOP) is tentatively planned for a four-month period,
November 1992 through February 1993. Prior to the IOP,
several pilot studies concerning key processes in the warm
pool region are proposed to be carried out in order to design
a well founded and optimal observational array for the IOP.
The proposed location of the TOGA COARE is shown in Figure 2.
It is located within the warm pool (the area with SST 28° C)
and in the region of maximum heating and convective activities
as well as maximum frequency of episodic wind events. The
area also straddles the equatorial wave guide. Table 1 shows
the list of the physical processes that will be studied in the
IOP. Note that the first three processes in the list are airsea
interactive fluxes processes. Details of the TOGA COARE
Composite program will be modified and finalized in the near
future.
IV.
Other Related International Climate Research Programs
There are several national and international climate
research programs closely related to the large scale air-sea
interactions. The World Ocean Circulation Experiment (WOCE)
has a strong air-sea flux component. The climate and Global

394
Change program under IGBP is a much expanded research program
closely associated with WCRP. Satellite programs such as the
Tropical Rainfall Measuring Mission (TRMM), International
Cloud Climatology program (ISCCP), Global Energy and Water
Experiment (GEWEX) and others will all contribute to the
formulation and measurement of interface fluxes in the airsea
interaction processes.
V. Summary
The primary scientific objective of all climate research
programs is to accurately predict climate variations in the
respective time scales of interest.When the time scales of
interest get longer, slower-varying processes have to be
included in the earth climate system. As evidence of progress
made in the TOGA program, large scale air-sea interactions are
the most prominent group of processes affecting climate
change. From the results obtained in TOGA studies, scientists
also have discovered a number of important gaps of knowledge
in the ocean-atmosphere coupled system, especially in the
interactive fluxes exchange between the two media in the warm
pool region over the Western Pacific, The TOGA COARE Program
is specifically designed to better understand of the
multiscale air-sea interactions in the warm pool region. The
success of this air-sea interaction study hopefully will lead
to better climate predictions, for the ENSO time scale in
particular.

395
References
Hsiung, J., 1985: Estimates of Global Meridional Heat
Transport, J. Phys. Oceanographer., 15, 1405-1413.
TOGA COARE Science Plan, 1989: A Coupled Ocean - Atmosphere
Response Experiment for the warm pool regions of the
Western Pacific UCAR publication, 111 pp.
Reed, R.K., 1985: An estimate of the climatological heat
fluxes over the Tropic Pacific Ocean, Clim. and Appl.
Metd., 24, 833-840.
World Climate Research Program, 1985: Scientific plan for the
Tropical Ocean Global Atmosphere Programme WCRP publication
#3, World Meteorological Organization, Geneva, 146 pp.
Weare, B.C., P.T. Stub, and M.D. Samuel, 1981: Annual means
surface heat fluxes in the tropical Pacific Ocean, J. Phys.
Oceanographer., 11, 705-717.
Wyrtki, K., 1965: The average annual heat balance of the
North Pacific Ocean and its relation to Ocean Circulation.
J. Geophys. Res., 70, 4547-4559.

396
TABLE 1:
PROCESSES TO BE STUDIED IN TOGA COARE
(ATM = atmosphere, OCE = Oceanography, INT = interface)
1.
2.
3.
4.
5.
6.
7.
PROCESS
MOISTURE FLUXES
(precipitation and evaporation)
MOMENTUM FLUXES
RADIATIVE AND SENSIBLE HEAT FLUXES
CONVECTIVE PROCESSES
OCEAN MIXING
HORIZONTAL ADVECTION
VERTICLE ADVECTION
(inferred from horizontal divergence)
IOP COMPONENT
ATM, INT
ATM, INT
INT, OCE
ATM
OCE
OCE, ATM
OCE
8. EQUATORIAL WAVE GENERATION AND PROPAGATION OCE, ATM

397
Modeling
Piiot Studies
Enhanced Monitoring
Intensive Observation Phase
Analysis
TOGA/COARE Schedule of Activities
Fig. 1: Time-line and planned schedule of activities for the TOGA COARE elements.

398
w / / / / / /
/ / /-"V / / Z
Fig.2: Location of the proposed TOGA COARJE domain (10'N and 1CTS, 14CTE and 180'E)
relative to the December, January and February (DJF, solid line, stippled area) and the June,
July and August (JJA, dashed line) warm pools defined by the 2S*C sea surface temperature
isotherms.

399
THE RELATIONSHIP BETWEEN CURRENTS AND WINDS
NORTHEAST OF TAIWAN
Wen-Ssn Chuang
Institute of Oceanography
National Taiwan University
Taipei, China
ABSTRACT
In the area northeast of Taiwan, the Kuroshio turns as it
encounters the continental shelf of the East China Sea. The turning
also triggers a vigorous exchange of water masses between the open
ocean and the marginal sea, which has been previously identified from
hydrographic surveys and satellite infrared images. Direct evidence
from current measurement, however, was rare. Because of this a
current meter mooring was placed at the shelf break (about 50 km from
the coast of Taiwan at water depth of 500 m) from September 9 to
November 21, 1988. Two current meters were attached to the moorings at
240 and 345 m below the surface.
There were quite significant velocity and temperature variations
in the three months the data covered. In September, the atmospheric
pressure was low, the winds were mild, and the current in both layers
were towards the south-southeast. In other words, there was a
continuous drainage of shelf water to the open ocean in the study
area. In October, the weather condition was characterized by frequent
tropical low (typhoon) passages, the currents fluctuated accordingly.
The dominant flow was still towards the ocean in both layers, but the
magnitude was reduced compared with that of September. The northeast
monsoon season began on November, and the study area was covered by
high pressure system most of the time. The currents at the 240-m depth
showed shoreward flow, while the currents at the 345-m stayed seaward.
Despite the air gradually cooled from September to November, the
water temperature increased during the same period. A sign which
suggests that either the path of the Kuroshio drew near the snelf, or
part of the Kuroshio water was brought to the shelf break. ^The
temperature difference between the 240 and 345-m depth remained
unchanged, in September and October, but increased quite significantly
in November. Together with the information provided by current
records, it is believed that the strength of the outflow from the East
China Sea weakened towards the end of the summer, and the whole
Kuroshio approaches In. In the early winter, the Kuroshio intrusion
occurs in the upper 250 meters or more. Conceivably, these open oceanmarginal
sea exchange processes are seasonally dependent and may be
closely related to the large scale atmospheric conditions.

400
CHINA-JAPAN JOINT RESEARCH PROGRAM ON THE KUROSHIO
Su JiIan
Second Institute of Oceanography
State Oceanic Administration
Hangzhou, Zhejiang, China
ABSTRACT: The China-Japan Research Program on the Kuroshio
is briefly reviewed. The collaborative aspect of this
program is commented upon. Finally/ the air-sea interaction
part of this program is described.
1. INTRODUCTION
On 28 June 1987 the State Oceanic Administration (SOA) of China and the
Science and Technology Agency (STA) of Japan signed an agreement in Beijing to
ratify the China-Japan Joint Research Program on the KurosMo (JRK). JRK is a
seven-year bilateral program which actually was launched in 1986. The program can
be extended by mutual agreement when it ends in March 19S3.
Oceanographers from both China and Japan have always paid special attention
to the Kuroshio which flows close to their countries. It is only natural for them
to pool their efforts together to study this strong current and its interaction
with the marginal seas along its left flank. On the initiative of SOA the
collaborative program was worked out between scientists fro* both countries in
1985. While waiting for the format approval by the governments both sides agreed
to initiate the program in 1986* the final year of the Japanese national program/
Kuroshio Exploitation and Utilization Research (KER). In China the Science and
Technology Department of SOA serves as the coordinator of JRK, while the Research
and Development Bureau of STA coordinates the participations in JRK of all
Japanese scientists. Subordinate departments and institutions of four Japanese
agencies are taking part in this joint study. The four agencies are,
respectively/ STA> Maritime Safety Agency/ Meteorological Agency and Fisheries
Agency. However/ so far in China only institutions from SOA, a total of seven/
are participating in JRK.

401
2. THE PROGRAM
The objectives of the program are divided into five aspects. The physical
aspect aims to study the Kuroshio-shelf water interaction, the origin of the
shelf warm currents ( Taiwan, Tsushima/ and Huanghai), and the seasonal and
short-term variability of the Kuroshio. The boilogical aspect concentrates on
the ocean productivity (both primary and secondary) and plankton distribution and
their relationships with the ocean environment, as well as on the plankton
compositions along the Kuroshio and other shelf currents. The chemical aspect
studies the distribution of various natural and anthropogenic substances in the
East China Sea and along the Kuroshio, the removal processes of these substances
from the water column, and the budget model of these substances. The air-sea
interaction aspects studies the transport of monentiua and energy across the sea
surface, the vertical structure of the atmospheric boundary layer over the ocean,
and the transformation of the air masses when they break out from the continent to
the East China Sea. Finally, JKR also conducts studies on the thermal energy
and kinetic energy in the Kuroshio for future expoitation.
As is well known, there were two large observation programs concerning the
Kuroshio, namely, the UNESCO sponsored 13-year (1965-197) Cooperative Study of
the Kuroshio and its Adjacent Regions (CSK) and the 10-year (1977-1986) Japanese
national program, KER, as mentioned earlier. CSK covered a wide area in the
Northwest Pacific west of 155° E and the marginal seas nearby, while KER paid
more attention to areas south of Japan. Because of its objectives the study area
of JRK included the western part of Huanghai Sea, the most part of the East China
Sea and large areas south of Japan.
3. EXTENT OF COLLABORATION
Exchange of scientists to visit laboratories and to board research vessels of
each other country has gone smoothly. Most of the observed data of the joint
program has also been exchanged. Annually an oceanographic atlas based on the
observed data is being published alternatively by each country. As to the
observation programs, JRK is, to large extent, similar to CSK, in which
participating countries run their own programs pretty much on their own without
close coordination with one another. One reason for the lack of concerted
observations stem from the fact that scientists from both sides hold* in such a
vast area with complicated hydrography/ different primary research interests. In
the past three years the Chinese scientists put more emphasis in two areas of the
East China Sea where the Kuroshio is thought to interact strongly with the shelf
water. The two areas are, respectively, northeast of Taiwan where the Kuroshio
enters the East China Sea and southwest of Kyushu where the Kuroshio banks sharply

402
eastward. In addition to maintaining monitoring sections south of Japan, the
Japanese oceanographers also pay close attention to both the area where the
Kuroshio exits the East China Sea and the shelf areas north and east of the
Changjiang river mouth.
The working group which coordinates this bilateral program has recognized the
deficiency of the research collaboration aspect and has taken steps to strengthen
it. In 1988 two joint research projects have been initiated, one on the Tsushima
current and the other on the circulation on both sides of the Ryukyus. This year
joint research projects have expanded to other disciplines/ including one on
micro- and meso-scale air-sea interaction to be conducted over the East China Sea.
4. AIR-SEA INTERACTION ASPECT
Since the start of JRK the Japan Meteorological Agency has been conducting
observations over the East China Sea to study, via indirect means, the energy and
momentum transfer across the sea surface. Vertical profiles of physical
characteristics of the maritime atmospheric boundary layer have also been measured
with tethered ballon and aerological observations. On the Chinese side, so far
research attention on air-sea interaction has been focussed on carbon dioxide and
aerosol measurements. Preliminary findings by the Chinese scientists are as
follows-. (1) Both the concentration and size spectrum of aerosol have apparent
seasonal characteristics, reflecting the difference in the transport of terrestial
substances under different weather systems; (2) The partial pressure of carbon
dioxide in maritime atmosphere was found to be around 350 ppmv, about the same as
the global average value. The corresponding value in ocean is in general higher
than the atmospheric value by 20 to 25 ppmv. There are targe variations in the
oceanic value, with a maxumum difference of more than 100 ppiv; (3) In the East
China Sea, the total inorganic carbon concentration in the Kuroshio water was
found to be around 1.8-2.0 lol/kg and the total inorganic alkali 2.2-2.4 mol/kg.
The winter values are higher than the summer values; (4) A one-dimensional air-sea
coupled boundary layer model which takes both convection and entrainment into
account shows promising agreement with results measured by the Japanese scientists.
As commented earlier, a joint resarch project on micro- and meso-scale
air-sea interaction will be conducted over the East China Sea later this year.
Tentative subthemes of this project are (1) characteristics of the maritime
atmospheric convective boundary layer and effects of cloud on this boundary layer,
(2) transfer of momentum and energy across the sea surface, and (3) mechanism of
air mass transformation.

403
INFLUENCE OF MICROSCALE AIR-SEA INTERACTION
ON CLIMATE RESEARCH
by
Jin Wu
Air-Sea Interaction Laboratory
College of Marine Studies, University of Delaware
Lewes, Delaware 19958
U. S. A.
Processes of microscale air-sea interaction are of interest to the
climate research in two Important, and also timely, aspects. One Is
associated with the coupling of atmospheric and oceanic systems, and
the other with the marine remote sensing. Throughout the years, the
atmospheric scientists have studied phenomena above the sea surface,
and have done well. For further improvements of their models, refined
inputs, however, are needed of fluxes across the air-sea interface,
which link dynamically the atmosphere and oceans. The same can be
said about oceanographers; they also need those fluxes for a further
upper grading of oceanic models. In other words, we have reached a
point to perform our research on one side of the sea surface alone is
not good enough; we have to start to couple our models. There is,
therefore, an ever-pressing demand of more accurate quantifications of
those fluxes in the domain of microscale air-sea interaction. As for
the marine remote sensing, clearly we have entered the age of satellite
meteorology and oceanography. Many of the parameters driving both
systems, atmosphere and oceans, can be sensed remotely with spaceborne
detectors. The remote sensing has now become an integrated part of
the research program of these systems, along with in—situ observations
and model development, with a emphasis on comparison and validation
among them. But, what is the connection between the remote sensing
and the microscale air-sea interaction To many, If not all, of the
remote sensors "the atmosphere Is transparent", In the words of the
remote—sensing community, meaning that these sensors can see through
the air. They sure cannot see through the water, but everything on
the sea surface. The marine remote sensing relies on the knowledge of
microscale air-sea interaction to interpret their signals from the sea
surface.

404
Processes of Hicroscale Air-Sea Interaction
Few examples of our own research are selected to demonstrate that
microscale air-sea interaction processes are important in linking the
atmosphere and oceans. The wind stress acting to the sea surface Is
definitely a critical parameter for the climate research. The wind
stress (r) can be referred to in different forms, such as the windstress
coefficient (C-i A) » the wind-friction velocity (u^), and the
roughness length (z ). These terms are related through the so-called
logarithmic wind profile (Wu 1968),
c io - [1]
where U is the wind velocity at the elevation z above the mean sea surface,
K, is the von Karman universal constant, p is the density of air,
and the subscript 10 corresponds to 10 m above the mean sea surface.
I would like to emphasize though that our research and your demand are
the same; it is to be able to associate all parameters mentioned above
with environmental variables. First of several environmental variables,
of course, is the wind velocity, we have established that the windstress
coefficient increases with the wind velocity as shown in Fig. 1.
T 1
Figure 1, Wind-Stress Coefficients over the Sea Surface. The data are
shown as open circles, the dashed line corresponds to Eq.
[1], and the solid line is a linear approximation. This is
reproduced from Wu (1982).

405
The results shown In Fig. 1 are for the open sea, with the data
illustrating that the proposed formulae can be applied to very high wind
velocities up to 50 or 60 m s~
(Wu 1980, 1982). We have also provided
modifications due to air-sea temperature differences; see Fig. 2, where
C
is the wind-stress coefficient obtained under unstable (or stable)
conditions, Z is the anemometer height, and L is the Monin-Obukhov
stability length (Papadiraitrakis et al. 1987). Such a figure, of
course, is familiar to you for results over land; but this is over the
sea surface. 2.0
Sable Conditions:
CJ/C10«*xpf0.70Z/U
}— C,/C10 » exp 1-0.95 2/U
Figure 2.
Variations of Wind-Stress Coefficient with Stability
Length. The shaded areas cover the range of variation of
results from various investigations. This is reproduced
from ttu (1986).
Speaking about the sea surface, we have to deal with waves. A
parameter, the so—called wave age, is used; it can be expressed in one
of two ways, either as the ratio between the phase velocity of dominant
ocean waves and the wind velocity c/ u i/y or with the wind-friction
velocity c/u^. For the open sea where dominant ocean waves propagate
in pace with the wind, the ratio C/U IQ has a value of nearly unity.
The corresponding value for c/u^ is about 25. At short fetches near
shore or fronts, the dominant waves are shorter and have a smaller

406
phase velocity; both ratios, therefore, have smaller values. We see,
in Fig. 3, for those short but choppy waves at smaller wave ages, or
short-fetches, the wind-stress coefficient is greater. These aspects
have yet to be incorporated into the models; but you must, as its
effects are seen very significant.
T I I
4 6 8 tO
Wav* A$«. c/u.
Figure 3. Variations of Wind-Stress Coefficient with Wave Age. This
is reproduced from Wu (1985a).
Ul0(mf 1 ):
10
0.2
to 1 id 2 to 3
Nondimwisionai
10* to* 10*
Figure 4,
Ratios Between Wave-Drag and Wind-Stress Coefficients at
Various Nondimensional Fetches. This is reproduced from
Wu (1988).
To oceanographers, the important thing is the wind-driven ocean
circulation, but the energy comes from the wind is partitioned between
currents and waves (Wu 1975) . We have studied this partition in the

407
laboratory,
momentum flux to waves.
and obtain the wave drag which represents the portion of
an estimation for the oceanic condition.
On the basis of these results, we have made
The results in terms of the
ratio between wave-drag and wind-stress coefficients, C /C- , are
shown in Fig. 4.
This ratio, therefore, represents the fraction of
momentum flux used in producing waves at different nondimensional
2
fetches gL/U.Q, and under different wind velocities. At open sea,
only 10% of the wind stress is used for wave generation; at short
fetches, this is shown in the figure to be very different, especially
under low winds.
Interpretations of Remote Sensing Signals
Let me give you two examples in the area of marine remote sensing.
You have used the sea-surface temperature (SST) for years and in many
cases interchangeably with the subsurface temperature.
The difference
between these two temperatures is generally considered to be less than
1°C, although you realize that there is a cool skin over the sea surface,
the so-called thermal sublayer (Wu 1985b) .
We find that the
general concept about the thermal sublayer adopted from studies over
the solid surface is not applicable over the air-sea interface.
nondimensional thickness of the thermal sublayer generally expressed as
* "" $ t u * w A' does not nave a constant value, where u^ is the friction
velocity of currents and v the kinematic viscosity of water; see Fig.
5. The difference between the sea-surface and the subsurface tempera-"
ture (AT) depends also on the heat flux across the sea surface (Q) as,
The
AT - AQz//c p pu^w [2]
where c p and p are respectively the specific heat and density of
water. The sea-surface-temperature deviation is seen to vary rather
strongly with the wind velocity, especially so at low winds. This
importance should be further emphasized because most of the heat flux
in the global sense is from the tropical area, where the wind is
generally very low.
Now let us turn to another area of remote sensing, which along with
SST excite greatly the atmospheric scientists and oceanographers. It is
the mapping of the sea-surface wind velocity from space. This technique
is closely related to Bragg scattering of microwaves from ripples not

T i l l !
o
I •
GO
00 J
Wind Velocity. U,Q (mi ')
Figure 5. Frequently Cited Oceanic Results on Coefficients of Thermal-
Sublayer Thickness. The line is fitted on the basis of
theoretical considerations (Wu 1984). This is reproduced
from Wu (1985b).
from long waves. These ripples can be represented best with the slope
spectrum, not the height statistics, of ocean waves (Haimbach and Wu
1986). We have also completed the feasibility study in this area as
illustrated In Fig. 6, with the radar sea returns correlating well
with the wind-friction velocity. Let me discuss with you, however,
I
I °' 2 20 40 60
Wind-Friction Velocity, u. ferns' 1 }
Figure 6.
Variation of Radar Returns with the Wind-Friction Velocity.
The line represents the power law. This figure is reproduced
from Wu (1989a).

some of the problems, which are backed up by our physical understanding,
but have so far been ignored by the remote-sensing community. According
to the current understanding, the radar returns with W polarizations
should be associated with ripples, the returns should increase contl-
1/2
nuously with the wavenumber of ripples, following roughly k ' ; the
actual returns, however, reach a saturated level, not Increasing with
the wavenumber.
409
On the other hand, the HH polarized results should be
associated with breaking waves, they surprisingly follow very well the
predicted trend.
In addition, we have studied the ripple structures
and turbulence characteristics, and expected that the radar returns
should be very much influenced by the stability condition.
The radar
returns are intensified under unstable conditions are suppressed under
stable conditions.
In other words, first we have to find AT, but we
have no means of measuring it remotely now. Finally, we have to deal
with sea-surface conditions.
Pictures taken during the space shuttle
mission indicate that under light winds, say 6 or 7 m s , the sea
surface is covered by natural slicks.
According to our studies of the
marine microlayer, the suppression of ripple structures should follow
the trend shown in Fig. 7, where E s (f) and E c (f) are signal densities
from slick and clean surfaces.
In any event, I hope that all these
have illustrated to you that the studies of microscale air-sea
interaction are important to the remote sensing which may be of interest
to you.
Surface-Waff Length, X (cm)
Figure 7 Summary of Available Slick Measurements with Microwave ^
" Sensors. The curves are fitted on the basis of wave-damping
measurements. This is reproduced from Wu (1989b).

410
Concluding Remarks
In order to understand and protect the environment, a team effort
is generally required. This also coincides with recent realization that
oceans may play a greater role on the climate than what we have
previously recognized. Much of the research effort has, of course,
been upgraded with wider coverage and quicker updating of our earth
through remote sensing. The atmosphere and oceans have now been
regarded as a true interacting system. We, in the air—sea interaction
area, are glad to contribute to the climate research.
ACKNOWLEDGEMENT: I am very grateful to the sponsorship of my work
provided by the Fluid Dynamics Program, Office of Naval Research
(N00014-89-J-1100), and the Physical Oceanography Program, National
Science Foundation (OCE-8716519).

411
REFERENCES
Haimbach, S. P. and Jin Wu, "Directional slope distributions of winddisturbed
water surface", Radio Sci., 21, 73-79 (1986).
Papadimitrakis, Y. A., Y.-H. L. Hsu and Jin Wu, "Thermal stability
effects on the structure of the velocity field above the sea
surface", J. Geophys. Res. 92, 8277-8292 (1987).
Wu, Jin, "Laboratory studies of wind-wave interaction", J. Fluid Mech.
34, 91-112 (1968).
Wu, Jin, "Wind-induced drift currents", J. Fluid Mech. £8, 49-70
(1975).
Wu, Jin, "Wind—stress coefficient over sea surface near neutral
conditions - A revisit", J. Phys. Oceanogr. 10, 727-740 (1980).
Wu, Jin, "Wind-stress coefficients over sea surface from breeze to
hurricane", J. Geophys. Res. .87, 9704-9706 (1982).
Wu, Jin, "Viscous sublayer below wind-disturbed water surface", J.
Phys. Oceanogr. 14, 138-144 (1984).
Wu, Jin, "Parameterization of wind-stress coefficients over the sea
surface", J, Geophys. Res. £0, 9069-9072 (1985a).
Wu, Jin, "On the cool skin of the ocean", Bound-Layer Meteorol. 31,
203-207 (1985b).
Wu, Jin, "Stability parameters and wind-stress coefficients under
various stability conditions", J. Atmos. Oceanic Tech. 1, 333-339
(1986).
Wu, Jin, "Momentum flux from wind to aqueous flows at various wind
velocities and fetches", J. Phys. Oceanogr. 18, 140-144 (1988).
Wu, Jin, "Dependence of microwave sea returns on wind-friction velocity
under varied atmospheric stability condition 11 , I.E.E.E. J. Oceanic
Eng. 14, 254-258, (1989a).
Wu, Jin, "Suppression of oceanic ripples by surfactant - Spectral
effects deduced from sun-glitter, wave-staff and microwave
measurement", J. Phys. Oceanogr. 19, 238-245 (1989b).

412
THE VARIATIONS OF THE SST IN THE EASTERN AND WESTERN
TROPICAL PACIFIC AND
THEIR RELATIONSHIP WITH THOSE IN THE WORLD OCEAN
Pan Yi-Hang
(NRCMEF/SOA,China)
Oort Abraham H.
(GFDL/NOAA, USA)
and
Richardson William
(NOS/NOAA, USA)
ABSTRACT
The seasonal change of the connections
between the SST in the EEP with those in the
world ocean are given by 12 correlation maps
of the 12 calendar months for the period of
1950-1979. The standard deviation map
calculated from these 12 correlation maps has
shown that the areas with larger values also
located in the western Pacific but mainly in
the tropical regions of south-western Pacific
and south-eastern Indian oceans. Discussions
about the seasonal change of the SST in the EEP
as well as those in the world ocean will be
given.
I. INTRODUCTION
In previous papers, the high correlation between the
SST in the eastern equatorial Pacific (EEP) with those in
the world oceans has been shown by the similarities of the
spatial distribution between the 11 correlation maps for 11
decades of 1870-1979 [ Pan and Oort 1989]. The independence
of the SST in the western equatorial Pacific (WEP) from
those in the EEP has been shown by the standard deviation
map of the 11 correlation maps [Pan and et al 1989], In
this paper, we'll further investigate how the connections
between SST in the EEP with those in the world ocean change
with the seasons.
II. DATA BASE
The data base for this study is Comprehensive Ocean-
Atmosphere Data Set (GOADS.) , the most extensive surface
marine data set in existence today. These data were
observed mainly from the international ships over the world
ocean containing various parameters as sea surface
temperature (SST), surface air temperature (SAT) and et al.
Under a joint project in the United State, they were
collected for the period of 1854 to 1979 and edited monthly
into 2°*2° boxes by ERL, CIRE, NCDC/NOAA and NCAR [Woodruff
etal 1987], Then; the global monthly fields of SST, as weil Q
as other parameters were objectively analyzed onto a 1*1
grid at GFDL/NOAA by using Levitus scheme [Lev!tus 1982].

413
The climatology of global SST fields for 12 calendar months
were averaged from the thirty years of 1950-1979. The
monthly anomalies of SST for every individual month were
obtained using this climatology.
III. HIGH CORRELATION BETWEEN SST IN EEP WITH THOSE IN
WORLD OCEAN FOR 1870-1979
As presented in previous paper [Pan and Oort, 1989],
the high correlation between the SST in the east equatorial
Pacific (EEP) with those in the world oceans have been
shown by the coincident variations of the annual mean
between the SST averaged from the world ocean (60°S-60 C N),
the tropical ocean (30°S-30°N) with the EEP(180 C> -80 0 W / 20 0 S-
20 D N) as well as the key region (130*W, 10 D S-^T) for the
period of 1870 to 1979, see Fig.l. Also, we calculated the
11 correlation maps of the SST in EEP with those in the
global ocean for every ten years from 1870-1879 to 1970-
1979. These eleven correlation maps have shown similar
patterns with each other (figures omitted). The averaged
mean map of them was given here, as Fig.2, to show the
typical pattern of high -positive correlations presented in
most areas of the tropical oce'ans with the maximum of the
value 0.9 at the east equatorial Pacific but with the
exception at the western Pacific and eastern Indian ocean
in the tropics.
IV. ANNUAL VARIATION OF THE CORRELATION BETWEEN SST
WITH THOSE IN GLOBAL OCEAN
IN EEP
In order to understand the seasonal change of the
connections of SST between the EEP with the global oceans,
the same correlation maps for the 12 calendar months for
the period 1950-1979 were calculated. Similar to those for
the 11 decades, the correlation maps for the 12 calendar
months also display a similar pattern as shown in Fig.2.
Fig. 3 gave parts of the correlation maps for the 12
calendar months, i.e. January, April, July and October.
Although the major pattern of the distributions of the
correlation coefficients are fairly steady during an annual
cycle, but in the areas of the western Pacific and the
eastern Indian ocean of the tropics, the connections of SST
with the EEP are more variable. Not only the magnitudes of
the correlation coefficients are very small, even the sign
of them has been changed between different calendar month.
V. STANDARD DEVIATION FOR THE 12 CORRELATION MAPS
Then, the standard deviation from the 12 correlation
maps were calculated as shown in Fig.4. It is clear to give
the distribution for the seasonal change of the
correlations. Where the magnitude of the standard deviation
is larger (smaller), the connection of SST with those in
EEP is weaker (closer) . Consistently with the 12 correlation

414
maps, there is an extensive area with the standard
deviation smaller than 0.1 located east of dateline in the
tropical Pacific, where the index of SST for EEP is
averaged from. It is worthy to note that the belt with
value of the standard deviation larger than 0.2 spreads
just east of New Genesia at Solomon and Coral seas and
emanated south-eastwards in the south Pacific. On the other
side of the north part of Australia in tropical Indian
ocean, the belt with value larger than 0.2 located round
Indonesia and crossed the equator northwards in Bengal Bay,
It implied that comparison with the other parts in the
world oceans, the SST in these areas do not much correlated
with those in the EEP and may have their own features in
seasonal changes. In other words, the seasonal changes of
SST in these regions do not much follow those in the EEP.
Of course, there are still some magnitudes larger than 0.2
spread in mid-latitudes of both hemispheres but their areas
are too small.
VI- DISCUSSIONS
The great interest in these areas is the location of
them. As well known, the tropical western Pacific and
eastern Indian ocean are the locations of the warmest water
in the world oceans. Importance is, from the view of ENSO,
these areas are the sources of the warm water for El Nino
events and also the origin fields for the upward currents
in Southern Oscillation. Here, in this paper, we use a
simple scheme of calculating the standard deviation from
the correlation coefficients between SST in the EEP with
the world oceans for 12 calendar months. The area with the
maximum value of the standard deviations exactly
corresponds with the warm water pool and the sources for
ENSO. Such as discussions by many studies [Rasmusson and
Wallace 1983; Berlage 1966; Troup 1965 and Trenberth 1976],
every ENSO event has its own characteristics. In the
regions of the tropical western Pacific and eastern Indian
ocean, either the zonal flows of warm water in the
equatorial oceans or the migrations of the upcurrents in
the tropical atmosphere all have evident seasonal changes
in every individual year and also interannual variations.
It might be the reason that these regions become important
members in the Asia monsoon system. Therefore, further
investigations about the air-sea interaction over these
areas both regional and global are important. Probably, it
might improve the predictions of El Nino.
Acknowledgements; The authors want to thank Mr. John Carey
for his support to pursue this joint study with GF.DL.at
NOS. One of the authors (Y.H.Pan) acknowledges the
financial support of NOS and OAR/NOAA during her stay at
NDS/NOAA to complish this research.

415
Reference
Berlage H.P., "The Southern Oscillation and World
Weather". Commun. Trans. No. 88, Roy. Netherlands
Meteor. Inst., 152 pp.(1966).
Levitus S., "Climatological atlas of the world ocean"
NOAA Prof. Pap. No. 13, 163 pp., U.S. Government
Printing Office, Washington, D.C.(1982).
Pan Y.H. and A.H.Oort, "Some correlation analysis on the
sea surface temperature variations over the eastern
equatorial Pacific with the world oceans",
(to be published on Climate Dynamics,1989)
Pan Y.H., W.Richardson and A.H.Oort, "Some different
characteristic of SST between the western and
eastern tropical Pacific ocean", (to be published on
"Symposium on Western Pacific Air-Sea Interactions"
Beijing, November, 1988)
Rasmusson E.M. and J.M.Wallace "Meteorological aspects of
the El Nino/Southern Oscillation", Science Vol.222
No.4629 1195-1201 (1983)
Troup A.J. "The 'southern oscillation" 1 , Quart. J.R. Met.
Soc. £1, pp. 490-506 (1965)
Trenberth K.E. "Spatial and temporal variations of the
Southern Oscillation. Quart. J.R. Met. Soc. 102 pp.
639-653 (1976)
Woodruff S.D.,R.J.Slutz,R.L.Jenne and P.M.Steurer,
"A Comprehensive Ocean-Atmosphere Data Set", Bull.
Amer. Meteo. Soc., 6.8 / 1239-1250 (1987)

416
Flg.1.
Time series of the annual mean SST anomalies (°C) for the EEP region (thin soiJd line),
the center of EEP (thin dashed line), the tropJcal oceans (heavy sdld line) and the world
oceans (heavy dashed line) for thel 870-1979 period.
(Taken from [Pan and Oort 1989])
SON
'120' " ' ' sow' ' 6 'id
Flg,2. The mean map from 11 correlation coefficient distribution between the SST in EEP
With those over the world oceans for the 11 decades from 1870-1879 through 1970-1979.
(Taken from [Pan and Oort 1989])

417
90 N
IE' 120' ' 'iio' ' ' ' '120' ' '
J
Ww'
i
Fig.Sa. Map of the correlation coefficient of the SST in the EEP (20 0 S-20 0 N,180-80°W) with
those in the world oceans for January.
90 Nr
28!' '-WE
Fig.3b. Map of the correlation coefficient of the SST in the EEP (20 0 S-20 ft N l180-80°W) with
those in the world oceans for April.

418
90 N
90 S|. . .
20E
20E
Fig.3c. Map of the correlation coefficient of the SST in the EEP {20 0 S-20 0 N, 18Q-80°W) with
those in the world oceans for July.
60 N
0 20 E
Fig.3d. Map of the correlation coefficient of the SST in the EEP (20°S-20 e N,180-80°W) with
those in the worid oceans for October.

Fig.4.
Map of the standard deviation from 12 correlation coefficient distribution between the
SST in EEP with those in the world oceans for the 12 calendar months of the thirty
years 1950-1979,
(unit of Isofine is 5* 10* 2 , shaded area is SD > 0.20)

420
ON LABORATORY SIMULATION OF SEA BREEZE
S.C Kot
Department of Mechanical Engineering
University of Hong Kong
INTRODUCTION:
The differential heating of land and large water bodies diurnally
caused sea/.lake breezes. The local circulation pattern set up by the
sea/lake breeze is of major concern to air pollution meteorologists on the
coastlines, LyonsW. Stack gas emission may be trapped within the local
circulation and may not be dispersed by the larger weather systems. Cases
of sea breeze observations near Hong Kong were documented by Zhou et
al(l^). The site was at Daya Bay, about 53 krn north of the centre of Hong
Kong. They estimated the occurrence of sea breeze, land breeze, onshore
and offshore gradient winds to be 26.5, 28.1, 28.7 and 16,7% respectively of
time in a year. The corresponding average wind speeds were 3.3» 2.4, 4.4
and 3.2 m/s respectively. The sea and land breezes had small depths
averaging 440 and 280 m respectively, while the onshore and offshore
gradient winds had average depths of more than 1 km. The sea breezes
turned back at 1-3 km inland. Since Daya Bay is so close to Hong Kong
similar sea breeze characteristics are expected here. The main difference
in actual flow field is of course due to the different detailed topography of

421
the rugged coastline terrain. A model study can be performed using the
observed sea breeze characteristics.
Mathematical model using equations of fluid motion had been
developed to model sea breeze fronts, Clarke( 2 ).
Three dimensional model is
needed for the complex terrains in Hong Kong. Three dimensional
calculations
for flow past non-rectangular bodies are notoriously difficult
and time consuming on the computer.
The other alternative is the use of
scale modelling. A wind tunnel with the capability for cooling a
significant portion of the test section floor is required for simulating sea
breeze, Meroney et al( 5 ). This type of meteorological wind tunnel is
expensive to build and run. There are only a few existing in this world. A
cheaper alternative has to be found for modelling sea breeze intrusion.
Fortunately
SimpsonW noticed the internal structure of sea breeze front
was similar to the gravity current formed when a lock-gate separating two
fluids of different density was opened suddenly.
This experiment can be
performed very inexpensively in the laboratory with a long water
channel.
To see whether the laboratory experiment can really model the
nature, a particularly complex situation of the crossing of two fronts was
simulated in the water tank. Sometimes, the interaction between a
thunderstorm outflow and a sea/lake breeze may occur. Observations of
such interactions were reported by Boyd(^), Moraz and Hewson(^). For
large islands or peninsulas, collision of sea breeze fronts had been also
observed. A well documented case near Stansted was reported by Rider and
S imp son (7). The movement of the fronts were followed by the radar echos
of swifts. Some qualitative and quantitative results were published in that
paper. Clarke(^) also performed a numerical computation on a two
dimensional double sea breeze collision.
In this paper the experimental procedure on the collision between
two gravity currents is described. Suggestions for performing a sea breeze
intrusion into complex terrain in a water tank is put forward.

422
EXPERIMENT:
Only simple and easily available equipments were needed, namely,
a large water tank, common salt, acrylic or wooden lock-gates, self-winding
camera, clock with fractional seconds showing, translucent paper and a
high wattage lamp. Briefly, the experimental set-up is described in Figure
1. The water tank used for the experiment on the collision of two sea breeze
fronts was of size 20 x 52 x 348 cm (width x depth x length). Two locks were
fitted 30 cm from each end. A saline solution of varying specific gravity
was prepared. The saline solution was introduced to the bottom of the tank
between the ends and the locks. The water tank was already filled with
fresh water to about 40 cm depth. The saline solution was introduced very
carefully and slowly so as not to enhance mixing. The specific gravity of
the saline solution was measured by a reflectometer. The salt solution was
filled up to 10 cm height so that the free surface of the fresh water was
sufficiently far away, to minimize its effect on the gravity currents.
To observe the movement of the gravity currents, a shadowgraph
was rigged up. A bright lamp was placed at the back of the tank and the
tank was illuminated through a translucent paper. The translucent paper
was marked by a grid and attached to the back side of the tank. As the
fluids were of different specific gravity, the interfaces could easily been
seen using the shadowgraph method. A digital stop watch showing
fractions of seconds was attached to the front of the tank. The locks were
raised one after another. The advancement and collision of the gravity
currents was recorded by a camera taking photographs at the rate of about
one second a frame. Quantitative parameters such as the velocity of the
fronts and the depths of the gravity currents were measured from the
photographs. Dye could be added for better identification of fluids of
different specific gravities.

423
The experiment was repeated many times with different specific
gravities of saline solution at both ends. Collisions between gravity
currents formed from specific gravity 1.01 with those formed from 1.01 to
1.04 at 1% increment were performed.
RESULTS AND DISCUSSIONS:
The experimentally determined velocities of the gravity currents
are shown in Figure 2.
shown in Figure 3.
The composite photograph of experiment run 6 is
In that run the specific gravity of saline solution on
the left side was 1.01 and the specific gravity of saline solution on the right
side was 1.03.
For collision of gravity currents of substantially different specific
gravities, the qualitative behaviours all showed similar features. The
velocity of gravity currents did not alter appreciably before collision.
After colliding, the denser fluid slid under the less dense fluid with some
mixing. A hook shape overspill of less dense fluid was always observed.
The maximum height of the overspill could reach to more than five times of
the gravity currents. This hook engulfed the ambient fluid (fresh water).
Since the ambient fluid was lighter, instability occurred resulting in
vigorous mixing. The lifting of a colder mass of air in the sea breeze or
thunder storm outflow into great height was the cause of severe weather
and the lumpiness seen on the radar.
In comparison with observation, two characteristics of sea breeze
collision were reproduced in the laboratory. After the violent collision a
reformed sea breeze moved forward at a slower velocity for the stronger
sea breeze and a reflected solitary wave (or bore) was observed. These were
modelled by the laboratory collision of gravity currents. For more
discussion on the comparison please refer to Kot and Simpson (1987).

424
With the ability to simulate complex phenomenon of collision of
sea breezes confirmed, the collision of one sea breeze with a hill should
create no new problems. The hill can certainly be idealized as a gravity
current of infinite specific gravity. In order to obtain quantitative results
meaningfully, the scaling parameters would need to be discussed.
First of all geometric similarity need to be observed. From
observation, the depth of sea breeze near Hong Kong is of the order of 400
m. If a model gravity current depth of 20 cm is used to simulate the sea
breeze, a 1:2000 scale model is required.
From the theory on gravity current, the velocity U of the
advancing front into a still fluid is
where
p- is the density of still fluid.
p« density of gravity current
g acceleration due to gravity and H height of gravity current.
With 1/2000 scale modelling, the model speed of gravity current, by
assuming full scale sea breeze at 3 m/s, is 6.7 cm/s. The density difference
for two fluids is then 1.12% assuming a model gravity current height of 20
cm. This density ratio is within the range of our experiment for collision
between gravity currents. A water tank of 3 x 0,6 x 4 m (width x depth x
length) can be used for simulation of a sea breeze intrusion of about 3 km
inland with three dimensional terrain, The sea breeze front need not be
straight, curved locks can easily be fabricated.

425
CONCLUSIONS:
The use of wind-tunnel for sea breeze study can produced a steady
flow pattern, so there is plenty of time for taking quantitative
measurements. But the cost is great. Using water tank experiments, the
flow pattern established is of a short time span.
tank experiment is the ease in flow visualization.
taken, they can be studied in good time.
operation is ridiculously low.
The advantage in water
After photographs are
The cost of equipment and
For air pollution meteorologist, the height of
the sea breeze is the most important parameter for stack height
determination. The water tank experiment can provide a quick and
inexpensive method of obtaining this information for rough terrain.
REFERENCES:
(1) Boyd, J.G., Observation of two intersection radar fine lines, Monthly
Weather Review, Vol. 92. No. 3, 188 (1965).
(2) Clarke, R.H., Colliding sea-breezes and the creation of internal
atmospheric bore waves: two-dimensional numerical studies, Aust.
Met. Mag., Vol. 32. 207-226 (1984).
(3) Kot, S.C. and Simpson, I.E., Laboratory experiments on two crossing
fronts, Proc. of the International conference on Fluid Mechanics, 73.1-
736, Beijing (1987).
(4) Lyons, W.A., Turbulent diffusion and pollutant transport in shoreline
environments, in Lectures on Air Pollution and Environmental Impact
Analysis, ed. D.A. Haugen, Chap. 5, AMS, (1976).

426
(5) Meroney, R.N., Cermak, J.E., and Yang, B.T., Modelling of atmospheric
transport and fumigation at shore-line sites, Boundary Layer
MeteoroL, Vol. 9. 69-90 (1975).
(6) Moraz, W.J., and Hewson, E.W., The mesoscale interaction of a lake
breeze and low level outflow from a thunderstorm, J. of Applied
Meteorology, Vol. 5. 148-155 (1966).
(7) Rider, G.C., and Simpson, I.E., Two crossing fronts on radar, Met. Mag.
Vol. 97. 24-30 (1968).
(8) Simpson, I.E., A comparison between laboratory and atmospheric
density currents, Q.J.R. MeteoroL Soc., Vol. 95. 758-765 (1969).
(9) Simpson, I.E., Gravity currents in the laboratory, atmospheric and
ocean, Ann. Rev. fluid Mech., Vol. 14. 213-234 (1982).
(10) Zhou, R., Wu» D. and Yan, Z., The evaluation of site characteristic for
Guangdong nuclear power plant, Proc. 6th Pacific Basin Nuclear Conf.,
306-312, Beijing (1987).
LOCK-GATE
LOCK-GATE
FRESH WATER
SALTWATER
SALTWATER
Fig. 1
Sketch of Water Tank

427
80
60
U mm/s. 40 *
20
0:02 0.04
(p - p )/p.
2 1 1
0.06
Fig. 2 Experimentally Determined Gravity
Current Velocity

428
Fig. 3
Composite Photograph of Gravity Current Collision

429
THE IMPLEMENTATION AND OPERATION OF AN ANALYSIS SCHEME
AND A LIMITED AREA MODEL FOR HONG KONG
Y.KChan
Royal Observatory, Hong Kong
ABSTRACT The Royal Observatory currently operates an optimal interpolation, data
analysis scheme and a very fine mesh limited area model. The physical, mathematical
and computational aspects of this system are briefly described, with emphasis on the
use of bogus data for tropical cyclone analyses and the use of time-dependent lateral
boundary conditions in short-range weather prediction. To extract the most from the
models, various derived elements are computed and presented in formats convenient
for use by forecasters in daily weather forecasting operations. The performance of
the model in the forecast of various weather elements and synoptic patterns around
Hong Kong is discussed.
1. INTRODUCTION
As an input to the production of weather forecasts, the Royal Observatory (RO) has been using
outputs from numerical weather prediction models of the European Centre for Medium Range
Weather Forecasts (ECMWF) and the United Kingdom Meteorological Office (UKMO) for some
years. Such information has proved valuable in the daily operation of the provision of weather services
to the public and in the development of forecasting aids. However, this information is transmitted over
the Global Telecommunication System at reduced resolution. Thus, it is adequate only for predicting
large-scale weather systems such as surges of the winter monsoon, but inadequate for meso-scale
features such as monsoon troughs which are often the cause of inclement weather conditions.
Thus, it becomes necessary to develop, for application in Hong Kong, fine resolution analysis scheme
and limited area model as an objective means of assimilating all the available data and predicting the
evolution of the meso-scale features. With such a view in mind, an attempt was carried out in the RO
to adapt the two dimensional multivariate optimal interpolation data analysis scheme (OI) and the
very fine mesh limited area model (VFM65) of the Japan Meteorological Agency (JMA) for
application in the domain around Hong Kong. Due to the limited capacity of the computer system in
the RO, a numerical weather prediction model in a domain larger than the present one cannot be run
to provide the lateral boundary condition operationally. Thus, arrangement was made with JMA to
transmit boundary data from their Global Spectral Model to Hong Kong everyday for use in the daily
operational runs of our models.
2. THE OPTIMAL INTERPOLATION DATA ANALYSIS SCHEME
The analysis scheme is a two dimensional multivariate optimal interpolation scheme performed on 10
standard pressure levels in the vertical. The horizontal domain of the analyses is shown in Figure 1.
The domain is covered by a 101 x 66 latitude-longitude grid of one degree resolution. The grid length is
111.2 km true at 22.5°N.
In JMA, the predicted values from the Asian Model serve as the first guess field. In the RO, such
values are not available and an extended area model is being tested to provide predicted values at the
first guess field. For the time being, the first guess field is a coarse resolution analysis using the
Cressman scheme.

430
A detailed description of the method of analysis and the determination of statistical coefficients
can be found in the Appendix to Progress Report on Numerical Weather Prediction, JMA, 1986.
As the JMA statistical coefficients used in the analysis scheme were determined to 'optimize' the
analyses in the domain around Japan, they may not 'optimize' the analyses around Hong Kong
simultaneously. Thus, past data are being run to determine statistical coefficients revelent to the
RO model domain. At present, all statistical coefficients for pressure/height and temperature have
been replaced. Effort is now concentrated on improving the coefficients for humidity.
3. TROPICAL CYCLONE BOGUS DATA
The genesis and development of tropical cyclones usually take place over oceans where few
observations are available in their proximity. Thus, it is necessary to incorporate bogus data in the
objective analysis to improve the representation of tropical cyclones.
The numerical scheme to generate bogus data for tropical cyclone in the JMA VFM65 (Ueno,
1989) covers these aspects: the low- and mid-level circulation, the warm core, and the upper-level
divergence field. The initial internal balance among the data is achieved by applying the limited
area normal mode initialization scheme.
As no initialization scheme, except smoothing, is applied hi the Royal Observatory models, only
bogus data of mean-sea-level pressure and winds at 850,700 and 500 hPa are generated empirically
based on a Rankine type model (Anderson and Hollingsworth, 1988). The generated fields are
spherically symmetric and computed from the following manually analysed data:
(i) central position,
(ii) central pressure,
(iii) radius and pressure at radius of strong wind,
(iv) radius and value of outermost closed isobar.
The internal balance among the bogus data is achieved during forward integration in time.
4. THE RO LIMITED AREA MODEL
4.1 Grid and Physical Processes
The RO limited area model is a three dimensional primitive equation model in sigma coordinate
and has 13 layers in the vertical. In the design of the model, emphasis is placed in the simulation of
small scale features of the atmosphere, especially those caused by the surface conditions of the
earth such as orography and land-sea contrast. Thus, the model has a high vertical resolution in the
lower troposphere (Figure 2) and a fairly sophisticated scheme for the parameterization of the
boundary layer.
Figure 3 shows the horizontal domain of the model. The domain is covered by a 51 x 36 one
degree grid in latitude-longitude projection. The grid-length is 111.2 km true at 22.5 °N. In the
horizontal, all the variables are defined at the same grid point. Thus, the grid is a non-staggered
one (Arakawa A-grid). Apparently, no decoupling or other problems are detected in the forecast
outputs of the model so far since the model became operational in mid->$eptember 1988.

437
42 Physical Processes
The physical processes included in the model are: surface flux (Kondo, 1975), radiation
(Smagorinsky, I960), vertical diffusion (Mellor and Yamada, 1974), large scale condensation
(Numerical Prediction Division, JMA, 1986), convection (Gadd and Keers, 1970) and horizontal
diffusion. The horizontal diffusion terms are given empirically as fourth-order Laplacians with
constant diffusion coefficient of lu m /sec. Each momentum equation includes two diffusion
terms. The first term acts on both the rotational and the divergent components of the wind while
the second term acts on the divergent component only.
42 Topography
As the domain of the RO limited area model is different from that of the JMA VFM65, a new
topography has to be constructed. In designing this topography, we have to bear in mind that our
boundary data come from the JMA Global Spectral Model. In order to avoid the generation of
inertial waves due to discontinuity across the boundary, the topography field employed in our
model should be one that could be joined smoothly together with that of the JMA model in the
lateral boundary region (Figure 3).
Our first attempt was to adapt the topography field of the JMA Global Spectral Model which is
computed by converting 10 minute latitude/longitude resolution grid-point topography values as
read from relief map to spherical harmonics truncated at wavenumber 63. These harmonics are
then smoothed by using the exponential filter (Hoskins, 1980). The grid-point resolution of the
global topography field is 1.875 degree. A bicubic spline is employed to interpolate this field into a
one degree resolution field (TG) for use in our model. The resultant field is too smooth for a one
degree resolution model. Nanling, which plays an important role in controlling the flow of the
northerly airstream over south China, is not accurately represented (Figure 4). Operational
experience showed that the model with this topography did not perform well in respect of the
timing of the arrival of surges of winter monsoon at Hong Kong.
As a second attempt, the model employs an envelope topography (TE) with one standard
deviation, derived from the 10 minute latitude/longitude resolution grid-point topography data.
The resulting field is much more realistic (Figure 5). However, the problem of'smooth transition in
the boundary region' exists and suspected inertial waves are spotted in the in-coming boundary
region occasionally. Further numerical experimentation will be carried out for a new topography
(T) = KB (TG) + (!-KB) (TE) for various values of KB in the boundary region.
43 Initialization
The mean sea-level pressure, wind, temperature and height at the ten pressure levels and the
dew-point depression at the four pressure levels analysed by the operational optimal interpolation
data analysis scheme are vertically interpolated to the thirteen sigma levels using cubic spline. The
specific humidity is computed from the dew-point depression and temperature. It is then modified
to eliminate supersaturation and instability. A nine-point filter is applied to all variables on the
sigma levels to remove horizontal two-grid waves.
Numerical experimentations have been carried out on dynamic initialization schemes proposed
by Sugi (1986) using the following as the forward-backward integration schemes:
(i) Tatsumi's economic explicit scheme;
(u) Okamura scheme.

432
However, the schemes failed to converge after a considerable number of iterations. Thus, as an
interim measure, the limited area model is now integrated operationally without the use of any
initialization scheme.
According to Satomura (1988), implicit schemes will be more effective as the forward- backward
integration scheme in dynamic normal mode initialization. Considerations are being given to
developing a dynamic initialization scheme to improve the performance of the limited area model.
4.4 Numerical Integration and Lateral Boundary Conditions
The TatsumTs economical explicit scheme is used to integrate the limited area model forward in
time. A detailed description of the scheme can be found in Tatsumi (1983). The convergence of the
Tatsumi scheme is not strong enough. This necessitates the introduction of more damping schemes
into the model. At present, the modified version of the Asselin (1972) time filter and the gravity
wave damping scheme are used with the damping parameters set to 0.1 in both schemes.
The formulation of the lateral boundary condition follows that proposed by Hovermale
(unpublished).
In the boundary region which is 7 grids in width from the lateral boundary (the dotted area in
Figure 9), the following terms are added to the momentum, the thermodynamic and the tendency
equations :
—~ V.* 2 (X - Xmu] (4.1a)
z
\ /
dt
X stands for the prognostic variable, LB the width of the boundary region and 1 the distance from
the lateral boundary.
The operator VJT is a non-dimensional finite difference analog of the Laplacian operator defined
by
Vx 2 X(x,y) * XQc+d,y+d) + Xfc+d,y-d) + X(x~d,y+

433
time cross-section at two grid-points near Hong Kong are also produced. The following is a simple
account on the performance of the model around Hong Kong in a limited data sample.
5.1 The Forecast of Surface Pressure and Temperature
The forecast surface pressure was the most accurate among all forecast elements: within 2 hPa
(Figure 7). The forecast surface temperature was not that accurate. However, the forecast trend
was still a useful forecasting aid. In most cases, the forecast overnight temperature drop from the
afternoon maximum was smaller than the actual value (Figure 8). Also, most of the forecast
changes in morning minimum temperature fell into the right categories (Figure 9). When applying
this to the forecast of surges of winter monsoon, it was found that the model failed under two
circumstances:
(a) on the arrival of northerly surge during fine days, the slight initial temperature rise before
dropping could not be predicted;
(b) when the surge arrived overnight, the forecast temperature drop in the morning was usually
too large.
5J2 Winter Monsoon Surges
The forecast temperature sequence was useful in the forecast of surges of winter monsoon. The
timing of the first arrival of surges was found to be correct in 2 cases out of 5. For the remaining 3
cases, the timings were all 12 hours too early. These cases occurred when the smooth topography
was in use. The bias has been reduced after a more realistic Nanlmg was introduced into the
model. Replenishment of cold air was accurately indicated in 7 cases out of 7, which resulted in
colder mornings to follow. The first warming-up after a surge was predicted in 4 cases out of 5.
For 850 hPa level, the change from northerly to southerly flow was predicted in 3 cases out of 5.
In the remaining 2 cases, the changes were also predicted but with a delay of 6 and 12 hours
respectively.
53 Changes in Vertical Profile
On several occasions, the model managed to forecast sudden drying up on the 850/700 hPa levels
which was not evident from the synoptic pattern and so would pose considerable problems for
forecasters. In general, the timing for the moistening up or drying up of the vertical profile was
accurate to within 6 hours.
6. ACKNOWLEDGEMENT
I would like to thank Dr. Yukio Kikuchi, Director-General of the Japan Meteorological Agency
for supplying the software of the optimal interpolation data analysis scheme and the very fine mesh
limited area model to the Royal Observatory for adaptation, as well as transmitting the boundary
data for daily operational runs of the models. I am also grateful to Dr. H.K. Lam and Mr. C.Y. Lam
of the Royal Observatory for their constant advice and encouragement. Thanks are also due to
Miss Kitty Chu for her excellent typing, and to Miss May Wai for her beautiful diagrams.

434
REFERENCES
Anderson, E. and A. Hollingsworth, 1988: Typhoon bogus observations in the ECMWF data
assimilation system. ECMWF Res. Dept. Tech. Memorandum No. 148.
Asselin, R., 1972: Frequency filter for time integrations. Mon. Weath. Rev.,100,487-490.
Businger, J.A., J.C. Wyngaard, Y. Izumi and E.F. Bradly, 1971: Flux-profile relationships in the
atmospheric boundary layer. J. Atmos. ScL, 28,181-189.
Gadd, AJ. and J.F. Keers, 1970 : Surface exchanges of sensible and latent heat in a 10-leveI model
atmosphere. Quart. J. R. Met. Soc., 96,297-308.
Hoskins, B.J., 1980: Representation of the earth's topography using spectral harmonics. Mon.
Weath. Rev., 108,111-115.
Kondo, J., 1975 : Air-sea bulk transfer coefficients in diabatic conditions. Bound. Layer Met., 9,
91-112.
Kondo, J., 1976: Heat balance of the East China Sea during the air mass transformation
experiment. J. Met. Soc. Japan, 54,382-398.
Kurihara, Y., 1961: Accuracy of winds-aloft data and estimation of error in numerical analysis of
atmospheric motions, J. Met. Soc. Japan, Vol. 39, No. 4,331-343.
Lenhard, R.W., 1970: Accuracy of radiosonde temperature and pressure height determination.
Bull. Amer. Met. Soc., 51,842-846.
Mellor, G.L. and T. Yamada, 1974: A hierarchy of turbulence closure models for planetary
boundary layers. J. Atmos. Sci, 31,1791-1806.
Numerical Prediction Division, JMA, 1986: Outline of operational numerical weather prediction
at JMA. Appendix to Progress Report on Numerical Weather Prediction.
Satomura, T., 1988: Dynamic Normal Mode Initialization for a limited-area model. J. Met. Soc.
Japan, Vol. 66, No. 3,261-276.
Smagorinsky, J., 1960: On the dynamical prediction of large scale condensation by numerical
methods. Geophysical Monographs No. 5, American Geophysical Union, 71-78.
Sugi, M,, 1986: Dynamic Normal Mode Initialization. J. Met. Soc. Japan, Vol. 64, No. 5,623-636.
Tatsumi, Y., 1983: An economical explicit time integration scheme for primitive model. J. Met.
Soc. Japan, 61,269-288.
Ueno, M., 1989:. Operational bogussing and numerical prediction of typhoon in JMA. JMA/NPD
Tech. Rep. No. 28.

435
55 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 U5 150 155 160 165_
65
60
55
50
45
40
35
30
25
20
15
10
5
$ ^«**
N"A\ \
^
-
/
v/ (
J
S~\
\
1
*
A
V
• i
^
s,
c
1
4,
J
^V
)
f
<
r^
^
i
f\
>
/
^j
n
is
,
V:
-( i
-^
5^
/
#
/^
>
x
V
/
•
/~
%
T
V 7
i^i'...*^
'
.
X*
X
I/
^
0
65 70 75 80 85 90 95 100 105 110 nS 120 125 130 135 140 145 150 155 160 165
60
55
50
45
40
35
30
25
20
IS
10
5
Figure 1
Horizontal domain of the Royal Observatory analysis scheme

436
Figure 2
Vertical structure of the Royal Observatory
limited area model
140.
Figure 3
Horizontal domain of the Royal Observatory limited
area model. The bold line is the model coastline.
The dotted region is the lateral boundary region

437
Nanling
Figure 4
Topography (TG) of
the JMA Global
Spectral Model
Figure 5
Envelope topography (TE)
of the RO limited area
model
c -
, ^'
Nanling
< 1010-
24 HOUR FORECAST MEAN SEA LEVEL PRESSURE
Figure 6 Combined topography (T) Figure 7 Accuracy of forecast mean
sea level pressure

438
36 HOUR FORECAST MEAN SEA LEVEL PRESSURE 48 HOUR FORECAST MEAN SEA LEVEL PRESSURE
Figure 7
(con't) Accuracy of forecast mean sea level pressure
Si
1!
I
.0 '*
FORECAST OVERNIGHT TEMPERATURE DROP
{ FROM AFTERNOON MAXIMUM )
VERIFICATION OF FORECAST TEMPERATURE TREND
(1711.-15.12.1986 )
-7 -6 -5 -4 -3 -Z -I 0 I 2 J ' 4 5
FORECAST CHANGE IN MORNING MINIMUM TEMPERATURE
VERIFICATION OF FORFCAST TEMPERATURE TREND
.(17n.-15.t2.l988 )
Figure 8
Accuracy of forecast
overnight temperature
drop from afternoon
maximum
Figure 9
Accuracy of forecast change
in morning minimum temperature

439
BMC LIMITED AREA MODEL: OPERATIONAL APPLICATION AND RESEACH
Guo Xiaorong, Yan Zhihui & Zheng Guoan
National Meteorological Center, State Meteorological Administration,
Beijing •
ABSTRACT
The limited area 5-layer primitive equation model, as the first
operational precipitation forecast model in China, has been run at Beijing
Meteorological Center (BMC) for more than five years. The operational
results show that this model gives continuous services without
interruptions, and the forecast skill is satisfactory for the forecast of
weather systems such as extratropical cyclone, cold front, meiyu front
and the precipitation associated with them. Forecast guidance based on
the model is widely used at local weather services nowadays.
Some experiments on the sensitivity of sub-grid scale convective
effects to horizontal resolution were carried out with a nested version of
this model. The results show that the contribution of the sub-grid scale
convective condensation, computed by the Kuo's scheme, to the total
precipitation decreases remarkably with the increase of horizontal
resolution. It is 62% of total precipitation in the 381km grid, but only
10% in the 47.625km grid. It seems that the precipitation produced by
grid scale condensation will take a dominate proportion in total
precipitation when the resolution is high enough (less than 50km).
I. INTRODUCTION
Since August 1983, a limited area 5-level primitive equation model
has been running once a day at BMC for the routine regional forecast.
The purpose of this model is to make short-range forecasts of synoptic
circulation systems as well as 24h total amount of precipitation over
China. Forecast guidance is sent to local weather services by facsimile
transmission and the grid variables in code. This model has proved to
be effective in forecasting the development of some synoptic systems,
such as cyclone, landing typhoon, front and the precipitation brought
about by them.
The original form of this model was designed by a NWP research
group in Peking University (NWP Group, Peking University, 1980), Some
improvements including the treatment of lateral boundary condition, a
modified Kuo's convective parameterization scheme and the initialization
were.made for the operational application ( Guo, et al., 1982 and Zhang.et

440
al.,1984). In this paper, a general description of the model physics and
the operational results are given.
The precipitation in the atmosphere is related to two kinds of
condensation processes: large-scale ascending motion and cumulus
convection. In most numerical models, the conve'ctive condensation is
considered as a kind of sub-grid scale physics and computed by the
so-called "parameterization" method. But calculating large-scale condensation
and cumulus convection separately would not simulate the
precipitation process correctly since the two processes interact with
each other. However, the interaction is too complicated for us to give an
exact explanation. For example, removal of cumulus convection does not
decrease would rather increase the rainfall amount sometimes(Guo,1987).
On the other hand, as a "sub-grid scale" physical process, cumulus
convection should play different roles in models with different
resolutions. It is universally recognized that increasing the resolution is
efficient in improving the precipitation forecast. It is also necessary to
examine the effects of these two kinds of condensation processes on
precipitation forecasts as resolution increases. A sensitivity experiment
of sub-grid scale convection effects (Kuo's scheme) on horizontal
resolution has been carried out, and the results are shown in the IV
section of this paper.
II. DESCRIPTION OF THE MODEL PHYSICS
The precipitation scheme consists of two processes. The first one is
large scale condensation, which is similar to that presented by
Smagorinsky(1965). If the relative humidity exceeds 80% of saturation, the
excess moisture will be condensed and droped down as precipitation. At
the same time the temperature is increased to conserve the moist static
energy. The variations in temperature and specific humidity produced by
the large-scale condensation are expressed as follows
1 C
where Cp is specific heat, L is latent heat, and Rv is the gas constant of
water vapour. The re-evaporation of falling rain-drops is neglected.
The second process is convective condensation, for which the
modified Kuo's scheme (1974) is used. Earlier tests with Kuo's scheme
resulted in predicted rainfall areas much wider than those observed
(Zhang, et al.,1981). In order to concentrate the predicted rain area,
criteria controlling the convective parameterization are required, viz. the
stratification must be conditionally unstable at the beginning of
convection; and both low level moisture flux convergence and velocity
convergence must exceed some given critical values, which are empirical
constants.
The rate of convective precipitation
Re is determined by
where Mt is the moisture flux convergence in the vertical column of
the modeling atmosphere, and it can be calculated from

1 - 00
(P S V
-j mV- —
JQ.4-6 m
where b is a parameter, indicating the fraction of moisture flux
convergence, which is not dropped out as convective rain. According to
Anthes (1977), it is dependent on the mean relative humidity RH of the
ambient air, i.e.,
0.3 , RH

442
where subscript x denotes u, v, T and q respectively. DH is the
diffusion coefficient, which is a spatial function and decreases from the
boundary to the interior. In the inner region (from fifth row), DH takes
a constant value of 10 s m 2 s~ 2
III. SUMMARY OF THE OPERATIONAL USE
Since the precipitation is difficult to predict, we focus the attention
on the verification of the precipitation forecast to assess the model
performance.
The evaluation of the precipitation forecast has been made
synoptically and statistically. Generally speaking, the forecasts for the
western part of China are not adequate, especially for the region close
to the Xizang (Tibet) Plateau. As regards eastern China, the forecast
guidance is a useful reference in the forecasting of the rainfall
associated with synoptic and subsynoptic scale systems.
1. Subjective Evaluation
Case 1: On 8-10 August 1984, an intensive typhoon which made
landfall brought about heavy rainfall over North and Northeast China
Fig.L The 24-hour precipitation forecasts and the observation
(F--prediction, Q~-observation, a—OOZ 8th-OOZ 9th Aug. 1984,
b—OOZ 9th~OOZ 10th Aug. 1984.C—OOZ lOth-OOZ llth Aug. 1984)

443
and caused severe flood. Local rainfall rates of lOOnim per 24 hours
were sustained during 8-10 August. Fig.l shows the 24h predicted and
observed accumulated rainfall. Both the area and the intensity of the
precipitation were predicted successfully. The area covered by the
predicted rainfall is nearly coincident with the observations. The
predicted maximum value of 24h rainfall is more than 100mm for the last
two days, although it is still less than the
observation.
Case 2: A southwest
vortex formed in the Sichuan
Basin on 14 June 1986. When
a cold front tied with a
Mongolian cyclone reached the
Changjiang valley on 15th, the
vortex moved eastward out of
the Basin and a coastal
cyclone was initiated at the
Changjiang Estuary on 16th.
The precipitation associated
with the vortex and the front
intensified and moved eastward.
A heavy rain belt with a
maximum rate of about 100 mm
per 24 hours occurred over the
north of the middle and lower
reaches of the Changjiang
River. The rain area located at
Northeast China was associated
with the Mongolian cyclone
moving eastward (Fig.2, right).
The corresponding 24-hour
precipitation forecasts are also
shown in Fig.2 (left). As
mentioned above, when the
southwest vortex stayed in
Sichuan east of the Xizang
Plateau (14-15th), the rain
associated with the vortex was
not well predicted. But the
movement and intensification Fig.2. The 24-hour precipitation
of the precipitation are predicted
quite well (15-16th),
diction, 0--observation, a—OOZ
for casts and observation (F—pre-
even though the predicted 14th~OOZ 15th June 1986, b—OOZ
amount of precipitation is not ISth-OOZ 16th June 1986)
as much as that observed.
2. Statistical Evaluation
Thirty weather stations in the middle and east of China are selected as
a base to verify the forecast skill of the model. Three scores, i.e.,threat
score (T), pass over score (PO) and swing with no hit (NH)

444
AM
AM
PO-
N3
A/i + N3'
and
NH--
N2
N\
are evaluated, where Nl, N2 and N3 are the total numbers of stations
corresponding to each class illustrated in Table 1. For example, N3 is the
number of stations where precipitation is observed but not predicted.
Whether 'precipitation' or 'non-precipitation' predicted at the stations is
defined as follows. For every selected station, there are four adjacent
grid points. It is considered that precipitation is predicted at that
station if precipitation is predicted at one or more of the grid points. T
score denotes the rate of the correct forecast. PO denotes the rate of the
observed but not predicted precipitation and NH the rate of the predicted
precipitation which did not occur.
Table 1. Illustration of the number of stations
Observed
Precipitation
Non-Precipitation
Predicted
Precipitation
Nl
N2
Non-Precipitation
Obviously, the closer to 1 T is and the smaller PO and NH are, the
better the forecast is.
The monthly mean values of score T,' PO and NH from July to
November 1984 and 1988 are shown in Table 2.
Table 2. Monthly Mean Values of Scores
Month
JuL
Aug.
Sep.
Oct.
Score
T
0.46
0.51
0.55
0.45
1984
PO
0.46
0.41
0.23
0.17
NH
0.30
0.22
0.33
0.49
1988
Nov. 0.29 0.20 0.69 0.15 .0.75 0.60
It can be seen that the T score is higher in summer than that in
autumn. Obviously.all of the scores are very poor and they are much
different in the different years. We do not think these scores can
express objectively the skill level of precipitation forecast. But,
unfortunely, there is no universally accepted method for use to verify
precipitation forecasts in the world,
IV.
SOME EXPERIMENTS ON PRECIPITATION
T
0.27
0.42
0,34
0.23
PO
0.61
0.43
0.49
0.70
N3
NO
NH
0.51
0.45
0.49
0.29
As one of the attempts to improve the forecast of rainfall amount, we
examined the effects of varying horizontal resolution on precipitation

445
forecast. Based on a nested version of this model, four grids with the
grid length of 381km (A), 190.5km (B), 95.25km (C) and 47.625km(D) were
designed and nested one by one. Because our attention was focused on
the rainfall amount forecast, the finest grid (D) was located covering the
heavy rain area in Case 1 mentioned above. Using the initial data at
OOZ of August 9, 1984, the 24-hour precipitation forecasts were made
Fig.3. 24-hour precipitation
forecasts
with these four grids (Fig.3). It can be seen that the predicted
precipitation both in the rain area and amount gets more and more close
to the observation (Fig. 1 O-b) as the horizontal resolution increases from
381 to 47.625km. The model with 381km resolution predicted too wide a
rainfall area and a maximum 24-hour precipitation amount of only about
46mm. With the 47.625km grid, however, a southwest to northeast
oriented rain belt with a maximum amount of 234mm were predicted which
compares well with the observation.
Fig.4 illustrates the maximum and areal mean amounts (at a 381x
381 km rounding the maximum value) of the predicted 24-hour total(RT),
large-seal condensation (RL) and convective precipitation (RC) by the
four grids. It can be seen that the maximum amount of RT and RL
increase remarkably with the decrease of the grid length. But for RC

446
there is only a minor difference among these four grids. For grid A
(381km), the maximum value of RC is larger than that of RL, while for
grid D (47.625km) the value of RL is much larger than that of RC and
approaches the maximum amount of RT. The contribution of the two kinds
of condensation to total
precipitation can be presented
by the areal mean
values of RT, RL and RC.
The proportion of RL in
RT increases with the
increase of resolution. It
is about 38% for the
381km grid and increases
up to 90% for the
47.625km grid. On the
contrary, the proportion
of RC in RT decreases as
the resolution increases.
It is about 62% for the
381km and only 10% for
the 47.625km grid. This
is identified by the
24-hour forecasts of RT,
RL and RC predicted by
the grid A and D(Fig.5).
For the 47.625km grid,
the distribution of RT
almost coincides with
that of RL. But the
distribution of RT predicted
by the 381km grid
is similar to that of RC
rather than that of RL. Fig.4. Maximum(top) and areal mean(bottorn)
amounts of RT, RC and RC
predicted by grid A, B, C and D.
V. Summary
1. The operational limited area model used at BMC is fairly simple.
But the operational results show that (1) the model is stable and a
continuous service can be provided and (2) the model has some merits in
forecasting the movement and development of some synoptic scale
systems, such as front, cyclone, vortex and the associated precipitation.
2. The precipitation forecast is effectively improved, in terms of both
rainfall area and amount by increasing the horizontal resolution.
3. The relative contribution of covective condensation, computed by
Kuo's scheme, in fine mesh grid models is not as much as in coarse grid
models. The grid scale condensation dominates in the total precipitation
when the resolution is high enough (less than 50km ).

447
Fig.5. 24-hour forecasts of RT t RL and RC for gid A and D
REFERENCES
Anthes, R. A. (1977), A cumulus parameterization scheme utilising an
one-dimensional cloud model, Mon. Wea. Rev. 105: 270-286.
Chen S.J.,Zheng L.J. and Zhang X.W. (1980), A method of initializing the
primitive equation model including real wind and height data, Ada
Meteor. Sinica, 38: 122-129 (in Chinese).
Cressman,G.P.(1959), An operational objective analysis system, Mon. Wea.
Rev. 87: 367-374.
Davies,H,C.(1976), A lateral boundary formulation for multi-level
prediction model, Quart. J. Roy. Meteor. Soc., 102: 405-418.

448
Guo X.R.,Yan Z.H.,Zhang Y.L. and Chen S.J.(1982), A limited area
primitive equation model for operational use, in Proceedings of the
Third National Symposium on Numerical Weather Prediction, Science
Press, Beijing, pp. 77-82 (in Chinese).
Guo X.R.and J.Hoke(1987), The impact of sensitive and latent heat release
on prediction of precipitation associated with an intensive cyclone,
KEXUE. TONGBAO, Vol.32 No. 10 683-688.
Krishnamurti.T.N. and Moxim,W.J. (1971), On parameterization of
convective and nonconvective latent heat releases, J. AppL Meteor.,
10; 3-13.
Kuo,H.L.(1974), Further studies of the parameterization of the influence of
cumulus convection on large-scale flow, J. Atmos. Sci., 31:
1232-1240.
Mesinger, F.and Arakawa, A. (1976), Numerical method used in
atmospheric models. Vol.I, OARP PubL Ser. No.17, 64 pp.
NWP Group of Peking University, (1980), A five-level primitive
equation model used on precipitation forecast, in Proceedings of the
Second National Symposium on Numerical Weather Prediction, Science
Press, Beijing, pp. 1-12 (in Chinese).
Zhang, Y.L.and Yan, S.Y. (1981), Experiment of convective parameterization
in the numerical prediction of heavy rainfall, Acta Meteor. Sinica,
39: 10-17 (in Chinese).
Zhang,Y.L.and Yan,Z.H.( 1984), The improvement of convective parameterization
scheme in BMC regional model, BMC Office Note No.84-13
(in Chinese).

449
THE OPERATIONAL GLOBAL FORECAST SYSTEM AT CENTRAL WEATHER BUREAU
Chuen-Teyr Terng
Computer Center, Central Weather Bureau,
Taipei, China
ABSTRACT
Because of the limitations of subjective weather forecasting and
the advantages of numerical weather forecasting (NWP), the Central
Weather Bureau (CWB) has begun a five-year project to establish a NWP
system. Due to the fact that Taiwan is situated at the boundary of a
large continent (the Euro-Asia continent) and a large ocean (the
Pacific), and also the fact that Taiwan is located in the sub-tropical
region, being affected by different weather systems in winter and in
summer, four different NWP models are required. The four models are :
Global (for 5 to 7-day forecast and 4-D data assimilation), Regional
(for 2-day forecast and weather patterns of the Asian area), Mesoscale
(for 1 day forecast and weather patterns around Taiwan), and Typhoon
Track (tracking typhoon movements around Taiwan). Besides these
models, there is a control system, which automatically monitors the
entire system, and a data-acquisition system.
A super computer (CDC-205) and a computer for the front end (CDC-
840) were installed in CWB in December 1986. Beginning in July 1987,
the global model underwent testing and it was declarea operational in
July 1988. The global model composes of three parts : analysis,
initialization and forecasting. The objective analysis portion uses
FIB and Barnes's schemes for the analyses of sea-level surface
pressure and upper air mass and wind field.. The initialization scheme
uses variational method. The UCLA general circulation model forms the
basis of the forecasting part.
The procedure of the global model is as follows : The global
observations are transmitted to CWB through GWDI and JWA's satellite
network. The data are then decoded and error-checked by CWB's frontend
computer. The data are then transmitted to the super-computer for
sea-level surface pressure analysis, upper air mass field and wind
field analyses, and initialization, all on isobaric coordinates. The
data are next interpolated to sigma-coordinates. Forecasting
computations are performed as the next step. The forecast results are
interpolated back to isobaric coordinates and drawn as weather charts
for use by the forecasters.
With the advancement of numerical methods, the enlargement of
computer memory and the arrival of faster computers, NWP is facing
rapid improvements. In view of this, CWB is planning its 2nd
generation of NWP system. This includes the introduction of optimal
initialization, normal model initializationj spectral forecasting
model and more advanced physical parameterization. We hope that in the
near future, our forecasting capabilities can be compared with those
of the most advanced operational centers of the world.

450
THE EAST ASIA HEAVY RAINFALL NUMERICAL FORECASTING AND
THE NUMERICAL NOWCASTING OF SEVERE CONVECTION WEATHER
Zhou Xiaoping
Institute of Atmospheric Physics,
China Academy of Sciences.
(IAP/CAS) Zhongguancun, 100080, Beijing, China
ABSTRACT
The highest 24-hour intensity of heavy rainfall in China nearly reaches the world maximummore
than 1000 mm/day because of the interaction of the East Asia Monsoon and tropical
systems. Meteorologists in China have concentrated their attention on this topic for a long time.
As far as numerical forecasting is concerned, there are three or four numerical limited area
models used in the Beijing Meteorological Center, in regional centers and In the Water
Conservency institutions of China. Some of them have been very successful.
The main dynamical condition for the 24-48 heavy rainfall forecast is the combination of a
large water vapour flux convergence (V.qV) at lower levels and a strong divergence at upper
levels. The point is how to correctly catch the convergence by using poor data.
An IAP heavy rainfall model is one of this kind of hydro-static meso-a numerical model which
has the capacity to give detailed heavy rainfall process with a 3-hour time resolution.
The 3 hour or 6 hour output from the model can be a background for severe storm nowcasting
use. But, to forecast the formation and movement of a storm or 3 MCS with diameter from 20 km
to 200 km displayed on radar screen and satellite picture needs a non-hydrostatic meso-8 model.
IAP now is developing such a kind of model for the future operational use at the time when we
shall have more powerful computer, sufficient data by Doppler radar and TOVS data from
satellite. We take the Klemp and Wiihelmson model (1978) as a prototype with some
modifications which can catch the most important physical processes through experiments and
make the forecasting easier in the nowaday technical condition. We have simplified the micro
physics in the clouds and the turbulance coefficients with only two orders of accuracy.
A 2-km horizontal grid-size and 1-km in vertical are used with the purpose so that we need
predict only the formation and movement of the storms larger than 10 km in diameter rather than
the fine structure of the storm. Twenty seconds is taken as the time-step due to computational
restrictions. This meso-8 non-hydrostatic storm model is nested in the IAP meso-a hydrostatic
model which provides the lateral boundary conditions and the initial data such as the
temperature deviation and the strong convergent area. The domain is 60x60 km and 13 to 20
levels. The computer time is ten minutes for one hour forecasting by using the Convex-cl
(multiple processors 40 MIPS) in IAP,
As a first step, the results are very encouraging and useful to mesoscaJe observation
designing and forecasting.

451
Numerical Simulation of Mesoscale Meteorological
Phenomena in Hong Kong
K.K. Yeung, W.L. Chang, B. Wan
Royal Observatory, 134A, Nathan Road, Kbwloon, Hong Kong.
1. Introduction
Traditionally, methods for the objective forecast of
temperature are based mostly on empirical relationships between temperature,
wind speed and direction, sky condition, dew point and other
meteorological parameters. Such methods have been developed by Chin
(1974) for use in Hong Kong. Other examples include the probabilistic
methods of Murphy and Winkler (1974), Winkler and Murphy (1979) and
the regression formula given by Roodenburg (1983). For applications in
complex terrain, McCutchan (1979) has given a method based on Fourier
analysis. As for climatic applications, one can refer to Hansen and
Driscoll (1977) for the generation of hourly temperatures.
With the advent of high speed computers, numerical models
have been employed to give temperature forecasts. An example of this
is a primitive equation model, by Druyan (1974), which give short
range forecasts of global temperatures. Models simulating the evolution
of the atmospheric boundary layer have also been used such as the
one described by Stull (1984) which can be applied in relatively level
terrain far away from cold fronts.
For operational forecasting of temperature, the method
which, in recent years, seems to have gained the greatest popularity
is Model Output Statistics (MOS). This method seeks to establish
statistical relationships between the prognoses from the numerical
models and the weather elements to be forecast. For temperature
forecasts, trials using the MOS method have been reported by several
researchers, amongst whom are Klein et.al.(1975) for the United
States; Francis et.al. (1982) for the United Kingdom; Woodcock (1984)
for Australia and Lemcke et.al. (1988) for the Netherlands. The relationship
between subjective and these numerical-statistical methods
has been examined by Murphy et.al. (1988).
To provide forecasts of temperature (as well as other
weather elements) for Hong Kong's fast growing new towns presents
special difficulties and challenges. This is because of the rugged and
inhomogeneous nature of Hong Kong's terrain (Fig v 1). The lateral
dimension of the topography averages 20 km or less so that if the
influence of the local features is to be taken into consideration, the
mesh size of the model should be less than 10 km. This paper reports
on our numerical experiments using such a local circulation model

452
(described in section 2) carried out to see if it can simulate spatial
and temporal temperature variations for the region and with the aim of
eventually using it to supplement the larger scale forecast from VFM
models or ECMWF.
Fig. 1
1 km GRID TERRAIN (150 x 150 grids )
AREA INCLUDES PEARL ESTUARY. HONG KONG & DAYA BAY
2. The Numerical Model
The local area model used in this experiment is a three
dimensional prognostic primitive equation model with Boussinesq approximation.
A more detailed description of the model may be found in
Kikuchi et al (1981) and Kimura (1983).
The heat budget calculation which changes the ground
temperatures follows Bhumralker's scheme (1975). However to simulate
the sky conditions, an extra factor which adjusts the amount of incoming
short wave radiation is employed. In the constant flux layer, the
vertical eddy diffusivity is obtained from the flux profile relationship
of Dyer and Hicks (1970) and of Kondo et al (1978) for the stable
and unstable stratification respectively. For the other layers, turbulence
is parametrized by Mellor and Yamada's level two turbulence

453
closure model (1974).
Radiation boundary conditions are imposed at the four
lateral boundaries. This is certainly not the best approach especially
for rapidly changing boundary conditions but is sufficient for the
present study which is aimed at providing indications up to 12 hours
under rather steady meteorological conditions. The ground is treated
as a no-slip boundary and the model top a material surface. The equations
are written in terms of the terrain following co-ordinates.
The model domain for this experiment is about 300 km x 300
km with grid size of 5 km. There are 16 layers for the vertical and
the model top is about 7 km above mean sea level.
3. Results and Discussion
Two typical cases of winter monsoon surges were simulated,
one under overcast conditions and the other having some bright periods
during the day. For each case, a simulation using "nearest to actual"
solar heating conditions and one with "hypothetically extreme" solar
heating conditions were made. The purpose is to highlight quantitatively
the solar effects as well as to establish the significance of
the results. In order to have a better feel of the terrain influences,
the overcast case was nested down to 2-km grid and the result was
compared with those obtained from the 5-km grid and the VFM with grid
size of 111 km.
3.1 Simulation of a Surge with Overcast Conditions
On January 11, 1989, the temperature and pressure gradients
were tight over South China. A surge of northerly winds arrived
Hong Kong around late evening. Temperatures over the territory were
between 19 to 21 degrees initially (12 UTC). The temperatures fell
steadily to around 15 degrees in the next 24 hours.
To simulate this event, the 12-UTC upper air data from
Hong Kong and a number of Chinese stations in Guangdong were used to
initialize the model.
The result of the simulation suggests a drop of about 4°C
throughout the coastal region. This compares favourably with the
actual temperature fall of the region. To simplify the validation of
this model result, time series of the actual observations (ACTUAL),
time series of the forecast by the full simulation (MODEL) and that
forecast by the control simulation using in this case clear sky
conditions (M-SUNNY) at three strategic locations, namely the Royal
Observatory (RO), Macau and Guangzhou (GZ) are plotted in Fig. 2a, 2b
and 2c. The root mean square (RMS) errors* the bias and the standard
deviation (std. dev.) are evaluated and listed in Table 1 below.

454
Fig. . 2. Validation of Temperature Forecasts at (A) te the Royal Observatory Hong Kong (R.O.);
(B)) Macau; and (C) Guangzhou (O.Z.) during a Northerly Monsoon Surge with Overcast Conditions.
Simulation results from the Same Model Having Clear Sky Conditions (& M-Sunny); and
from m a Larger L Scale Model with Orid Length about 100 km (x 1 1 1 km) are included for Comparison.
FIG. 2 a CASE 1 — R.O.
FIG. 2 b
CASE 1 — MACAU
CASE 1 —
G.Z.
24 hr RMS error
bias
std. dev.
1st 12 hr RMS error
bias
std. dev.
RO
1.4
•0.3
1.4
1.4
•1.2
0.7
GZ
2.2
1.7
1.4
1.6
1.0
•1.3
Table 1
Macau
1.3
0.5
1.2
0.8
-0.4
0.7
M-SUNNY
4.8
2.0
4.4
2.1
-1.8
1.0
The RMS error is about 1.4 degrees for RO and Macau. It is
higher for the Guangzhou area (2.2 degrees) but is to be expected
since it is located nearer to the northern boundary and hence, is more

455
prone to boundary contamination. This is supported by the smaller mean
error (1.6 degrees) for the initial 12 hours. As for the trend, the
predicted initial temperature fall over the Hong Kong area for the
first 6 hours is not observed. This may be due to the initial surface
temperature field being ill-represented as there are only two upper
air stations within the domain to influence the initialization.
The results from the control run are apparently poor in
terms of both absolute values and trends in comparison
simulation
period. The
with the full
run. This is particularly so for the latter half of the
12-hour and 24-hour RMS errors are 2.1 and 4.8 degrees
respectively for Hong Kong. The control run predicts a sharp rise in
temperatures after day break to a maximum value of 25 degrees while
the observed temperatures actually fell steadily to 18 degrees. This
result clearly demonstrates that the solar effect is an influential
factor in local temperature forecasts.
The forecast for Hong Kong by the VFM model for the same
period are also plotted in Fig. 2a for comparison. This model predicts
a trend which is close to the actual observation but is about 3°C
lower. In view of (i) the tight temperature gradient over South China,
i.e.
about 6°C per 100 km, and (ii) the fact that VFM grid temperatures
are only mean values over an area of 111 km x 111 km, a deviation
of 3°C of the grid value
from any point measurement should be
regarded as reasonable. Nevertheless, this comparison suggests that
model resolution also plays a significant role in local temperature
forecasts.
3.2 Simulation of Surge with Bright Periods during the day
To supplement the findings in Section 3.1,another event
was selected. On March 4, 1989, a cold front reached the South China
Coast bringing the temperatures of Hong Kong down from about 18°C in
the morning to around 11°C in 24 hours' time. But, during the early
part of the day, there were some bright periods and temperature over
the Hong Kong Harbour area had even risen to a maximum of 20 degrees.
This combined solar and advection effects are simulated
using the 00-UTC upper air data from Hong Kong and a few stations in
Guangdong. The short wave radiation terms are multiplied by 0.4 to

456
simulate the partly cloudy condition of the day. A control run using
overcast sky conditions to contrast the difference were also made. As
before, the time series
of the actual observation (ACTUAL), the full
simulation (MODEL) and the control (M-CLOUDY) are plotted in Fig. 3a,
3b and 3c. The root mean square errors are listed in table 2 below.
Fig. 3. Validation of Temperature Forecastt at (A) the Royal Observatory Kong (R.O.);
FIG. 3b CASE 2 - MACAU
a MOtm. * Acnwa- HOU " <
Table 2
RO GZ Macau M-CLOUDY
24 hr RMS error 2.0 3.0 2.4 2.8
bias 1.5 2.9 2.3 -0.7
std. dev. 1.3 0.8 0.4 2.7
1st 12 hr RMS error 0.8 3.3 2.6 3.4
bias 0.5 3.1 2.6 -2.9
std. dev 0.7 1.1 0.3 1.7

457
RMS errors for RO is relatively small, 2.0°C and 0.8°C for
the 24 hours and 12 hours respectively. Trend-wise, the afternoon peak
for the Hong Kong area is well predicted but not the sharp drop in
temperatures after 22 hours local time. The forecast values for GZ and
Macau are not so acceptable, both have RMS errors above 2.5°C. The
trend for Macau is however very close to the actual (the bias is large
but the std. dev. is only 0.3°C). The large discrepancies are due to
an inaccurate initial temperature field near the Macau area as attested
by the 3°C lower initial temperature. As for Guangzhou, the deviations
are likely to be the result of wrongly assumed solar condition
because Guangzhou was actually overcast for the day.
The results of the control run for RO and Macau are obviously
less accurate in comparison with the full simulation results.
However, to the contrary, results for GZ compares better with observation
which supports the finding in the last paragraph.
The VFM model outputs are plotted in Fig. 3a for comparison.
It has greater error values for the first 12 hours but decreases
towards the end of the period. The trend prediction is less successful.
To test the case further, another run was made using the
12-UTC upper air data of the same day as the initial data. Only the
results for RO are plotted here on Fig. 4 . The forecast results again
compares favourably with observations especially for the first 12
hours. Both the minimum and the maximum temperatures
degree of accuracy.
are within one
Fl«, 4.
(R.0.1:
Further Validation of Temperature Forecast at (A) the Royal Observatory Hong Kong
FIG. 4
CASE .3— P.O.

458
3.3 Nesting Run with 2 km Grids
The overcast case was simulated again using
a 2-km grid
system nested from the 5~km simulation to see if more accurate terrain
description will improve the results. The time series for the three
locations are shown in Fig. 5a, 5b and 5c. The 5-km results and the
VFM outputs are plotted together for easy comparison. The figures
indicate that the 2-km forecast is only marginally better than the 5-
km forecast.
Fig. 5. Comparison of Temperature Forecasts Obtained from Models Using Grid Size of 2 km,
5 km and 111 km for the Locations (A) the Royal Observatory Hong Kong (R.O.); (B) Macau
and (C) Guangzhou (G.2.).
FIG.
5d — R.O.

459
Comparison of the surface temperature fields (results not
shown here) at 5-km and 2-km shows that the overall patterns are
similar. Both simulations predict that the Pearl Estuary will hold
back the advancement of the cold air and temperatures at Macau will be
four degrees lower than that at RO. There are some evidences of warm
air being held back behind the hills over the Hong Kong territory for
the 2-km run but not for the 5-km run. However, with only limited
test results, this difference may not be statistically significant. A
smaller grid size test was not made by reason of the much longer
computer time required.
4. Conclusion
This experiment
shows that dynamic numerical models at
fine enough grid resolution (in this case, 5-km) under near steady
meteorological conditions are able to augment synoptic scale models in
providing more detailed temperature field forecast for local regions.
As the current model uses only one sky condition for the
whole domain and forecast period, local variation in cloud cover will
cause disparity between forecast value and observation.
Experiment with 2-km nested grid yields no marked differences
against the 5-km grid results and more test runs are needed to
elicit the influence of terrain.
The experiment also shows that forecasts longer than
twelve hours are affected by inadequate boundary treatments. Therefore,
unless accurate boundary conditions are available, say from
larger area models, increasing the domain size to about 1000 km may be
needed to achieve a better 24-hour forecast.
5. Acknowledgement
The authors would like to thank the Director of the Meteorological
Research Institute
of Japan for making the model available
for the present study. The authors also wish to thank Mr. K.Y. Tarn and
Mr. K.L. Tang for the preparation of the graphs and the Data Processing
Division of the Royal Observatory
and support.
for their computational services

460
REFERENCES
Bhumralker, C. M., 1975: Numerical experiments on the computation of ground surface
temperature in an atmospheric general circulation model. J, Appl. Meteor., 14,
1246-1258.
Chin, P. C., 1974: Prediction of daily minimum temperature in Hong Kong during the cool
season. Technical Note (local) No.19, Royal Observatory, Hong Kong.
Dyer, A. J., and B. B. Hicks, 1970: Flux-gradient relationships in the constant flux
layer. Quart. J. Roy. Meteor. / Soc., 96, 715-721.
Druyan, Leonard M., 1974: Short-range forecasts with the GISS model of the global atmosphere.
Man. Wea. Rev., 102, 269-279.
Francis, P. E., A. E. Day and G. P. Davis, 1982: Automated temperature forecasting, an
application of Model Output Statistics to the Meteorological Office numerical
weather prediction model. Meteorol. Mag., Ill, 73-87.
Hansen, James E., and Dennis M. Driscoll, 1977: A mathematical model for the generation
of hourly temperatures. J. Appl. Meteor., 16, 935-948.
Kikuchi, Y., S. Arakawa, F. Kimura, K. Shirazaki and Y. Nagano, 1981: Numerical study on
the effects of mountains on the land and sea breeze circulation in the Kanto
district. J. Meteor. Soc. Japan, 59, 723-738.
Kimura, F. 1983: A numerical simulation of local winds and photochemical air pollution,
(1): Two dimensional land and sea breeze. J. Meteor. Soc, Japan, 61, 862-878.
Klein, W. H., and G. A. Hammons, 1975: Maximum and minimum temperature forecasts based
on model output statistics. Mon. Wea. Rev., 103, 796-806.
Kondo, J., 0. Kanechika and N. Yasuda, 1978: Heat and momentum transfers under strong
stability in the atmospheric surface layer. J. Atm. Sci., 35, 1012-1021.
Lemcke, C., and S. Kruzinga, 1988: Model output statistics forecasts: three years of
operational experience in the Netherlands. Mon. Wea. Rev., 116, 1077-1090.
McCutchan, Morris H., 1979: Determining the diurnal variation of surface temperature in
mountainous terrain. J. Appl. Meteor., 18, 1224-1229.
Mellor, G. L., and T. Yamada, 1974: A hierarchy of turbulence closure model for planetary
boundary layer models. J. Atmos. Sci., 31, 1791-1806.
Murphy, Allan H., and Robert L. Winkler, 1974: Credible interval temperature forecasts:
some experimental results. Mon. Wea. Rev., 102, 784-794.
Murphy, Allan H., Yin-Sheng Chen and Robert T. Clemen, 1988: Statistical analysis of
interrelationships between objective and subjective temperature forecasts. Mon.
Wea. Rev., 116, 2121-2131.
Roodenburg, J., 1983: Forecasting urban minimum temperature from rural observations.
Meteorol. Mag., 112, 99-106.
Stull, Roland B., 1983: Integral scales for the nocturnal boundary layer, Part 1: Empirical
depth relationships. J. Climate Appl. Meteor., 22, 673-686.
Winkler, Robert L., and Allan H. Murphy, 1979: The use of probabilities in forecasts of
maximum and minimum temperatures. Met. Mag., 108, 317-329.
Woodcock, F., 1984: Australian experimental model output statistics forecasts of daily
maximum and minimum temperature. Mon. Wea. Rev., 112, 2112-2121.

461
AN OVERVIEW OF PRESENT TYPHOON FORECAST OPERATION IN TAIWAN
Shinn-Liang Shieh
Central Weather Bureau, Taipei, China
ABSTRACT
In Taiwan, typhoons are classified into four intensity classes
according to the central maximum surface wind speed, namely, weak,
moderate, intense and super typhoon. The current operational
techniques employed in typhoon forecasting at the Central Weather
Bureau will be discussed in this paper. Analysis of satellite
imageries and radar observation data are routinely used to determine
the central location of typhoons. The results have indicated quite
reasonable on positioning moderate and intense typhoons, however they
may produce large errors for weak typhoons, and for typhoons whose
centers are not vertically aligned. After long term verifications of
forecast methods used for track predictions, it is found the wellknown
HURRAN and CLIPER methods provide better forecasts and yield
mean 24 hour vector errors of 170 and 173 km, respectively, as
compared with the 177 km of the official forecasts. However, it is
also revealed that the empirical subjective forecasts still play
important roles for predicting typhoons that turn suddenly, loop, or
move erratically.

462
The Impact of the Termination of Aircraft Reconnaissance on
Tropical Cyclone Warnings and Forecasts
in the Western North Pacific
by
Johnny C. L. Chan and K. P. Wong
Royal Observatory, Hong Kong
Abstract
In August 1987, the reconnaissance mission into tropical cyclones over
the western North Pacific was terminated. Since then, operational forecasting
of these cyclones has been made without this piece of near-groundtruth
information. This paper compares the warnings and forecasts made by
two operational centres in this ocean basin prior to and after the termination
of reconnaissance to determine its impact.
It is found that the initial position errors of both centres are larger
in 1988 (when no reconnaissance information was available) than those in the
period 1983-87 by about 30%, especially for weaker systems. Differences
between warning positions issued by the two centres are also larger. Uncertainties
in the warning positions have led to the 24-hour operational forecast
errors for 1988 being larger than in previous years even after the
'forecast difficulty' of individual years has been taken into account.
However, no impact is apparent for the 48- hour subjective forecasts
although forecasts from a persistence-climatology type method are still
affected. The effect of the lack of reconnaissance information is also
apparently felt in the warning intensity as the warning intensity errors in
1988 are the highest within the entire data set for one of the centres.
1. Introduction
Since the late 1940s, the United States military had been responsible
for sending reconnaissance planes into tropical cyclones over the western
North Pacific (WNP). The main objective of such missions was to determine
the location and intensity of these cyclones. However, the reconnaissance
mission was terminated during August 1987. Since then, forecasters in the
WNP region, have been making tropical cyclone forecasts without this type of
information. The consensus among operational forecasters is that the determination
of the position and intensity of tropical cyclones has been more
difficult without data from reconnaissance. If this is true, a larger variability
in the initial position error might result. Forecast errors of
prediction methods which utilize persistence information might also be
larger.
The first study to analyze the possible impact of the loss of aircraft
reconnaissance was done by Martin (1988). He compared the initial position
and forecast errors made by the Joint Typhoon Warning Center (JTWC) for
cases with and without reconnaissance during the first 12 hours before warning
time for the period 1980-1986. Results from his study suggest that
reconnaissance information appeared to have helped the initial positioning
The initial position error is the magnitude of the vector difference
between the warning and the best-track positions.

;
463
and intensity estimation. However, forecasts beyond 12 hours were, on the
average, not significantly different between cases with and without reconnaissance
except for recurving cyclones.
The present study represents an extension of that of Martin (1988) in
the following respects. The analyses cover the period 1983-1988, the last
year of which was completely devoid of reconnaissance information. This
avoids the complication that reconnaissance data at some previous time might
have influenced the forecasts (Martin (1988) combined cases with reconnaissance
data more than 12 hours before warning time and those without any
reconnaissance). In this study, only tropical cyclones within the Hong Kong
Area of Responsibility (10-30°N, 105-125°E) are analyzed. Cyclones within
this area are often affected by the terrain of Taiwan and the Philippines
well before they make landfall (Brand and Blelloch, 1973, 1974; Chang, 1982;
Bender et al. t 1987). Therefore, reconnaissance data may be more important
in determining the short-term movement of the cyclone. On the other hand,
ship and coastal radar reports are also more frequent in this region, which
may reduce the forecaster's dependence on reconnaissance information. Comparisons
are also made of the warning and forecast positions issued by the
Royal Observatory (RO) and JTWC. Because each centre has a different set of
objective techniques and prognostic products, how the forecasters assimilate
the reconnaissance information into the overall forecast scheme may be different.
This may result in a different impact when such information is not
available.
This paper will present results of such a study with an overall objective
of identifying the impact on the operational forecasts of tropical
cyclone motion and intensity as a result of the termination of aircraft
reconnaissance. Because such reconnaissance is not expected to be revived
in the near future, how forecasters can make the best use of the currentlyavailable
information will also be discussed.
2. Data set
The warning and forecast positions consist of those issued by the RO
and the JTWC (as received by the RO in real time) for tropical cyclones
within the area bounded by (10 N, 105°E) and (30°N, 125°E) during 1983-88.
To compute the initial position and forecast errors, the best-track
positions of the RO are used as the standard. To compare the errors made by
the RO and the JTWC, a homogeneous data set is necessary. Therefore, unless
otherwise stated, all comparisons are made using a data set which consists
of warning or forecast positions issued by both centres.
3. Initial position errors
The initial position (IP) errors made by the two centres are shown in
Fig. 1 with the values listed in Table 1. The most obvious result is that
the IP errors in 1988 for both centres are larger than those in all the
previous years in the data set. In some cases, the differences are very
large. For example, the IP errors made by both centres in 1988 are about
twice those in 1987. For the period 1983-86 (during which reconnaissance
data were available), the weighted mean IP errors of RO and JTWC are 33.6
2 Note that the reason for the JTWC errors to be usually larger than those of
the RO is probably that the IP errors are determined based on the RO besttrack.
' ' . . • ' " ' '
• ' . • . ' • . . • ' .' • , ' , • ' . • .-'' ' • ' • •'' ' ' •' • ' -

464
and 45.8 km respectively, which are only about two- thirds of those in 1988.
These results appear to confirm the feeling among operational forecasters
that the determination of the warning position has been more difficult
without reconnaissance information.
Because some of the warning positions between 1983 and 1987 were also
issued without any reconnaissance information, it may be more meaningful to
compare the IP errors in 1988 with only those cases in these other years in
which reconnaissance data were available. To do this, a method similar to
that used by Martin (1988) was adopted. That is, the warning positions
between 1983 and 1987 are divided into three groups: (a) those in which
reconnaissance data were available within 6 hours before the warning time (D
< 6), (b) those in which these data were available within 6 to 12 hours (D <
12) and (c) the rest of the data sample. Comparisons are then made between
the IP errors among groups (a), (b) and the average for all cases in 1983-87
with those in 1988 for different intensity categories (Table 2).
The most notable result in the comparison is that the IP errors in 1988
of both centres for all intensity categories (except for the STS category of
RO) are larger than those in previous years in which reconnaissance data
were available within 12 hours of warning time. The difference is the largest
for tropical depressions (TD) and tropical storms (TS). This result
therefore suggests that without reconnaissance data, the operational estimate
of the position is much more difficult especially for weaker systems.
Another observation from Table 2 is that when reconnaissance was available,
the IP errors of both centres decrease with the intensity of the
cyclone (except for tropical depressions). However, this decrease only
occurs in 1988 between the TS and severe tropical storm (STS) categories. A
possible reason why the IP errors for tropical depressions are smaller is
that estimates of the best-track positions at this stage are often based on
the warning positions as not much other information is usually available
even during post-analysis.
It is also interesting to note that when reconnaissance information was
available within 6 hours before warning time, the IP errors of both centres
are comparable (other than the TD category). However, these errors differ
more when reconnaissance data were only available within 6 to 12 hours. In
1988 when no reconnaissance information was available, differences in the
IP errors between the two centres are quite large even for the two most
intense categories of cyclones (STS and T). This suggests that by relying
almost entirely on satellite or synoptic information, forecasters from different
centres can have very different ideas on where the cyclone might be.
4. Forecast errors
Because the RO does not issue 72-hour forecasts, only the 24- and 48-
hour forecasts of the two centres are compared. In addition, a climatologypersistence
forecast computed by the RO is also included for comparison.
This method averages the predictions made using the persistence and climatology
methods and is therefore known as (P+O/2. Neumann (1981) suggested
that forecasts from this type of technique can be used as a baseline to
evaluate the performance of other forecast methods and as an indication of
the difficulty of the forecasts in a particular year. Because this technique
depends on persistence, the presence or absence of reconnaissance data
may also affect the performance of this technique.

465
a. 24-hour forecast errors
The 24-hour forecast errors of the two centres and the (P+O/2 method
are shown in Table 3. Notice that the errors of (P+O/2 appear to go
through a two-year cycle. Applying the concept of forecast difficulty of
Neumann (1981), 1988 could be considered as a more difficult year. Though
the error of 232 km in 1988 is higher by about 10% than the average for the
period 1983-86 (216 km) during which reconnaissance information was available,
it is comparable with those of the other two "difficult" years of 1984
and 1986. If this two-year cycle is genuine, it appears that the lack of
reconnaissance data in 1988 did not have an adverse effect on the performance
of the (P+O/2 method in its 24-hour forecasts. This may partly be due
to the compensatory effect of the climatology forecasts in the (P+O/2
method.
The mean 24-hour forecast errors in 1988 for both centres are also the
highest within the data set, with that of JTWC having the largest value. If
these errors are plotted relative to those of (P+O/2 (Fig. 2}, it can be
seen that the JTWC performance is the worst in 1988, suggesting that the
lack of reconnaissance information had a significant impact on the 24-hour
forecasts of the JTWC. On the other hand, the RO forecasts in this year is
comparable to those of (P+O/2. Although 1988 can be considered as a
difficult year, the RO forecasts in the past two "difficult years" (1984 and
1986) were both lower than those of (P+O/2. Therefore, the relatively poor
RO performance cannot be attributed to forecast difficulty alone. Rather,
the uncertainty in the warning positions due to the lack of reconnaissance
information must have had a contribution.
As in the last section, comparisons between the two centres can also
be made for different intensity categories and using only cases in which
reconnaissance information was available. The results in Table 4 show that,
on the average, for all tropical cyclones during 1983-87 for which reconnaissance
data were available within 12 hours, the 24- hour forecasts of the
two centres and the (P+O/2 method are lower than the corresponding forecasts
for 1988. A similar observation can be made for tropical cyclones in
different intensity categories with the exception of a few. Further, for
severe tropical storms and typhoons, the sybjective forecasts in 1988 are
only comparable or worse than the (P+O/2 forecasts but those in previous
years are better especially for cyclones in the D < 6 category.
Differences in the 24-hour forecast errors between the two centres are
also much smaller when reconnaissance data were available. In fact, even
for typhoons, this difference in 1988 is 38 km compared with ~ 10 km for
previous years with reconnaissance data.
These results are consistent with those for the IP errors in that
uncertainty in the warning positions has led to erroneous estimate of the
persistence component and thus the 24-hour forecasts. This, is true. even, for
the most intense systems,
b, 48-hour forecast errors
A similar two-year oscillation also occurs in the 48- hour (P+O/2
forecast errors, with that in 1988 being the highest (Table 5). The error
of 550 km is about 100 km greater than the mean of the errors during 1983-
1986 (448 km). This is at least partly related to the uncertainty in the
persistence component due to the lack of reconnaissance information. However,
because of the small number of cases in 1988, such a conclusion should
be considered only as preliminary.
The 48-hour forecast errors of the RO is again the highest in 1988
while that of the JTWC is the second highest within the data set. However,

466
relative to (P+O/2, they actually decrease compared with 1987 with the JTWC
having the lowest relative error within the data set (Fig. 3). Two possible
reasons may account for this result. Because the (P+O/2 forecast errors in
1988 are higher, any increase in the absolute errors of the two centres will
be reduced when compared with this method. It is also possible that persistence
does not play a very significant role in the subjective forecasts
at 48 hours so that uncertainties in the warning positions do not have a
large impact on these forecasts, at least on the average. More data are
necessary to determine which of these two reasons is correct.
If the forecasts are categorized by the intensity and availability of
reconnaissance data, the number of cases in each category for 1988 is very
small (between 1 and 14) which makes the sample not representative. Therefore,
these results will not be shown.
In any case, it appears that the termination of reconnaissance has an
impact on the (P+O/2 forecasts even at 48 hours. However, forecasters in
both centres seem to be utilizing other information such as synoptic data
and prognoses in making predictions for this time period so that persistence
may not play as an important role as the 24-hour forecast.
5. Warning Intensity errors
Besides the determination of the centre of a tropical cyclone, the
reconnaissance mission would also provide an estimate of the intensity of
the cyclone. Without this information, such an estimate will be based
almost exclusively on satellite analysis using, for example, the Dvorak
(1975, 1984) technique. Martin (1988) has shown that satellite analysts
analyzing the same picture could come up with very different estimates of
the intensity. Therefore, the warning intensity could differ from the besttrack
estimate by a significant amount.
To study this, the absolute differences between the warning intensities
issued by the two centres and the best- track estimate of intensity made by
the RO are computed. It can be seen from Fig. 4 (values listed in Table 6)
that, for the RO, these differences (warning intensity (WI) errors) in 1988
are higher than those in almost all the previous years, though not by a very
large amount. However, the WI errors of the JTWC in 1988 are smaller than
those in two previous years (1985 and 1987).
A further analysis of the WI errors can be made by categorizing the
sample according to the best-track intensity. Because the WI given by JTWC
is a one-minute average while that of RO is a ten-minute average, this
analysis is only done using the RO data. The results (Table 7) suggest that
the WI errors are the largest in 1988 for all classes of tropical cyclones
except typhoons. This means that because of the lack of reconnaissance
information, estimates of the Intensity of weaker cyclones have become less
accurate. For more intense systems, the convective features are usually
better defined on satellite imageries and hence the reconnaissance reports
of the intensity are less critical.
6. Summary and discussion
Because of the termination of reconnaissance missions, forecasters in
the western North Pacific had to rely mostly on satellite analyses in issuing
warnings and forecasts of tropical cyclones In 1988. Comparisons of the
warning and forecast errors during this year and those of previous years
show a definite impact of the termination of the reconnaissance mission.
Specifically, the uncertainty of the warning positions increased by > 30 7*

467
compared with previous years. Such an increase in uncertainty occurs
regardless of the intensity of the tropical cyclone, and especially so for
tropical depressions and tropical storms. With no reconnaissance as "ground
truth", differences between warning positions issued by the Royal Observatory
and those by the Joint Typhoon Warning Center are larger in 1988 compared
with those differences in previous years. Because of the increase in
the uncertainty in determining the warning positions, the 24-hour forecast
errors have also increased for both subjective and objective methods. The
impact on the forecast errors extends to the persistence-climatology type
technique but apparently not as much to the subjective forecasts.
Without reconnaissance information, estimates of the intensity of tropical
cyclones in 1988 had to rely almost exclusively on the analysis of
satellite imageries. This large subjectivity led to a decrease in the
accuracy of the warning intensities compared with previous years.
While these results are only valid for a small area of the western
North Pacific, they do highlight the impact of the lack of reconnaissance
information and substantiate the findings of Martin (1988). Other types of
information available within the area of the present study such as synoptic
and radar observations and ship reports do not appear to be able to take the
place of reconnaissance data. To substantiate these results further, the
warnings and forecasts for the entire ocean basin should be analyzed. Such
an analysis can even be considered as a routine in future years to determine
if forecasters have adapted to this new situation of operating without
reconnaissance data.
As the reconnaissance mission is not expected to be revived in the near
future, these results suggest that operational centres should now devote
more attention to the development of techniques to improve the estimate of
the warning position of tropical cyclones as well as their intensities.
With no reason to expect an increase in land- based observations or ship
reports, these techniques will almost have to be based exclusively on satellite
analyses. As Chan and Holland (1989) have suggested, researchers in
satellite applications for tropical cyclone forecasting should also concentrate
their effort in this area in order to help the forecasting community.
Only through the cooperation between these two groups can an improvement in
tropical cyclone warning and forecasting result.
Acknowledgement. The authors would like to thank Mr. T. C. Chu and Mr. T.
F. Lee for their help in the data reduction.
References
Bender, M. A., R. E. Tuleya and Y. Kurihara, 1987: A numerical study of
the effect of island terrain on tropical cyclones. Mon. Wea. Rev.,
115, 130-155.
Brand, S. and J. W. Blelloch, 1973: Changes in the characteristics of
typhoons crossing the Philippines. J. Appl. Meteor., 12, 104-109.
Brand, S. and J. W. Blelloch, 1974: Changes in the characteristics of
typhoons crossing the island of Taiwan. Mon. Wea. Rev., 102, 708-713.
Chan, J. C.L. and G. J, Holland, 1989: Observing and forecasting tropical
cyclones: Where next Accepted for publication in Bull. Amer. Meteor,
Soc. (December issue)

468
Chang, S. W., 1982: The ore-graphic effect induced by an island mountain
range on propagating tropical cyclones. Mon. Wea. Rev., 110, 1255-1270.
Dvorak, V. F., 1975: Tropical cyclone intensity analysis and forecasting
from satellite imagery. Mon. Wea. Rev., 113, 420-430.
Dvorak, V. P., 1984: Tropical cyclone intensity analysis using satellite
data. NOAA Tech. Memo. NESDIS 11, US Dept. of Commerce, Washington,
DC 20235, 47pp.
Martin, J. D., 1988: Tropical cyclone observation and forecasting with and
without aircraft reconnaissance. Atmos. Sci. Paper No. 428, Colo.
State Univ., Ft. Collins, CO 80523, 108pp.
Neumann, C. J., 1981: Trends in forecasting of the tracks of Atlantic
tropical cyclones. Bull. Amer. Meteor. Soc., 62, 1473-1485.

469
Table 1. Initial position errors (km) of the RO and the JTWC. a = standard
deviation.
YEAR
1983 1984
1985 1986 1987 1988
124
174
159
218
149
121
RO
error
a
30.4
28.5
39.7
43.5
23.9
28.7
37.6
44.0
23.3
22.6
45.0
42.9
JTWC
error
a
36.2
31.4
61.9
66.4
41.4
54.0
41.7
44.8
34.3
33.4
62.2
54.0
Table 2. Average initial position errors (km) made by RO and JTWC during 1988
and 1983-87 categorized by availability of reconnaissance data and
best-track intensity.
D < 6 = data available within 6 hours before warning time, D < 12 -
data available within 6 to 12 hours before warning time. The group
'1983-87 Average 1 includes all the available data for these years.
TD « tropical depression, TS = tropical storm, STS = severe tropical
storm, T = typhoon; N = number of cases, a = standard deviation.
Legend:
No. of cases (N)
RO error JTWC error
RO a JTWC a
Data
Group
D < 6
D < 12
1983-87
Average
1988
INTENSITY
C A T E G O R Y
TD TS STS Total
N=9
48 .5 59 .4
43 .4 70 .5
N»9
46 . 2 52.7
40
0
.1 27
N=106
41 .1 77 ,3
40 .8 70 .1
N=18
64 .5 110 .1
39 .4 72 .7
53.2
60.9
N=
.v-
50.0
35.6
N-
48.8
51.7
N=
66.5
47.3
41
48.6
40.6
24
65.8
34.8
224
62.0
57.1
31
68.9
42.3
N=51
28.6 27 .2
21.9 22 .5
N»27
29.2 36 .9
24.1 19 .6
N=189
27.5 35 .2
26.3 25 .3
N=35
25.8 47 .3
23.3 31 .3
19.3
14.6
23.6
15.6
tfs.
N-'
18.6
14.4
89
17.1
9.0
47
27,2
18.9
N=305
22.5
15.8
N«-17 35.7 ^ 47.4
43.3 54.5
N=196
31 .2 29.2
37 .3 31.5
N=107
32 .8 40.4
28, .2 28.7
N=824
31, .7 43.2
36, .5 46.6
N=121
45. 0 62.2
42. 9 54.0

470
Table 3. 24-hour forecast errors (km) of the (P*C)/2 method, the RO and the
JTWC. a = standard deviation.
Y E A R
1983
1984
1985
1986
1987
1988
No, of cases
87
128
116
174
116
71
(P+CJ/2
error
a
171
135
238
145
189
92
239
154
192
132
232
143
RO
error
a
199
123
213
116
170
83
213
148
200
110
231
133
JTWC
error
o
159
96
211
127
194
109
209
152
197
93
254
138
Table 4. 24-hour forecast errors (km) made by the (P+O/2 method, the RO and
the JTWC during 1988 and 1983-87 categorized by availability of
reconnaissance data and best-track intensity.
D < 6 = data available within 6 hours before warning time, D < 12 =
data available within 6 to 12 hours before warning time. The group
'1983-87 Average 1 includes all the available data tor these years.
TD = tropical depression, TS = tropical storm, STS = severe tropical
storm, T = typhoon; N = number of cases, o = standard deviation.
Legend:
No. of cases (N)
RO error
JTWC error
(P+O/2 error
RO 0
JTWC o
(P+O/2 o
I N T E N S I T Y
C A T E G O R Y
Data
Group
TD
TS
STS
T
Total
D < 6
N=7
318 151
289 123
395 189
N=34
248 115
212 135
234 109
199 91
179 88
229 132
N=74
185 98
174 99
206 102
N=162
208 108
188 109
227 124
D < 12
N=8
209 82
312 110
251 135
N*15
283 144
302 204
289 183
N=23
196 156
175 118
215 140
N=39
179 152
191 156
194 127
N=85
205 151
217 162
222 147
1983-87
Average
N=67
226 111
251 129
232 165
N*139
237 132
221 149
233 141
K=154
189 121
184 112
219 145
N=261
182 113
179 106
190 121
201 122
197 124
211 138
1988
N=10
278 142
310 152
287 147
N=ll
301 168
354 120
243 144
147 108
232 122
185 137
K=29
250 97
212 125
243 135
N=71
231 133
254 134
232 143

471
Table 5. 48-hour forecast errors (Jon) of all the four operational centres and
the (P-t-a/2 method, o = standard deviation, r = range of the largest
1 Aft. ^# » W _ .__ _•_
iV*
• .. _ _ t
its of o.
1983
1984
1985
Y E A R
1986
1987
1988
No. of cases
(P+O/2
error
a
38
308
282
77
477
285
62
400
176
129
495
324
73
306
170
25
550
321
RO
error
0
418
234
421
260
329
168
449
314
354
204
501
236
JTWC
error
0
366
220
392
271
335
241
496
394
342
153
446
337
Table 6. Warning intensity errors
o = standard deviation.
(kt)
of the RO and the JTWC.
Y E A R
1983
1984
1985
1986
1987
1988
No. of cases
124
174
159
218
149
121
RO
error
0
4.8
7.4
3.9
5.3
4.5
6.2
3.3
4.2
4.3
4.6
5.2
5.0
JTVC
error
0
7.2
6.9
S.7
5.7
8.0
7.0
e o
4.9
8.5
8.4
7.5
8.1
Table 7.
Warning intensity efrors (kt) of RO for individual years categorized
by best-track intensity.
TD - tropical depression, TS = tropical storm, STS - severe tropical
storm, T = typhoon; N = number of cases, o = standard deviation.
Year
1983
error
0
1984
error
0
1985
error
o
1986
error
0
1987
error
o
1988
error
0
TD
JMB"
1.8
2.3
H*27
1.4
2.2
N=30
0.8
1.9
H*13
1.6
2,0
N=18
1.5
.2.5
N*18
3.5
3.4
I N T E N S I T Y
TS
N='34
3.6
4.0
N=53
3.1
3.9
N=31
5.2
4.5
N*85
2.3
3.4
N=21
4 . 2
2.8
• ; . .. N*31 •
6.5
5.2
C A T E G O R Y
STS
H=35
4.9
6.2
N=62
3.1
4.9
N=36
4.2
4.5
N*31
5.1
4.8
N=25
4.2
3.7
NO5
5.3
. . ' • 4.3. , • • • •
T
N*37
7.3
10.8
N*32
8.8
6.8
N=62
6.0
8.1
N=89
3.8
4.4
N*85
5.0
5.2
N«37
4.7
5.7
Total
N«124
4.8
7.4
N*174
3.9
. • - 5.3 ;
N^159
4.5
6.2
K*2I8
3.3
4.2
N«149
4.3
4.6
N*121
5.2
5.0

472
Initial Position Errors
1983-1988
1983
O RO o JTWC
Fig. 1. Initial position errors of the RO and the JTVC for the years 1983-
24-hr Forecast Errors Rel. to (P + O/2
(1983-1988)
A
a RO o JTWC
Fig. 2. 24-hour forecast errors of the RO and the JTWC relative to those of
the (P+O/2 method for the years 1983-88. X positive (negative) value
means that the forecast is vorse (better) than that of the (P+O/2
method.

473
48-hour Forecast Errors Rel. to (P + O/2
1980-1988
RO
Year
O JTWC
Fig. 3. 48-hour forecast errors of the RO and the JTWC relative to those of
the (P+O/2 method for the years 1983-88. A positive (negative) value
means that the forecast is worse (better) than that of the (P+O/2
method.
Warning Intensity
1983-1988
Errors
1983 1964 1985 1966 1987
a
RO
Year
O JT W
Fig. 4, Warning intensity errors of the RO and JTVC'.for the years 1983-

474
A SPECTRAL MODEL FOR MEDIUM RANGE WEATHER
FORECASTS AND ITS PERFORMANCE
Ji Liren Chen Jiabin Zhang Daomin Wu Wanli
Shen Rujin Sheng Hua Huang Boyin
(Institute of Atmospheric Physics, Beijing)
ABSTRACT
A global spectral model developed by Lab II, IAP and its performance are
introduced in this paper, The model has some merits in the treatment of large
scale orography. By introducing a standard stratification, the prognostic variables
are separated into a basic part and a deviation part respectively. The spectral
expansion is conducted only for the deviation, thus improving the convergence
rate. Recently, the scheme has been generalized with a latitudinally dependent basic
state and a still higher forecast score expected. The parameterization schemes
for diabatic physical processes included in the model are also briefly discussed.
Finally, as an example, a five—day forecast for a real case is given, in which the
setting in of bai-u rain belt in the Yangtze Valley and the northward shift of the
subtropical high over the west Pacific are well reproduced. Comparative forecast
experiments show the crucial influence of diabatic factors on the medium range
variation of the subtropical high.
1. INTRODUCTION
Over the past ten years, substantial progress has heen achieved in medium range
numerical forecast. As an example, the practical preditability of ECMWF model has
reached about six and a half days on average. Efforts for this purpose in China may be
traced back to the mid seventies when a hemispheric model including various diabatic
physical processes (Zhu et al, 1979) was developed and a number of numerical experiments
on medium range weather processes carried out (Jiang et al. 1982). However, being
restricted by computer resources and other conditions, a system for operational medium
range forecast was developed only in the last two years. The global spectral model reported
here is a part of the joint efforts.
In section two of this paper, the dynamic framework of the model will be described.
The discussion will be focused on the use of standard stratification approximation in
spectral formulation. Section three presents a brief description of all components of
diabatic processes included in the model. Section four illustrates the performance of the
model by forecast experiments for a real case. And finally, the conclusion.
2. MODEL EQUATIONS AND METHOD OF SOLUTION
The model equations and their spectral formulation, and the semi-implicit time in-

4/5
tegration scheme basically follow the ECMWF model (Baede et al., 1980; Louis, 1984).
Some improvement has been made to reduce truncation error. As is well known, a widely
used approach to include the effect of large— scale orography in numerical model is to introduce
the so-called sigma coordinate. In spite of the merits of this kind of coordinates
there are some attendant difficulties, such as the calculation of pressure gradient over
mountainous area, etc. In the spectral formulation, there are additional problems involving
orography, such as Gibbs phenomenon in the vicinity of steep mountain, spurious
orography on the oceans, etc. (Machenhauer, 1979). Since all prognostic variables
are now defined on sigma surfaces, their smoothness and convergence rate of spectral
expansion may be degraded if steep mountains are included in the model.
To deal with these problems, the so-called standard stratification approximation
method is adopted(Chen et al., 1987), which was first suggested by Zeng (see, for example,
Zeng, 1963). For spectral formulation, the basic idea of this method can be addressed
as follows. If there is a function /(A) with jumps and discontinuities and if a
known auxiliary function (p(X) could be found which has discontinuities similar to
/(A), then let /(A) » /\ (A) +

476
where
I p u = FC-

477
'' / »
where 1 d . .....
where J# stands for the vertical turbulent flux of \I/,K^ exchange coefficient.
Boundary conditions for (10) are
=0, for p = 0,
-*,).
for z -+°>
where \l/ 3 denotes if/ at surface; M(Z) and \l/(z) are values at Z near surface, which
may be taken as the lowest model level. And C^ is the drag coefficient.
Given the vertical distribution of iKz), which is predicted by the model, and
parameterization formulua for C^ and K^ one can calculate the contribution of turbulent
transfer from (10).
Based on the Monin— Obukhov similarity theory, the explicit form of C m (for momentum
transfer) and C h (for sensible heat) are given by (Louis, 1979)
where K is Von Karman constant, Z 0 roughness length given by climatic data and Mi
Richardson number.
The expressions of f m and f h are given according to stratification conditions, i.e.
near neutrality, highly unstable and highly stable.
Extending the similarity theory to the Ekman layer and the free atmosphere, exchange
coefficient K m and K h may be given by
dZ (13)
* * dZ
where l m and l k are the mixing lengths given by Blackadar (1962).
3.3 Surface Processes
In the calculation of vertical diffusion, the variables at the surface •#, are yet to be

478
given. This is treated separately for sea and land surface.
For sea, the surface temperature is prescribed by monthly climatological data, and
kept constant during time integration. ,
For land, diffusion equations are used to describe the variation of temperature and
moisture in the soil. Using a three-layer model, the equation for soil temperature may be
written as
dT. IF K(T 4 -T S )
8t pCD { ^Q.SD^Dt
_ K(T d -T s ) K(T d -T d )
dt Q.5Z> 2 U>i -
where ^F represents all surface heat fluxes including solar radiation, long-wave radiation,
sensible and latent heat fulxes, K is the diffusion coefficient, p g and C g the density
and specific heat capacity of soil, and Dl, D2 and D3 are the depths of the layers
respectively.
For soil moisture, there are similar equations.
3.4 Precipitation and Latent Heat Release
To simulate the condensation process in stratiform cloud, the so-called
"saturation-condensation" method is adopted, i.e. wherever q > q SAT (T) occurs at certain
time— step, then an adjustment for T and q is conducted. The adjusted temperature
T * and specific humidity q * should satisfiy
where q SAT is saturated specific humidity and the excess of q becomes precipitation.
For cumulus convection, the modified Kuo (1974) scheme is adopted. The conditions
for convection to take place is a covergence of moisture of background fields (7>0)
•3/1
and a conditional unstable laspe rate (-—- >0). Besides, to avoid spurious and random
op
convection due to computational error in moisture convergence and to consider the fact
that systematic convection is often related to synoptic systems, we impose an addional
critirion CAT > 0, f # being vorticity at the lowest model level.
The heating and moistening due to cumulus convection is given by
dP
C p (T e - T)— ,
J p t
where * C" denotes variables in cloud. The moistening parameter b is taken following
Anthes with (RH) C = 0 and n = 3. (Anthes, 1977; Das etal., 1988)
4. RESULTS OF COMPARATIVE FORECAST EXPERIMENTS
Here comparative forecast experiments for a real case is given to illustrate the performance
of the model. The case selected is 12GMT, 14-19 June 1979, the time when the
6

ainy season of Yangtze valley (bai-u) set in.
Fig.la shows the initial geopotential field of 500 hpa (12GMT, 14 June ). The
circulution at mid and high latitudes over Asia resembles one of typical bai—u patterns.
The major trough and ridge are stable, moving slowly eastward thereafter. However the
key system, the subtropical high over the west Pacific is still far to the south. The contour
of 588 dam, which is usually used to define the extent of the subtropical high, is located
to the south of 20 °N at 110 -120 °E. On day 2 (Fig.lb), the high begins to move northward,
the 588 dam contour reaching 22 °N at 120 °E. Also noted is a minor trough over
the northwestern part of the Tibetan plateau which will continue to move eastward becoming
the first weather-producing system that triggers bai-u. Fig. 1C shows the 500
hpa patten of 12GMT,18 June, the day that the rainy season of bai-u starts. The salient
features are the northward shift of the subtropical high, the 588 dam contour being at
27 °N, 120 °E and the minor trough at 110 °E, which causes the confluence of cold and
warm humid air and the corresponding rain band over the Yangtze Valley (Fig.2).
Starting from 12GMT, 14 June and using the T21L5 version of the model i.e. triangular
truncation with 21 zonal waves and 5 layers, several 5-day forecasts have been
made for various situations. The results are as follows.
Table 1 gives the forecast scores for various situation. It may be seen that the
adiabatic forecast by standard stratification approximation (SAD) is better than that by
original ECMWF scheme (EC), evaluated by both tendency correlation coefficient
(RDC) and RMS. When condensation processes are included (LC) the gain in forecast
skill is apparent. When all the physical processes other than condensation are also added
(NAD), further improvemant of skill is also clear. For day 4 and day 5 an increase of
preditability by more than one day is obtained, showing the crucial role played by
diabatic processes in medium range forecast. For comparison the adiabatic run by a high
resolution version T42L9 ( HAD ) is also listed in the table. One can see that for day 4
and day 5, the gain of skill from diabatic effect for T21L5 is even higher than that from
an increase of resolution.
Fig.3a shows day 1 forecast 500 hPa map by diabatic run (NAD). Comparing with
Fig.lb, one can see it captures well the northward shift of the subtropical high over the
west Pacific, though the position is a bit farther to the north. The trough moving along
the northern fringe of the plateau and its associated rain area are fairly well predicted.
Besides, the large area of rainfall at northeast China agrees well with the observed. On
day 4 (Fig, 3b), the model sucessfully reproduces the pattern typical of bai-u. The positioning
of trough and ridge at mid and high latitudes, and subtropical high over Asian
area are in agreement with observation.
The weakness of the forecast is also apparent. The predicted subtropical high is situated
north of its real position. This discrepancy occurs on day 1 forecast and remains so
through 5-day forecasts, which is related to the underestimate of the development of the
major trough over the eastern coast of Asia. One may also see that the amplitude of
troughs and ridges at high latitudes are considerably reduced, showing that the low resolution
of the model is not adequate in describing weather systems in high latitudes.
Fig,4 shows the time evolution of the mean position of 588 dam contour at 110-
120 °E, which is crucial to the positioning of the major rain band at the eastern part of
China. It illustrates that the trend predicted by the run with all model physics is in
agreement with the observed, viz northward shift from 15th to 18th and retreating on
19th. The retreat resulted from the development of the trough over the seaboard to the
north, which is captured by the diabatic forecast. However it is also apparent that the
479

480
predicted location is too far to the north by about three degrees latitude on average over
the five days compared with the observed. As for the adiabatic run, it also predicts the
northward shift of the high but the position is even further north. Besidies, it completely
fails to predict the turning of the track.
5. CONCLUDING REMARKS
We have made a brief description of a global spectral model intended for medium
range weather forecast and introduce some features of the model, which differ from other
models. The model resolution is relative low and only a few case prediction experiments
have been conducted. Keeping these in mind, some preliminary conclusions may
be given. The numerical scheme suggested seems to have some merits in the treatment of
model orography. The practical preditability of the model may reach about five days, if a
tendency correlation coefficient of 0.6 is used as a criterion. It is capable of simulating the
behaviour of some major weather systems which are crucial to the weather of China.
And the results clearly show the important role played by diabatic physical processes in
medium range forecasting.
Acknowledgment. This work has been sponsored by the Medium Range Numerical
Weather Forecast Research Project.
REFERENCES
Anthes, R.A., Mon. Wea. Rev., 105 (1977), 270.
Baede, A.P.M., Jarraud, M. and Cubasch, U., ECMWF Technical Report No. 15,1980.
Blackadar, A.K., J. Oeophys. Res., 61 (1962), 3095.
Chen, J.B., Ji, L.R. and Wu, W.L., Advances in Atmospheric Sciences, 4 (1987), 156.
Das, S. et al, Mon. Wea. Rev. 116 (1988), 715.
Jiang, D.Y., Wang, Z.H. and Ji, L.R., Collected Papers on Medium Range Numerical Weather
Forecast, China Meteorological Press, 1982.
Kuo, HX-, J. Atmos. Sci., 31 (1974), 1232.
Louis, J.—F. (editor), ECMWF forecast model physical parameterization, Research Manual 3,
ECMWF Meteorological Bulletin, 1984.
Louis, J~F., Boundary Layer MeteoroL, 17 (1979), 187.
Machenhauer, B., GARP publications series No. 17. "Numerical methods used in atmospheric
models." Vol.2,124.
Zeng,Q.C., Acta Meteorologica Sinica, 33 (1963), 472.
Zhao, G.X. et al., Kexue Tongbao (Science Bulletin), 32 (1987), 1479.
Zhao, G.X. et al., KexueTongbao (Science Bulletin), 33 (1988), 847.
Zhu, B.Z., et al., A eta Meteorologica Sinica, 38 (1980), 130.

Fig. 1 ECMWF -FGGE analysis of 500 hPa geopotential for (a) 12GMT 14 June 1979
(b) 12 GMT 15 June (c) 12 GMT 18 June, referred to as day 0, 1 and 4 respectively. The
contour interval is 8 dam.
481

482
Fig.2 Distribution of 24h precipitation for 00 GMT 18 June to 00 GMT 19
June (in mm).
Fig.3 Forecast 500 hPa geopotential for (a) day 1 (b) day 4. The contour interval is 8
dam. Dashed Ime denotes the isoline of 10 mm rainfall.

483
40}
Fig.4 Daily position of the 588 dam geopotential isoline averaged for 110
-120 °E, solid line for observation, dashed line for non-adiabatic run
(NAD) and dotted line for adiabatic run (AD).
TABLE 1
500 hPa geopotential forecast scores of the model for 14 June 1979 case. EC,
adiabatic run (T21L5) with common formulation; AD, adiabatic run (T21L5)
with standard stratification approximation; LC» AD plus condensation and
heat release process; NAD AD plus all model physics; HAD, same as AD except
for higher resolution (T42L9). RDC, correlation coefficient calculated by
differences from initial field. RMS, in meters.
Day
1
2
3
4
5
RDC
0.80
0.79
0.66
0.61
0.56
EC
RMS
33
52
77
87
97
RDC
0.80
0.81
0.71
0.66
0.62
SAD
RMS
32
47
68
79
85
RDC
0.81
0.81
0.72
0.68
0.65
LC
RMS
32
48
67
67
80
RDC
0.81
0.82
0.74
0.71
0.69
NAD
RMS
32
44
62
68
71
RDC
0.82
0.86
0.79
0.73
0.68
HAD
RMS
30
40
58
68
75

484
LONG-RANGE FORECASTING OF TAIWAN MEI-YU
Ming-chin Wu
Department of Atmospheric Science,
National Taiwan University
ABSTRACT
Recent studies on the possible mechanisms that cause the seasonal
rainfall variation in Taiwan Mei-Yu are reviewed. Those on the various
potential long-range forecasting techniques applied to this specific
problem are evaluated. A method of forecasting the rainfall departures
based on the regression scheme is developed. Some forecast experiments
are done, and the forecast verifications for May and June rainfall
index in southern Taiwan are examined. The forecasting studies show
that forecasting using rainfall record only is not promising no matter
whether frequency domain approach or time domain approach is applied.
The diagnostic studies show that the inter annual variability of the
Taiwan Mei-Yu is connected to the large scale circulation anomalies.
Warm ocean surface, low sea level pressure, more low level convergence
around Taiwan, deep East Asia main trough at 500 hPa, negative phase
of the Southern Oscillation and westerly wind regime in the tropical
atmosphere in the months before the Mei-Yu onset herald an abundant
Mei-Yu season. Following the diagnostic studies, plausible predictors
of the Taiwan Mei-Yu are identified. Regression models are constructed
and used to predict May and June area rainfall. The results show that
forecasting the southern Taiwan May rainfall is skillful. However,
those for June are not significantly skillful.

485
THE ROYAL OBSERVATORY LONG RANGE RAINFALL FORECAST METHODS
Robert Lau and M.Y. Chan, Royal Observatory Hong Kong
A REVIEW
The Royal Observatory Hong Kong was primarily engaged in day to day weather forecasting
before the sixties. In 1963, the worst drought in history occurred in Hong Kong and gave the
Observatory the necessary impetus to step into long range rainfall forecasting. On and off over a
period of twenty five years, many statistical relationships have been attempted to predict summer
rainfall (May to October) for Hong Kong, a sizable proportion of which was fruitless and has long
since been discarded. Two relationships currently in use, just manage to withstand the test of time
so to speak and are described below.
(1) Based on the possible relationship between the summer rainfall in Hong Kong and the Intensity
of the winter monsoon in East Asia the previous winter.
In 1963, GJ. Bell found that the mean January surface pressure in Hong Kong correlated well
with the Hong Kong rainfall the following summer (June to September) at a level of significance
better than 5%. This and later research pointed to the possibility that summer rainfall in Hong Kong
might be related to the intensity of the winter monsoon occurring in east Asia the previous winter
as the immense Asian continental anticyclone invariably surges south to affect the south China
coast every January. A lot of efforts was then devoted to the selection of appropriate parameters
attempting to measure the average strength of the winter monsoon. The parameters that are
plausible or showed promise at one time or another are briefly outlined as follows:-
(a) Pressure difference between Irkutsk and Tokyo (DPIT)
Bell (1976) used the mean January surface pressure difference between Irkutsk (5216N 104
21E) and Tokyo as a measure of the intensity of the winter monsoon in east Asia. To predict the
summer rainfall for a particular year, the regression equation used would be derived from the
correlation between the corresponding pressure differences and the summer rainfall over the
previous 15 years. A 15-year correlation window was chosen by Bell to achieve a reasonable
statistical significance and at the same time minimizing complicat Jons from trends and secular
changes.

486
monsoon in east Asia and for correlating with the summer rainfall in Hong Kong. Regression
equations are again derived from a 15-year correlation window and summer rainfall forecast for
a particular year predicted from the regression.
(c) Mid-latitude 700 hPa zonal index (7240)
A 700 hPa mean January zonal index for the zone 40N-60N 65E-165E is computed from
generated grid point field using a specific algorithm and is used to measure the winter monsoon
intensity in east Asia. Forecast summer rainfall is obtained from regression equations derived
from correlation of 15 pairs of 700 hPa January index and subsequent summer rainfall in Hong
Kong.
(d) Mid-latitude 500 hPa zonal index (5240)
Instead of 700 hPa level, zonality at 500 hPa level is used as predictor.
(e) 200 hPa Wind Shear between Kagoshima and Singapore (KSSH)
The January mean difference between the westerly component of the 200 hPa Kagoshima
wind and the easterly component of the 200 hPa Singapore wind is correlated with the summer
rainfall and a regression equation obtained from 15 pairs of values from data over the past 15
years. The mean January 200 hPa wind shear between Kagoshima and Singapore for the current
year is used as a predictor to obtain the summer rainfall forecast from the regression equation.
(f) Mid-latitude 500 hPa anomaly (A5NT)
The absolute extreme value of the 500 hPa mean anomaly in January in the region 30N-5QN
120E-150E as read from the Japan Meteorological Agency charts is correlated with the
subsequent Hong Kong summer rainfall and again a regression equation is derived for the
15-year period prior to the forecast year. Using the current extreme anomaly as the predictor, the
forecast Hong Kong summer rainfall is obtained as the y-estimate of the regression equation.
(g) Mid-latitude 700 hPa anomaly (A7NT)
Same as A5NT except that the 700 hPa level Royal Observatory charts are used,
(h) Mean 500 hPa height at specific grid point (M5HA)
Mean 500 hPa heights in January at grid point 35N 135E are obtained from analysed charts
and are correlated with summer rainfall in Hong Kong for the 15-year period prior to the forecast
year and a regression equation derived. Summer rainfall for the year is then predicted using the
mean 500 hPa height at 35N 135E that January as the predictor.
(i) Mean 500 hPa height at specific grid point (M5HB)
Same as M5HA but with a different reference grid point at 55N 80E.
These methods (a) to (i) purporting to indicate the strength of the northeast monsoon, are
definitely interrelated and essentially one and the same. The Irkutsk/Tokyo pressure difference is
closest to a direct measure of the pressure gradient. The logic in using the zonal indices is that
the more meridional the mean flow at upper levels, the stronger and more persistent should be
the resultant surface monsoon. The rationale behind the Kagoshima and Singapore 200 hPa wind
shear method lies in the fact that the strength of the westerly jet above Kagoshima and the
easterly jet above Singapore should be a reflection of the Hadley cell in winter. As to the methods
involving January anomalies, the hope is that the extreme values can serve as tell-tale.

487
(2) Based on possible relationship between El Nino and Summer Rainfall in Hong Kong (ELNN)
The effect of El Nino on global weather is by now undisputed. The teleconnection between
sea surface temperature (SST) in the equatorial Pacific and summer rainfall along the south China
coast was little more than a conjecture in the seventies. Commencing mid-seventies, sea surface
temperature trend in the equatorial Pacific and the activity of tropical cyclones is used to correlate
with summer rainfall in Hong Kong. Sea surface temperatures at/near Canton Island (2 46S 171
43E) are used. The trend is obtained by subtracting the SST, meaned over the 3-monthly period
September, October and November of the preceding year from the SST, meaned over the
following 3-monthly period December, January and February. This is correlated with the Hong
Kong summer rainfall total using a 15-year correlation window as in the previous methods.
Forecast rainfall is also obtainable using the latest SST trend as the predictor.
PERFORMANCE EVALUATION
Since 1975, quantitative rainfall totals to be expected for Hong Kong during the months from
May to October have been computed for the Hong Kong Water Supplies Department every
February using a subjective weighting of the results of all of the methods mentioned above. The
performance of the various methods from 1975 to 1988 is displayed in Table 1. It has to be said
that none of the methods yields particularly satisfactory results. The mean absolute errors are
mostly in excess of 300 mm, or more than 15% of the summer (May-October) rainfall in Hong
Kong, and the standard deviations are 200 mm plus. Overall the ELNN method is the best
performing method with a mean absolute error of 257.2 mm and a standard deviation of 230.6
mrn. However, a quick look at the correlation coefficients of the various methods in Table 2
reveals that the ELNN method is the worst correlated with coefficients generally not even
significant at the 5% level. The DPIT, A5NT and A7NT methods are, on the other hand, significant
at a level of better than 0.1 % (r > 0.7604) for the first three years. Although on the downward trend,
a number of methods are still significant at a level of better than 5% (r> 0.5145) by 1988. This
foregoing observation surprises and suggests that regression equations with good correlation
coefficients do not necessarily give better rainfall forecasts. It vindicates Reynolds (1978) who
pointed out that in order that a regression equation may have a useful predictive value, a
correlation coefficient approaching 0.90, with 81% of the variability explained, is required.
OTHER METHODS TRIED DURING 1989
Despite the obvious shortcomings of the methods concerned and the unnecessary confine of
applying only simple regression analyses, it was thought that the prospect of greatly improving
the accuracy of the techniques to forecast quantitative seasonal rainfall for a small territory like
Hong Kong, was poor, and the task too daunting. Also for most of the eighties, the water supply
situation in Hong Kong has not been pressing. The situation abruptly changed in winter 1988
when water storage became once again low and the summer rainfall forecast resumed its
importance.
A number of methods were tried in early 1989 and are discussed below:
(1) Spectral analysis of Hong Kong summer rainfall
Instead of relating the point rainfall in Hong Kong to the large scale circulation in East Asia, it
was thought that an examination of the time series of summer rainfall in Hong Kong may be
useful. A spectral analysis of Hong Kong summer rainfall with a view to finding out any significant
rainfall periods has therefore been initiated. Should distinct peaks be identifiable, rainfall amounts
corresponding to these periods can be auto-correlated to arrive at a set of regression equations
for predicting purposes without going into the physical causes. The result of analysis illustrated
in Figure 1 suggests rainfall peaks at periods of 2.2, 3.2 and 12.5 years. These periods

488
unfortunately are not integral and it became extremely complicated to set out the rainfall data for
auto-correlation.
(2) Multiple regression with Hong Kong summer rainfall
The spectral analysis of Hong Kong rainfall having seemingly little practical value, attention
was diverted to salvaging ideas through multiple regression. Using the periods of the spectral
peaks as the starting point, phenomena with approximate periodicities were merged together
correlating with summer rainfall using multiple regression. They are -
(a) Quasi-biennial oscillation (QBO)
The 2.2 year peak was conveniently linked to the Quasi-Biennial Oscillations. Bell's findings
(1964,1974), using a plot of 12-month running totals of rainfall in Hong Kong from 1853-1962 to
demonstrate the existence of a succession of two to three 26-month cycles in three different time
intervals, has lent some support to this assumption. Annual mean winds at 50 hPa in Singapore
is then selected as a measure of the QBO, westerly winds are counted as negative and easterlies
positive.
(b) El Nino (ELNN)
The 3.2 year period is difficult to account for as there is no known meteorological phenomenon
exhibiting such a periodic cycle. Sea surface temperature data presented by Fritz (1985) and sea
surface temperature trends near Canton Island in the central Pacific as computed by the Royal
Observatory Hong Kong, although aperiodic, do exhibit an 'average' recurrence interval of 3-4
years for abnormally high sea surface temperatures. The sea surface temperature trend
computed in ELNN is therefore chosen as another parameter for multiple correlation.
(c) Sunspot number
The remaining spectral period of 12.5 years is also difficult to explain. The known phenomenon
having a cycle closest to this period is the sunspot cycle of 11 years. Although Bell (1977) had
pointed out earlier that relationships between sunspots and rainfall/other meteorological
parameters are probably more complex than generally supposed, record dry and wet years
(Table 3) in Hong Kong do suggest the existence of an approximate 11-year rhythm. The annual
sunspot number is thus accepted as the third parameter for correlation.
(d) DPIT
Rainfall, like any other meteorological parameters, also shows annual variations. The intensity
of the winter monsoon measured by DPIT is hence selected as the fourth parameter for
correlation with summer rainfall.
Results of multiple regression with Hong Kong summer rainfall using the four parameters of
QBO, ELNN, Sunspot Number and DPIT (MR4X) are presented in Table 4, Again a 15-year period
is selected to facilitate comparison with simple regression methods. The correlation coefficients
show that summer rainfall in Hong Kong is well correlated with the four parameters at a level of
significance of better than 5%, the majority cases are better than 1 % (r« 0.84).
Tables 5 and 6 show similar results with three parameters (MR3X, sunspot number removed)
and two parameters (MR2X, sunspot number and QBO removed) respectively while a summary
of the performance of all methods (both simple and multiple regression forecasts) are presented
in Table 7. Apparently the multiple regression method with three parameters gave the best results
with mean absolute error of 247.7 mm and a standard deviation of 197.7 mm.

489
CONCLUSIONS
Seasonal rainfall forecasts using simple regression equations yield relatively unsatisfactory
results with mean absolute errors in excess of 15%. The only exception is the ELNN method
which gives a mean absolute error of about 13%. However, a sea surface temperature trend in
ELNN is not well correlated with the summer rainfall in Hong Kong, with r < .5 in most cases and
are not significant at the 5% level.
The multiple regression method has improved the forecast accuracy to a mean absolute error
of slightly less than 13% and a standard deviation of about 10%. At the same time a high
correlation with values of 'r' ranging from .72 to .94 can be achieved. Despite this, there is still a
great deal left to be desired on quantitative rainfall forecasting methods as errors of over 35% still
appear in 2 out of the 14 forecasts during the period 1975-1988.
The multiple correlation methods predict summer rainfall in Hong Kong for 1989 to be in the
region of 1948.1 mm - 2161.2 mm, a value which has yet to be verified but appears at the moment
to be in the right ball park.
ACKNOWLEDGEMENT
The authors are most grateful to their colleague, Mr. K.W. Li for his devoted effort in all the
statistical computations involved in this study.
REFERENCES
Bell, G.J. 1964 Preliminary Report on Hong Kong Rainfall
1974 Quasi-biennial Variations in the Tropical Troposphere,
Royal Observatory Occasional Paper No. 28
1976 Seasonal Forecasts of Hong Kong Summer Rainfall,
Weather 31 No. 7 pp 208-212
1976 Seasonal Forecasts & Northern Hemisphere Circulation
Anomalies, Weather 31 No. 9 pp 282-292
1977 Changes in Sign of the Relationship between Sunspots
and Pressure, Rainfall and the Monsoons, Weather 32 No. 1
pp 26-32
Fritz, S. 1985 The Aleutian Low in January & February - Relative to
Tropicl Pacific Sea Surface Temperature, Monthly Weather
Review February 1985 pp 271-275
Reynolds, G. 1978 Two statistical Heresies, Weather 33 No. 2 pp 74-76

Figure 1.
SPECTRAL ANALYSIS FOR MAY-OCT RAINFALL
1884 - 1988
. 01
*g a 2
40
60 80
PERIOD ( year )

491
(verifying against the actual May to October totals i n millimetres
over the period 1975-1988)
YKAft DP IT A5NT A7NT 57,40 77,40 KSSII J5ZI M5HA
1975 -469.3 -603.9 -490 . 5 -604 .3 NA -361.8 -058. 3 -402.6
1.970 -317.0 -472.9
-210 .0 -313 .3 -235.3 -80.8 -310.5 -202.9
1977 19.4 -97.0 358 .0 13 .8 49,7 224,4 58.4 304 :5
1978 -98. 1 - 530. 1 -107 .2 ,-lt 1 .5 -34.7 -433.2 - 141.5 -375.8
1979 88.3 -376.7 250 .7 -5 .7 -00,7 63,0 -93.4 11.4
1980 2(M .5 510.7 458 .4 471 .3 470,1 503.0 4 67s 8 472.3
1981 84 .3 298.0 558 .0 354 .0 397.0 282.9 423.0 322.8
1982 -808.0 -698.8 -453 .8 -759 .0 -771 . 1 -871 .6 750.3,- - 905.8
1983 439. 2 565.0 020 .2 45 .2 122,3 135,0 -10.7 204.4
1984 - 59 . .1 91.2 234 ,0 -Hi .3 -316.5 -392.1 - 288. "2 -81.0
1985 039.7 407, 7 284 ,4 430 .5 4 HI. 2 343.2 375.4' 329.6
1980 -114.8 48.2 -224 .0 -74 .3 -11,2 -255. ] 34.8 -280, I
1987 0 4 1 . 5 710..3 001 ,0 705 .5 788.8 154.7 720.1 428.5
1988 090.1 1012.9 305 .4 685 .8 740.1 595.4 858.5 892.9
MKANCABS. ) 333.8
STIXADS. ) 205.5 261 403.2
.6
Table 2
YKAK
1970
1977
1978
1979
1980
1981
1982
1983
1984
1985
1980
1987
1-988
1
2
3
4
5
6
7
8
9
10
378 .2
157 .4
356 .1 345.2 335.9
261 .1 279.0 211.7
370.8
277.6 •
C o r r cvl a t i on cooP P i c ionts of various simple
regression methods
DP IT A 5 NT
-.8J10 . 8202
-.8055 . 7890
-.8002 . 7941
-.0875 . 6941
-.6900 . 0901
-.7373 . 6738
-.7892 , 7034
-.7485 . 6680
-.7283 . 6052
-.7116 . 0 I 5 1
-.0550 . 6122
-.0948 . 6205
-.5794 . 4012
A 7 NT
.8131
.8134
.8009
.7123
.7105
.7058
.6852
.6888
.6377
.6035
.6428
.6291
.4910
5Z40
5 4 3 1
52 11
5297
3019
. 3239
3 OH 1
. 4620
5774
6349
0089
, 6787
7 1 0 2
5873
77,40 KSSH J57.J
.5048 -.7451 .4907
.4950 -.7508 .4772
.5227 -,74. r >8 .4868
.2930 -.5099 .2415
.3072 -,5741 .2575
,3503 -.5787 .3212
.4403 -.6297 .4147
.6173 -.5233 .6117
.6550 -.6000 .6324
.6912 -.6322 .7147
.6838 -.6322 .7147
.7395 -.6353 .7703
.5946 -.6081 .6500
M51IA
,75.10
.7514
.7351
.5063
.5043
.5114
.5595
.4551
. 4 5 1 1
.5458
.5458
,5030
.4764
381 .0
254.8
Table 3 Record dry and wet summers (May to October
in millimetres) in Hong Kong
Urj£HL Wettest
1903 834. 5
1973 2894. 0
1954 968. B 1982 2772. 0
1895 1018. 9I 1891 2714. 0
1938 1071 .
1957 2656. 1
1901 1078. 7
1889 258C. 4
1912
1 150. 5
1894 2520, 8
1887 1163, 0 1972 2478. 2
1907 1240. I 1975 2464. 0
1898 1242. 6 1923 2451. 3
1905 1338. 1 1947 2356. 4
M5HB
-.0094
-.5674
-.5404
-.3084
-,3110
-.3363
-.4081
-.2653
-.2665
-.4330
-.4010
-.5098
-.4151
M5HII
-930, 1
-492,1
320.2
-309.7
-180,8
5 U5 . 5
412.5
-913.4
-50.3
86.5
372.3
-140.0
401.8
529.4
404.0
258.4
KL.NN
.4804
.4848
.4647
.5845
.5950
.5742
,5109
.4035
.4145
.4410
.4604
.4179
.4693
KI.NN
12 .4
7 .8
309 .7
115 .5
-04 .7
544 .8
672 .0
-723 .8
-38 .3
174 .0
378 .3
191 .9
132 .3
234 .5
257 .2
230 .6

492
TaUj.g_ji
Multiple rnfjross i on analysis results
(f our parameters : WIT, Kl.NN ,QIK).SUNSPOT versus M»y~Oct rainfall)
YKAlt
1075
1976
1977
1978
l!>79
I9KO
19H1
1982
19B3
I9«4
1985
19R6
1987
1988
1989
XI
DHIT
- 7 1 . 6
-72. 7
-71 . {')
-71 . 9
-78. f>
-70. 6
-80. 3
-83. 9
-80. 5
-82. 2
-84 .
r}
-R-l . 1
-85. •1
-71 . 4
-48. 2
COEKPICJKNT
X2 X3 X4 CONST
KLNN QUO SUNS POT
lf>O.G
154.8
1 d 5 . 5
1 6 (5 . 4
1 0 3 . 2
I f> 0 . 5
4 9 . 0
41,4
3 7 . 4
4 3 . 4
4 -1 . 0
55.9
n r> , i
1 8 -1 . 0
205.1
-4 .2
-5.0
-3.4
-3.6
-4 .4
-.4
-3.8
'-3.0
-6.7
-7.9
-8.0
-10.5
-10.4
-7.0
-8.2
.8
1 .9
2.2
2. 1
2. 3
1.2
-.:»
-1. 1
. 1
-.2
-.2
.0
. 1
.9
.7
32 19 . 1
3218 .3
3217 . 1
3 22 3.3
3330 .9
3219 .8
3508 .2
3C.3-1 .4
3 fi f i 0 .4
35 « 4 . 7
3021 7
3 f> H 0 0
3577 9
3
2
3 2 1 3
2721
11 POST ACT KfMOU
(mm) (mm) (mm)
.92 2470.5 2464.0 6.5
.93 1776.8 2047.9 -271.1
.94 1565.4 1592.0 -26.6
.93 2157.6 2102.0 55.6
.90 2706.6 2242.4 464.2
.88 1752.1 1425.9 326.2
.90 M89.fi 1354.4 135.2
.92 2038.0 2772.0 -734.0
.86 2219.5 1.970.0 249.5
.85 1814.9 1 7-67.0 47.9
.84 1907.3 J564.5 342.8
.84 2027.9 1942.8 85.1
.84 2'! 05.0 1716.8 688.2
.76 1657.7 14 16,. 9 240.8
.79 2161.2 ******* ******
MEAN (ABSOLUTE) 262. 4
STANDARD DEVIATION 225. 7
ERROR
(X)
.3
-
-1
13 .2
.7
2.6
20 .7
22 .9
10 .0
-26 .5
12 .7
2 .7
21 .9
4 .4
40 . 1
17 .0
*****
Table 5
Multiple regress i on analysis results
( t h ree pa ramr te rs: DPJT,ELNN,QDO versus Mny-Oct rainfall)
YEAR
1 9 7 r>
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
XI
DP IT
-71 . 1 157.7
-7J . 4 152.2
-70. f> 1 58,9
-71 . (} 1G0.5
-75. fi-153.0
-71 . 5 147.0
-80. 6 149.8
-86. 4 143.7
-8(5. 1 137.1
-82. 5 144. 6
-8-1. 7 145.2
-84. 7 155.8
-85.
154.5
-69. 3 182.fi
-45. 9 205.8
COKKFTCIKNT
X2 X3
El.NN QUO
CONST
-3.5 3299. 1
-3.7 3307 .4
-2.fi 3314 .2
-2.2 3315.4
-3.5 3393.7
-.2 3303.7
-3.9 3491.4
-3.3 3()()3.9
-6.7 3 f 1 0 2 . 8
-8.0 357 5. R
-8.1 3610.7
-10.5 '3581 .0
-10.4
-6.7
-8.1
3585.3
3233. 1
2721 .1
R
.91
.92
.92
.92
.88
.87
.90
.92
.86
.85
.84
.84
.84
.75
.79
POST ACT ERROR EHKOR
(mm) (mm) (mm) (%)
2527.4 2464.0 63.4 2. 6
1864.4 20 '17. 9 -183.5 -9. 0
164G.R 1592.0 54.8 3. 42
2233.0 2102.0 131.0 6.
2567,8 224 2-. 4 325.4 14. 5
1629.7 1425.9 203.8 14. 31
1518.2 1354.4 163. 8 12.
21 15.5 2772.0 -656.5 -23. 7
2216.0 J 970.0 246.6 12. 5
1812.H 1767.0 45.5 2. 6
1898.5 1564.5 334.0 21. 3
2029.0 1942.8 86.2 4. 4
2412.8 1716.8 696.0 40. 5
1694.4 1416.9 277.5 19. 6
2109.8 ****** ****** *****
MEAN (ABSOLUTE) 247.7
STANDARD DEVIATION 197.7

YEAR
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
Multiple regression analysis results
(two parameters : DPIT,ELNN versus May-Oct rainfall)
COEFFICIENT
XI X2 CONST
DPIT ELNN
-68.4
-68.4
-68.7
-69.0
-71.4
-71,4
-79.5
-85.8
-85.2
-81.9
-78.5
-74.5
-75.9
-65.7
— 40 . 8
167.5
157.4
162.4
163.3
152.7
147. I
151,2
144.2
136.7
140.6
138.1
153.9
139,5
165.5
194 . 8
3241 .7
3241 .0
3271 .3
3274 .5
3306 .3
3303 .0
3460 . 7
3586 .9
3636 .0
3542 .3
3492 .7
3368 . 5
3406 .2
3174, 1
2628 . 4
R
FCST
(o»n)
.91 2570.3
.92 1873.0
.92 1604 .8
.92 2290.3
.87 2474.2
.87 1631.4
.90 1525.8
.91 2062.7
.84 2303.7
.82 1627.7
.81 2104.1
.78 1950.4
. 7 7 2193. 1
.72 1802.2
. 72
ACT
(mm)
2464. 0
2047. 9
1592. 0
2102. 0
2242. 4
1425. 9
1354. 4
2772. 0
1970. 0
1767. 0
1564. 5
1942. 8
1716. 8
1416. 9
MEAN (ABSOLUTE)
STANDARD DEVIATION
ERROR ERROR
(mm) (%)
106. 3 4.3
-174. 983 -8.5
12. .8
138. 9.0
231. 8 10.3
205. 5 14.4
171. 4 12.7
-709. 37 -25.6
333. 16.9
-139. 3 -7.9
539. 6 34.5
7. 6 .4
476. 33 27.7
385. 27.2
0
7
263.
197.
Relative performance of all methods : both simple regression and multiple regression
(verifying against the actual May to October totals in millimetres)
YEAR DPIT A5NT A7NT 5Z40 7Z40 KSSH J5ZI M5HA M5HB ELNN MR4X MR3X MR2X
1975 -469,,3
-663 .9 -496 .5 -604.
NA -361 .8 -658. 3 -402 .6 -930 .1 12 .4 6. 5 63 .4 106.8
1976 -317,,0
-472.9
-216 .0 -313. 3 -235 .3 -86 .8 -310. 5 -202 .9 -492 . I 7 .8 -271 . 1-183
.5 -174.9
1977 19, ,4 -97 .6 358 .0 13. 8 49 . 7 224 .4 58. 4 304 .5 320 .2 309 .7 -26 .6 54 . 5 12.8
1978 -98,,1
--530 .1 -167 ,2 -111. 57 -34 , 7 -433 .2 -141. 5 -375 .8 -309 . 7 115 . 5 55 .6 131 .0 188.3
1979 88, .3 -376 . 7 250 .7 -5 . -60 . 7 63 .0 -93. 4 11 .4 -180 .8 -64 , 7 464 .2 325 .4 231.8
1980 204, ,5 510 .7 458 .4 471. 3 476 .1 503 .6 467. 8 472 .3 516 . 5 544 .8 326 .2 203 .8 205.5
1981 84 .3 298 .6 558 .6 354. 0 397 .0 282 .9 423. 0 322 .8 412 .5 672 .6 135 .2 163 .8 171.4
1982 -808 .0 -698 .8 -453 .8 -759. 0 -771 . 1-871
.6 -750. 37 -965 .8 -913 .4 -723 .8 -734 .0 -656 . 5 -709.3
1983 439, .2 565 .6 626 .2 45. 2 122 .3 135 .0 -10. 264 .4 -50 .3 -38 .3 249 . 5 246 .6 333.7
1984 -59 .1 91 .2 234 .0 -411. 3- -316 .5 -392 .1 -288. 2 -81 .0 86 .5 174 .6 47 .9 45 .5 -139.3
1985 -639,.7
407 .7 284 .4 430. 535 484 .2 343 .2 375. 4 329 .6 3-72 .3 378 .3 342 .8 334 .0 539.6
1986 -114 .8 48 .2 -224 .0 -74. -11 .2 -255 .1 34. 815 -280 . 1 -140 ,0 191 .9 85 .1 86 .2 7.6
1987 641 .5 710 .3 661 .6 705. 788 .8 154 .7 720. 428 .5 401 .-8 132 .3 683 .2 696 .0 476.3
1988 690 .1 1012 .9 305 .4 685. 8 740 . 1 595 .4 858. 892 .9 529 .4 234 . 5 240 .8 277 . 5 385.3
MEAN(ABS. ) 333.8
463 .2 378 .2 356. 1 345 .2 335 .9 370. 8 381 .0 404 .0 257 .2 262 .4 247 . 7 263.0
STD(ABS, ) 265. 5 261 .6 157 .4 261. 1 279 .0 211 .7 277. 6 254 .8 258 .4 230 .6 225 .7 197 . 7 195.4

494
THE NONLINEAR INTERACTION OF INTERNAL
AND TURBULENCE
WAVE
Liu Shi-da
Department of Geophysics, Peking University
Beijing, China
ABSTRACT
We analyse the behaviour of nonlinear dynamical
systems in the form of Galerkin of turbulence equations.
These equations are presented by the models of Burger's
and Chao Jiping.
On the parameter plane (R-j»*O» "t^e
region is
divided into five different behaviour by bifurcation
point curves and limit point curves* They are mean
field 0, periodic internal-gravity wave P, convection
motion T, pulse wave S and turbulence behaviour C.
It is shown that these behaviour can transform into
one another through pitchfork bifurcation and Hope
bifurcation. Clearly, our result is better than that of
linear Miles theorem. In the case of stratified
stability, turbulence can also occur. If the initial
Richardson number R.is slightly larger than 0.25 and the
internal-gravity wave propagates through such a system
in which mean velocity is decreased and mean temperature
lapse rate is increased, then the new Richardson number
R.may become smaller than 0.25 and turbulence may occur.
This is also the reason that intermittency occurs
between the turbulence and internal-gravity wave. The

495
computational trajectories are consistent v/ith the
qualitative analysis.
In order to show quantitatively the truth of these
behaviours, Lyapunov exponents LE are computed.
I. Introduction
The mutual interaction among mean motion, periodic
(v/ave) motion and turbulent motion constitutes the basis
of atmospheric motion. The wave and turbulence extract
energy from mean flow fields and mean temperature
fields. On the other hand, wave and turbulence modify
the mean field.The interaction between wave and
turbulence is also a common phenomenon in the atmosphera
boundary layer at night and in the troposphere. Many
observations such as ocean themocline, strong atmosphere
turbulent thin Iayer 0 Intermittency and inhomogeneity
of atmosphere turbulence can be explained by waveturbulence
interaction*
Fua and Chimonnas (1) have discussed the interaction
of wave-turbulence by using M 1-J th order scheme" for
the turbulence* But, this analysis did not obtain the
occurrence condition of turbulence and internal-gravity
wave. And the scheme is rather complex. Chao (2) has
discussed the convective behaviour by extending Burgers
(3) turbulent nonlinear model, but he did not discuss
the periodic behaviour and turbulence. In this paper,we
investigat the transform criterion of mean field,
internal—gravity wave and turbulence. And we obtain the
critical curve to delineate wave and turbulence region
by mean of bifurcation theory, dearly, our result Is
better than that of Miles theorem (4) ..
II. Basic Equations

496
Burgers(l948) had put forward a simple turbulence
model [3 J . Chao(l962) extended Burgers model to include
temperature advection. The basic equations of the
Burger s-Chao model are as follows:
-a
„,
where U(t) and T(t) are horizontal velocity and
temperature averaged for height respectively, w(z,t) and
0(z,t) are perturb ational vertical velocity and
temperature respectively, P and Q are pressure gradient
force and heat source respectively, both of which are
necessary for the maintenance of initial mean motion UQ
and mean temperature T Q . The terms involving 3>
represent viscous dissipation. Eqs. (1)~(4) represent
the nonlinear dynamical system. The first terms on the
right hand of Eqs* (2) and (4) represent the mean field
acting on the perturbation fields. The third terms on
the right hand of Eqs. (1) and (3) represent the
perturbation field acting on the mean fields. £0 denotes
buoyancy force. 1iT^ and ^tfl^c
are advection terms.
^-.JL ^ ig ^g^ temperature, taken as a constant), j^
is the adiabatic lapse rate and H is the vertical
thickness.
The advection terms are omitted in Chao's model. The
temperature advection term ^M is included in our
model. Let

497
Substituting Eq. (5) into Eqs. (l)-(4) and taking
dimensionless form, we obtain a trunc ated- spec trum
model :
3 " •*-• * 3:*4
x, « c c x, - )a: a - n% -ir af,^
where the " • " represents a derivative with respect to
dimensionless time T « X = "fftt. y Xi-JJ ^Cj 3 *

498
From Eqs,(8), the necessary condition for the
occurrence of state T is
We analyse the stability of state 0 on the parameter
plane (R.,R ), the pitchfork bifurcation curve 1 is(6,J
Hope bifurcation curve 4 is
t/o;
The Hope bifurcation curve 3 and limit point curve 2 of
state T are shown in Fig.1,
0*00 0.10 0.20 0.3Q
RI
0.50
FIGURE 1, The bifurcation point curves 1,3,
and limit point curves 2 in plane(R. ,R ).

499
These curves divide the plane (R.,R ) into six regions:
0,T,S,C,P and R. They denote different motion behaviours.
IV. Trajectory In Phase Space And LE Exponent
Eqs.~~(6) can be integrated numerically to locate
trajectory in the phase space by the Runge-Kutta
procedure. The typical periodic behaviour in S and P
region, projections of orbite of C region on plane(xp,x/)
are given in Fig.2.
crH «
O
(C)
O
fO
FIGURE 2* The motion behaviour in region
S f C and P region. Re=50.
(a). Xavaries as ~t (Ri=0.10) f in region S.
(b).The projection on plane (x 2 ,x 4 ), in C region
(Ri«0.20 andRi=0.25)
(c). x^varies as -c (Ri«0.40),in region P

500
In order to show quantitatively the truth of these
behaviours, Lyapunov exponent LE are computed (6,7,8) ,
they are given in table 1.
TABLE 1, Lyapunov exponent spectrum for chaotic
and periodic attractor.
Be^ 1 071
p
c
S
R;
0-40
0*5
Q.IO
R 6
so
50
so
LE, -
0.00
o.&8>
0.00
LE,
-O.S4
0-00
-0.75
LE,
-;.S8
-/-o3
-/.76
Lt+
-4.il
-7:26
-4.65
LE 5
-*8.&f
-3i.8a
-36-63
V, The Interaction Of Internal Wave And Turbulence
From Fig. 1, the internal gravity wave and turbulence
may appear intermittently when R. approaches i. For
example, the internal gravity wave extracts energy from
the mean field to form a velocity field with a strength
of 10 order for R =50,R. =0.40. But the internal wave
may disappear until R.=0.20, and a chaotic field is
formed.
After the internal waves have been formed, the mean
internal energy increases Tt "^^o when the temperature
stratification is stable. Therefore 7-7^:> y«-y a >
That is -CV-VaX-C Vo-Va) >
or
C*

501
REFERENCES
(1] .Fua,D. ,Chimonas,C 0 , J 0 Atmos 0 Sci. ,39, 2451-2463(1982)
(2) .'Chao,J.P« f Acta Meteorologica Sinic'a, 32,11-18(1962)
(3j. Burger's, J.M., Adv. Appl.Mech 0> 1,171-201 (1948)
(4]. Van zandt, Radio Sci. ,13,819-829(1978)
^0Yih,C.H^,Stratified_flow, Science Press(in Chinese,
1983)
(6) . Arnold, V. I. , Hathematic al^Method^of ^Classic al
^nS er ( 1 978 )
. ,Holmos,P. ,Nonlinear_Oscillation,
Springer (1983)
(8) .Haken f H. , Adv§nced_§Ynergetics, Springer (1984)
(9).Frish,U. ,J.of Fluid Mech. ,87,719-736(1978)
tD. ,Scientia Atmospherica Sinica,JJ^373-380(l987)

502
Multiple Equilibria Of a Thermally Forced
Baroclinic Atmosphere
Chi-Sann Liou
Naval Environmental Prediction Research Facility
Monterey, CA 93943-5006, USA
ABSTRACT
In an attempt to explain weather regimes and blocking phenomena,
multiple equilibria of an idealized, thermally forced atmosphere are
studied using a quasi-geostrophic two-level model. The model is a
channel model on a $-plan with a sinusoidal bottom topography. The
stationary solutions (equilibria) of the nonlinear system are solved
by severely truncating the spectral representation of dependent
variables.
With truncation retaining one zonal wave and one meridional
mode, three equilibria are possible in the baroclinic model which
are similar to Charney and DeVore's (1979) results found in a
barotropic model. The high-index and low-index equilibria are
orographically stable, while the intermediate equilibrium is
orographically unstable. With truncation retaining one zonal wave
and three meridional modes, two extra equilibria exist in the model.
One of the additional equilibria is orographically stable with a
circulation pattern similar to a typical blocking situation.
Time integration of the model shows that the truncated orographically
stable equilibria directly link to weather regimes of the
model. A 200 year simulation with annual cycle for the external
forcing shows that model climate is in bimodal distribution with two
distinct types of winter. The predictability of the winter type is
very high.

503
1. INTRODUCTION
The earth topography has two primary effects upon large-scale
circulation of the atmosphere: (1) it interacts with zonal mean flow
to generate stationary waves, and (2) it interacts with the generated
stationary waves to feed back the zonal mean flow with mountain
torque (form drag) and eddy heat and momentum fluxes. Earlier
studies such as those by Charney and Eliassen (1949), Bolin (1950),
Murakami (1963) and Derome and Wiin-Nielsen (1971) were concentrated
in the generation of stationary waves, ignoring their feedback upon
the zonal mean flow. In their studies, therefore, a linearized
system of equations with a prescribed zonal mean flow was used in
whichthe topography was only an eddy-independent stationary forcing.
However, when Charney and DeVore (1979) considered the mountain
torque feedback in a severely truncated barotropic spectral model,
they discovered the existence of multiple flow equilibria (stationary
solutions) in the model. Within a certain range of the external
mechanical forcing, there could be three stationary solutions under
a given mechanical forcing: a high-index (nearly zonal) equilibrium,
a low-index (strong meandering) equilibrium and an intermediate
equilibrium.
The study of multiple flow equilibria was extended to baroclinic
models by Charney and Strauss (1980), Raods (1980), and Yoden (1983).
Although the models they used were somewhat different from each
other, their results were similar such that multiple flow equilibria
appeared in the baroclinic models but in a range of unrealistically
large thermal forcings. In this study, we examine multiple flow
equilibria in baroclinic models that are different from the above
models in the formulation of surface friction and mountain forcing,
and in spectral truncations. Moreover, we look into the reasons of
forming multiple flow equilibria in details.
The application of multiple flow equilibria to general timedependent
solutions is based upon the hypothesis that some equilibria
may sustain instability and synoptic-scale wave forcings for a long
period of time to form attractor basins of the time-dependent solutions.
If multiple attractor basins are indeed a character of the
atmosphere, the knowledge of transition between attractor basins
and the stickiness of each attractor basin may provide a new vehicle
to study the variability and long-range predictability of the climate.
2. TWO-LEVEL SPECTRAL MODEL
The model used in the present study is similar to the model used
by Phillips (1956) for general circulation study, except topography
and interface friction are included in the present, model and a
Newtonian type heating is used to thermally force the system. The
model is a two-level quasi-geostrophic model applied to a channel on
a $~plane. The boundary conditions are cyclic in the zonal direction
and rigid at the two lateral walls. An idealized sinusoidal topography
is given as the lower boundary (Fig. 1).

504
For the two-level quasi-geostrophic model, the governing equations
can be written as,
- 7§ M = -- J( M , v M ) - — J(4 T/ T - -
at f 0 f 0 3x
K *
T/ h s ) ---
2
(1)
o
-
3t
> T — ^
X z = -- J(4> M , ^ 4-T^ - —
f 0 f 0
where
- * 3
: vertical difference of geopotential
2 (2)
2
2
2
* T - -=- J(

505
In this study, the following parameter values are chosen to represent
typical atmospheric conditions:
f Q = 10~ 4 sec" 1 , 3 = I.G^IO" 11 m"" 1 sec" 1 ,
R 287 m 3 sec" 3 sec 2 °K I r S « 0.33 m 3 sec" 2 rob" 2
K s = 4 K.J. = {6.5 day)""" , p = (20 day)"
T^ = 280°K, which gives £ = 1.22018*10~ 8 m" 1 sec" 1
Following Lorenz (1963) and Yao (1977), we further simplify
our model by using a truncated spectral representation for each
variable. In a rectangular domain, trigonometric functions are
the most convenient expansion functions. The truncated Fourier
expansions for all variables are chosen as follows:
(KM ~

506
The coefficients Z, T, H, and T are spectral coefficients of the
vertically averaged and vertically differenced geopotential, the
topography and the radiative equilibrium^temperature, respectively.
In the above spectral representations, T is represented by mode Al
only. Only one zonal wave number, n=3, is retained in the model and
two kinds of meridional truncation m=l and m=l,2,3 are used in this
study. In the case m=l, the topography is represented by mode LI
with H LI = 250 m. In the case m=l,2,3, the topography is represented
by a superposition of LI and L3 modes, with H LI = -3 H L3 = 250 m.
3. MULTIPLE EQUILIBRIA (STEADY SOLUTIONS)
With only one meridional mode (m=l), this is the simplest case
that includes nonlinear interaction between the mean zonal flow,
waves, and topography. With this truncation, the governing equations
are reduced to six ordinary differential equations with six dependent
variables. By requiring all tendency of the six equations to be
zero, we can solve the six nonlinear algebra equations with six
unknowns. As we searched for a forcing range of AT = 20°C to 80°C,
we found more than one (three) stationary solutions (equilibria) whe
AT* is greater than 76°C.
Fig. 2 shows the AT vs AT plot for the equilibria states.
There are three branches in the region 20°K < AT < 80°K, Branches
1, 2, and 3, which correspond to the high-index equilibrium,
intermediate equilibrium and low-index equilibrium, respectively.
A linear stability analysis has been performed by perturbing the
spectral coefficients of the equilibria in the form of cj>' = cj) e .
For Branch 3 all equilibrium states are stable, for Branch 2 all
equilibrium states are monotonically unstable (real 0) t and for Branch
1 all equilibrium states are oscillatorily unstable (complex O ).
We identify the monotonic instability as the baroclinic version of
orographic instability. It can be inferred qualitatively in the AT
vs. AT diagram (Fig. 2) . The orographic instability is due to the
existence of the resonance that makes a perturbation in AT deviate
from its equilibrium state through a change of the eddy heat flux
divergence larger than the change of Newtonian heating. Again,
unlike the barotropic case, a change in eddy heat flux rather than
mountain torque is responsible for this instability. We find that
all equilibria are unstable with respect to m=2 perturbations,except
for the low-index equilibrium at very low AT , AT < 29°K.
To allow eddy momentum flux and possible double jets pattern
zonal flows in the model, we have relaxed the spectral truncation to
additionally include second and their meridional modes (m—1,2,3) in
the model. The thermal forcing is kept the same with only mode Al,
while a third mode component is added in the topography spectrum. By
choosing H^ = -1/3 H LI , the topography is flattened near the lateral
boundaries. There are 18 variables in the system, 6 for mean zonal
components and 12 for wave components, with 18 ordinary differential

507
equations. The 18 ordinary differential equations are reduced to 18
algebra equations as we set the tendency to zero searching for
stationary solutions.
By using the iteration method, we can identify seven equilibrium
branches in the forcing range 20°K < AT < 80°K. For a given AT ,
however, there can be a maximum of five equilibria. Fig. 3 shows
these branches on the AT vs. AT* diagram. Branches 4 and 5 appear
when AT > 57.5°K, and Branches 1 and 2 appear when AT* > 79.5°K.
Branches 1, 2 and 3 correspond very closely to the equilibria of the
highest truncation case, m=l. On the other hand, Branches 4 and 5
are two new^branches. They appear even for a weaker AT than the
critical AT required for Branches 1 and 2. For these two new
branches, the first mode components still dominate in the mean zonal
flow but all three meridional modes are equally important in the wave
components. Branches 6 and 7 are also two new branches that appear
in very weak thermal forcing (AT < 28.5°K) . For these branches
second mode components are zeros and the third mode wave amplitudes
are larger than amplitudes of the first mode wave.
Linear stability analysis shows that all equilibrium states
are unstable except for Branch 3 under a very weak thermal forcing
of AT < 29°K, and for Branch 7 which appears only when AT < 28.5°K.
Branch 2, 4 and 6 are orographically unstable.
4. MULTIPLE STATISTICAL EQUILIBRIA (TIME MEAN SOLUTIONS)
As we have found all equilibria are unstable unless in a range
of very weak thermal forcing, the results of stationary solutions
cannot be directly applied to interpret the atmospheric phenomena.
We need to find how these (stationary) equilibria are modified by the
instabilities. To answer this question, long-term numerical time
integrations of the spectral equations with three meridional modes
are conducted for fixed values of the thermal forcing a variety of
initial conditions. The Runge-Kutta fourth-order scheme with a onehour
time step is used in the integrations. For fixed values of AT ,
four types of initial conditions are used: a baroclinic zonal flow
(that corresponds to the Hadley equilibrium) , a barotropic zonal
flow, a baroclinic wavy flow, and a barotropic wavy flow.
A statistical equilibrium state is defined as the time-averaged
state over the last 3 1/2 years of a 4-year integration. Several
additional experiments with slightly perturbed equilibrium states as
the initial conditions have also been conducted to examine the
relationship between the statistical equilibria and stationary
equilibria 1 . ' ' '.. ' - ' - , . ' • . • . • ' • • • '•/ ' ' • ,, '. . . ..•• •" . . • • • .•': .. .•
From the numerical integrations, four statistical equilibrium
branches, Branches A, B, B 2 and C, can be identified in the range
20°K < AT < 80°K. However, for a given thermal forcing the model
can produce a maximum of three statistical equilibria. Branch 1S>2 *-&

508
just the stable Branch 7 for stationary equilibrium and appears only
when AT* < 28.5°K. These four statistical equilibria are statistically
stable in the sense that their statistical characters are
persistent at least for four years. We consider them as a multiple
attractor basins of the system.
— *
Fig. 4 shows the AT vs, AT diagram for the statistical
equilibria with outlines of their typical characteristics. The mean
circulations of the branch C equilibria consist of meridional pairs
of high and low pressure centers that resembles a typical blocking
pattern (not shown). This result demonstrates that an attractor
basin exists, in this severely truncated system that blocking type
solutions are maintained even. Under the modification of various
instabilities, the mean features of Branch B are similar to those of
Branch 3, except that the former has a narrower meridional scale. The
mean features of Branch C are similar to the features of Branch 5.
The mean features of Branch A, however, are not similar to those of
Branch 1. Waves for Branch A have different phase characters from
those for Branch 1, and Branch A has double jets unlike Branch 1.
Details of those equilibrium circulations can be found in Liou
(1982).
5. INTERANNUAL VARIATIONS AND PREDICTABILITY
To investigate the interannual variations and predictability
of those equilibria, the model with a periodically changed AT is
integrated over two hundred years. A sinusoidally varying AT
between 30°K and 70°K is used to simulate the seasonal change in the
thermal forcing. For computation convenience, twelve equal length
months, 30 days for each, are used for a model year.
The integration is started from the "summer solstice,"
AT = 30°K, with the corresponding Hadley equilibrium as the initial
condition. In this two-hundred year history, we found two distinctly
different types of model winters. One type of the winter solutions
is very similar to the solutions for Branch B in which the first and
third mode waves are coupled and locked to the topography, while the
second mode wave is traveling eastward. The other type of winter
solutions is very similar to the solutions for Branch C in which all
waves are coupled and locked to the topography. These two types of
solutions appear stochastically in different years. Fig. 5 is a
sample of the time history which shows these two types of solutions.
We choose T as a representative quantity in showing the time
history. We can clearly see that during the winters of these six
sample years there are two years in which the solutions show a
smaller T with a weaker mean zonal flow and a locked second mode
wave. The solutions of the other four years show a larger T with a
stronger mean zonal flow and a traveling second mode wave. The two
types of solutions are clearly separatable throughout the whole
winter even to the following spring.

509
To investigate the statistical property of the quasi-stationary
solutions, we take a monthly running mean with a 20-day overlap and
regard it as a sample of monthly-mean model climate. Traveling waves
are almost averaged out in the monthly means, with locked waves
remaining. To examine the distribution of their probability density
functions, we evenly divide a range of each monthly-mean variable
into twenty intervals and count the frequency of appearance for the
variable in each interval. The number of frequency divided by two
hundred is the probability of this variable in each interval. Fig. 6
shows the contours of the probability distributions for monthly-mean
AT, 4>£i' anc * ^W2" There are two well-separated peaks in the probability
distribution, i.e., it is distinct bimodal distribution. The
well-defined bimodal distribution demonstrates that the system has
two preferred states for winter which resulted from two different
attractor basins pulling the solutions toward them.
In an attempt to understand how the system is attracted by
Branch B and Branch C throughout the__seasons, we traced the annual
cycles of the two sub-ensemble mean AT (starting from summer in the
AT) vs. AT diagram of statistical Equilibria (Fig. 7). During
summer the two mean AT are very close and lower than AT for Branch B.
During early fall, the system has two possible ways to go: one is to
follow a branch possessing a larger AT, and the other is to follow a
branch possessing a smaller AT. In the later part of the year (i.e.,
winter), the smaller AT solution stays near Branch B, while the
larger AT solution can no longer be maintained in late fall and is
attracted into Branch C. We interpret the split routes for the
system solutions are due to transition behavior of Branch B solutions
around AT = 40°K (not shown) . One type of the present solutions
follows the upper AT branch of the transition solution, and the other
type of solutions follows the lower AT branch of the transition
solution. As we have found in the fixed AT timeintegrations that
the lower AT branch of the transition solutions remain in Branch B
as AT increases, the smaller AT solutions are stable and__stay near
Branch B. On the other hand, as AT increases the upper AT Branch of
the transition solutions cannot be maintained and thus the solutions
attracted into Branch C in the following season. The self-transition
from Branch B to Branch C are found in fixed AT integrations which
indicates that Branch C is more stable than Branch B, Therefore, the
system is attracted into Branch C rather than Branch_J3 when it is
unstable. Whether the system will follow the upper AT branch or
the lower SF branch in the transition range determines the system
behavior in the following fall, winter and early spring seasons. We
consider that the choice by the system to follow one of those two
branches during the transition range is stochastic. However, the
predictability of different winter types is at least one season
earlier as the zone mean flow of the two type solutions is distinctly
separated in fall (see Fig. 6).

510
REFERENCES
Bolin, B., 1950: On the influence of the earth's orography
on the general character of the westerlies. Tellus,
2,184-195.
Charney, J.G. f and A. Eliassen, 1949: A numerical method
for predicting the perturbations of the middle latitude
westerlies. Tellus. 1, 38-54.
r and J.G. DeVore, 1979: Multiple flow equilibria
in the atmosphere and blocking, J, Atmos. Sci.,
36/ 1205-1216.
, and D.M. Straus, 1980: Form-drag instability,
multiple equilibria and propagating planetary waves
in baroclinic, orographically forced, planetary wave
systems J., Atmos. Sci. , 37, 1157-1176,
Derome, J., and A. Wiin-Nielsen, 1971: The response of a
middle-latitude model atmosphere to forcing by topography
and stationary heat sources. Mon. Wea. Rev., 99,
546-576.
Liou, C.-S., 1982: Multiple stationary and statistical
equilibria of a thermally forced baroclinic atmosphere.
PH.D. thesis, University of California, 224 pp.
Lorenz, E.N., 1963: The mechanics of vacillation. J. Atmos.
Sci.. 10, 448-464.
Murakami, T., 1963: The topographic effects in three-level
model of the -coordinate. Pap. Meteoro. Geophvs., 14
Tokyo, 144-150.
Phillips, M.A., 1956: The general circulation of the atmosphere:
A numerical experiment. Quar. J._ Roy. Meteor.
Sci., 82, 123-164.
Roads, J.O., 1980: Stable near-resonant states forced by
orography in a simple baroclinic model. J. Atmos.
Sci., 37, 2381-2395.
Yao, M.S., 1977: Thermally and topographically forced
general circulation regimes of a two-level quasigeotrophic
atmosphere, Ph.D. thesis, University of
California, 161 pp.

511
0 ^^S08%#g!%J!J#^
a), T, Q
3 Ap
B
Fig. 1 Vertical structure of the two-level quasi-geostrophic
model with only one wavelength shown in bottom topography.
STATIONARY SOLUTIONS WITH TOPOGRAPHY
m » 1
20
30 40 50 60 70 80
Fig. 2 cross channel temperature difference AT (thick lines) o£ each
equilibrium state as functions of thermal forcing AT for m*l. ..
The shaded area Is regions a< AT)/at < 0 when all other variables
are in equilibrium, and the thin dashed line is for equilibrium
without topography (Hadley solution).

512
STATIONARY SOLUTIONS
WITH TOPOGRAPHY
«n-l, Z, 3
80
70
eo r
ATCTC)
30 j-
y
/-
^
40 -
30 -
20
i
t
30 40 50 60 70 80
Fig. 3 As in Fig. 2, except for m-1,2,3.
80
WITH TOPOGRAPHY
70
60
I: EASTWARD
U; COCKED
(U: WEAKC.Y LOCKED
m, m': mAMO m'COUPLSO
50
30 -
20
20 30 40 50 60 70
Fig. 4 AS in Fig, 3^ except for time averaged AT of statistical
equilibrium states with outlines oi their characteristics.

FREQUENCY DISTRIBUTION OF 30 DAY RUNNING MEAN AMPLITUDE
IN THE 200 YEAR INTEGRATION
I'M
««.* -
«.« H
».•
Fig. 5 Sample of six year 4T history from the
200 year annual cycle simulation run.
Fig. 6 Probability distributions of monthly mean
amplitudes in the 200 year annual cycle simulation;
(a) AT , (b) &,(") , first zonal component of .
qeopotential height at the upper level, and (c) p wa * u
second mode wave amplitude at the upper level.
CTJ
CO

514
ANNTTAL CYCLE OF
2TTT
Fig. 7 Annual cycle of the ensemble mean ~averaged.over "block"
year (dotted line) and "normal" year (cross line) from the 200
year annual cycle simulation run. The background solid lines are
statistical equilibria with error bars showing standard deviations.

515
EFFECTS OF VERTICAL WIND-SHEAR ON KELVIN WAVE-CISK MODES
Hock Lim, Tian-Kuay Lim and C.P. Chang
ABSTRACT
This paper consists of two parts : a basic study of the Kelvin
wave-CISK modes followed by an examination of the effect of vertical
wind shear on these modes with a view of interpreting the observed 30-
60 day oscillations in terms of such Kelvin wave-CISK modes.
Linear analyses of CISK have invariably led to the conclusion
that the instability mechanism strongly favours cumulus scale rather
than synoptic scale circulations. This is often considered a serious,
if not fatal, defect of the CISK theory. On the other hand, numerical
CISK models had experienced no great difficulty in generating
disturbances which have reasonably realistic structures. It is a
little surprising that this apparant contradiction between theoretical
prediction and simulation results has so far not been thoroughly
investigated.
Lau and Peng (1987) reported that when they used a linearized
CISK parameterization, the disturbance generated in their numerical
model indeed had its energy concentrated in the shortest wavelength
components. However, when they used a "positive-only" CISK where
latent heating was allowed in the ascending regions but cooling was
not allowed in the descending regions, the disturbance tended to take
on a wavenumber-one structure.
In this paper, we examine the dynamics of wave-CISK using a 5-
level spectral model in p-coordinate. The model has no damping
mechanism and conserves energy in the absence of CISK. In the
preliminary experiments with linearized CISK, short waves grow rapidly
to dominate the flow field, in agreement with theory. However, with
"positive-only" CISK, the strong nonlinear interactions due to the
CISK heating mechanism was found to lead to organised wave-CISK modes
which propagates forward steadily without change of shape. The wave-

516
CISK modes have narrow axes of ascending motion and broad regions of
associated descending motion. For modes with small growth rate, the
descending motion covers the whole equatorial belt. For modes with
large growth rate, the descending motion becomes more concentrated
around the axis of ascending motion. The axis of ascending motion
tilts backward with height and the circulation ahead of the ascending
axis is deeper and more intense than the circulation behind the
ascending axis. Depending on the vertical heating profile, the
propagation speed varies from about 14 to 28 m/s in the absence of a
mean zonal wind. The structure and propagation speed of the wave-CISK
modes are in good agreement with the analysis of Chang and Lim (1988).
When damping mechanisms are included, they are found to reduce
significantly the growth rate, but to only slightly modify the other
characteristics of the wave-CISK modes.
The effect of a mean wind with a constant shear (dU/dp -
constant) is then investigated. An easterly shear is found to enhance
the instability of the wave-CISK modes while a westerly shear is found
to stabilise them. The wave-CISK modes are "Doppler shifted" in their
propagation speed, but the shift is often only about half the average
mean wind over the depth of the wave-CISK mode.
Based on the average mean zonal wind of the tropical belt (Newell
et al., 1972), the propagation speeds of the Kelvin wave-CISK modes at
different longitudinal band are estimated. After incorporating the
mean-wind effects, the Kelvin wave-CISK modes take about 40 days to go
round the equatorial belt.
This study strongly indicates that Kelvin wave-CISK modes are the
basic mechanism driving the observed 30-60 day oscillations first
observed by Madden and Julian (1972).

517
ANALOGOUS RHYTHM PHENOMENON OF CLIMATIC
ANOMALIES ON A SEASONAL SCALE
Chou Ji~~fan
Department of Atmospheric Sciences,
Lanzhou University, Lanzhou, Gansu, 730001 .China
ABSTRACT
The study of analogous rhythm of climatic anomalies on a seasonal
scale is summarized in this paper. The objective and meaning
of this problem are also introduced. The analogous evolution
of monthly mean circulation is analysed with strict statistical
method and observational data of long time series by Huang, Gao
and the author recently. The results show that the analosous
rhythm exists significantly. Some research has been done on the
comprehensive interpretation of the physical mechanism of the
rhythm, such as Marchuk (1979), Musaelyan (1980) and Wang
(1984). Recently the dynamical mechanism of analogous rhythm
phenomenon is studied theoretically with a greatly simplified
air-sea coupled model by Huang and the author. The result is given
in this paper. In addition, the numerical simulations and various
sensitivity studies are made by a more sophisticated global
air—sea coupled dynamic—statistical model to simulate the analogous
rhythm phenomenon in the evolution of the monthly mean
circulation anomaly. The results not only prove the results of the
theoretical analysis, but also provide a basis for utilizing the model
further to do seasonal long-term numerical forecast.
L Introduction
It is well known that the theoretical limit of daily weather forecasting is

518
about one to three weeks. Prediction over a month can only be based on
weather statistics, i.e., climate forecasting in which the monthly averaged
temperature and monthly total precipitation are main predictands. If the
monthly averaged temperature and total precipitation during the period
from May through September could be well predicted in March or April for
a certain year, then great benefits for agriculture, transportation, etc., could
be derived. This is especially true in the East Asia, which is under the influence
of the southwest monsoon at that time of the year. Such predictions
depend on the prediction of the monthly averaged circulation. At present,
GCMS are widely used for this purpose by averaging the daily forecasts.
Shukla (1981) showed that, for a given boundary forcing, the effect of atmospheric
transients in the initial field on the monthly averaged circulation
is in the range of 45 days. Using an eleven—layer GCM and observational
data of surfce sea temperature (SST), Owen (1987) (9) simulated the summer
rainfall over Sahel successfully. Also with a GCM, Dickinson
(1987) (2) made a test on EINino during the 1982/83 winter. The results
show that the prediction for the tropics is greatly improved with the use of
observational SST while little improvement appears in the mid—latitudes.
These and some other tests suggest that the external forcing is closely related
to the monthly averaged circulation. From another point of view,
Namias and Cayan (1984) (8) pointed out that external forcing cannot determine
all the statistical properties of atmspheric parameters. It was also
suggested by Lau (1981) (3) , in his time integration results over 17 years with
a GCM, that the monthly averaged circulation over two months was
unpredictable to a certain degree. Such unpredictability was called by Madden^
981 ) (5) as ''climate noise 77 . Then he introduced the concept of
"Signal—to—niose ratio (SNR)" and found it to be dependent on season and
geographical region. However, we should not forget that the observational
data of the atmospheric initial field are different from those used in the
simulations. One is the adjustment between the atmosphere and the external
forcing. The other is that the atmospheric evolution could influence the
boundary forcing, and in reverse, this forcing could influence the atmosphere.
To determine the future external foring, a coupled air—sea
(land)model is needed, in which the initial condition should include the
3-dimensional distribution of some physical quantities not only in the atmosphere,
but also in the uppermost layers of the earth's crust. Then the
seasonal climate is determined by the boundary condition at the earth's surface
and the atmospheric initial field. The authof believes that, excluding
the climatic noises, the initial field (in the atmosphere and in the uppermost

519
layers of the earth's crust) should include information of monthly averaged
circulation in the coming 6-11 months. The theoretical predictability on
monthly averaged circulation should be 6-11 months {chou, 1986) (1) . But
the difficulty is that no sufficient 3-dimensional data in the uppermost layers
of the earth's crust can be used. Fortunately, the 3—dimensional data
can be derived from general circulation anomalies in previous integration.
Therefore, we can use this circulation instead of the 3—dimensional data.
Nowadays, the routine long-range forecast performed with synoptic methods
or statistical methods is mainly concerned with the evolution theory of
general circulation. In these works, it is found that the teleconnection of
general circulation exists not only in space, but also in time. Some theories
have been suggested to explain the spatial teleconnection, but little attempt
is being made to investigate the temporal teleconnection, though the study
is of great significance for simulating the atmospheric general circulation
and predicting the climate numerically. In this paper, we shall introduce
some preliminary researches in this area, which consist of some unpublished
works of the author and his cooperators. Considering the limited space of
the paper, we shall give only a simple summary.
2. Analogous Rhythm Phenomenon
When atmospheric or oceanic variables are analyzed, a close asychronous
connection with a large time phase difference can be found. The relation is
very helpful for a forecaster to ensure the validity of the forecasts at the valid
time. Therefore, a lot of research has been carried out. However, some of
these relations presented are nothing but wrong ones due to sample error.
Thus the problem becomes complicated.
The original concepts for solving this problem can be traced back to
1920's. Mylbralovsk found a kind of temporal teleconnection for some
weather phenomena by analyzing daily surface weather maps in the USSR.
This teleconnection has a quasi—periodic time interval (90 days or 150
days), and he called it ''rhythm". The forecast accuracy with rhythm is fairly
high, between 65% and 80%. The difficulty is that the condition for beginning
the rhythm is so strict that few cases can be found during a month
(Wang and Zhao, 1.987) (13) ,. Of course, this cannot meet the need of prediction.
Wang (1984) (12) defined the /'rhythm'' as the time connection in the
atmosphere with a certain period (especially, quasi-semiannual interval),
He emphasized that the rhythm appears only in a fixed season and did not
repeat continuously, and no second phenomenon can be analyzed from the

520
first one. In his work, the long-term weather process is divided into 3 kinds
of time scales, in which rhythm is the one with a scale of 3—6 months. Zhao
et al. (1982) (15) pointed out that there are certain rules on the spatial distribution
of the rhythm index. The significant indices mainly locate in three
regions: the North Pacific, the North Atlantic and the region from the
South Asia to the West Pacific. In contrast, few high rhythm indices are
found in the continent of middle— high latitudes. Most works related to the
rhythm phenomenon are based on the calculation of lag correlation
coefficients. An restriction here is that the relation must be linear. However,
if the analogous method is used, the restriction will disappear and the complicated
properties and nonlinear interactions of the real system can be studied.
This analogy was originally used for assessing predictability by Lorenz
(1969) (4) in his investigations of middle-range weather forecasting. It has
also been found that this approach can be successfully applied to researching
the rhythm in long-term circulation variations (Wang, 1984) (12) .
Using monthly averaged surface pressure field, SOOhpa geopotential hight
field over the Northern Hemisphere and the maps of SST anomalies over
pacific and Atlantic during 1951-1975, Wang (1984) (12) examined their similarities
between any two of them. It is found that the similarities do not decrease
monotonicsely with time, but increased in some specific interval of
time, for example, from five to seven months after the initial month. It
means that there is rhythm phenomenon in the atmospheric circulation variation
as well as in the oceanic variations. However, in Wang's paper, the
category of similarity is a sign correlation r
where, N is the total number of points in the two anomaly maps used;
n + and n_ the number of points of the same and opposite signs,
respectively.
The analogous evolution of 500hpa monthly mean circulation is analysed
with strict statistical method by Huang, Gao and Chou (unpublished), recently.
The similarity index is
•R^-dnqVyl-yW) (2)
where c= 16 /In2, 0^ is the similarity coefficient between two anomaly
fields in different years, The average value of 0y is given by
cT = l-r(l-^) (3)
(1)

521
where r 4j is the correlation coefficient, Ey the Euclidean distance between
two anomaly fields, S i Sj the mean squared deviations of the two fields,
respectively, S = S. + S .: m the number of mesh points. The results show
m i j
that a quasi—semiannual rhythm exists significantly, i.e. the anomalous
fields between two different years are analogous in a starting month, the
similarity will rapidly become poorer in time, but after about six months, it
will become a little analogous again. It appears that a quasi-semiannual
rhythm phenomenon does exist in the evolution of monthly averaged circulation
anomalies and SST anomalies. Further research on this area will
greatly encouraged us to realize climate anomalies of a seasonal scale and
design numerical climate forecasting models correctly.
3. The Mechanism Of Analogous Rhythm
Some research has been done on the comprehensive interpretation of the
physical mechanism of rhythm. Marchuk(1979) (6) tried to explain the
rhythm with air—sea interaction. In a further study, Musealyan
(1980) (7) suggested that, during the summer half year, solar radiation is
obsorbed and stored in the deep layer of the sea, which may be called a
"memory 77 process; and during the winter half year, the stored energy is
transfered to the air through different processes and then propogated
downstream with the flow. In this way, he explained the formation of temporal
teleconnection between time-mean summer cloud cover over the
North Atlantic and the winter temperature in European parts of Russia.
Zhao and Wang (1982) (15) found that the rhythm indices depend closely
on geographical regions and the most significant indices appear in oceans.
This is another evidence of the relation between rhythm and air-sea
interaction.
In a study of long—term weather processes on air—sea interaction, Zang
and Wang (1983) (14) pointed out that there is. an obvious rhythmic relation
between summer SST and winter SST in the west drift region. They explained
that, during summer the ocean active layer is warm in the up-layer
and cool in the down-layer; solar energy and heat energy transferred by
atmospheric turbulent exchange are absorbed by the ocean, increasing the
temperature in the surface layer; in winter, ocean discharges the energy
stored in summer and the thermodynamic structure of the ocean active layer
becomes an isothermal one. Therefore, the influence of summer SST
anomalies on the air can be detected only in winter.

522
The works mentioned are qualitative and only thermodynamic processes
are taken account. Recently, Huang and Chou (unpublished) studied the
mechanism of analogous rhythm on monthly averaged circulation anomalies
over the Northern Hemisphere dynamically and numerically. In their
work, the evolution of monthly averaged circulation is taken to be a disturbance
supperimposed on the historical analog. The state of air and sea
can be devided into the basic and disturbed state, such as the monthly averaged
value in a historically analogous years for the latter, which may be called
analogous deviation disturbance (hereafter refer to as ADD). Thus the
study of establishment and variation of the rhythm can be transformed to
that of the evolution and stability problem of the ADD. With the
3—dimensional structure character of general circulation anomalies (i. e. the
baratropic property, stationary wave and space teleconnection property of
monthly averaged circulation anomalies), the model can be greatly simplified
and the amplitude equations of ADD for the air-sea coupled system
can be deduced. Using these equations, in an uncoupled or coupled system,
the time evolution charateristics of the ADD amplitude are studied. The results
show that in a uncoupled system with a constant SST, ADD may increase
rapidly due to linear instability. This is consistent with the analogous
evolution of circulation anomalies at initial time in the real atmosphere.
However, because of the nonlinear interaction in the atmosphere, the deviation
disturbance does not increase infinitely but tends to a steady state. This
will not bring out an analogous phenomenon. But in a coupled system,
things are different. The SST ADD can not only force the atmosphere but
be influenced by it as well and nonlinear feedback processes in an air—sea
system may produce a nonuniform fluctuation of ADD. Being forced
seasonally by the monthly averaged circulation, the coupled fluctuation of
ADD in the air-sea system is intensified and a large amplitude appears.
The result shows that the analogous rhythm phenomenon is an uneven vacillation
of ADD which is caused by nonlinear coupled interaction of air-sea
system forced by seasonal changes of the monthly mean circulation.
Considering the simplicity of the model, more sophisticated numerical
models are needed to compare the actual data with theoretical results. We
design another global air-sea coupled model wi£h a predictor of. ADD
which describes the similarity of monthly averaged circulation anomalies. It
is to input the three—dimensional structure characteristics of anomalies of
the monthly mean circulation as facts into the model, and filter the so-called
'''Climate noise 77 partly. In addition new processing methods are given on
the parameterization of diabatic heating and numerical solving process of

the equations and so on. Then performing coupled or uncoupled model test
with the model are possible.
In the uncoupled model test, SST deviation is stationary, which implies
linear and nonlinear response of analogous evolution of monthly averaged
circulation to stationary SST forcing. In the coupled tests, control test is
done at first. On the basis of this, various sensitivity studies are made. The
results not only prove the result of the theoretical analysis, but also provide
a basis for utilizing further this model to do seasonal long—term numerical
forecast.
4. Discussion
It has been shown in the analysis of actual data that there is a temporal
teleconnection in the general circulation, one of which is the analogous
rhythm of monthly averaged circulation. The author believes that it is true.
In spite of this the truth of this phenomenon can still be doubtful because of
the vagueness in this phenomenon and the limitation of the statistical methods
and sample error appearing in the study.
As to the mechanism of the establishment of rhythm, the current study is
still preliminary. Although the rhythm has been simulated with a numerical
model, the physical mechanism is still ambiguous. To some extent the problems
solved far less than that those produced. Further study should be done
to realize the mechanism of the formation of climate anomalies on a seasonal
scale and present a numerical climate forecasting method.
REFERENCES
[I] Chou, J.F., Long-range numerical weather prediction, Meteorological press,
Beijing, PP.329 (1986)(m Chinese).
[2] Dickinson, A,, WMONWP Annual Progress Report for 1987,195-196(1988).
[3] Lau, N.C., Mon. Wea. Rev., 109,2287-311(1981).
[4] Lorenze. E.N., J.Atmos. ScL, 26,636-646(1969).
[5] Madden, R.A., J.Geophys. Res., 86, No.lO(1981).
[6] Marchuk, G.L, World Climate Conference, W.M.O. Geneva, 12-32, February,
1979.
[7] Musealyan, C.A.Meteorology and Hydrology, 3,1980 (in Russian).
[8] Namias, J., and D.Cayan, Oceamis, 27,40-45(1984).
[9] Owen, J., WMO NWP Annual Progress Report for 1987,195-196(1988).
[10] QhvC.and J.Chou, Acta Meteor.Sinica 1,32-42(1987).
II1] Shukla,!, J, Atmos. Sci. 38,2547-72(1981).
523

524
[12] Wang, S.-W., Adv. in Atmos. ScL, 1, 7-18(1984).
[13] Wang, S.-W. and Zhao Z.-C., Basic for Long-Range Weather Prediction,
Science-Technique Press, Shanghai, PP. 201, (1987)(in Chinese).
[14] Zang,H.-f. and S.-W. wang, Acta Ocean Sinica 5, 163-171 (1983) (in
Chinese)
[15] Zhao, Z.-C., S.-W. Wang and Z.-H. Chen, Acta Meteor. Sinica, 40,
464~474(1982)(in Chinese).

525
An Inquiry into the Nature of Regional Cyclogenesis
by
Mankin Mak
Department of Atmospheric Sciences
University of Illinois, Urbana-Champaign,IL,61801,USA.
Abstract
Extratropical cyclones tend to originate or rapidly develop over certain preferred
longitudinal sectors of the atmosphere as clearly evidenced by the pronounced
longitudinal variability in the local time mean circulation statistics. The regions over
north-western Atlantic and Pacific Oceans off the east coast of North America and Asia
respectively are the regions of frequent cyclogenesis. Those are also known to be the
downstream locations of the two major troughs in the winter mean flow throughout the
troposphere. It is therefore reasonable to anticipate that regional cyclognesis is closely
related to the localization of the background flow itself.
This paper reports the results of the modal and non-modal instability of a basic
flow with a general three-dimensional shear structure in a two-layer quasi-geostrophic
model. This analysis confirms that not only the local baroclinic and barotropic shear, but
also the zonally uniform part of the basic flow have a direct influence upon all the
instability properties of the disturbances. There exist several branches of travelling global
and local unstable modes for a range of parameteric values relevant to the observed
atmosphere. The global modes are mainly associated with the zonally uniform part in the
baroclinic shear. The local modes are highly concentrated at the exit region of the jetstreak.
In addition, one branch of unstable modes is stationary mainly arising from the
barotropic process at the level of the jet. The shorter this jet is, the more dominant would
be this mode.
A complete local energetics analysis has been found particular^ instructive in a
recent study of the local barotropic instability (Mak and Cai, 1989). A generalisation of
such a local energetics diagnosis for this baroclinic model reveals that there are two local
energy generation processes and three redistribution processes. These energy processes
in a typical unstable disturbance are found to have comparable values and counterbalance
one another to a high degree. It is the compensating and yet accumulative effects of these

526
processes that give rise to a distinct downstream localization in the unstable disturbance.
Results of a corresponding non-modal analysis will also be reported. They reveal how
such a local disturbance could naturally emerge for different initial perturbations. The
details of these results will be reported in an article by Cai and Mak (1990)
1. Introduction
Extratropical cyclogenesis does not occur uniformly at all longitudes. For
example, winter cyclones preferrably form over the east coast of North America
(Roebber, 1984). There are zonally elongated zones of maximum high-frequency
variance in the geopotential data downstream of the North Pacific, North Atlantic , and
south Indian Ocean jet streams. Such zones are generally referred to as the "storm
tracks" ( Blackmon, 1976, Trenberth, 1981). There have also been observational
studies about the feedback effects of the transient disturbances onto the time-mean flow
(e.g.Lau and Holopainen, 1984). The storm tracks appear, in turn, to be closely related
to the low-frequency atmospheric variability, The theoretical understanding of this set of
interlocking phenomena is still very incomplete.
The problem of regional cyclogenesis is, in essence, an instability problem of a
zonally inhomogeneous atmospheric flow. This talk is mainly based on the results
reported in Mak and Cai (1989) and Cai and Mak (1989). An integral part of both
analyses is a complete diagnosis of the local energetics. Our specific task is to
investigate the basic dynamics of local instability of a baroclinic jet-streak in the context
of a two-layer quasi-geostrophic beta-plane channel model. For lack of space, the
derivation of the formulae is not shown. Only some highlights of the results in CM are
presented.
2.a Model
A two-layer quasi-geostrophic (3-channel model centered at 45° is used with a
width of TtL = 4500 km and a length of 2nL/y =30,000 km, whereby y = 0.3. We use
OU Poo = 1000mb, D=15 m/s, 1)* L=l day and LD) as units to measure the horizontal
distance, the pressure, the velocity, the time and the streamfunction of the motion. The
nondimensional governing eqs. can be written in terms of a barotropic (\|/) and a
baroclinic (0) streamfunction of the perturbation flow and those of the basic state OF, ®)
,v V)..+J(e,v 2 e) •+J(0,v 2 e) + p ||+rvV=o
3 2. 2 2 2 2 36
•v (V 0) + J(\jr,V 0) -f JOF,V 8) + J(6,V ¥) + J(0,V w) + B ^
01 . T r fjv

527
(1)
- 2F 0 ( + %,0) + JCF.9)) + rV0 = 0.
fnL 2
The Froude number FQ is related to the static stability S by FQ = 4 -^ . We set FO =
Sp§o
5.0, p = p* L 2 !)" 1 = 2.5 and a frictional coefficient r = r^L/D' 1 = 0.1 . A spectral
method is used to obtain the solution with 660 components.
2.b Basic state
We consider the following basic flow
W+8 = Wi(x ; y)= -(Uo+Ur)y + A exp(-a| 1-x/xo|) sin 2y
W-8 - W2(x,y)= -(Uo-Ui)y
(2)
with xo = TC/Y. Fig. 1 shows the upper level streamfunction of such a basic state with A
= 1.5, a = 4.0, and U 0 = U T = 0.3. The basic flow at the lower level is at rest in this
case. The length of the jet is one eighth of the channel length (3750 km). The maximum
zonal wind speed is about 54 m/sec and the maximum local vertical shear of the zonal
wind is 27 m/sec/500mb. UQ is the vertical mean constant zonal wind and Uj
represents the vertical mean shear of the basic flow. The results for different
combinations of the two parameters, A and a, have essentially the same characteristics.
The uniform zonal wind shear U T is set to be either Uj = 0.0 which corresponds to no
background temperature contrast, or U T = 0.3. The constant zonal wind U 0 is allowed to
vary from 0.0 to 1.0.
27T/7
Fig. I Nondimensional streamfunction at the upper level Tj of the basic flow for A - 1.5,
a « 4.0, and U T = U 0 * 0,3. The contour is 0.21. T 2 is identically zero for U 0 « U T .

528
3 Local energetics analysis
The vertically-integrated local energy in the two-layer model is
2
{£} ={K] + {P} where (K) = Zj( u + v ) and {P}=2F 0 0 , {K } is the
i =1
vertically-integrated local kinetic energy and {P}is the potential energy. The subscript "i"
stands for the layer index. The equation for the total local energy can be shown to be
- -
-f {E«D} + {F h *T h } - 2r {K}
at
(3)
The first term on the r.h.s. is the advection of K and P by the mean flow. The second
term represents two processes of local redistribution of energy by the ageostrophic
component of the pressure and flow components. {E.D} represents the local barotropic
mechanism of energy generation. The E vector is a measure of the local shape and
horizontal orientation of the disturbance field. The D vector is a measure of the
deformation field of the basic flow. { F h vT h } measures the local baroclinic mechanism of
energy generation. The Fh vector measures the horizontal heat flux and the Th measures
the local slope of the basic isentropic surface. The last term on the r.h.s. measures the
frictional dissipation rate.
Furthermore, the feedback tendencies of basic state, -— and -— , due to the
heat and vorticity fluxes of a disturbance can be obtained by solving the following
equations with the r.h.s. to be treated as given,
2 BVJ> 2 2
V ( ) = -J(y,V V)-J(6,V 8)
(V - 2F 0 )( ^~) = - J(y,V 6) - J(9,V Y) + 2F o Jfy, 9). (4)
All terms on the r.h.s. of (3) and (4) can be readily evaluated once the solution of the
strcamfunctions, \|/and 8, are known.
4. Results
Only part of the results of the normal mode unstable disturbances are discussed.

529
4.a Dependence of the instabiiity properties upon IL
The normal mode instability calculation confirms that in general there exist both
local and global unstable modes for a jet-streak with a background shear IL,. A local
mode consists of a group of dominant waves which jointly contribute to a maximum of
local energy in the downstream of the jet core. The growth rate of a local mode is
strongly dependent upon the vertical mean of the uniform part of the zonal wind, U 0 . A
global mode, on the other hand, mainly consists of a single wave and its growth rate is
only weakly dependent upon U 0 . Moreover, there exist only unstable local modes in the
absence of a background baroclinic shear (i.e., U T = 0.0).
Fig. 2 shows the variations of the eigenvalues of all unstable monopole modes
with respect to U 0 for the case of U T = 0.3. Panel (a) shows that the growth rates of the
local modes (modes #1, #2, #3, and S) decrease with U 0 after they reach their peak
values and eventually become zero for large U 0 . Mode #1 is again the most pronounced
local mode. Among the three travelling local modes, mode #1 is the most unstable
mode for U 0 < 0.38; mode #2 becomes most unstable in the range of 0.38 < U Q <
0.75; and mode #3 is most unstable when U 0 is sufficiently large. It is found that the
maximum energy center of a local mode gradually shifts to further downstream region of
the jet stream and the maximum value of the local energy also tends to decrease as U Q is
increased (not shown here). The detailed structures as well as the results of the various
diagnostic analyses of a representative travelling local mode are presented below. Mode
S also has a strongly localized structure but its growth rate is the smallest among the local
modes for this value of U T . The growth rate of this mode is also much smaller than that
for U T = 0.0. Hence, the most favorable condition for this mode to grow is when there
is no zonally uniform background baroclinic shear.
The growth rates of the global modes are shown in panel (b). In general, they
are much less sensitive to U Q and tend to gradually increase with U Q . Comparing panel
(b) with panel (a), we find that a local mode is the most unstable mode for U Q < 0.45
and the global mode #6 becomes the most unstable mode for U 0 > 0.45. Hence, a large
value U Q is favorable for a global mode but not for a local mode in the presence of a
background shear U T . This indicates that the constant zonal wind U Q not only changes
the instability properties of an individual mode but also alters the structure of the most
unstable mode (from a local mode to a global mode).
Fig. 2c shows the variations of the frequency of each branch of the unstable
monopole modes. The result reveals that it is possible to separate the travelling local
modes from the global modes according to their frequencies. The frequencies of the
travelling local modes are higher than those of the global modes. It is also found that the

530
dominant waves of a local mode are shorter than the dominant wave of a global mode.
These are Rossby waves individually. This is why a local mode travels eastward much
faster than a global mode.
«- (a)
LOCAL MOOES
Fig. £
Variation of the instability properties of the unstable monopole modes with respect to
U 0 for A - 1.5, a ~ 4.0, and U T =0,3. (a) Growth rates of the local modes, (b)
Growth rates of the global modes, (c) Frequencies of all the unstable modes.

531
4.b Properties of the unstable travelling local modes
We present the results of various diagnostic analyses of the most unstable local
mode for IL = U T = 0.3 (i.e., mode #1) as a representative of this class of modes. That
mode is indicated by a dot in panels (a) and (c) of Fig.2 The growth rate of this mode is
a r = 0.22 (e-folding scale « 5 days) and the frequency is aj = 1.62 (period » 4 days).
(i) structure
Shown in Fig. 3 are the normal mode perturbation streamfunctions in each layer
at the two extreme phases (t = 0 and t = K /(2

532
slightly longer than the length of the jet itself (3750 km). The westward vertical tilt of
these meridionally elongated eddies is well observed in both phases. The meridional tilt
at the upper level is not favorable for the eddies to extract kinetic energy from the basic
flow. The perturbation flow at the lower level has a weaker meridional tilt. The eastward
propagation of the disturbances can be identified by comparing the longitudinal position
of an individual eddy at the two phases.
(ii) Local energetics
The results of the budget calculation for {8} are shown in Fig. 4. The local
generation rate of potential energy {F h * T h }is localized in the immediate-downstream of
Fig. 4-
Local energetics of the unstable mode shown in Hg.3. (a) Potential energy
generation rate- (b) Kinetic energy generation rate, (c) Energy advection by the basic
flow, (d) Work done by ihe gcosirophic pressure acting on the ageostrophic wind,
(e) Work done by the ageostrophic pressure acting on the geostrophic wind.
(0 Net local energy change rate.

the jet core. The maximum value is 1.63 and is found at the longitude "Xj = 1.0" (the
first grid point to the right of the center in the plot) (Fig.6a). The spatial distribution of
the generation rate is fairly symmetric about its maximum. The domain integrated
potential energy generation rate is 2.1. The local kinetic energy generation rate {E*
D}is found to have a localized negative center at the longitude "xj = 2.0" with a peak
value of - 0.26 (Fig. 4b). Therefore, the instability of this mode is attributable to the
baroclinic process. Advection (Fig. 4c) removes energy from the immediate-downstream
region to a region further downstream of the jet core. The same is true for the work done
by the ageostrophic geopotential (Fig. 4e). On the other hand, the mechanical work
done by the geostrophic geopotential on the ageostrophic wind tends to redistribute
energy from the exit region to the entrance region of the jet stream (Fig.4 d). The
contribution to the local energetics from the baroclinic ageostrophic component is much
stronger than that from the barotropic ageostrophic component (Fig. 4d vs 4e). It has
been verified that the time-mean local rate of change of the energy is exactly twice the
product of the local energy and the growth rate of this mode (Fig. 4f). The effective
local energy generation rate {G} is equal to the sum of panels (a), (b), (d), and (e) plus
the dissipation term. Its maximum value is centered at longitude "x { = 0.5".
Therefore, the maximum local energy growth rate is located east of the maximum
effective local energy generation rate (longitude "xj = 1.0").
(iii) Feedback effects on the basic flow
533
The feedback tendencies of the perturbation normal mode on the basic flow are
shown in Fig. 5. The feedback effects are quite localized as one might expect from the
structure of the normal mode. The disturbance at the upper level (Fig. 5a) tends to
reinforce the basic jet stream by having a positive tendency of *F on the south side of the
jet and negative tendency on the nortli side. The disturbance at the lower level (Fig. 5b)
Kg, B Feedback effects of the unstable mode shown in Fig.3 on the basic flow.
(a) Geopotential tendency at the upper level (b) Geopotential tendency at the lower
• • ' • • •• level' • • • ' " • • ' ' • . ' • • ' • • ' : • • . • . ' .-•• ' • • - • • • • - • • . ". .

534
tends to induce a westerly jet at the middle of the domain. The feedback at the lower level
is even stronger than that at the upper level. Since there is no basic jet stream at the lower
level, the feedback effect amounts to reducing the vertical shear of the basic flow. This is
the same as saying that the feedback on the basic temperature field 0 has a wanning
tendency on the north side of the jet and a cooling tendency on the south side. This
destructive feedback on the basic temperature field is attributed to the strong localized
northward heat flux of the disturbance near the jet core. As a consequence of the upgradient
transportation of the vorticity, the disturbance has a constructive feedback on the
barotropic part of the jet-streak. These theoretical results of the tendency calculations are
in good agreement with the observational results of Lau and Holopainen (1984).
5. Conclusions
Regional cyclogenesis is, in essence, a local instability of a 3-D atmospheric
baroclinic flow. It usually occurs at the downstream of the baroclinic jet-core in the
winter season. There are typically many local and global unstable modes even for a
idealized basic flow. They can be however classified into three classes, analogous to
Frederiksen and Bell (1987). The local energetics of a representative local (monopole
cyclogenesis) mode reveals that the potential energy is extracted at a maximum rate from
the baroclinic shear in the near-exit region of the basic jet stream. Most of this potential
energy is simultaneously converted locally to the kinetic energy through the mechanism
of warm air rising and cold air sinking. The barotropic process destroys kinetic energy
of this mode. The ageostrophic pressure work is quite weak compared to the geostrophic
pressure work. The latter acts to remove some energy from the exit region to the
entrance region of the jet-streak. The advection term, however, acts to remove energy
from the region where the energy generation is strongest to a region further downstream.
As a consequence of the various energetic processes, the maximum local energy change
rate is found at the exit region of the jet stream and so is the local energy.
The tendency calculations reveal that the localized disturbance has a baroclinically
negative effect and a baiotropically positive effect on the basic jet stream. For the other
results of the modal and particularly the non-modal instability analyses of regional
cyclogenesis, readers are referred to Mak and Cai (1989) and Cai and Mak (1989).
References:
Blackmon, M. L., 1976: A climatological spectral study of the 500 mb geopotential
height of the Northern Hemisphere. J. Atmos. Sci., 33, 1607-1623.
Cai, M. and M. Mak., 1989: On the basic dynamics of regional cyclogenesis. J.Atmos.
ScL (Submitted)

535
Frederiksen. J., and R. C. Bell, 1987: Teleconnection patterns and the roles of
baroclinic, barotropic and topographic instability. J. Atmos. Sci., 44, 2200-2218.
Lau, N.-C, and E. O. Holopainen, 1984: Transient eddy forcing of the time-mean flow
as identified by geopotential tendencies. J. Atmos, Sci, 41, 313-328.
Mak, M. and M. Cai, 1989: On local barotropic instability. J. Atmos. Sci., 46, 3289-
3311.
Roebber, P. J., 1984: Statistical analysis and updated climatology of explosive cyclones.
Mon. Wea. Rev., 112, 1577-1589.
Trenberth, K. E., 1981: Observed southern hemisphere eddy statistics at 500 mb:
Frequency and spatial dependence. J. Atmos. Sci., 38, 2585-2605.

536
TYPHOON FORMATION AND DEVELOPMENT - AN OBSERVATIONAL POINT OF VIEW
Cheng-Shang Lee
Department of Atmospheric Sciences,
National.Taiwan University
The physical processes which lead to the formation and
development of tropical cyclones are still not well-understood. The
current observational analysis attempts to advance our knowledge
towards fully understanding these complicated processes by using the
rawinsonde composite and individual case analyses. Results have
indicated that the large scale momentum surges, which can provide
inward eddy vorticity flux, are conducive to the development of a
cloud cluster to a tropical cyclone. The cumulus heating efficiency is
very low due to the weak vorticity field associated with the system
during this formation stage. The most common large scale forcing
includes the cross-equatorial surges due to strong cold out-break flow
in the counter-hemisphere and the trade wind surges.
After the formation of a tropical cyclone, the heating efficiency
increases due to the increased vorticity associated with the system. A
tropical cyclone can develop much faster if the spiral rain bands are
active, because they can provide the needed inward moisture and
angular momentum flux for cyclone to develop. However, the cyclone can
also increase its developing trend through the CISK mechanism. At this
stage, it might be more realistic to pay attention to the processes
which can hinder the cyclone developing trend, such as the strong
middle to upper level wind shear or the land masses influence.
The intensity limit which a typhoon can reach is often posed by
underneath sea surface temperature. However, the existence of TUTT can
often redirect the upper level outflow to form outflow channel (or the
vacumn effect) which is often favorable for typhoon development.
In addition, results using satellite data analysis also show that
a convection burst often occurs before or during the major intensity
increase. The maximum inner region convection often leads the maximum
intensity of typhoon.

537
DYNAMICS OF VORTEX MOTION ON TROPICAL /3-PLANE
Melinda S. Peng and R. T. Williams
Department of Meteorology
Naval Postgraduate School
Monterey, CA 93943
ABSTRACT
A linear nondivergent barotropic model is developed to understand the dynamics associated
with the asymmetric structure and to obtain the asymmetric circulation for a vortex moving on
the /-plane. The total system is transformed to a coordinate system moving with the vortex. The
direction and speed of movement is specified from full nonlinear model results. Stability analysis
for a commonly used axisymmetric wind profile shows that it is barotropically unstable. Two
different wavenumber one gyres are obtained from the asymmetric vorticity equation. The inner
gyres correspond to the free unstable mode from barotropic instability. The outer gyres, whose
orientation is always along the track direction specified by the movement, correspond to the /-gyres
obtained in the nonlinear numerical models. In steady-state solution, the outer /-gyres can be
isolated by placing the inner boundary some distance from the center or by reducing the resolution
so that the intensity of the inner gyres is reduced. The asymmetric circulation obtained from the
time-dependent linear model has the correct orientation and magnitude when compared to the
solutions of the full nonlinear model. The nonlinear dynamics associated with the asymmetric
circulation proposed by previous studies is confirmed.
1. Introduction
The prediction of tropical cyclone movement is difficult in part due to a lack of knowledge of
the dynamics of vortex motion in a flow field with an absolute vorticity gradient. To understand the
basic dynamics of cyclone movement, analytical studies are carried out with simplified equations
on a /~plane. Considerable disagreement exists in the literature between various analytical and
numerical studies. An extensive review can be found in the introduction section of Willoughby
(1988), Only those studies that are of immediate relevance to the present study are mentioned
here.
In the Chan and Williams (1987) nondivergent barotropic model, a westward distortion of
an initially symmetric vortex occurred due to the linear Rossby wave dispersion. However, this
asymmetry results in a northward movement of the cyclone center when nonlinear advection of
the vorticity is included. The combined effect causes the cyclone center to move toward the northwest.
Nonlinear numerical studies by Fiorino and Elsberry (1989) confirmed the northwestward
movement on the /-plane without mean flow. The emphasis of Fiorino and Elsberry was on the

538
influence of the mean wind profile of the vortex on its movement. As in DeMaria (1985), Fiorino
and Elsberry found that the intensity (maximum tangential wind) has little influence on the vortex
movement, while the strength of the outer wind profile has a pronounced influence. The asymmetric
part of the flow was dominated by a wavenumber one structure, which Fiornio and Elsberry
called the /-gyres The gyres on the /-plane were oriented in the direction of the movement of
the vortex, and the average speed of the gyres within a radius of 300 km from the center nearly
coincided with the vortex speed of movement.
The purpose of the present study is to use a simple linear model to obtain the asymmetric
circulation that is associated with the movement of a vortex on a /2-plane, when the direction and
translation speed of the cyclone are specified. The speed and direction are specified using the
information obtained from the nonlinear numerical results such as Chan and Williams (1987) or
Fiorino and Elsberry (1989). Another goal is a better understanding of the dynamics associated
with the vortex motion on a /-plane. The time evolution as well as the steady-state structure of
the asymmetric flow will be investigated. If the correct structure of the asymmetric circulation
can be obtained from the linear model, it can be used to improve bogusing of tropical cyclones
into numerical track forecast models.
The model and the solution procedure are described in section 2. The steady-state solutions
are presented and discussed in section 3. In section 4, we perform the linear stability analysis of the
basic wind profile that will help us explain the results in section 3. The time evolution solutions
are presented in section 5. Summary and conclusions are given in section 6.
2. The Model
The dynamics of the vortex are described by the nondivergent barotropic equation on the
/-plane. Based on the quasi-linear, quasi-steady movement of the cyclone vortex on a /-plane as
observed in the full numerical model, the total system of the vortex is transformed to a coordinate
system that moves with the vortex. The dependent variables, however, remain in the original coordinate
system. The extra advective term thus obtained acts as an additional forcing to the vortex
structure. Given the direction and translation speed from the numerical model, our hypothesis
is that this additional forcing along with other mechanisms will produce the correct circulation
associated with the asymmetric part. The nondimensional equation is
(2.1)
If L and U are the characteristic length and speed for tropical cyclones, c = LPftfU
smallness of the / term. For appropriate values, e = 0.12.
measures the
This equation is expressed in polar coordinates (r,0) and is transformed to a constant moving
frame of reference, which gives
= 0- ( 2 - 2 )
where u r is the radial velocity, vg is the tangential velocity, c and tj> represent the speed and
direction of the moving coordinate system, respectively,

In Fiorino and Elsberry (1989), the basic symmetric vortex changes only a small amount during
the time integration. Therefore, the system can be linearized with respect to the axisymmetric part
of the vortex and all the dependent variables can be expanded in terms of the / term coefficient,
e. Based on the numerical model results, the translation speed is typically 2-4 m/sec, which is an
order smaller than the characteristic speed in the vortex, and therefore can be expanded in terms
of e as well. Therefore, we expand the variables as:
539
v e = V Q (r)
ti r = «n(r,0,0 + », (2-3)
c = eci + • • .
Substitutong (2.3) into (2.2) and collecting terms of the same order of e, the order one balance
yields only the arbitrary structure of the basic vortex profile that is independent of the azimuthal
direction. The order € balance gives the tendency equation for the asymmetric part of the vortex
% = -*%-%+«*"*-*-*">«>

540
3. Steady-State Solutions
The purpose of obtaining steady-state solutions of (2.4) (or (2.6)) is to understand the balanced
state from this model. First, let us review the basic dynamics for the wavenumber one structure
from Chan and Williams (1987) and Fiorino and Elsberry (1989).
The origin of the wavenumber one gyres is the linear / term that forces the initial vortex
to become asymmetric with positive vorticity tendency west of the vortex center and negative to
the east. Without a mean flow, the vortex can move through nonlinear advection only if there
is a wavenumber one asmmetry. More specifically, the circulation in this asymmetry crosses the
center of the vortex so that it can advect the total vortex. Therefore, the asymmetry is oriented
in the direction of the vortex movement as observed by Fiorino and Elsberry (1989). This effect
is represented by term 2 on the R.H.S. of (2.4), which is the advection of the basic vorticity by
the asymmetric flow. The direction of the maximum amplitude of this term is therefore in the
direction of the asymmetric circulation when the basic vorticity gradient is negative. The advection
of the asymmetric vorticity by the basic symmetric flow (term 1 on R.H.S. of (2.4)) will not move
the vortex center, but it helps to rotate the gyres so that they are aligned with the direction of
movement.
To show the balanced state, vectors representing the maximum magnitude of the terms in the
azimuthal direction at different radii are given in Fig. 2. The prescribed direction of movement
is 120° and the speed is 0.6 (2.8 m/s). The /-term (term 4) always points toward west. The
direction of term 3, which is the forcing term due to advection of the coordinate system, depends
on the direction specified and the basic vorticity gradient. In the inner part of the vortex (Fig. 2
a) where the basic vorticity gradient is negative (vorticity decreases outward from center), term
3 points toward the southeast when the specified direction is to the northwest (opposite to the
specified direction). Term 2, which represents the advection of the basic symmetric vorticity by
the asymmetric flow, is a maximum in the direction of the orientation of the gyres. Term 1, which
represent the advection of the asymmetric vorticity gradient by the symmetric flow is almost in
the opposite direction of term 2. The residual of these two terms balances with the sum of term 3
and term 4.
In the outer part of the vortex (Fig. 2 b), the basic vorticity gradient is positive (vorticity
increases outward). In this situation, term 3 points exactly to the direction of specified movement.
The / term (term 4) continues to point westward. Term 2 is now in the opposite direction of
the gyre's orientation (d^/dr > 0) and term 1 is approximately in the same direction of term 2.
Therefore, from Fig. 2 b, the gyres are oriented in the same direction as the specified movement
of 120° in the outer part. In between, the balance changes gradually from the inner part to the
outer part (not shown).
The steady-state streamfunction corresponding to Fig. 2 is shown in Fig. 3 where the resolution
is Ar == 5 krn. Note the gyres structure in Fig. 3 correspond exactly to this inner balance
(Fig, 2 a)i From previous discussion, one can see that to obtain the gyres oriented in the direction
of the specified movement, the gyres must be determined by the balance in the outer part of the
vortex; If the inner boundary is placed some distance from r =r 0, then the overall asymmetric
streamfunction can be determined by the balance in the outer part instead of the inner part of the

541
vortex. Solutions obtained by placing the inner boundary 50 km (Fig. 4) from r = 0, produces
the gyre structure in the right orientation. In this case, the balance turns to that of the outer part
quickly and the overall structure is determined by the outer balance. The intensity of this outer
gyres is not sensitive to the innery boundary. However, if the inner wall is moved closer and closer
to the center, the intense inner gyres as in Fig. 3 gradually emerge and finally dominate so that
the outer gyres can not be identified.
The boundary condition used for the steady-state solutions is that # = 0. In the outer part
of the domain, the advection terms are very small and the scale analysis in Section 2 is not valid.
As demonstrated by Carr and Williams (1989), it takes an infinite amount of time to reach steadystate.
While the magnitude of the streamfunction for the /-gyre is very close to that obtained in
the full nonlinear model, the streamfunction near the outer boundary is packed close to the outer
boundary which is unrealistic.
4. Instability of the Basic Velocity Profile
Because the basic radial profile satisfies the necessary condition for barotropic instability, it is
natural to consider whether the inner gyres for the steady state could be a result of the barotropic
instability associated with the basic state. To study the linear instability, the forcing terms are
omitted in (2.4) and normal mode solutions are obtained. The resulting eigenvalue problem is
solved both by the full matrix method method and by the shooting method. Unstable modes are
indeed found for wavenumber one in the azirnuthal direction. The radial structure (eigenfunction)
for the unstable mode is given in Fig. 5. The solid line is the amplitude and the dashed line is
the phase angle. The maximum amplitude is at the radius of the maximum wind. If the radius
of maximum wind is altered, the location of the maximum amplitude follows. The phase angle
is increasing outward, which is consistent with the condition for barotropic instability because
the basic symmetric wind profile is decreasing outward beyond 100 km. Notice the similarity of
this radial structure with that for the inner gyres (Fig. 6) as obtained from the two-dimensional
structure of Fig. 3. This comparison indicates that the inner gyres are actually the unstable mode
of the system.
5. Time-Dependent Linear Solutions
With a basic profile that is unstable, it is not appropriate to consider the linear steady-state
solution without some modification to the basic state, In this section, the time-dependent linear
equation (2.6) is solved with the direction of the vortex motion fixed and the translation speed
increased linearly with time from zero to 0.6 (2.88 m/sec) at 1.5 days and then kept constant
(Fig. 7). These translation speeds are similar to those in the numerical run. For comparison, the
asymmetric streamfunction for the same basic wind profile from the full nonlinear numerical model
is given in Fig. 8. During the early evolution, the gyres in Fig. 7 developed due to the j3 term.
The size of the gyres is the same as the / forcing from the symmetric flow. As time increases,
the gyres become oriented toward the northwest with the ventilation flow in the direction of the
specified 1 movement direction (Fig. 7 b). Since the radial domain of Fig. 7 is 2000 km, the maxima
of the streamfunction are located between 400 to 500 km. The physical domain of Fig. 8 is 1600
km. The magnitude of the maximum streamfunction is within 20% of the magnitude of the gyres
in the full model. Further integration shows the emergence of the inner gyres located at the radius

542
of maximum wind (Fig 7 c). This inner gyres develop as a result of the barotropic instability of
the basic wind profile. The outer gyres remain quasi-steady, while the inner gyres rotate with the
time.
The streamfunction fields in the nonlinear numerical models of Chan and Williams (1987) and
Fiorino and Elsberry (1989) show little indication of the inner gyres with same resolution. One
major reason is that the nonlinearilty in the full model will modify the basic vortex profile. The
axisymmetric flow of the vortex in the nonlinear numerical model at a later state during integration
is retrieved. The shear damping study by Carr and Williams (1989) suggests that these changes
are small. However, this small amount of change is critical. If the profile at this later state is put
into the linear model, the inner gyres are also eliminated.
As stated before, the truely steady-state assumption is not appropiate near the outer boundary.
The time-dependent solution does not have this problem and the maximum of the magnitude is
closer to the nonlinear model's result. Thus, the time-dependent solution successfully simulates
the asymmetric structure of the vortex in the nonlinear model.
6, Summary and Conclusions
The nondivergent barotropic vorticity equation linearized with respect to the basic symmetric
vortex in a moving coordinate system is used to understand the dynamics of the asymmetric
circulation in a cyclonic vortex moving on a /5-plane. The model is not closed in that the speed
and direction of the movement of the cyclone must be specified. The purpose of this study is to
obtain correct asymmetric circulation associated with the moving vortex using a linear model.
In previous studies using nonlinear numerical models, the asymmetric gyres of a cyclonic
vortex moving on a /3-plane have been oriented mainly in the direction of the track movement. In
the present study, different gyres were obtained in the inner and the outer part of the vortex due
to different signs of the basic vorticity gradient in the radial direction. The inner gyres have very
large magnitude and with an orientation that has no relation to the specified movement of the
vortex. Linear stability analysis of the axisymmetric wind profile for the normal mode solutions
has shown that the basic symmetric flow is barotropically unstable. Agreement of the structure of
the inner gyres and the unstable eigenfunction indicates that the inner gyres are the free unstable
mode in the system. Maximum amplitudes of both are at the radius of maximum wind.
The outer gyres were found to be oriented in the same direction as the specified track movement
and correspond to the /-gyres observed in the numerical model. However, their existence may be
shielded by the large magnitude of the inner gyre in the present linear model. In the steady-state
solution, the outer /-gyres can be isolated by placing the inner boundary some distance from the
center of the vortex so that the strength of the inner gyres would not obscure the outer gyre. The
/-gyres can also be obtained in the present model by reducing the resolution of the model so that
the instability of the inner gyres is reduced. Another way of reducing or removing the unstable
inner gyres is to put strong damping in the inner region to reduce the instability as suggested by
;
Willpughby (1988). '.\.',^ . ^ /.Y ' ' V. Y YY .. ...' .- " Y;-" YY • ' , '
When the speed and direction of the vortex motion from the full nonlinear model are used in

543
the present linear model, the structure and the magnitude of the streamfunction obtained from the
linear model corresponding to the outer gyres are very close to wavenumber one structure in the
nonlinear model. In a numerical tropical cyclone track forecast model, since the speed and direction
of the cyclone can usually be analyzed from observational data at the initial time, the asymmetric
gyre structure can be obtained from the simple linear model and superposed on the symmetric
vortex when bogusing the tropical cyclone. It is hoped that the asymmetric structure inserted this
way to the track forecast model will be capable of including the /-drift more effectively.
Since increasing the maximum wind of the vortex will only enhance the unstable inner gyres,
it has little effect oft the total movement of the vortex. When the tangential wind in the outer
part of the vortex is increased, the strength of outer /-gyres is enhanced and the ventilation flow
through the vortex center is increased. Therefore, the vortex movement is increased. This is the
reason that the track movement is very sensitive to the outer part of the wind field and not to the
maximum wind as observed by DeMaria (1985) and Fiorino and Elsberry (1989).
The barotropic instabilitu is not unique to the basic wind profile chosen in the present study.
Other profiles such as in Smith et aL (1989) also support barotropic instability and have results
consistent with those presented above. In the presence of instability, steady-state solutions would
not be appropriate without any modification. As the inner gyres appeared to be very weak in the
nonlinear model, some nonlinear feedback to the basic mean wind may be considered. There may
also be interaction of the inner gyres and the /-gyres which cannot be included in the current
model. The closure problem of determining the translation speed and direction of the vortex as
well as the others will be pursued in the future studies.
References
Carr, L.E., and R. T. Williams, 1989: Barotropic vortex stability to perrurbations from axisymmetry.
J. Atmos. ScL , 46 , 3177-3191.
Chan, J. C., and R. T. Williams, 1987: Analytical and numerical studies of the beta-effect in
tropical cyclone motion. Part I; Zero mean flow. J. Atmos. Sci. , 44 , 1257-1265.
DeMaria, M., 1985: Tropical cyclone motion in a nondivergent barotropic model. Mon. Wea. Rev.
, 113,1199-1210.
Fiorino, M., and R, L. Elsberry, 1989: Some aspects of vortex structure related to tropical cyclone
motion. J. Atmos. Sci.. 46,975-990.
Smith, R.K.> W. Ulrich and G. Dietachmayer, 1989: A numerical study of tropical cyclone motion
using a barotropic model. Part I. The role of vortex asymmetries. Quart. J. Royal Meteor. Soc.
(in press).
Willoughby, H. E., 1988: Linear motion of a shallow-water, barotropic vortex. J. Atmos. Sci. ,
45 , 1906-1928.

544
(A) b- l.O.(B) b- 0.8.(C) b- 0.6
8-
-1.0,(B)b= 0.8,(C)b= 0.6
T3 *.
in
"£g
Figure la. Tangential wind profile from Eq. Figure Ib. Vorticity gradients corresponding to
(2.8) for the basic vortex with different values Figure la.
ofb;
-30 -20 -10 10 20 30 -0.03-0.02 -0.01 0,00 0.01 0.02 0.03
Figure 2. Vectors representing term 1 (

545
Figure 3, Steady-state asymmetric streamfunc- Figure 4. Steady-state asymmetric streamfunction
using b = 1.0 profile, ci = 0.6 and = tion as in Fig. 3 except the inner boundary is
120°. Resolution of the grid is Ar = 5 km. 50 km from the center.
Figure 5. Complex eigenfunction of the unstable
wavenumber one for b = 1.0 profile. Solid
line is the amplitude and dashed line is the
phase angle.
Figure 6. Complex streamfunction in the radial
direction for the steady-state solution shown in
Fig. 3. Solid line is the amplitude (nondimensional)
and dashed line is the phase angle.

546
b= 1.0,C= 0.03,0 = 120., T = 50. b= i.o,V= 40.0,hr= 5.
b= 1.0,0 0.60,0= 120., T = 1600
(a)
b= l.O,V= 40.0,hr= 45.
(b)
b= 1.0,C= 0.60,0= 120., T*= 2400.
b= 1.0,V= 40.0.hr= 65.
(c)
Figure 7. Time-dependent solution for asymmetric
streamfunction with Ar = 20 km. Scale
of the field is 2.4 x 10 6 m 2 /sec. T = 800 in 7b
corresponds to 2.32 day.
(0)
Figure 8. Time evolution for the wavenumber
one asymmetric streamfunction from nonlinear
numerical model with the parameter settings
the same as for Fig. 7, Scale of the field is
W 7 m/sec.

547
EFFECT OF THERMAL AND DYNAMIC FORCING ON THE
SECONDARY CIRCULATION OF TYPHOONS
Ding Yihui
(Academy of Meteorological Science, State
Meteorological Administration, Beijing)
Liu Yuezhen Sun Ziping
(Institute of Atmospheric Physics,
Academia Sinica, Beijing)
ABSTRACT
A non-dimensional equation for typhoons has been
derived and then 11-yr compositing typhoon data were used
to estimate the forced secondary circulation associated
with typhoons. The main results are as follows:
1. The diabatic heating and cumulus vertical heat,
mixing are major thermal forcing factors. They have the
same order of magnitude. They both may force vigorous
typhoon secondary circulation;
2. The effects of eddy flux and cumulus horizontal
mixing of heat are of minor importance;
3, Cumulus vertical momentum mixing and eddy
horizontal momentum flux are major dynamic forcing factors.
They play an important role in the development and
maintenance of typhoons. They may force a stronger
secondary circulation. Between the two of them, the former
has a more significant effect than the latter. It can help
enhance large scale low-level convergence, thus
intensifying the positive feedback process relevant to CISK
mechanism;
4, Vertical and horizontal turbulent fluxes of
momentum and cumulus horizontal flux of momentum can also
force positive circulation cells, but with less significant
effect than the above-described two terms.
T. INTRODUCTION
Secondary circulation exists in numerous weather
systems, and plays an important role in their maintenance
and development. Eliassen (1951) first discussed the
problem of secondary circulation systematically. Later,
Sawyer (1956) and Eliassen (1962) derived the equations of
the secondary circulation for front and jet stream-frontal
systam, independently. Shapiro (1981) derived a more
complete equation of the secondary circulation taking into
account the effects of eddy vertical transport of momentum
and heat, and diabatic heating. Recently, secondary
circulation for typhoons has been studied. For instance,
Willoughby (1979) estimated the secondary circulation of

548
tropical cyclones. Then, Shapiro and Willoughby (1982) set
rhe foroi r.g term to be a point source and estimated the
secondary circulation forced by heat and momentum sources
by using an ideal fi^ld.
A secondary circulation, relative to the basic
circulation or primary circulation is defined as a
transverse circulation superposed on the basic circulation
that is controlled by the physical process in the system.
The basic circulation or primary circulation is a
circulation which satisfies some balance relationship (for
example, geostrophic balance and gradient wind balance),
Once this balance relationship of the primary circulation
is destroyed, a secondary circulation would be induced and
acts as a restoring mechanism of the equilibrium state of
the basic flow. Therefore, the secondary circulation is a
result of the continuous adjustment of the basic flow with
a general tendency to maintain its balance relationship.
The physical factors affecting the secondary
circulation of typhoons may foe divided into two types:
thermal forcing and dynamic forcing. The present paper will
deal with these two types of forcing based on observed
typhoon data. The thermal forcing includes the large-scale
and cumulus-scale condensation heating, cumulus heat
transport, turbulent heat transpor, radiative heating (or
cooling) and re-evapotation in clouds. Among them, the
release of culumus convective condensation latent heat
plays an important role in the maintenance and development
of typhoons. Culumus heat vertical mixing and radiative
heating have been stressed only recently , The dynamic
factors include the horizontal and vertical turbulent flux
of momentum, culumus horizontal and vertical flux of
momentum and large scale eddy horizontal flux of momentum.
ATflong them, the last two terms have been emphasized in
recant years (Challa and Pfeffer, 1984; Gray, 1979) . In
this paper we estimate the secondary circulations forced by
various thermal and dynamic factors and discuss their
relative importance*
IT. EQUATIONS
The set of primitive equation in the cylindrical
coodinate system with Z=-Hlnp/p 0 as the vertical coordinate
can be expressed as follows:
CIO

549
where b=g(T-Tr)/Tr buoyance force, N=g (^J -V") /Tr the
squared buoyance frequency, is gee-potential height, f is
the Coriolis force, T and 0 are temperature and potential
temperature in cloud, respectively, Tr and 6t* are the
environmental temperature and potential temperature,
respectively, y- the temperature lapse rate, and Yd tne ^
adiabatic lapse rate. Also, ^*fcj, (v*--l/r* }u+k z^^^ ^ A^Uk
~!4*)ir+fec^«X«o , AOM^c^/r*>^^^
and H is the scale height, i.e., H=RT 0 /g. Other symbols are
conventional ones.
Averaging the equation set (1) over an approriate
area,- one may obtain those terms which respresent the
effects of culumus convection. Then, in order to transform
the resulting set of equations into dimensionless form, one
may introduce the following variables;
-t* ,
absolut vorticity
inertial parameter'
vertical shear 1%
Fronde number F r ss
aspect ratio A*.=s -
quantities.
* J «
5^* ,. Rossby number
/• Richardson number Rj
» "*" denotes dimensionless
We further expand each variable in terms of a small
parameter d.=U/v. After some mathematical manipulation, we
can derive a diagnostic equation of the secondary
circulation for the tropical cyclone:
where Vis the stream function,
and "VJ5*« /k*j^* d 54|^" While introducing the above streamfunction,,
the isothermal atmosphere with the temperature T
is assumed. In this case, the height Z defined with z--
Hlnp/p is just eqxaal to the actual height,
The elliptic condition for the equation (2) is D
--CS"^>*>O.The observed data were used to estimate D* and we
find that all the grid points of the domain under study
satisfy the elliptic condition of Er>0.

550
fine equation (2) will be used to estimate the
secondary circulation associated with the tropical, cyclone.
The terms on the right hand side of this equation represent
thermal and dynamic forcing for the secondary circulation,
including the diabatic term, the horizontal and vertical
turbulent flux of heat, the culunms horizontal and vertical
flux of heat, the horizontal and vertical turbulent flux of
momentum, the culumus horizontal and vertical flux of
momentum, and the eddy horizontal and vertical flux of
momentum. It is very important to evaluate their relative
importance for forcing the secondary circulation in the
tropical cyclone.
ITT. DATA AND COMPUTATIONAL ASPECT
The data used in the present study are teken from the
data set of composite typhoons for 1966-1977 compiled by
Prof. Gray's group of the Department of Atmospheric
Sciences. Colorado State University. The characteristic
values of the composite typhoons are given below: V=30m/s,
TialOro/s, W=l,4m/s, R=lllkm, Z=15km and$ =1. 83X10 T t/s .
The ten layers in vertical are taken from lOOOhPa to
lOOhPa with the vertical interval of lOOhPa* The horizontal
grid length AZ* =0*5, i.e., half a latitude degree. Along
the horizontal direction, we have 14 grid points. In the
calculation, the boundary condition of JP =0 were taken.
This assumption of boundary condition may cause some
uncertainty for the computational results.
IV. RESULTS AND COMPARISON OF RELATIVE IMPORTANCE OF VARIOUS
FORCING TERMS
(l)The Thermal Foring
1)Diabatic heation
The diabatic heating term is here taken to be Q=Q k +Q c
+ Q^ , .' where Q k is large-scale condensation heating, Q c
convective heating and Q^the radiative heating (cooling).
The flux of sensible heat from the underlying surface is
neglected.
The modified Kuo-scheme of cumulus parameterization
adopted by Krishnamurti et al (1980) is used to estimate Q
reasonable partition between heating and moistening as well
as meso-and small scale moisture convergence. Q k is
a'stiwated by the expression Q =-Lu%i, No direct estimate of
Q^ in the present paper has been attempted and only its
nljmatological values were taken from Ding and Liu (1985),
On the average, the mean cooling rate in the clear region
•U"-, to -3'*c"/d while that in cloudy region is —1 to l*C/d,
The diabatic heating may force a very strong secondary
rHrrulation (Fig, 1,)•/ showing the center of the positive
cell located at ^00 hPa at a distance of 270 km

from the center of cyclone. Tn addition, at r* = 5 and 800
hPa, there Is a negative circulation. The reason of its
formation is not clear.
xlOOhPa 5.5
551
1 2 3 + 5 6, 7 r*
Pig, 1 The secondary circulation forced by the diabatic
heating term, with interval of stream function being 0.8*10"
!.) Turbulent flux of heat
Tt is generally accepted that the vertical flux of
heat (the 3-rd. term on the right hand side of Eq. (2)) is
much wore important than the horizontal flux (the 2-nd
t.errn)in the development of typhoons,
For the secondary circulation forced by the horizontal
turbulent flux of heat, there is a weak positive
circulation with two centers at about 750 hPa in the lower
and middle troposphere (not shown). A negative circulation
is located in the upper troposphere, with two stronger
centers found in the layer of 300-200 "'hPa .
The secondary circulation forced by the vertical
turbulent flux of heat shows a complete pair of positive
and negative circulation cells, with the center of the
former found, at 550 hPa and the center of the latter at 750
hPa (not shown), Above 300 hPa, one may observe a weak
positive circulation,
forced
greater
of heat
3)
Tt
Tn comparison, the magnitude of secondary circulation
by the vertical turbulent flux of heat is much
than that forced by the horizontal turbulent flux
Cumulus horizontal flux of heat
is generally recognized that the transport of heat
by cumulus convection (the 6-th term on the right hand side
of Eq, (2}) is much greater than that by turbulence. Kuo
(1974) suggested that this term may foe estimated according
to K-th.eory, but it is very important to -take K value
appropriately. As a rough approximation f we take this'
coefficient to be three times as great as the eddy
viscosity coefficient. The secondary circulation thus
obtained has a pattern similar to that forced by horizontal
turbulent flux of heat, only with greater magnitudes of
.(not", shown ) , •"." •' ' ... : . • ' . - . ' . .. ;'./•' • ' • . . • .; •.•;' '• • • . v ' • ' • • ' • • "

552
4) Cumulus vertical flux of heat (the 7-th term on the
right hand side of FIq,(2})
Based on the parameterization scheme proposed by
Schneider and Lindzen (1976) , this term may be further
rewrithen as follows:
He is the dimensionless cumulus mass flux which is a
crucial, parameter to determine the cumulus mixing.
Fig. 2 is the secondary circulation forced by this
r.firm. There is a complete positive circulation in the
ant-ire domain, with the center located at 300 km and 500
hPa .
x 100 hPa
7 r'
Fig.-2 The secondary circulation forced to 1 cumulus vertical
flux of heat (the interval of ^ is 0.4*10*).
5)Comparison among various thermal forcing terms
The intercomparison of various forcing terms has
Indicated that, for the thermal forcing of the. secondary
circulation of typhoons the horizontal turbulent flux of
heat, is insignificant; the vertical turbulent flux, of heat
is relatively large; the cumulus horizontal flux of heat is
relatively small,
only with the same order of magnitude as
the horizontal turbulent flux of heat, while the cumulus
vertical flux of heat is more significant, with the order
of magnitude of 10~*and the rather complete circulation
ceii.. • • . , '. . . .- . . ' , • . • • . . . ' ;. , . •. . . . . •••. •
The di.abatic heating is the primary forcing factor
among them. with the order of magnitude of the secondary
circulation being 10"*and a greater magnitude of than that
for cumulus vertical flux of heat. This vigorous positive
circulation dominates nearly the entire domain under study.
Fig. 3 is the secondary circulation forced by all the
thermal forcing terms (including the above five
factors).showing features similar to Fig. 1. A positive
rircul at.ion dominates the entire domain with a weak
negative circulation at low level in the outer region of

553
typhoon, indi rating the primary importance of diabatic
heating. The difference between Fig. 3 and Fig. 1 shows up
in the*. enhancement of the positive circulation, the descent
of the position of the circulation center and a weakening
of the negative circulation of the former, implying the
considerable role of cumulus vertical mixing of heat.
Therefore. the thermally forced secondary circulation is
mainly controlled by these two terms.
x 100 hPa
Fig. 3 The secondary circulation forced by all the thermal
factors (five terms).with interval of ^p being 0,8X10"
(2) The Dynamic Forcing
1.) Horizontal turbulent flux of momentum (the 8-th
term on the right hand side of Eq.(2))
The secondary circulation forced by this term
indicates a negative circulation in the lower troposphere
with mult.iple centers and a positive circulation in the
middle and upper troposphere {not shown).
2) Vertical turbulent flux of momentum.
This term is expressed by the 9-th term on the right
hand side of Rq, {.) , Fig. 4 shows the secondary circulation
forced by this term. One may observe a positive circulation
i r« the whole troposphere, with the gradient concentrating
in r.he lower and middle troposphere and center of
circulation found, at 900 hPa r about 500km. from the centr of
tropical cylone.
• x 100 hPa
7 r

554
Fig. 4 Same as Fig, 1, but for the vertical turbulent flux
of moment!]. Interval of ^ i s 0.8X10*",
3) CuTm.il us horizontal flux of momentum
K"-theory may he used to estimate this term
approximately (the 10-th term, on the right hand side of
Eq.(2) ). The secondary circulation forced by this term is
mainly a positive circulation located in the middle and
upper troposphere {not shown).
4}Fddy horizontal flux of momentum
When there is a significant inward momentum flux in
the tropical cyclone, it is mainly caused by the asymmetry
of large-scale eddy. McBri.de {1981} calculated this kind of
momentum flux (the last second term on, the right hand side
of Flq. (P.)}. Normally; one may use actual data to estimate
the perturbation quantities along azxmuthal direction and
then to obtain the secondary circulation forced by this
asymmetric distribution of physical quantities. But the
data set used here is two-dimensional by which it is
impossible to derive the average perturbation along \ .
instead, an empirical function is se3.ect.ed to rpresent the
distribution of —r * [U&.* ViL*] that is very close to the
results calculated based on real data by McBride, -r * [Uo,*
Vk*] is assumed to be a, function of r * and Z * and is set
to be zero at boundaries. At 200 hPa, it has a maximum of
2.68*l£T*st r * =4(about 20 latitude degree m* s~* in the
dimensional case).
This distribution is equivalent to a quadratic
surface. Below 200 hPa, it is part of an ellipsoid whereas
above 200 hPa it is a sloping surface that intersects with
the ellipsoid, at 200 hPa,
HxlOOhPa *-0-2 .._
—.0
1 2 3 4 5- 6 , 7 J
Fig, B Same as Fig. V, but for the eddy horizontal flux of
momentum; with the interval of 3P being Q.2x10
Fig, 5 is the secondary circulation forced by the eddy
horizontal flux of momentum defined with the abovedescribed
empirical function. It can be seen that there is
basically a positive circulation in the entire domain with
^the center at 700 hPa and r *=3,5, and the maximum
*•'. At the top there is a negative circulation, with

555
the center at 200 hPa and the maximum of -0.33X10*** . The
order of magnitude of ULj * is 1 O" 1 : W * 30""*.
Tt is generally accepted that the vertical transport
of momentum i.n typhoons is accomplished by cumulus
convection while the large-scale eddy vertical transport of
momentum is less important.. Therefore.. the latter has not.
estimated in the present, paper (see the last term in F,a
(2 ) ) -
5) Cumulus vertical flux of momentum
This^ term in Eq. (2) is written as its form of
parameters ^ation:
where "UL* the dimension! ess vertical wind shear in cloud, llj*
the dimension! ess vertical wi.nd shear in the surrounding.
The secondary circulation forced by cumulus vertical
flux of momentum. (Fig. 6} is characterized by a positive
and negative secondary circulation. The center of the
negative circulation is located at p=800 hpa and r *=1,
with the maximum being -2.23XlO"*and the center of the
positive circulation at p=500 hPa and r *=2 with the
maximum being 2.53X10"*", The problem as to how the negative
circulation can occur remains to be further studied.
x 100 hPa
2
Fig.6 vSame as Fig, 1., but for the cumulus vertical flux of
momentum, with the interval of y 0.8X10"
6) Comparison among various dynamic forcing terms
From the above results, we can draw an important,
conclusion: cumulus vertical flux of and eddy horizontal
flux of momentum are basic dynamic forcing factors that
determine the secondary circulation. Tn contrast, the
turbulent flux.-of momentum (horizontal and. vertical) and
cumulus horizontal flux of '.momentum are less important.
Fig, 7 shows the secondary circulation forced by all the
dynamic factors. A complete positive circulation is located
in the middle and upper troposphere with a center of 2.97X
10""** at. p-500 hPa and r *-2.5. In the lower troposphere, a
negative circulation exists in the region from the eyewall
of the typhoon to 280 km/ with the center of -1.99xlO" a ' at

556
n=800 bPa and r **1. Now it is not clear yet why this
e circulation may occur,
1 1- x 100 hPa
J 2> 3 4 5 6
'7 r
Fig. 7 The secondary circulation forced oy the sum of
dynamic factors. The interval of isolines is 0.8*10.
(3) Comparison Among Various Dynamic And Thermal
Forcing Factors .
I
8 'The secondary circulation forced by combined thermal
and dynamic factrs, , The order of magnitude of is 10 with
the interval 0.3X10"
Fig, 8 shows the secondary circulation generated by
the combined dynamic and thermal factors. It can be seen
that there is a complete positive circulation in the entire
aomain under study. The center of the circulation is
located at p«500 hPa and r •**2.5-> with the maximum at the
center being 1,43*10"
. Ta'hle 1 demonstrates the relative importance of the
various dynamic and thermal factors, in descending order.
The maximum of "JM in Table 1 denotes the central value
of the forced positive circulation* It can be seen that
cHahatio hRating is the most important forcing followed by
cumulus vertical flux of heat. Cumulus vertical flux of
•momentum is the major dynamic forcing term. But,- it is less
important., relative to the thermal forcing. Tn addition, the
7 r

557
vertical turbulent flux of momentum should also be taken
into account, for the forced secondary circulation. Table 1
also indicates that the effect of cumulus convection is
very important., due to the fact that the first three terms
related to it.
Table 1 A 1 i st of realti.ve importance of
dynamic and thermal forcing in descending order.
Forcing terms
Order of magnitue
of forced decondary
circulation
di.abatic heating
lO*^
CM vertical flux of heat 10"**
Cu vertical flux of momentum
ID"*
eddy horizontal flux of momentum 10"*
vertical trubulent flux of momentum. 10"*
eddy horizon tal. flux of heat 10"^
Cu horizontal flux of momentum 10"*
horizontal, trubulent flux of momentum 10"*
Cu horizontal flux of heat 10* 7
horizontal turbulent flux of heat 10"""*
total thermal forcing 10"*
total dvnamic forcing 1.0**
sum of thermal and dynamic forcing 10"'
various
i
j Maxi mum
\ of
j
9.15x10'*
3.0 X1Q-*
2.53*10*
0.95x10*
7.72*10"*
4.6 XlO"*
2.58X1CT*
0.9 XLO"*
4.5 *10" 7
1.5 xi cr 7
11.6x10"*
2.97X10*
1.43X10*'
V.CONCLUSIONS
The present paper used non-dimensional equations to
study secondary circulation in typhoons and 11-yr composite
typhoon data to estimate the relative role of thermal and
dynamical forcing in these secondary circulations. The main
results are as follows:
1 . The effect, of turbulent horizotal flux and cumulus
horizontal mixing of heat are of minor importance and may
be neglected.
9.. The trubulent vertical flux of heat may force a
positive circulation in the lower and middle troposphere
with the order of magnitude ofj'being 10.
3. The dlabatlo heating and cumulus vertical mixing of
heat are major thermal forcing factors, Thet both may force
the vigorous secondary ci rculation with y being of 10"t
4. Cumulus vertical momentum mixing and eddy
horizontal momentum flux play an important role in the
development and maintenance of typhoons. They may force a
stronger secondary circulation. The former has more
significant effects than the latter, Tt can help enhance
large-scale low-level convergence, thus intensifying
positive feedback processes relevant to CTSK me.chani.sm.
6. Vertical and horizontal turbulent fluxes of
momentum and "/cumulus horizontal flux of momentum can force
nosi tive ci rculation eel Is, but with less significant
effect than the two terms discussed above.

558
RRFKRRNCF
Challa, M., and R. T,, Pfeffer ; 1984, The effect of
cumulus momentum mixing of the development of a symmetric
model hurricane. J. Atmos. Sci-, 41, 1312-1319.
ning Yihui and Tii.u Yuezhen, 1985, A study of budget of
kinetic energy in typhoon, scienitia sinica. Series B, Mo,
11. 1045-1054"
FH -i assfin. A,, 1951 slow thermal ly or friotionally
controlled meridional circulations in a. circular vortex,
Astrophys.. 5 19-60.
Kliassen. A., 1962 on the vertical circulation in
frontal zones, Geofys. Publ. Geophys. Norv., 24 147-160.
Gray. W, M., 1979, Hruuican.es: their f ormation,
structure and likely role in the tropical circulation.
Meteorology over the Tropocal oceans. D.b» Shaw. Kd. , ^oy.
Meteor,, BB-21R.
TCri shnawurti . T.N. . et al 1984, Cuiw.ilus
parameterization and rainfall rates, T: Mon. Wea. Rev, ,
108 r 465-477,
TCuo. H.T,,., 1974, Further studies of the
parameterization of influence of cumulus convection on
large scale flow. vT.Atmos, Sci * , 31, 1232-1.240,
T.indzen, R,s., 1980 Keport of workshop held at NASA
Goddard Institute for space studies, 42-51, Now Youk,
October 29-31.
McRride, J., 1.981.. Observational analysis of tropical
cyclone formation. ]t :Budget analysis, J, Atrnos. Sci., 38,
1152-1166.
Sawyer. J. 5., 1956. The vertical circtJlation at
meteorological fronts and its relation to fron.togenesis.
Proc. Roy, $00,. Tiondon. A234, 346-362.
Schneider, K. K., and R. s. liindzen, 1976 A discussion
of the parameterization of momentum exchange by cumulus
convection. J.Geophys. Res- ., 81, 3158-3160.
Shapiro, M, A,, 1981. Frontogenesis and
geostrophically forced. secondary circulation in the
vicinity of jet-stream-frontal
Sci., 38 ; 954-973.
zone systems, J. Atmos,
Shapiro. T,. v7., and H.K, Willoughby; 1982, The
responses of balanced hurricanes to local sources of heat
and mome-ntum. J, Atmos, sci, 39, 37-394*
Wi 11 ou.gh by, H. --K. , 1979 , Forced secondary circulation
in hurricanes, J, Geophys. Res. 84, 3173-3183.

559
RECENT RESULTS IN LIMITED-AREA NUMERICAL WEATHER PREDICTION
Simon Wei-Jen Chang
Rangarao Venkata Madala
Keith Denis Sashegyi
Atmospheric Physics Branch, Space Science Division
U. S. Naval Research Laboratory, Washington, D. C. 20375
ABSTRACT
Several improvements to the Naval Research Laboratory (NRL)
limited area weather prediction system have recently been incorporated.
A multi-layer planetary boundary layer parameterization based on the
tubulent kinetic energy and dissipation rate has been tested and
implemented to the system. This results in an improvement in the
forecast of finer mesoscale sturctures. The tropical cyclone version of
the model is used to study the outflow layer structure of tropical
cyclones. It is found there exists secondary circulations around the
outflow jet which may play an important role in the interaction between
tropical cyclone and upper tropospheric large scale systems, A new
vertical mode initialization is incorporated to the prediction system,
which proves to be an effective way to suppress oscillations in the
early hours of model integration due to initial imbalance. And finally,
a nested grid version of the NRL weather prediction is being developed.
I. INTRODUCTION.
Simulation studies with the newly improved U. S. Naval Research
Laboratory (NRL) limited-area numerical weather prediction system have
been carried out recently at NRL. The NRL prediction system is
maintained on the CRAY X-MP/24 at NRL. Advancements and improvements
are continuously periodically incorporated into the system. Basic and
applied research and studies are conducted with the model at NRL and
other research institutes.

560
The limited-area forecast system at NRL contains four major
components: the analysis, the initialization, the dynamic model, and
the output package. The Barnes successive correction scheme is used in
the regional and mesoscale analyses. For the the initialization, the
system has options to start predictions from uninitialized fields,
nondivergent wind field, nondivergent static balanced fields, or
vertical mode initialized fields. As reported in Chang et al. (1989),
the governing primitive equations of the model are in surface-pressureweighted
flux form written for curvilinear horizontal and

561
PEL scheme consists of a similarity surface layer based on the Monin-
Obkhov stability, and a mixed layer (Gerber et al., 1989), in which the
mixing is governed by prognastic equations of the turbulent kinetic
energy (TKE) and the dissipation rate (e) . To test this new
parameterization for the regional model, the effects of different PEL
parameterizations are compared in the NRL regional model with a coastal
frontogenesis/cyclogenesis case from the Genesis of Atlantic Lows
Experiment (GALE). The following three PEL configurations are
considered: (a) a 10-layer model with a one-layer bulk parameterization
of the PBL, (b) a 16-layer model with a drag coefficient surface layer
and a mixing-length mixed layer, and (c) a 16-layer model with a
surface energy balanced soil slab and the new PBL parameterization.
Major results suggest that for regional and turbulent scale
structures of complex atmospheric processes such as the sea-land
thermal distribution, a warm ocean current and strong vorticity
advection, the model version with the improved TKE parameterization
yields more accurate and realistic simulations. The results also
indicate that the increased vertical resolution alone does not lead to
better forecasts. All three versions of the model produce good
forecast performance scores, reflecting the sound design of the basic
dynamics, numerical techniques and the lateral boundary treatment in
the regional model (Fig 1). A detail report on these simulations can be
found in Holt et al. (1990).
III. Tropical Cyclone Outflow Jet.
Recent observational and theoretical studies have suggested that
for mature tropical cyclones, the asymmetric anticyclonic outflow layer
is more dynamically unstable than the symmetric cyclonic circulation in
the mid- and lower-troposphere. It has been found in many cases that
the interactions between the tropical cyclone outflow layer and uppertropospheric
systems lead to behavior changes in tropical cyclones.

562
Our study is to obtain and analyze the tropical cyclone's outflow
structure as simulated by the NRL model, and to examine whether the
hypothesized interactions between the upper-tropospheric systems and
the outflow are likely. For this purpose, a realistic steady-state
tropical cyclone structure is obtained by initializing the NRL regional
model with the mean tropical sounding and a Rankine vortex. In the
simulated tropical cyclone, it is found that the outflow layer is
dominated, in angular momentum transfer, by one or two jets, depending
on the zonal wind shear. It is also found that there are secondary
circulations around the outflow jet (Fig. 2). The secondary
circulations in the jet entrance region is a thermally direct
circulation, i.e., the rising branch is located on the anticyclonic
shear side of the jet toward the storm center, while the subsiding
branch is located on the cyclonic shear side of the jet away from the
center. In the jet exit region, the direction of the secondary
circulation is reversed and thermally indirect.
We found that these circum-jet secondary circulations play
important roles in the interaction of the outflow jet with uppertropospheric
systems. In a numerical experiment we found that when the
jet is accelerating, simulating the confluence of the outflow jet and
an approaching westerly trough, the secondary circulation is enhanced
and initiates deep convections in the area of upward branches near the
jet (Fig. 3). These findings will be reported in Shi et al. (1989).
IV. VERTICAL NORMAL MODE INITIALIZATION.
In order to filter out the spurious oscillations in the first few
hours in the forecast of the regional model, a new initialization
scheme is devised based on the vertical normal modes of the model. The
spurious oscillations are mainly caused by the existence of fast-moving
gravity waves, which arise due to the incapability of observing
atmospheric state on all scales and errors in the regional analysis.
The vertical normal mode method is to project the analysis into the
model's vertical eigen space* A steady state solution of the

563
divergence, vorticity and generalized potential can be obtained for
each mode by solving a set of Helmholtz equations. A fast, exact
Helmholtz equation solver, the stabilized error vector propagation
solver, is used with either Neumann or Dirichlet boundary conditions.
Currently, only the fastest three modes in the model are so treated. We
found that the model forecast is void of spurious oscillations when the
initial regional analysis is initialized with this method (Fig. 4), as
reported in Sashegyi and Madala (1989).
V. NESTED GRIDS.
To better resolve certain meteorological phenomena over a limited
area within the regional model without severely straining the available
computer resources, a nested grid version of the NRL model is
developed. In this version, the grid interval ratio is 3:1 and only two
grids are considered at present. The communication during the model
integration between the two grids is one-way, from the coarse grid to
the fine grid. At the interface, tendencies of all prognostic
variables on the coarse grid are interpolated onto the seven grid
points of the fine grid within the interfacial zone. These
interpolated tendencies will then be blended with the computed
tendencies of the seven grid points of the fine grid in the interfacial
zone. The two grids have their respective land/sea/ice tables, terrain
heights, ocean surface temperatures, initial analyses and normal mode
initialization, except where there is blending in the interfacial zone.
The nested model is tested with a real data case of IOP-2 in GALE,
during which a coastal front was formed and developed into an intense
cyclone with a deepening rate in excess of one Bergeron. The coarse
grid covers a domain of 115°-65°W and 25°-65°N, with a spatial
resolution of 2° in longitude and 1.5° in latitude. The fine grid
covers the area of approximately 85°-62

564
coarse grid (Fig. 5). The structures of the coastal front and the low
pressure system are much better defined.
A two-way interacting nested grid system will be incorporated in
the near future. The effect of the two-way interaction will be
evaluated.
ACKNOWLEDGMENTS
We /thank Dr. Teddy Holt, Naval Postgraduate School, and Mr. Jiann-
Jong Shi, North Carolina State University, for their contribution to
the efforts reported here.
REFERENCE
Chang, S, K. Brehme, R. Madala and K. Sashegyi, 1989: A numerical study
of the east coast snowstorm of 10-12 February 1983. Monthly Weather
Review, 117, 1768-1778.
Gerber, H, S. Chang, and T.
Holt, 1989: Evolution of a marine boundary
layer jet. Journal of the Atmospheric Sciences, 46, 1312-1426.
Holt, T, S. Chang and S. Raman, 1990: A numerical study of the coastal
cyclogenesis in GALE IOP 2: Sensitivity to PEL parameterization. To
appear in Monthly Weather Review, February issue.
Sashegyi, K, and E, Madala, 1989: Test of initialization procedures
with the NRL limited-area numerical weather prediction model. NRL
Technical Report.
Shi, J.-JV, S. Chang, and S. Raman, 1989: A numerical study of the
tropical cyclone outflow jet. Submitted to Monthly Weather Review.

565
Fig. 1. Observations and 24 hour forecasts for the GALE coastal
frontogenesis/cyclogenesis case. Sea level pressure (solid lines, mb),
surface winds (vectors, m/s) and temperature (dashed lines, C) valid at
12Z 26 January 1986 for (a) NMC/RAFS analysis, (b) the 10-layer model,
(c) the 16-layer model with mixing-length PEL, and (d) the 16~layer
model with TKE PBL parameterization.
30 j
60
MB
p
I
30
120
180
210
240
270
.300
Fig. 2. The normal wind
components (in isotachs with 10
ms" 1 interval) and tangential
wind components (in vectors)
relative to a vertical crosssection,
which cuts through the
outflow jet in the entrance
region and along a south-north
plane located to the north of the
tropical cyclone center between
30-300 mb.

566
ACCUMULATED PRECIPITATION
Fig. 3. Accumulated precipitation
patterns (cm), storm centers (*),
and the outflow jet (shaded) in
simulated tropical cyclones:
(upper) 24 h precipitation between
48-72 h occurs mainly near the
storm center in the control case,
(middle) 12 h precipitation between
48-60 h shows jet-related
convection on the anticyclonic
shear side of the jet, (lower) 12 h
precipitation between 60-72 h shows
the expansion of the jet-related
convection and reduced
precipitation in the inner region.
150 160 170
LONGrruos
SURFRC£ PRESSURE at SSN. sow PQXSIC00T flT LEVEL S RNQ 3SM. SOW
Fig. 4. Surface pressure variations (left panel) and vertical velocity
p s and for
initialized initial conditions (curve C), Curve E shows the surface
pressure variation with time for initialized initial conditions but
with an alternative treatment at the lateral boundaries.

567
62. S
56.5
-70.0 -60.0
58.5 or
a f t f f t
55.4 ~
52.2 R
49.1
36.6
33.5 -84.7-81. 6-78.6-75.5-72.5-69* 4-6o. 4-63.3
Fig 5 The predicted surface pressure and wind fields valid at
8601251500UTC on the coarse (upper panel) and fine (lower panel) grids
of the nested model. In this experiment, more details of the forecast
on the fine grid are resolved in spite of the same terrains and initial
analyses used on both grids.

569
AUTHOR INDEX
ANTHES, Richard A.
CHAN, Johnny C.L
CHAN, Y.K.
CHANG, Long-Nan
CHANG, Simon Wei-Jen
CHAD, Jiping
CHEANG, Boon Khean
CHEN, Qiushi
CHEN, George Tai-Jen
CHOU, Jifan
CHOU, LC.
CHUANG, Wen-Ssn
DING.Yihui
GUO, Xiaorong
HONG, Slu-Shung
HSU, Sheng-l
HUANG, Joseph C.K.
Recent applications of the Penn State/NCAR mesoscale 274
model to synoptic, mesoscale scale and climate studies
The impact of the termination of aircraft reconnaissance 462
on tropical cyclone warnings and forecasts in the western
North Pacific
The implementation and operation of an analysis scheme 429
and a limited area model for Hong Kong
On temporal variations of low level jets associated with 38
Asian summer monsoon
Recent results in limited-area numerical weather prediction 559
Monthly and seasonal forecasts and tropical 354
ocean-atmosphere interactions
Wind and moisture fields during the periods of enhanced 69
and suppressed convective activity over the Malaysia-South
China Sea region during the northern winter monsoon,
November-December 1986
A cloud wave theory and its application to the 30-50 day 226
oscillation in the equatorial atmosphere
Overview of Mei-Yu research in Taiwan 14
Analogous rhythm phenomenon of climatic anomalies on 517
seasonal scale
A numerical simulation of the Mei-Yu front 68
The relationship between currents and winds northeast of 399
Taiwan
Effect of the thermal and dynamic forcing on the secondary 547
circulation of typhoons
BMC limited area model: operational application and 439
research
On dynamical studies of orographically induced mesoscale 313
phenomena
Revival of the tipping-bucket raingauge 199
Large scale air-sea interaction in the western Pacific region 385

LIN,Y.J.
Structural features of a squall line over the Taiwan 150
570
HUANG, Shisong
Jl, Liren
JIANG, Huo-Ming
KAU, Wen-Shung
KOT, S.C.
KRISHNAMURTI, T.N.
KYLE; William J.
LAI, ST.
LAM, Hilda
LAU, Alexis Kai-Hon
LAU, Robert
LAU.K-M.
LEE, Cheng-Shang
LIM, Hock
LIN, Hai
LIN, Hai
LIN,Pay-Liam
Influence of variations of the circulation system over the 105
South Indian Ocean on the East Asia summer monsoon
and the northern hemispheric general circulation of the
atmosphere
A spectral model for medium-range weather forecasts 474
and its performance
Study on the frontal cyclone system in southern China 333
and the vicinity of Taiwan area during late-winter and
early-spring
A simulation of lee-cyclogenesis over Yun-Gui Plateau 80
with the use of a hemispheric spectral model
On laboratory simulation of sea breeze 420
Predictability of low frequency modes 220
The impact of urbanization on climate in Hong Kong 261
and its implications for human energy exchanges
Impact of hourly S-VISSR satellite imagery on 210
operational forecasting in Hong Kong
Comparison of radar estimates and surface rainfall 180
during rainstorms in 1987-88
Observed structure and propagation characteristics 48
of summertime synoptic scale disturbances over the
tropical western Pacific
The Royal Observatory long range rainfall forecast 485
methods
Seasonal and intraseasonal variations of the East Asian 94
summer monsoon
Typhoon formation and development - an observational 536
point of view
Effects of vertical wind shear on Kelvin wave 515
.-CISKMod.es
Introduction to Heine Basin Field Experiment (HEIFE) 251
- atmosphere-land surface interactions program
Remote sensing of atmospheric compositions and 170
optical characteristics
Radar observation of precipitation system in Taiwan 160

571
LIOU, Chi-Sann
LIU, Cho-Teng
LIU, Chung-Ming
LIU, S.K.
LIU, Gin-Rong
LIU, Shida
LO, K. Kenneth
LUO, Huibang
MAK, Mankin
PAN, Yi-Hang
PENG, Melinda S.
PU, Shuzhen
SHIEH, Shinn-Liang
SOONG, Su-Tzai
SU, Jilan
SUN, Wen-Yih
TAG, Shiyan
TERNG, Chuen-Teyr
TSAY, Ching-Yen
TSENG, Hsien-Yuan
Multiple equilibria of a thermally forced baroclinic 502
atmosphere
The teleconnections of equatorial SST in Taiwan area 365
The microphysics of a Mei-Yu case : data analysis 304
A three-dimensional model of the China Seas and 365
aspects of typhoon surge predictions
Applying tropopause data observed by VHP radar to 139
improve satellite temperature sounding
The non-linear interaction of internal wave and turbulence 494
The microphysics of a Mei-Yu case : theory 343
The mean heat sources over Asian monsoon region 58
during the period from April to October of 1980-1983
An inquiry into the nature of regional cyclogenesis 525
The variations of the SST in the eastern and western 412
tropical Pacific and their relationship with those in the
world ocean
Dynamics of vortex motion on tropical ^-plane 537
Interannual variabilities of the western tropical Pacific 375
Ocean and low frequency response of the subtropical
high over the northwest Pacific Ocean
An overview of present typhoon forecast operation in 461
Taiwan
Numerical simulations of topographical effects on airflow 323
and precipitation
China-Japan joint research program on the Kuroshio 400
Baroclinic instability of Modified Eady waves 284
The changes in circulations during the transition period 2
from winter monsoon to summer monsoon in the
northern hemisphere
The operational global forecast system at Central 449
Weather Bureau
The coupling of upper-level and low-level jet streaks 3
during Taiwan heavy rainfall period in Mei-Yu season
Doppler weather radar observation and aviation 190

572
WANG, Bin
WANG, Jough-Tai
WU, Jin
WU, Ming-Chin
WU, Rongsheng
XU, Jianmin
YE, Duzheng
YEUNG, K.K.
ZHANG, Zuojun
ZHOU, Xiaoping
ZHOU, Xiuji
ZHU, Qiangen
Some dynamic aspects of the equatorial intraseasonal 119
oscillations
A study on the radiative balance simulated by a general 241
circulation model
Influence of microscale air-sea interaction in climate 403
research
Long range forecasting of Taiwan Mei-Yu 484
Influences of orography on flow in boundary layer 294
Chine.se polar orbiting meteorological satellite FY-1 131
The thermal structure and convective activities over 1
Tibetan Plateau in summer and their interactions with
large-scale circulation
Numerical simulation of mesoscale meteorological 451
phenomena in Hong Kong
Normal modes of climatological mean flow and their 240
roles in the atmospheric general circulation
The East Asia heavy rainfall numerical forecasting and 450
the numerical nowcasting of severe convection weather
The mesoscale monitoring system of the rainstorms 140
and severe convective weather events in China
A numerical study on the effect of the Plateau upon cold 4
surges in East Asia