Low (web) Quality - BALTEX

78
Impacts of the spectral nudging technique on simulation of the East Asian
summer monsoon in 1991
Jianping Tang and Shi Song
School of Atmospheric Sciences, Nanjing University, Nanjing, China, 210093 jptang@nju.edu.cn
1. Introduction
Traditionally, the regional climate models (RCMs) are
nested in the global models to provide detailed regional
information. The method of Davies (1976) has been widely
applied to the formulation of lateral boundary conditions
(LBCs). However, the relaxation does not handle larger
scales correctly, and distorts the long waves reflecting and
interfering within the domain. One of the most substantial
issues in regional climate modeling is that the RCM
integrations are limited by the errors induced by the LBCs
(Risbey and Stone 1996; Christensen et al. 1998; Menendez
et al. 2001; Misra et al. 2003), especially the synoptic-scale
systematic errors during the long-term RCM integrations.
To improve the downscaling performance of the RCMs, the
method of adding nudging terms in the spectral domain has
been developed to incorporate large-scale regulations inside
the RCM domain (Waldron el al 1996; von Storch et al
2000). The basic idea of spectral nudging is that the regional
model should not modify the large-scale field in the regional
domain, which is from the coarse-resolution base field and
considered accurate in regional downscaling. The spectral
nudging method has been widely assessed in the
downscaling experiments in Western Europe (Feser 2006),
contiguous U.S. (Miguez-Macho et al. 2004) and North
America (Kanamaru and Kanamitsu 2007) and has been
proved a demonstrative ability in improving RCM’s
performance.
The purpose of this paper is to evaluate the ability of the
spectral nudging method in the regional climate simulation
over China.
2. Model Description and numerical experiments
The regional climate model used in this study is NCAR-
MM5v3, a three-dimensional, limited-area, primitiveequation
model. The simulation domain centered at 35°N
and 105°E with 107×93 horizontal grid points and 50-km
grid spacing covering most of East Asia. Initial and
boundary conditions are provided by the
2.5°×2.5°NCEP/NCAR reanalysis (NNRP) data, and sea
surface temperature (SST) are from the NOAA weekly OI
SST data.
The experiments are initialized on 16 May 1991 and
integrated until 1 Sep 1991. The control (CTL) integration is
the regular MM5 runs without any spectral nudging
applications, the spectral nudging method is applied to the
wind fields at and above 850 hPa in the spectral nudging
(SN) run.
3. Results
3.1 Large-scale circulation
Figure 1 shows the differences of JJA-averaged geopotential
height (GPH) field at 500hPa between model simulations
and the NNRP data in 1991. Clearly there is a dominant
negative bias in the CTL run in most part of the domain,
with a maximum locates at the southwest, and a less severe
positive bias in northeast of the domain. The SN run is able
to reduce the dominant negative bias to much smaller values
although slightly increase the positive bias in the north of
the domain. In addition, differences of the JJA-averaged
500hPa temperature field between the model simulations
and the NNRP reanalysis are analyzed. Similar to those of
the mass field, the SN run largely reduces the northern
positive bias and the southern negative bias (Figure not
shown).
Figure 1. Difference of JJA-averaged geopotential
height at 500hPa between model simulations and NNRP.
3.2 Regional simulations of daily precipitation events
Figure 2 presents the model simulated and the
observational time series of regional-averaged daily
precipitation rates in Northeast China, North China, the
Yangtze River basin, and South China. Obviously, the SN
run successfully reproduce the observed individual
precipitation events with day to day variations, while the
CTL runs show relatively larger discrepancies. Note that
the three heavy rain periods of 1991 over the Yangtze
River basin are captured markedly better by SN runs than
CTL runs. And the SN run diminishes most false
precipitations generated by the CTL run, i.e. in the North
China in late August. The correlation coefficient between
the observation and the SN run is 0.69, which is
remarkably better than that of the CTL run (0.42).
Figure 2. The model simulated and the observational
time series of regional-averaged daily precipitation rates.