World Meteorological Organization Symposium on Nowcasting - WMO

36
P1.15
2.3 A Bayesian Hierarchical Particle Filter model for storm cell motion forecasting
Operational multi-sensor nowcasting of severe convective storms in the Swiss Alpine
area
A. M. Hering, U. Germann, P. Ambrosetti, I. Giunta, L. Clementi and L. Nisi
MeteoSwiss, ML, Locarno-Monti, Switzerland
For the automatic detection, tracking and analysis of intense convective cells MeteoSwiss
uses operationally the real-time nowcasting system TRT (Thunderstorms Radar Tracking), as
part of its severe thunderstorms nowcasting and short-term warning facility. The presentation
will focus on the current version of the TRT system including the latest operational
improvements in the hail products, and show first preliminary results from two new projects
that can be included into TRT.TRT is a multi-sensor system that uses volumetric reflectivity
data of multi-radar composites (mosaic) to analyse the 3D storm structure and compute
severe weather attributes based on heuristic and centroid-methods, as well as cloud-toground
lightning flashes. Gridded fields and cell-based attributes of individual storms like e.g.
VIL, Echo Tops, altitude of maximum reflectivity, hail probability, are computed every 5min
and are available in real-time to the forecasters as 2D animations and time series. TRT was
used in real-time during MAP D-PHASE (a FDP of the WWRP of WMO) and was available to
end-users on the project visualisation platform. A “cell severity ranking” product, developed
and tested for MAP D-PHASE, was introduced operationally to find and highlight the most
dangerous and strongest storms based on their severity. It includes also the computation of a
60 minutes position forecast, based on the motion of individual cells. The uncertainty of the
expected position is taken into account increasing the area of the forecasted cell
proportionally to the spread (standard deviation) of the velocity vectors from the last three
5min time steps. The cell ranking is successfully used by forecasters and allows them to
focus on the most severe cells maintaining situational awareness and to speed-up the
decision process of thunderstorms warnings.The latest operational improvement in TRT is the
computation of the estimated probability of hail of any size, as well as the maximum expected
hail size. These products are based on the difference between the altitude of the Echo Tops
at 45 and 51 dBZ, and the freezing level, determined from the high-resolution, nonhydrostatic
NWP model for the Alpine region COSMO-2.
For a future improvement of the TRT-system we plan to include the results of two initiating
international projects. The first will develop an improved extrapolation technique that
combines Lagrangian persistence of radar precipitation fields with orographic forcing in
complex orography. This information can be used to extend the lead-time of the nowcasting
warnings. The second aims to merge severe convection predictors, retrieved from satellites,
with the properties of evolving thunderstorms. The satellite based anticipation of convection
initiation will be available in form of continuously updated probability maps. This will allow it to
improve nowcasting, combining satellite information during the whole thunderstorm cycle,
from early detection until dissipation, with the TRT system and to better characterise the
different phases of a convective storm.
Neil I. Fox [1] Yong Song [2] Christopher K. Wikle [2]
[1] Department of Soil, Environmental and Atmospheric Sciences University of Missouri
Columbia, MO 65211 [2] Department of Statistics University of Missouri Columbia, MO
65211
A new storm motion forecasting method is proposed that uses a particle filter concept within a
Bayesian Hierarchical framework to better predict nonlinear cell motion. The motion of
individual cells is predicted using a spatio-temporal model within which the cells are
characterized by as having mass, parameterized by maximum radar reflectivity, and are
allowed to interact with one another. The model is based on Euler-Lagrangian dynamics and
uses a Markov Chain Monte Carlo (MCMC)) based particle filter to assess dynamic state and
parameter estimation. This estimation is completed by finding the Bayesian posterior
distribution of state locations and model parameters. There are opportunities within the
parameterization to utilize additional information such as wind velocity. Tests were conducted
on selected cases featuring both linear and nonlinear motion of convective cells where
interaction between those cells may or may not be evident. Results show that the scheme is
capable of predicting storm cells moving both linearly and nonlinearly while allowing one cell
to affect the motion of its neighbors.
P1.16
Quality Control of Aircraft Moisture Soundings
Seth Gutman [1] and David Helms [2]
[1] NOAA Earth System Research Laboratory, Boulder, Colorado USA [2] NOAA National
Weather Service, Silver Spring, Maryland USA
The growing availability of moisture observations from Tropospheric Airborne <strong>Meteorological</strong>
Data Report (TAMDAR) and Water Vapor Sensing System (WVSS-II) sensors installed on
commercial aircraft and the ability to rapidly assimilate them into numerical weather prediction
models such as the Local Analysis and Prediction System (LAPS) and the Rapid Refresh (RR
next generation RUC) provides forecasters with a greatly improved severe weather
nowcasting capability. All observing systems, including these, are subject to systematic and
random errors. Characterization of systematic errors is a necessary precondition for effective
data assimilation, but even mature observing systems such as radiosondes have systematic
errors that have largely gone undetected until recently. The detection of random errors is
even more difficult unless independent observations with high precision and comparable
accuracy are available. This presentation describes a technique to assess the accuracy of
aircraft moisture soundings in near real-time. The technique was developed to assess the
accuracy of Radiosonde Replacement System moisture soundings made by GPS
radiosondes for the U.S. National Weather Service. The method integrates the (calculated or
measured) mixing ratio during aircraft ascent or decent and compares it with a Global
Navigation Satellite System (GNSS) estimate of total precipitable water vapor made at a
105