Basic Convective and Mesoscale Research
NSSL Project 1 – Severe Weather Warning Applications and Development
Funding Type: CIMMS Task II
Objectives
Develop storm-scale data assimilation methods to be used for “warn
on forecast” applications; examine performance of the ensemble
Kalman filter for multiple convective modes and for assimilating multiple
types of observations into numerical models; examine the sensitivity
of the analyses to changes in the implementation of the EnKF algorithm
for both practical considerations and to understand better how the assimilation
procedure works on the storm scale with a complex convective situation.
Accomplishments
This study builds on past work by CIMMS research scientists by presenting
a successful application of the ensemble Kalman filter data-assimilation
technique using real Doppler radar observations of a mixed mode convective
event, which has yet to be presented in the literature. It is shown
that the quality of the storm-scale analysis produced by the technique
can be at least maintained, if not improved, by using a relatively
small localization radius and by implementing a threshold correlation
to limit small, presumably noisy, covariances from influencing the
analysis. The smaller localization radius has the added benefit of
lowering the computation cost, substantially in some cases. It is also
shown that the analysis system can accurately reproduce the reflectivity
and velocity structures of the observed storms, but the low-level thermodynamic
structures are very sensitive to the design of the system.
This project is ongoing.
Publications
Coniglio, M. C., D. C. Dowell, and L. J. Wicker, 2007: Ensemble Kalman
filter assimilation of Doppler radar data: Analyses of a developing MCS.
22nd Conf. on Weather Analysis and Forecasting/18th Conf. on Numerical
Weather Prediction, Park City, UT, Amer. Meteor. Soc., 3B.3.
Comparison of the reflectivity from KDDC observations on the 1.3° elevation scan (left column), the model reflectivity from the mean analysis of the control ensemble experiment (middle column), and the model reflectivity from a simple baseline experiment in which no assimilation of radar data takes place (right column) at 20 min [(a)-(c)], 40 min [(d)- (f)], 60 min [(g)-(i)], and 80 min [(j)-(l)]. Pertinent geographical features are identified in the upper-left panel.