The impact of radar data assimilation on a squall line in Mississippi

The potential for improving short-term model forecasts of a squall line using mesoscale observing networks (or mesonets) and radar data is addressed in this study. On 30 April 2005 between 00Z and 12Z, a squall line with severe winds and a few tornadoes passed through the Mississippi Delta. The Delta is the distinct northwest section of the state of Mississippi that lies between the Mississippi and Yazoo Rivers. This region, created by regular flooding over thousands of years, is remarkably flat and contains some of the most fertile soil in the world. Because of this region's agricultural importance, improving severe weather forecasts are of great interest to the state.

The Delta contains a number of stations in the Soil Climate Analysis Network (SCAN). SCAN uses meteor burst telemetry to obtain remote site information reliably on a near real-time basis, and focuses on the agricultural areas. The data include hourly and daily recordings of precipitation, air temperature, solar radiation, wind speed, relative humidity, soil moisture, and soil temperature. This study seeks to investigate the impact of this data on the analysis and prediction of spring squall lines.

Another research tool with tremendous potential for improving short-term thunderstorm prediction is the assimilation of Doppler radar data (radial velocity and reflectivity) into WRF. Xiao et al. (2005, 2006) show this technique improves rainfall forecasts, and better defines rainband structure, especially if used in a 3DVAR cycling mode. The overall goal of this research is to improve - using variational data assimilation techniques on radar and SCAN data with WRF - the prediction of the Quantitative Prediction Forecast (QPF) for this case.

The first step involves defining the background error. The background error will be constructed using the "NMC method" with two weeks of 12-h runs. It will be shown this produces generally the same background errors as 6-h runs. Both runs produce different results than the default background error in WRF, showing the importance of constructing your own background error for high-resolution, regional runs. Background errors were also computed using ensemble runs. It will be shows this produces reasonable results as well, suggesting that this easier method could replace the NMC method.

Four runs are conducted, using 1) standard weather data, 2) SCAN data, 3) radar data, and 4) radar and SCAN data. QPF totals will be shown for all four runs based on threat scores, as well as a qualitative comparison of the squall line simulations in terms of propagation speed, intensity, and appearance. It will be shown some aspects of the squal line were handled better using the radar data. However, the SCAN data made little impact.

While deterministic forecasts are the goal for rainfall situations, nonlinear dynamics suggest that simulating the detailed evolution of individual convective cells may not be possible beyond a few convective overturning times - on the order of a few hours - except perhaps in special cases such as quasi-steady supercell convection. On the other hand, there is a great deal of evidence suggesting that ensembles of convective cells are in statistical equilibrium with mesoscale or synoptic scale circulations, offering hope that ensemble-average precipitation might have much longer predictability timescales.

The value of a probabilistic approach using ensemble WRF modeling will also be shown. Probability plots for different rain thresholds will be computed and validated against observations, as well as the ensemble consensus. The multiple runs will also be overlaid for an intercomparison of structure, rain rate, and phase speeds. The overall usefullness of a "consensus" forecast from the ensemble runs will be addressed.