Assimilation of Radar Data to Improve High-Resolution Numerical Simulations of Severe Hail Producing Thunderstorms

Hail causes over 1 billion dollars in damage annually in the United States, despite this explicit hail prediction using convection resolving models remains relatively understudied. In this study an optimal data assimilation (DA) configuration for explicit hail prediction is developed. An ensemble Kalman filter (EnKF) is used to assimilate surface and radar observations into ensembles run using either a double moment (DM), triple moment (TM), or variable density rimed ice DM microphysics scheme. The DA configuration developed in this study is used to improve the representation of the microphysical state of a simulated severe hail producing supercell thunderstorm on 19 May 2013. Surface hail forecasts launched from these ensembles are verified against surface based hail reports and radar derived hail products.

Preliminary results indicate skillful surface hail size forecasts cannot be produced until hail growth and decay is properly represented. The ensemble run using the variable density rimed ice DM scheme produces surface hail size forecasts with the most skill both in terms of the spatial extent and size of hail. The variable density rimed ice DM scheme has improved representation hail within the melting layer, and explicitly predicts the density of both hail and graupel, allowing the scheme to represent a spectrum of rimed ice particles. Forecasts produced by ensembles using the DM and TM schemes have less skill in predicting maximum hail size than the variable density rimed ice DM scheme. Both schemes poorly represent hail in the melting layer, causing the schemes to overestimate the size of hail within a hail core.