Ensemble Probabilistic Prediction of a Mesoscale Convective System and Associated Polarimetric Radar Variables using Single-Moment and Double-Moment Microphysics Schemes and EnKF Radar Data Assimilation

Abstract

nsemble-based probabilistic forecasts are performed for a mesoscale convective system (MCS) that occurred over Oklahoma on 8-9 May 2007, initialized from ensemble Kalman filter analyses using multi-network radar data and different microphysics schemes. Two experiments are conducted, using either a single-moment or double-moment microphysics scheme during the one-hour long assimilation period and in subsequent three-hour ensemble forecasts. Qualitative and quantitative verifications are performed on the ensemble forecasts, including probabilistic skill scores. The predicted dual-polarization (dual-pol) radar variables and their probabilistic forecasts are also evaluated against available dual-pol radar observations, and discussed in relation to predicted microphysical states and structures.Evaluation of predicted reflectivity (Z) fields shows that the double-moment ensemble predicts the precipitation coverage of the leading convective line and stratiform precipitation regions of the MCS with higher probabilities throughout the forecast period compared to the single-moment ensemble. In terms of the simulated differential reflectivity (ZDR) and specific differential phase (KDP) fields, the double-moment ensemble compares more realistically to the observations and better distinguishes the stratiform and convective precipitation regions. ZDR from individual ensemble members indicates better raindrop size-sorting along the leading convective line in the double-moment ensemble. Various commonly used ensemble forecast verification methods are examined for the prediction of dual-pol variables. The results demonstrate the challenges associated with verifying predicted dual-pol fields that can vary significantly in value over small distances. Several microphysics biases are noted with the help of simulated dual-pol variables, such as substantial over-prediction KDP values in single-moment ensembles.