Retrieving precipitation microphysical state of convective storms using radar data and the ensemble kalman filter

An important piece of the puzzle for improving numerical weather prediction (NWP) microphysics is accurately estimating the drop size distribution (DSD) of the hydrometeors used for the initial condition in a forecast. Basically, two approaches currently exist for estimating DSDs from radar data: i) direct observation-based retrieval, and ii) NWP model-based retrieval (i.e., data assimilation (DA)). In direct retrieval, the DSD parameters are estimated from radar reflectivity and/or differential reflectivity. This method normally pre-assumes hydrometeor type and needs at least the same number of independent radar measurements as the number of DSD parameters sought. In reality, however, multiple species of hydrometeors exist in a convective system and there are more parameters desired than available independent measurements, specifically for multi-moment microphysics schemes in NWP models. In this case, model-based retrieval using the Ensemble Kalman Filter (EnKF) yields promising results. The EnKF uses ensemble covariances between variables to spread information from observed reflectivity to the microphysical state variables. These state variables can then be used to prognose the DSD parameters. For this study, the EnKF was applied to a mesoscale convective system (MCS) that passed over western Oklahoma early on May 9 2007 as well as a tornadic supercell thunderstorm in central Oklahoma on 10 May 2010. Both a single moment (SM) Lin three-ice microphysics scheme and a Milbrandt and Yau two-moment (DM) scheme were used. Both events were observed by KOUN, a polarimetric WSR-88D radar, and the Engineering Research Center for Collaborative and Adaptive Sensing of the Atmosphere (CASA) Integrated Project One polarimetric X-band radar network. The polarimetric radar measurements provide additional information on the hydrometeors present to assess the accuracy of the estimations. The final EnKF analyses will be compared to these observations. The degree of improvement of the DM scheme over the SM will also be assessed. Since the quality of the model forecast is directly related to the initial condition, the structure, evolution, and microphysical state of forecasts using the final analyses will be considered as well.