Probabilistic Analysis and Prediction of Convective Storms using the Local Particle Filter

Monte Carlo data assimilation techniques provide a means of representing the probabilistic evolution of simulated weather events conditioned on atmospheric measurements. These methods typically operate under Gaussian approximations for the underlying error distribution, which may be inappropriate for highly nonlinear applications. For the case of severe convective storms, nonlinearity in the dynamical model and measurement operators that map satellite and radar observations to prognostic model state variables both pose challenges for Gaussian based ensemble Kalman filters (EnKFs) and ensemble-variational hybrids. The current study applies a newly developed data assimilation method called the local particle filter (PF), which avoids the assumptions of methods currently used for this application. Theoretical advantages of the local PF are explored using idealized and real convective-scale data assimilation experiments performed in the Weather Research and Forecasting model. Results from this study mark the first successful application of a PF-based data assimilation strategy in a weather model and provide insight into how remotely sensed data can be used most effectively for real-time numerical weather prediction.