A Comparative Convective Study between the Local Particle Filter and Ensemble Kalman Filter with the Gridpoint Statistical Interpolation System

The local particle filter (LPF; Poterjoy 2016, Poterjoy et al. 2019) is a recent development in geophysical data assimilation. Like standard particle filters, the LPF does not assume a specific distribution for prior errors, which should make it more amenable for atmospheric and oceanic applications. In contrast, ensemble Kalman filters (EnKFs), assume a Gaussian distribution. While Gaussian errors may not always be appropriate for geophysical data assimilation, EnKFs are less sensitive to sampling error than particle filter-based techniques, owing to the parametric representation of errors. Therefore, potential benefits of the LPF over EnKFs must be considered alongside other factors, such as limitations on ensemble size and model errors. This research project is a study on the performance of the LPF versus an EnKF for a convective scale data assimilation problem. Previous research (Poterjoy et al. 2017) tested the performance of the LPF for an idealized squall line in the Weather Research and Forecasting (WRF) model within the National Severe Storm Laboratory Experimental Warn-on-Forecast System (NEWS-e) framework for ensembles. Results show a strong potential for applying the LPF to severe convective storms. We will present results comparing the LPF to the EnKF for several cases during May 2018 using the NEWS-e system. The presentation will detail our efforts with tuning the LPF to match the EnKF system, and whether we can actually improve upon the EnKF forecasts using this non-Gaussian data assimilation approach.