Localized Adaptive Inflation in Ensemble Data Assimilation: Application to a Radiation Belt Model

The Ensemble Kalman Filter (EnKF) has become an important data assimilation tool for numerical models in the geosciences. Recently, the EnKF has been applied to radiation belt models to accurately estimate Earth's radiation belt particle distribution. A particular concern in data assimilation for radiation belts is model deficiencies, due to lack of appropriate source and/or loss terms for trapped particles, which can adversely impact the solution of the assimilation. In this work we present a localized adaptive covariance inflation technique used to account for model uncertainty in EnKF. A one-dimensional radial diffusion model for phase space density, together with observational satellite data, is used in EnKF with the purpose of accurately estimating Earth's radiation belt particle distribution. Numerical results from identical-twin experiments, where data is generated from the same model, as well as the assimilation of real observational data, are presented. The results show improvement in the predictive skill of the model solution due to the proper inclusion of model errors in the data assimilation.