Kalman filter framework for a regional mass change model from GRACE satellite gravity

Abstract

In this study a regional modelling framework for water mass changes is developed. The approach can introduce geodetic observation types of varying temporal and spatial resolution including their correlated error information. For this purpose a Kalman filter process was set up using a regional parameterisation by space-localising radial basis functions and a process model based on stochastic prediction. The feasibility of the approach is confirmed in a closed-loop simulation experiment using gridded water storage estimates derived from simulated monthly solutions of the GRACE satellite gravimetry mission and considering realistic error patterns. The resulting mass change time series exhibit strongly reduced noise and a very high agreement with the reference model. The modelling framework is designed to flexibly allow a future extension towards combining satellite gravimetry with other geodetic observations such as GNSS station displacements or terrestrial gravimetry.