An ensemble Kalman filter with rescaling disaggregation for assimilating terrestrial water storage into hydrological models

Abstract

Abstract Assimilating satellite-based Terrestrial Water Storage (TWS) observations can improve the vertical summation of water storage states in hydrological models. However, it can degrade individual storage compartments or hydrological fluxes, limiting the applicability of TWS Data Assimilation (DA) for water management and flood monitoring. This issue arises from the ensemble-based TWS update disaggregation approach used by DA techniques like the Ensemble Kalman Filter (EnKF). Thus, this study makes two key contributions. First, we introduce a novel analysis method that provides quantitative and qualitative insights into how individual storage compartments are affected during TWS DA, by examining the sign and magnitude of the individual storage updates and their responses. Second, we propose a new disaggregation approach, EnKF-R, which “rescales” the individual storage of model compartments to match the updated TWS, avoiding the use of ensemble statistics within the disaggregation process. The EnKF-R approach was tested in two climatologically different river basins and validated against both synthetic and real independent data. Our results show that EnKF-R produces similar TWS estimates to the classical EnKF while reducing degradations in individual water storage compartments and with lower computational cost, making it a promising alternative. Limitations regarding spatial continuity and uncertainty estimation require further developments.