Assimilating GRACE Data for Improved Modeling of the Terrestrial Hydrological Cycle and Its Response to ENSO With the Noah‐MP Multi‐Physics Ensemble

Abstract

Assimilating the Gravity Recovery and Climate Experiment (GRACE) terrestrial water storage (TWS) has been proven a promising approach to improving terrestrial hydrological simulation, but its impact under different climatic regimes and on hydrological responses to important climatic oscillations (i.e., El Niño‐Southern Oscillation (ENSO)) remains unclear. This study constructed a land surface data assimilation (DA) system based on a multi‐physics ensemble constructed by the Noah‐Multiparameterization land surface model (Noah‐MP) and the Data Assimilation Research Testbed over the southern contiguous United States (CONUS). The system features a set of high‐skill and low‐interdependence parameterization configurations as the assimilation ensemble, based on which the daily‐scale GRACE TWS was assimilated using the Ensemble Adjustment Kalman Filter. Results show that DA enhances the performance of TWS anomaly in the southeastern/south‐central CONUS, the Rocky Mountains, the Sierra Nevada, and the Pacific Northwest. For soil moisture, DA offers more skill gains in the deep layer than in the surface layer. For snow water equivalent, DA improves the performance in the northeastern/central CONUS and the western mountains, but offers limited gains in the southern CONUS, where the snow cover is shallow and seasonal. DA improves the runoff simulations in arid and semi‐arid regions where Noah‐MP performs poorly, but slightly degrades the performance in humid regions where Noah‐MP performs well. DA also improves the fidelity of runoff–ENSO linkages in semi‐arid and arid regions by eliminating spurious correlations. Our findings confirm the effectiveness of GRACE DA in reducing hydrological simulation uncertainties and demonstrate its potential for improving water management planning under climatic oscillations.