Hydrological models play a crucial role in tracking and predicting terrestrial water storage, yet they face challenges due to uncertainties and inaccuracies caused by various factors such as meteorological processes and data limitations. To refine these models, data assimilation has emerged as a valuable tool, utilizing a new source of data to update model states while considering associated uncertainties, thereby enhancing our comprehension and predictive capabilities in hydrological processes. In this context, satellite data are receiving increasing attention because they can cover large areas and are useful in detecting spatial and temporal variability of water.

In most existing studies related to satellite data assimilation in hydrological models, one source satellite data is used in the analysis. This study focuses on improving subsurface water storage model accuracy by assimilating data from different satellite sources. In particular, we used data from the Soil Moisture and Ocean Salinity (SMOS) satellite and terrestrial water storage data from the Gravity Recovery and Climate Experiment (GRACE). The data are assimilated into a fully coupled subsurface-surface hydrological model, developed with ParFlow-CLM (PARallel FLOW-Community Land Model). The investigation is conducted in Iran. Employing an Ensemble Kalman Filter, three assimilation scenarios are explored: (i) GRACE, (ii) SMOS, and (iii) the combined assimilation of both GRACE and SMOS data (joint). Findings are validated against the Soil Moisture Active Passive (SMAP) and in-situ groundwater data by using a novel probabilistic reliability framework, demonstrating the advantages of joint data assimilation.

The study highlights the influence of assimilated remote sensing data type on the effectiveness of data assimilation. Assimilating GRACE data enhances groundwater level estimations.

However, SMOS data positively impacts topsoil moisture estimations, but adversely affects groundwater level estimates. Importantly, the assimilation of both GRACE and SMOS data through multivariate (joint) data assimilation significantly improves accuracy for both soil moisture estimation and groundwater level estimation.