Review of assimilating GRACE terrestrial water storage data into hydrological models: Advances, challenges and opportunities

Global climate change and anthropogenic impacts lead to alterations in the water cycle, water resource availability and the frequency and intensity of floods and droughts. As a result, developing effective techniques such as hydrological modeling is essential to monitor and predict water storage changes. However, inaccuracies and uncertainties in different aspects of modeling, due to simplification of meteorological physical processes, data limitations and inaccurate climate forcing data limit the reliability of hydrological models. Satellite remote sensing datasets, especially Terrestrial Water Storage (TWS) data which can be obtained from Gravity Recovery and Climate Experiment (GRACE), provide a new and valuable source of data which can augment our understanding of the hydrologic cycle. Merging these new observations with hydrological models can effectively enhance the model performance using advanced statistical and numerical methods, which is known as data assimilation. Assimilation of new observations constrain the dynamics of the model based on uncertainties associated with both model and data, which can introduce missing water storage signals e.g., anthropogenic and extreme climate change effects. Assimilation of GRACE TWS data into hydrological models is a challenging task as provision should be made for handling the errors and then merging them with hydrological models using efficient assimilation techniques. The goal of this paper is to provide an in-depth overview of recent studies on assimilating GRACE TWS data into hydrological models and shed light on their limitations, challenges and progress. We present a comprehensive review of some challenges with GRACE TWS data assimilation into a hydrological model including GRACE TWS errors e.g., the correlated noise of high-frequency mass variations and spatial leakage errors, and how to work with GRACE TWS data errors to use the potential of GRACE TWS data as much as possible. We provide a review of the benefits and limitations of available data assimilation techniques with emphasis on the capability of sequential methods for hydrological applications.