Assessment of two GRACE-assimilated terrestrial water storage datasets across 44 river basins using GPS observations

The Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-on (GFO) gravity observations have significantly improved models of the terrestrial water cycle globally. However, GRACE-assimilated models of terrestrial water storage still show differences amongst the models, and studies to determine their ability to predict the state of terrestrial water storage in different regions are ongoing. This paper uses Global Positioning System (GPS) data to assess two global GRACE-assimilated datasets: GLWS2.0 and CLSM-DA. From 2004 to 2019, the mean annual amplitude of water thickness of these datasets differs by more than 25 mm over 40% of the land area. Additionally, the models predict the timing of maximum water storage with difference in phase of 30-days across 50% of their domain. We compare the modeled hydrological loading vertical displacements predicted from these models with GPS uplift data as a measure of the model quality. We cluster 5,983 global GPS stations, each with at least three years of daily data, based on river basin borders. This segmentation allows for better detection of how hydrological conditions, e.g. precipitation patterns, soil characteristics, etc., and model calibration (applied in each river basin) influence the model-GPS agreement.    

Our comparison demonstrates that compared to GLWS2.0, CLSM-DA generally agrees better with GPS and GRACE data across more river basins. We find that the 100-300 mm larger annual water variation of CLSM-DA to GLWS2.0 accounts for CLSM-DA’s better agreement with GPS in Africa, Southeast Asia, and some parts of South America. For regions like the Western United States and Eastern Europe, where the two models propose a similar range of annual water variation, the 30-60 days phase delay of CLSM-DA improves its alignment with GPS. Our findings highlight the need for regional improvement in these models, particularly in areas where they significantly deviate from GPS observations of the terrestrial water variation.