The Three Gorges Reservoir (TGR) in China, with the largest dam in the world, stores a large volume of water and may influence the Earth’s gravity field on sub-seasonal to interannual timescales. Significant changes of the total water storage (TWS) might be detectable by satellite-based data provided by the Gravity Recovery and Climate Experiment (GRACE) mission. To detect these store water changes, effects of other factors are to be removed first from these data due to band-limited representation of near-surface mass changes from GRACE. Here, we evaluated three current popular land surface models (LSMs) basing on in situ measurements and found that the WaterGAP Global Hydrology Model (WGHM) demonstrates higher correlation than other analyzed models with the in-situ rainfall measurement. Then we used the WGHM outputs to remove climate-induced TWS changes, such as surface water storage, soil, canopy, snow, and groundwater storage. The residual results (GRACE minus WGHM) indicated a strong trend (3.85 ± 2 km3/yr) that is significantly higher than the TGR analysis and hindcast experiments (2.29 ± 1 km3/yr) based on in-situ water level measurements. We also estimated the seepage response to the TGR filling, contributions from other anthropogenic dams, and used in-situ gravity and GPS observations to evaluate dominant factors responsible for the GRACE-based overestimate of the TGR volume change. We found that the modeled seepage variability through coarse-grained materials explained most of the difference between the GRACE based estimate of TGR volume changes and in situ measurements, but the agreement with in-situ gravity observations is considerably lower. In contrast, the leakage contribution from 13 adjacent reservoirs explained ~74% of the TGR volume change derived from GRACE and WGHM. Our results demonstrate that GRACE-based overestimate TGR mass change mainly from the contribution of surrounding artificial reservoirs and underestimated TWS variations in WGHM simulations due to the large uncertainty of WGHM in groundwater component. In additional, this study also indicates that reservoir or lake volume changes can be reliably derived from GRACE data when they are used in combination with relevant complementary observations. View Full-Text
