How well do the CMIP6 climate models capture terrestrial water storage variations in data-scarce basins originating from the high mountains of Asia?

• Evaluating the performance of GCMs to simulate TWS over HMA basins with scarce data. • Simulated TWS from the BMA agrees better with observations than single simulations. • Future TWS shows decreasing trends in 67% of HMA basins with scarce data. • Increased TWS is projected for all seasons in basins of Qaidam and Inner TP. Understanding the terrestrial water storage (TWS) change across high-mountain Asian (HMA) basins is critical to enhancing our capability to predict and adopt to future climate change impacts on water resources. Meanwhile, it is critically important to accurately represent the dynamics of the terrestrial water storage for global climate models. This study, for the first time, explored the modeling and prediction skill in TWS change across HMA basins with scarce data for CMIP6 climate models. TWS was generally overestimated in the south and underestimated in the northwest of the study area. Nonetheless, high positive correlation coefficients (CC, above 0.6) between most of model simulations and monthly GRACE observations were detected over the above regions. Climate models reproduced well the seasonal variation of the observed TWS in most basins. However, it was difficult to capture interannual variability in TWS for the individual model, with CC lower than 0.6 in most basins. Then a Bayesian model averaging (BMA)-based multi-model ensemble framework was constructed to predict TWS change across HMA basins with scarce data by 2060 under three scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5). Our BMA-based TWS change estimation decreased the areal-mean normalized root mean square errors by 0.35–0.77 and increased the areal-mean CC by 0.32–0.44 across HMA basins with scarce data for 2002–2020. Future projections of TWS under most scenarios show decreasing trends in two thirds of HMA basins with scarce data, where consistent sign of trends for TWS in the historical period and future scenarios was detected except for the Yangtze River basin. By contrast, consistent increases of TWS are projected for all seasons in basins of Qaidam, Inner Tibetan Plateau and Yellow River under future scenarios, where significantly increasing trends of projected TWS are also detected. The decreasing trend in projected TWS over a majority of the HMA basins with scarce data suggests the risk of water shortage is likely to be aggravated and adaptive water resources management is needed. This study enriches the information for TWS change over HMA basins and offers a helpful direction for local water resource protection.