Evaluation of runoff variability in transboundary basins over High Mountain Asia: Multi-dataset merging based on satellite gravimetry constraint

Runoff variability in glacierized transboundary river basins over High Mountain Asia (HMA) directly affects the stability of water supply for more than one billion people in Asia. However, limited by insufficient in-situ gauges and imprecise hydrological model output, it is still a challenge to accurately monitor and comprehensively analyze the HMA runoff change. In this paper, we construct a water budget closure test of water balance equation based on satellite gravimetry constraints to assess the accuracy of hydrological dataset outputs, and propose a multi-dataset merging method to evaluate runoff variability in ten HMA transboundary basins over the past two decades. Results show that the runoff quantified by the hydrological dataset has relatively maximum uncertainty compared to precipitation and evapotranspiration. The performance of the reconstructed terrestrial water storage change (TWSC) from hydrological dataset varies with basins, and the maximum Nash-Sutcliffe Efficiency (NSE) value ranges from 0.31 to 0.94. Nevertheless, the current hydrological dataset struggles to accurately reconstruct the interannual and annual variability of TWSC, with the maximum cyclostationary NSE (NSEc) value ranging from −1.07 to 0.24. Runoff change in HMA exhibits both overall stability and regional climatic condition-related spatial heterogeneity. A significant downstream change-driven increase trend of runoff occurs in Indus Basin (0.2 ± 0.1 mm/mon/yr), while Brahmaputra Basin (−0.5 ± 0.4 mm/mon/yr) and Salween Basin (−0.4 ± 0.2 mm/mon/yr) show significant runoff decrease trends driven by upstream and downstream changes, respectively. Climate change has exacerbated the instability of runoff in the arid basins over northern HMA, leading to evident increase in annual amplitude. Furthermore, negative correlation is found between temperature and runoff at the interannual scale, especially in Ganges Basin (−19.73 ± 12.53 Gt/month per °C) and Mekong Basin (−17.46 ± 9.43 Gt/month per °C). Our multi-dataset merging methodology can improve the reliability of using global hydrological datasets to quantify runoff variability in poorly in-situ gauged regions, and may also be applicable to the evaluation of precipitation and evapotranspiration.