The change of groundwater storage (GWS) on the Tibetan Plateau (TP) is vital for water resources management and regional sustainability, but its estimation has large uncertainty due to insufficient hydrological measurements and diverse future climate scenarios. Here, we employ high-resolution land surface modeling, advanced satellite observations, global climate model data, and deep learning to estimate GWS changes in the past and future. We find a 3.51 ± 2.40 Gt yr⁻¹ increase in GWS from 2002–2018, especially in exorheic basins, attributed to glacier melting. The GWS will persistently increase in the future, but the growth rate is slowing down (0.14 Gt yr⁻¹ for 2079–2100 under a high emission scenario). Increasing GWS is projected over most endorheic basins, which is associated with increasing precipitation and decreasing shortwave radiation. In contrast, decreasing GWS is projected over the headwaters of Amu Darya, Yangtze, and Yellow river basins. These insights have implications for sustainable water resource management in a changing climate.

青藏高原 （TP） 地下水储量 （GWS） 的变化对于水资源管理和区域可持续性至关重要，但由于水文测量不足和未来气候情景的多样性，其估计具有很大的不确定性。在这里，我们采用高分辨率陆地表面建模、高级卫星观测、全球气候模型数据和深度学习来估计过去和未来的 GWS 变化。我们发现，从 2002 年到 2018 年，GWS 增加了 3.51 ± 2.40 Gt ^yr-1，尤其是在流出盆地，这归因于冰川融化。GWS 在未来将持续增加，但增长率正在放缓（在高排放情景下，2079-2100 年为 0.14 Gt ^yr-1）。在大多数内流盆地上投影的 GWS 增加，这与降水增加和短波辐射减少有关。相比之下，预估的 GWS 下降发生在阿姆河、长江和黄河流域的源头。这些见解对不断变化的气候中的可持续水资源管理具有重要意义。