Response of terrestrial water storage and its change to climate change in the endorheic Tibetan Plateau.

• Lake effect on TWSC was estimated using a newly developed long-term lake volume data. • Contributions of water storage components to TWSC were estimated using both RF model and water balance approach. • The RF method was used to estimate both TWSA and TWSC. • Regional TWSA decreased in the pre-mutation period and then quickly increased in the post-mutation period with an intensified hydrological cycle. Terrestrial water storage (TWS) and its change (TWSC) are extremely sensitive to climate change, and are good indicators of the response of the hydrological system to climate change in the Tibetan Plateau (TP). However, long-term variation in TWS and its components, as well as contributions of these component changes to TWSC, are still not well documented. In this study, we extended monthly TWS anomaly (TWSA) and TWSC from 1989 to 2019 using the random forest (RF) method, analyzed the long-term contribution rates of each component change to TWSC using a mass balance approach, and compared them with the rates estimated by the RF method. Our major findings are: (1) the warming and wetting trend in the entire analysis period (PA) and in the pre-mutation period (P1) was slowed down or even reversed in the post-mutation period (P2), which was caused mainly by the decrease in temperature and precipitation in the north-eastern Inner Basin (IB) and southern Qaidam Basin (QB); (2) the RF method is capable of estimating monthly TWSA and TWSC (0.872 ≤ R2 ≤ 0.957 in training data, 0.528 ≤ R2 ≤ 0.860 in test data). Variable selection has an important impact on the method, and TWSC can be used as an alternative when TWSA estimate is poor; (3) Significant annual TWSA increase (2.20 mm/a, P < 0.05) in the IB during PA comes from a higher increase during P2 (3.62 mm/a, P < 0.05) and a small variation during P1 (−0.65 m/a, P > 0.1). Spatially, the increase is attributed to the higher increase trends (> 4.50 mm/a, P < 0.05) in north-eastern IB during PA and P2, and slight decrease in western and southern IB during PA. The spatial and temporal heterogeneity was mainly caused by the differences in precipitation amount, glacier proportions, and permafrost types. (4) During PA, TWSC in the IB was mainly determined by lake water storage change (LWSC) with a contribution rate of 62 %, while in the QB it was mainly determined by soil water change (−43 %), and LWSC had a contribution rate of −27 %. LWSC impact on TWSC was more significant in the IB, especially in the southeastern sub-regions with regional water storage surplus (TWSC > 0). The overall increase in water storages and fluxes, and exchanges imply an intensified hydrologic cycle in the region. [ABSTRACT FROM AUTHOR]