image: Satellite-based assessment of recovery after severe drought in large river basins. (a) Cumulative recovery probability measured from the time of maximum water storage loss. The black curve represents events in all basins, and the red curve represents events in basins with at least one event remaining unrecovered for more than five years. Shaded ribbons indicate uncertainty. (b) Median recovery time among recovered events in each basin. White circles indicate that recovery times differ by more than one year between events. (c) The longest observed period without recovery in each basin through 2025. Squares indicate that another severe drought prolonged the deficit.
Credit: ©Science China Press
The return of rain does not necessarily mean that a river basin has replenished its water reserves. A new satellite-based study shows that recovery after severe drought is widespread across large river basins, but highly uneven. Some recover relatively quickly, while others experience water storage deficits that persist for years.
Published in National Science Review on September 30, 2026, the study was led by Yongqiang Zhang at the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences. The international team combined observations from the Gravity Recovery and Climate Experiment (GRACE) and its Follow-On mission (GRACE-FO) with climate and hydrological datasets. These satellites detect changes in Earth’s gravity that reveal changes in water stored in rivers, lakes, soils and aquifers. Across 174 river basins larger than 100,000 square kilometers, the researchers identified 468 severe drought events with detectable water storage responses during 2003–2025.
Measured from the time of greatest water storage loss in each event, the estimated cumulative probability of recovery reached about 84% within five years and 89% within ten years. Among events that recovered, basin-level median recovery times were generally no more than three years. Medians in the Amazon and Yangtze basins were typically within one year, whereas some events in Australia’s Murray–Darling basin took longer. Persistent deficits were concentrated in drylands and some inland basins that do not drain to the ocean.
Recovery followed diverse trajectories, both across basins and between drought events within the same basin. Water storage sometimes rose steadily, sometimes stalled after an initial increase, and sometimes recovered only after a long delay. Larger initial losses did not necessarily mean longer recovery. The consistency and timing of water gains were more closely related to recovery time than the initial loss alone.
These differences were closely linked to water availability after drought: precipitation minus evapotranspiration, the loss of water through evaporation and plant transpiration. Precipitation was the dominant predictor of water storage changes during recovery. Under drier conditions, evapotranspiration and its covariation with precipitation became increasingly important, highlighting the need to consider both water supply and water losses.
Long-lasting deficits were more common where dry climates, recurrent severe droughts and intensive human water use coincided. Central Asia’s Amu Darya illustrated this pattern, showing persistent depletion under limited water supply, repeated droughts and substantial water-use pressure. Across basins with persistent deficits, storage losses occurred mainly during drought, while changes during normal or wet periods were generally much smaller. Repeated losses could therefore accumulate without sufficient replenishment between events. An analysis of storage components indicated that persistent losses were generally dominated by subsurface water, including groundwater and water in deeper unsaturated soils.
Persistent deficits do not mean that a basin can never recover. They indicate that water storage had not achieved sustained recovery to the reference level for the drought event within the observation period. Longer-term changes in water storage can also influence this outcome. The findings support extending drought risk management beyond the return of rain to track whether water reserves are replenished. Strategies should reflect local recharge conditions and water demand, with particular attention to groundwater abstraction and persistent deficits in basins where recovery remains incomplete.
Zhenwu Xu, a postdoctoral researcher at the same institute, is the first author, and Yongqiang Zhang is the corresponding author. The study involved collaborators from TU Wien, the University of Potsdam, the University of California, Irvine, and other institutions.
The research was supported by the National Natural Science Foundation of China (Grant no. 42330506 and 42361144709), the Talent Program of the Ministry of Science and Technology of China, the Postdoctoral Fellowship Program (Grant no. GZC20250236), and the General Program of the China Postdoctoral Science Foundation (Grant no. 2025M770359).