Precise Ecological Restoration under Water Diversions-Groundwater-ecosystem Interactions in Drylands

Ecological Water Diversion Projects (EWDPs) have been employed worldwide to restore degraded ecosystems in drylands, but precise restoration locations and planting species under EWDP remain unknown, even if it has changed ecohydrological dynamics. Therefore, this study proposed a water diversions-groundwater-vegetation coupling framework based on Machine Learning for deciding where to restore in drylands. The Shapley Additive Explanation method was used to interpret the framework and disentangle the role of the local environment in defining vegetation patterns from a data-driven perspective. This approach was tested in a mega EWDP in the lower Tarim River Basin (TRB), Northwest China, in present and future restoration locations, and planting species were quantitatively estimated. The results showed that the terrain and groundwater table critically impact the vegetation communities in arid regions. Currently, 4.3% of the existing land should be trees and grass, while they account for 0.1% of the lower TRB in total. Within the next decade, the potential reforestation of areas with steadily growing natural vegetation would decrease by 5.3% and 12.6% if the yearly watering volume of EWPD decreased by 28.6% under climate change scenarios SSP245 and SSP585, respectively. Our results confirm the substantial variability in restoration areas under varying diversion volumes and underscore the importance of incorporating EWDP into future restoration management. The proposed approach can form a scientific basis for dryland precision restoration to facilitate water conservation and maximize ecological benefit.