Unraveling the Dynamics and Driving Forces of Desert and Riparian Vegetation in the Lower Reaches of a Typical Inland River Basin, Northwestern China

The lower reaches of inland river basins in northwestern China are highly vulnerable to water scarcity, ecological degradation, and climate change, yet vegetation responses and their environmental controls remain poorly understood. Using the terminal delta of the Heihe River basin (HRB) as a case study, we analyzed vegetation dynamics during 2001–2024 using the April–October growing-season median normalized difference vegetation index (NDVI; SMN) and net primary productivity (NPP) across desert, unstable riparian vegetation (URV), and stable riparian vegetation (SRV). Multi-source datasets were used, including MODIS products for vegetation dynamics, TerraClimate data for climatic variables, and GLEAM data for soil moisture. An RF–SHAP model integrating Random Forest (RF) and SHapley Additive exPlanations (SHAP) was used to identify dominant drivers and their interactions, considering precipitation (Pr), maximum and minimum temperature (Tmmx and Tmmn), vapor pressure deficit (VPD), downward surface solar radiation (Srad), wind speed (Vs), surface soil moisture (SMs), and root-zone soil moisture (SMrz). SMN and NPP increased significantly (p < 0.001), with the strongest increases in SRV; increasing trends covered approximately 84% and 77% of the study area, respectively. SMrz dominated NPP dynamics across vegetation types, whereas Tmmn, SMrz, and Srad dominated SMN dynamics in desert, URV, and SRV, respectively. SMrz showed thresholds of 0.070 and 0.080 m3 m−3 for rapid vegetation responses in desert and riparian vegetation, respectively. High SMrz under low Srad and Tmmn further enhanced riparian vegetation recovery. These findings highlight vegetation-specific controls and thresholds, supporting differentiated ecological restoration and water management in arid inland river basins.