Multi-Scenario Simulation and Driving Factor Analysis of Carbon Storage Based on PLUS-InVEST and XGBoost-SHAP Models: A Study from Weihe River Basin, China

Against China’s dual-carbon strategy and watershed ecological high-quality development initiatives, exploring land-use-driven carbon storage variations is crucial for watershed spatial governance, ecological restoration and carbon sink improvement. As the Yellow River’s largest tributary, the Weihe River Basin straddles the ecotone of arid–semiarid Northwest China and the eastern monsoon region, with fragile ecosystems, intense human activities and dramatic land-use evolution. Clarifying its spatiotemporal carbon dynamics and nonlinear factor response patterns can support low-carbon ecological regulation and coordinated watershed sustainability. This study integrates PLUS and InVEST models to quantify the spatiotemporal patterns of land use and carbon storage across 2000–2020 land use data. Four development scenarios are established to predict 2030 carbon storage spatial heterogeneity, and an interpretable XGBoost-SHAP framework was employed to disentangle the nonlinear responses of carbon storage to natural and socioeconomic drivers. The results show that basin carbon storage exhibited an overall increasing trend, with a net increase of 452.29 × 104 t over the 2 decades. Spatially, high-carbon areas presented patchy aggregation in the northeast, sporadic distribution in the west and zonal expansion in the south–central basin, while low-carbon areas were concentrated in the downstream Guanzhong Plain urban agglomeration. Land-use transitions represented a major source of regional carbon variability, with forest expansion driven by the Grain-for-Green Program generating the largest carbon sequestration increments. The 2030 scenario simulations revealed elevated carbon storage under natural development, cropland protection, and ecological protection scenarios, with the ecological protection scenario yielding the most prominent carbon gain. Conversely, unconstrained economic expansion coincided with prominent carbon storage declines, which implies that targeted ecological conservation and restoration could help alleviate carbon depletion across the watershed. NDVI, slope, and population density are the primary determinants of carbon storage, exhibiting typical staged nonlinear influencing effects. This study provides reliable scientific references for refined ecological restoration, spatial optimization, and carbon sink enhancement in ecologically fragile river basins.