Spatiotemporal dynamics of water–carbon synergy and its driving mechanisms in the Kuye River Basin on the Loess Plateau of China

Understanding the coupling between water and carbon processes is critical for ecosystem sustainability in water-limited regions, yet the underlying mechanisms remain insufficiently quantified under concurrent climate and land use change. This study developed an integrated framework combining multi-source remote sensing data, ecosystem service modeling, and interpretable machine learning to investigate the spatiotemporal dynamics, driving mechanisms, and threshold responses of water–carbon synergy, represented by the ecosystem services benefit (ESB) index, in the Kuye River Basin from 2005 to 2020. Results showed that precipitation and temperature exhibited increasing trends, accompanied by vegetation restoration, cropland decline, and urban expansion. Water yield, net primary productivity (NPP), and ESB displayed consistent spatial gradients, increasing from north to south, with a marked enhancement after 2012. Precipitation was identified as the dominant driver, exerting strong independent and interactive effects with vegetation and climatic factors. Nonlinear and threshold-dependent responses were evident, with critical thresholds identified for precipitation (386–421 mm), temperature (8.19–8.55 °C), NDVI (0.358–0.374), soil moisture (0.148–0.151 m³/m³), and evapotranspiration (239–245 mm). These results revealed that water–carbon synergy is governed by complex trade-offs between vegetation growth and water availability. By integrating ESB with SHAP, Geodetector, and piecewise regression with bootstrap-based confidence intervals, this study provides new insights into the nonlinear mechanisms and threshold behaviors of ecohydrological coupling, offering a robust basis for optimizing vegetation restoration and water resource management in arid and semi-arid regions.