Supply-demand coordination of soil conservation services and its driving mechanism in grassland ecosystems of northern China.

Soil erosion poses a persistent threat to grassland ecosystems and sustainable land management; previous large-scale assessments focused on soil conservation supply but ignored its alignment with long-term spatially heterogeneous erosion pressure. This disconnect limits the ability to diagnose where and why soil conservation services fail to meet ecological needs. In this study, a 24-year long-term dataset that integrated multi-source remote sensing, land-use data, and erosion modeling was leveraged to quantify the soil conservation supply and erosion pressure and analyze their coordination across the grasslands of northern China. A coupling coordination degree (CCD) framework was developed and the supply-demand ratio was used to reveal persistent spatiotemporal mismatches in the soil conservation performance in the study region over a period of 24 years, capturing both gradual trends and interannual variability. To uncover the mechanisms underlying these patterns, we combined an interpretable machine learning (ML) model-extreme gradient boosting (XGBoost) with SHapley additive explanation (SHAP) analysis-with the constrained line analysis and geographically weighted regression (GWR) methods, enabling the identification of nonlinear driver thresholds and spatially nonstationary controls. Slope, precipitation, vegetation cover, and temperature emerged as the dominant drivers; however, their effects exhibited clear nonlinear turning points, reflecting long-term interactions between climate variability, vegetation dynamics, and landscape structure. The novelty of this study lies in explicitly linking soil conservation supply to erosion pressure over an extended temporal scale, moving beyond static or supply-only ecosystem service assessments. By integrating long-term data with interpretable modeling and spatial heterogeneity analysis, this study presents novel insights into the mechanism of deteriorating effectiveness of soil conservation over a long-term period. This approach delivers an operational and evidence-based framework for identifying priority areas and designing spatially targeted and context-dependent soil conservation strategies, with the aim of enhancing the resilience of grassland ecosystems under the constraints of global warming and land-use change.