Topographic and economic drivers of agricultural–ecological coupling in Gansu Province: spatial differentiation and threshold effects

Improved understanding of agricultural–ecological coupling is critical for sustainable development in semi-arid regions. This study integrated the Remote Sensing Ecological Index (RSEI), Agricultural Green Development Index (AGDI), Coupling Coordination Degree (CCD) model, and XGBoost–SHAP to examine the spatiotemporal evolution and nonlinear drivers of agricultural–ecological coupling in Gansu Province from 2010 to 2023 at the county scale. A four-quadrant framework was applied to identify differentiated management types. Results showed that farmland ecological quality exhibited a stepped north–south pattern, improving during 2010–2018 but declining after 2019. Agricultural green development displayed a “high southeast, low northwest” pattern, increasing by 0.0036 yr −1 ( p < 0.01). The coupling coordination degree increased from 0.57 to 0.63, with significant spatial clustering (Moran's I = 0.438–0.528, p < 0.001), characterized by stable “high–high” clusters in southern Gansu and “low–low” clusters in northern Gansu. Relief amplitude (40.9%) and GDP per capita (36.1%) jointly explained 77.0% of coordination variation. Elevation above 2000 m constrained coordination, whereas GDP per capita between 15,000 and 30,000 CNY promoted green transformation, with diminishing effects beyond 40,000 CNY. Severe drought conditions (SPEI < −1.5) caused agricultural–ecological imbalance. Counties were classified into four management types (H H, L–H, L–L, and H–L), requiring differentiated governance strategies. This study provides quantifiable thresholds for improving agricultural–ecological coupling in semi-arid regions.