Conducting agricultural land suitability assessments (ALSA) scientifically is crucial for ensuring food security and fostering sustainable agricultural development. This study assessed the suitability of agricultural land in Taiyuan using a geographic information system (GIS) and the analytic hierarchy process (AHP), integrating factors such as topography, soil, water sources, and social conditions at a 1 km spatial resolution. The primary aim was to map the spatial distribution of agricultural land suitability and understand county-level variations. Given the irreversible impact of urban development on land use and the critical importance of ecological conservation, corresponding subtractions for urban and natural protected areas have been applied in this study during the assessment of agricultural land suitability. The findings revealed that Taiyuan’s agricultural land suitability generally falls within an intermediate range, without areas classified as completely unsuitable (lowest rank) or suitable (highest rank). The agricultural land suitability does not reach the extreme conditions of being “unsuitable” (lowest rank) nor “suitable” (highest rank), reflecting an overall intermediate potential for agricultural production across the entirety of Taiyuan. The spatial distribution indicates higher suitability in the east and lower in the west, with 33.1% of Taiyuan’s territorial area deemed relatively suitable, 61.3% moderately suitable, and only 5.6% generally suitable for agricultural production. Recommendations include focusing on high-economic-return crops in suitable areas, adopting drought-resistant varieties and enhancing agricultural infrastructure in moderately suitable areas, and prioritizing ecological conservation in generally suitable areas. Additionally, county-level strategies suggest differentiated agricultural models: agritourism and boutique agriculture in urban conflict areas like Qingxu and Wanbailin; cultivation of cold-resistant crops in ecologically fragile areas like Loufan; and sustainable agricultural practices like planting drought-resistant crops in water-scarce regions like Yangqu. This comprehensive assessment offers valuable insights for optimizing agricultural land allocation in Taiyuan, balancing economic development with ecological sustainability.
TENS constitutes a critical ecological barrier on the southeastern margin of the Qinghai–Tibet Plateau, providing essential services such as water conservation and biodiversity protection and helping to safeguard water security in the upper reaches of the Yangtze and Yellow Rivers. Thus, elucidating its vulnerability dynamics is paramount for regional security. Integrating multi-source spatiotemporal data with an interpretable XGBoost–SHAP framework, we quantified interannual variation in vulnerability and the nonlinear threshold responses of key drivers. The results showed pronounced nonlinear phase changes in vulnerability, with 47.96% of the area experiencing abrupt shifts. Notably, 37.89% of TENS reversed from decreasing to increasing vulnerability. TENS underwent an intensive transition during 2010–2015. Interannual variability was dominated by the coupled influence of human disturbance, soil moisture, and atmospheric water, accounting for nearly 60% of the variation, and showed distinct thresholds. Grazing intensity < 0.90 SU/ha was a moderate disturbance, reducing vulnerability, but it became a stressor above this level. Soil moisture showed an inflection point at 79 mm, while vapor pressure deficit (VPD) < 0.39 kPa enhanced resilience, revising the view of VPD as solely a stress factor. Different ecosystems exhibited distinct driving mechanisms. Grasslands were controlled by shallow soil moisture and grazing, forests by hydrothermal balance, and wetlands by low-intensity anthropogenic disturbance (NTL as a proxy; e.g., tourism development or urban expansion). These findings highlight the risk of abrupt shifts in vulnerability regimes (turning points and trend reversals) and support management that emphasizes quality improvement and threshold-based risk management.