Ecosystems in arid and semi-arid regions are highly sensitive to environmental changes and disturbances. Water scarcity, land degradation, and intensified human activities have increased ecological pressures. Therefore, a comprehensive assessment of ecological vulnerability is essential for developing effective response strategies. However, existing evaluation systems remain incomplete, particularly in terms of analyzing driving factors, with insufficient exploration of critical states in ecological vulnerability changes. This study, using Ningxia as a case, integrates the Sensitivity-Resilience-Pressure (SRP) model with the Remote Sensing Based-Ecological Index (RSEI) and Principal Component Analysis (PCA) to construct an Ecological Vulnerability Index (EVI) and analyze its spatiotemporal evolution. The XGBoost-SHAP model was employed to explore driving factors and their critical thresholds. The results show that from 2000 to 2020, 73.7% of Ningxia's area experienced a significant decline in ecological vulnerability, with an overall moderate vulnerability level. Approximately 60% of the region exhibited clustered "high-high" and "low-low" vulnerability distributions. Vegetation cover was identified as the most significant factor affecting ecological vulnerability changes, followed by climate conditions, ecological quality, and soil moisture, with minimal influence from human activities. These factors interact with vegetation cover, driving ecological vulnerability evolution. A vegetation cover threshold of 38% was found to be a critical point for positive and negative changes in ecological vulnerability. This study enhances the ecological vulnerability framework for arid and semi-arid regions and identifies critical thresholds of vulnerability change. The findings offer scientific support for ecological management in Ningxia and similar regions, providing new insights for future research.