Karst landscapes are characterized by distinctive geological and hydrological features, supporting fragile terrestrial vegetation ecosystems. Understanding vegetation dynamics and their driving mechanisms is crucial for ecological restoration and conservation in such regions. This study focuses on Guizhou Province, a typical karst area in southwestern China, and analyzes MOD13A1 NDVI time-series data from 2000 to 2021. By integrating multiple analytical techniques, including Theil–Sen and Mann–Kendall trend analyses, the Breaks for Additive Seasonal and Trend (BFAST) algorithm, and the XGBoost–SHAP interpretable machine learning framework, we systematically examine spatiotemporal patterns, nonlinear trends, and underlying driving factors. Results indicate a statistically significant increasing trend in NDVI. BFAST detected monotonic greening across 76.82% of the study area, while certain regions exhibited interrupted growth or transitions from greening to browning, with breakpoints primarily occurring in 2012 and 2014. SHAP analysis identified soil moisture as the principal natural driver and population density as the most significant anthropogenic factor. All driving variables displayed distinct threshold behavior and nonlinear responses in relation to NDVI, accompanied by notable interaction effects. This study enhances the understanding of vegetation dynamics and their drivers in karst mountainous regions, providing new insights for optimizing ecological management policies.
Xie et al. (Sun,) studied this question.