Abstract: In central Yunnan province,there are numerous areas of mining damage,and the study of their vegetation restoration is beneficial for the protection and management of the regional ecological environment.In this study,it focus on Anning city,Mile city and Malong district of Qujing city.The research utilizes Landsat remote sensing image data from 2000 to 2022,processed via the GEE platform.The annual average kNDVI index is employed to characterise the vegetation cover.Then combines with the spatial distribution of mining-affected land in the county area in 2023.The XGBoost-SHAP interpretable machine learning model,Theil-Sen and Mann-Kendall trend test,coefficient of variation and Hurst index are employed to systematically analyse the changes in vegetation cover and its influencing factors in the mining-damaged land in the study area.The study reveals that the vegetation of mining-affected land in the study area exhibites a marked degradation trend,characterised by diminished stability and the presence of positive persistence characteristics.The analysis indicates that topographic factors (such as elevation and slope) are emerged as the predominant influences on vegetation evolution,followed by soil chemical property factors (including nitrogen,phosphorus,and potassium) and soil physical property factors (such as porosity and clay content) to a least extent.The driving factor analysis method based on the XGBoost-SHAP model proposed in this study can effectively identify the key influencing factors of regional vegetation change,and provide a reference for ecological restoration research in similar regions.

Key words: mining-induced land degradation, mid-inner area of Yunnan province, vegetation coverage, XGBoost-SHAP