Objective Coal has been the predominant energy source in China, providing necessary energy to stimulate the national economy and create a context of overall wealth and development. However, coal mining has impacted the ecological environment, particularly vegetation ecosystems in and surrounding mining areas. Large-scale open-pit mining operations have generated extensive industrial brownfields, characterized by dramatic reduction in surface vegetation cover, sharp decline in biodiversity, and severe degradation of ecosystem services. These impacts extend beyond mere quantitative vegetation loss, fundamentally altering the structure and functionality of regional ecosystems, thereby affecting critical ecological processes including carbon sequestration and hydrological cycles. Although substantial research has been conducted globally on brownfield remediation, primarily focusing on vegetation restoration or soil remediation, there exists a significant knowledge gap about the ecological effects during brownfield formation stages. In particular, systematic investigations into the mechanisms through which mining activities affect vegetation cover remain particularly inadequate. Although sgnificantly contributing to empirical understanding, existing research has not quantitatively established the complete causal chain linking mining disturbances, vegetation responses, and subsequent ecological effects, and is limited to a single time-point measure, or small-scale observation or treatment plot, without emphasizing measuring multi-scale response or long-term observation of related vegetation change. In ecosystems especially fragile and harsh climate such as arid and semi-arid grassland mining areas in eastern Inner Mongolia, vegetation response to mining disturbance is important in understanding local ecosystems and the ecological impacts of mining activities, as they are sensitive to disturbance. This means that vegetation can serve as an indicator of ecosystem health, and an entry point for exploration of the ecological effects of an industrial brownfield. This research aims to, through an in-depth analysis of the interaction mechanisms between mining activities and vegetation cover, establish a scientific foundation for ecological restoration in mining areas, while offering theoretical and practical support for sustainable brownfield redevelopment.

Methods This research analyzes vegetation cover dynamics in the Shengli Mining Area (1992 − 2022). First, vegetation cover is calculated using the pixel dichotomy model, and temporal trends are assessed via the Sen slope estimator and Mann-Kendall test. Results are classified into five change categories: Significant degradation, mild degradation, no significant change, mild improvement, and significant improvement. The spatial distribution pattern of each category is analyzed to identify potential causes. In addition, Spearman’s correlation analysis is used to screen the possible drivers of vegetation cover in the research area, and the screened drivers include temperature, precipitation, geographic elevation and grazing intensity, based on which a driver model between drivers and vegetation cover is constructed using the MGWR-XGBoost fusion model in combibnation with the drivers and vegetation cover data from the non-exploitation period (1992 − 2003). Expected vegetation cover under the influence of the drivers over the mining period (2004 − 2022) is predicted using the developed model, with mining impacts during this period quantified as residuals between predicted and observed values, allowing for a comprehensive analysis of spatial distribution patterns, cumulative effects over time, and stage-specific changes. The research innovatively combines the spatial characterization of MGWR with the machine learning capabilities of XGBoost to effectively isolate the impacts of mining from the effects of natural factors, providing a robust quantitative framework for assessing the mechanisms of vegetation disturbance in mining-affected ecosystems.

Results This research has three significant findings regarding the ecological impacts of mining in the Shengli Mining Area. 1) Extended mining operations have caused significant vegetation degradation, with areas affected by degradation increasing 20.95% during mining operations compared to the period under research (1992 − 2022), which is important for characterizing a significant expansion of ecological disturbance from mining activities. 2) The MGWR-XGBoost fusion model indicates strong predictive capabilities for Vegetation Cover. The model has a coefficient of determination (R²) of 0.81, a mean absolute error (MAE) of 0.058 and a root mean square error (RMSE) of 0.077, and thus can be deemed as a reliable model capable of accurately capturing the relationships between vegetation and drivers, and enhancing the reliability of research results. 3) The spatial pattern of mining vegetation disturbance shows a “core − edge” dispersion pattern. Direct disturbance extends roughly 1 km from the mining area. Potential degradation trend zones show degradation over 3 − 4 km extending outwards from the mine area. Temporally, prolonged mining may directly affect cumulative ecological impacts over time, thus sustaining long-term vegetation stress. Spatially, concurrent mining impacts at neighbouring mines may lead to amplified degradation trends within their overlapping impact zones.

Conclusion This research demonstrates that open-pit mining in the Shengli Mining Area causes significant cumulative vegetation degradation, both temporally and spatially. To mitigate these impacts, future mining operations should adopt optimized spatial planning and phased extraction strategies. The developed MGWR-XGBoost fusion model shows satisfactory predictive accuracy for vegetation cover (R² = 0.81), although further improvements could be achieved through enhanced factor screening and refined model fusion techniques. Crucially, brownfield remediation efforts must prioritize spatiotemporal hotspots (areas with prolonged mining exposure and overlapping impact zones), where vegetation degradation is most severe. These findings may provide practical guidance for sustainable mining management and targeted ecological restoration in arid and semi-arid regions.