Spatio-temporal evolution characteristics and driving factors of soil salinization in the Zhundong Mining Area

Open-pit coal mining in arid areas, while disturbing regional hydrological processes, soil physical and chemical environments and vegetation patterns, intensifies the risk of secondary salinization. However, its spatio-temporal evolution characteristics and driving mechanisms are still poorly understood. This paper takes the Zhundong mining area in Xinjiang, an extremely arid region, as the research object. By using Landsat remote sensing images from 2000 to 2024 and 220 measured soil salinity data, and combining methods such as BP neural network, support vector machine and random forest, multiple salinity inversion models are constructed. The sensitive spectral variables were screened through the variable projection Importance method (VIP) to improve the model accuracy, and the driving effects of climate and human activities on the degree of salinization were explored. The results show that: (1) The VIP-RF model performs best in the complex surface background of extremely arid mining areas. (2) Spatially, the central desert and the northern alluvial plain are mainly non-saline or slightly saline soil, the junction zone is mainly moderately saline, and the northwest and southeastern gravelly deserts and Gobi are concentrated areas of severe salinization. (3) In terms of time, mild salinization was dominant from 2000 to 2010. After the mining area was developed in 2010, the areas of moderate and severe salinization increased by approximately 79% and 84% respectively, while the area of non-saline soil decreased by 62.2%. (4) In terms of driving mechanisms, from 2000 to 2010, it was mainly controlled by natural factors such as climate and topography, while from 2010 to 2024, human activities will gradually become dominant. Comprehensive analysis shows that the integration of spectral variable screening and random forest methods can effectively enhance the accuracy of salinization inversion in arid mining areas, and effectively reveal the evolution law of salinization under the interaction of natural and human factors. It has significant reference value for ecological environment monitoring and sustainable utilization of land resources in arid mining areas.