Species‐Level Vegetation Classification to Assess Mining Impact on Arid and Groundwater‐Dependent Ecosystems

ABSTRACT

Mining activities in arid regions cause significant disturbances to the groundwater system, which can adversely affect the growth, distribution, and composition of groundwater‐dependent vegetation (GDV). As a result, remote sensing techniques for monitoring vegetation dynamics offer an effective method for assessing groundwater fluctuations induced by mining in such ecosystems. To enhance the classification accuracy of GDV, this study introduces a Species‐level UAV‐Satellite Synergistic Classification (SUSSC) framework for vegetation species identification and spatial feature analysis. This method follows a two‐stage framework, first establishing a large‐scale sample library via UAV multispectral imagery coupled with the enhanced YOLOv8n model to provide ample training labels, and second conducting species‐level classification of GDV in mining‐disturbed areas by synergistically utilizing WorldView‐2 satellite data. The findings indicate that: (1) Species training labels extracted from UAV imagery effectively augmented the ground‐truth samples and substantially improved the classification performance based on WorldView‐2 data, resulting in a 20.1% increase in overall accuracy. (2) The SUSSC method enables accurate identification of species‐level GDV, including Caragana microphylla , Haloxylon ammodendron , Achnatherum splendens , Hippophae rhamnoides , and Populus tomentosa Carr , achieving an overall accuracy of 89.6% and a Kappa coefficient of 0.870, with classification accuracies for individual species exceeding 82%. (3) The results reveal a significant relationship between the spatial distribution of GDV with different root‐depth characteristics and groundwater levels. These findings indicate that GDV distribution can serve as an effective indicator of groundwater fluctuations in mining areas, offering new insights for ecological monitoring and groundwater risk assessment.