Figure 1 Study area.
Prior to the commencement of underground mining, the study area was characterized by a diverse distribution of species and some grazing vegetation communities, consisting mostly of sandy plant communities with perennials and shrubs (Liu et al., 2020; Zhang et al., 2022). The dominant trees were poplar, and the dominant shrubs were Salixcheilophila , Artemisia desertorum , and Caragana korshinskii ; the dominant herbaceous included Stipa capillat., Bothriochloa ischaemum and Setaria viridis . A total of six dominant vegetation species and bare land were selected for this study, including Poplar, Artemisia, Salix, Caragana, Herb, Corn and Bare ground. Grass was not specifically classified since it occurred under tree canopies and shrubs or was very sparsely distributed on bare land, making it challenging for UAVs to detect or identify specific grass species. Therefore, all types of grass were classified as ‘Herb’ without detailed evaluation.

2.2 UAV Surveying

A DJI P4-Multispectral UAV equipped with six 1/12.9 inch CMOS image sensors collected spectral data in five bands (blue, green, red, red edge, and near-infrared). The data collection occurred during the peak growth period of vegetation in the Ordos region, from 10:00-14:00 on September 10, 2020. Data were acquired during three separate flights covering a flight area of 38.57ha. Twelve ground control points were used to ensure accurate geopositioning of the data. Standard reference whiteboard correction with a reflectance value of 50% was performed before each flight to avoid overexposure or underexposure due to cloud changes. Flight planning included capturing photos with a 75% forward and side overlap ratio, evenly spaced intervals, a flight height of 120m, a pixel resolution of 6.4cm, and a speed of 9.8m/s. Pre-processing, including image stitching and radiometric correction, was performed by Agisoft PhotoScan (Figure 2).