Spatio-temporal analysis of coal mining impacts on land cover using remote sensing: a case study of Northern...

Open-cast coal mining can alter land cover rapidly, yet annual assessments may overlook the episodic nature of disturbance and partial vegetation recovery. This study examined decadal land-cover dynamics within the Northern Coalfields Limited operational landscape of the Singrauli-Sonbhadra coal belt, India, from 2016 to 2025. Temporally consistent January Landsat 8-9 Collection 2 Level-2 Surface Reflectance imagery was classified into four NDVI-based land-cover classes: barren/mining, sparse vegetation, moderate vegetation, and dense vegetation. Annual class distributions were evaluated using the Mann-Kendall test and Sen's slope estimator, while a pixel-based 2016-2025 transition matrix quantified gross endpoint transitions. SRTM-derived slope data provided additional topographic context. Across the common mapped domain of approximately 322.17 km², the barren/mining class increased from 114.28 km² in 2016 to 158.05 km² in 2025, equivalent to an increase of 43.77 km² or 13.59 percentage points of landscape coverage. None of the four classes showed a statistically significant monotonic trend (p > 0.05). The largest off-diagonal endpoint transition was from sparse vegetation to barren/mining land (59.80 km²), compared with 16.20 km² in the reverse direction. Slope analysis showed systematic differences in the terrain distribution of land-cover classes, although the largest absolute barren/mining extent occurred on near-level ground. The findings demonstrate the value of combining annual NDVI mapping, non-parametric trend analysis, transition matrices, and terrain context for long-term monitoring of mining-affected landscapes.