Elevation-dependent divergence of vegetation dynamics over the Anyemaqen Mountain revealed by reconstructed Sentinel-2 time series

Reliable monitoring of alpine headwater vegetation is constrained by persistent cloud cover, seasonal snow, and complex terrain. This study developed an indicator-oriented framework for the Anyemaqen Mountain in the Yellow River Headwaters using 122 observations of 10-m Sentinel-2 Level-2 A images acquired from 2015 to 2025. After quality screening, continuous Normalized Difference Vegetation Index (NDVI) and Normalized Difference Snow Index (NDSI) trajectories were reconstructed using temporal linear interpolation, spatial neighbourhood filling, and climatological mean replacement. Vegetation and snow-cover dynamics were then evaluated using the Theil–Sen trend estimation, Mann–Kendall test, pixel-wise NDVI–NDSI correlation, three elevation bands (< 3800, 3800–4600, and > 4600 m), and geographically weighted regression with elevation, slope, and aspect as explanatory variables. The results show that significant greening occurred across 7.2% of the study area and was concentrated mainly at high elevations, whereas significant browning affected only 0.1% but was localised in low-elevation valleys and riparian corridors. The remaining 92.7% showed no significant trend. Within the analysed snow-covered domain, 8.3% of pixels showed significant NDSI decline, while pixel-wise NDVI–NDSI correlations were spatially mixed, with a spatial mean r of 0.003; none remained significant after false-discovery-rate correction. The geographically weighted model represented spatially varying topographic controls, with local R 2 ranging from 0.001 to 0.752 and a median of 0.162. Together, these results define an elevation-dependent vegetation dipole characterised by significant high-elevation greening and localised low-valley browning. The reconstructed high-resolution framework provides a fine-scale ecological indicator for elevation-specific monitoring and management in alpine headwaters.