COMPARING UAV AND SATELLITES ACCURACY IN SIMULATING FINE-SCALE ALPINE GRASSLAND HETEROGENEITY IN PERIGLACIAL ENVIRONMENT

Coarse resolution imagery poses significant monitoring challenges in fragile alpine ecosystem.Thus, addressing underlying surface heterogeneity offers a promising approach to enhance the accuracy of alpine grassland characteristics retrieval.This study investigates the potential of fixed-wing unmanned aerial vehicle (UAV) for large-area monitoring of alpine grassland characteristics in high-altitude and low-temperature permafrost environments, where fine-scale heterogeneity poses challenges for traditional remote sensing techniques.By integrating multispectral imaging and Structure-from-Motion (SfM) photogrammetry, the vegetation index (VIs), fraction vegetation cover (FVC) and vegetation type maps were generated in the Wudaoliang region of the central Qinghai-Tibet Plateau (QTP).We also qualitatively compared the UAV-based vegetation information at sub-meter spatial resolution with that provided by satellites (including Sentinel-2, Landsat-8 and MODIS) at coarser resolutions.The results show that, compared with the other four VIs, including Kernel Normalized Difference Vegetation Index (kNDVI), Enhanced Vegetation Index (EVI), Soil-Adjusted Vegetation Index (SAVI) and Excess Green Index (ExG), the calibrated UAV-based Normalized Difference Vegetation Index (NDVI) achieved the best accuracy for alpine grassland FVC estimation, significantly outperforming in detecting low-stature alpine grassland.UAV imagery unveiled pronounced fine-scale heterogeneity in both FVC and alpine grassland types, strongly correlated with periglacial landforms and microtopography, while these patterns are largely invisible to satellites.Satellites data underestimates the spatial heterogeneity within alpine grasslands, which may lead to inadequate monitoring of fine-scale ecological variations.Our results provide a detailed assessment of spatial heterogeneity within alpine grassland in high-altitude periglacial environments, which offers valuable insights for ecological models predicting alpine variation and permafrost stability feedback.