Contributions to Satellite-Based Land Cover Classification, Vegetation Quantification and Grassland Monitoring in Central Asian Highlands Using Sentinel-2 and MODIS Data

The peripheral setting of cold drylands in Asian mountains makes remote sensing tools essential for respective monitoring. However, low vegetation cover and a lack meteorological stations lead to uncertainties modeling, obstruct uncovering driving degradation factors. We therefore analyzed the importance promising variables, including soil-adjusted indices high-resolution snow metrics, quantification classification Afghanistan’s Wakhan region using Sentinel-2 field data with random forest algorithm. To increase insights on remotely derived climate proxies, we incorporated temporal correlation analysis MODIS (NDSI) compared measured MODIS-NDVI anomalies. Repeated spatial cross-validation showed good performance (80–81% overall accuracy) foliar model ( R 2 0.77–0.8, RMSE 11.23–12.85). Omitting approach led positive evaluation bias 0.1 accuracy 25% models, demonstrating that studies not considering structure environmental must be treated caution. 500-repeated Boruta-algorithm highlighted MSACRI, MSAVI, NDVI short-wave infrared Band-12 as most important variables. This indicates that, complementary traditional indices, variables are modeling grasslands. Snow also high but they did improve models. Single-variable which were restricted areas very (<20%), resulted poor prediction better Our provides evidence proxies by showing highly significant correlations spring during 2001–2020 (Pearson’s r 0.68) 2006, 2007, 2016 2018 (R 0.3). Strong differences visible higher alpine grasslands (MODIS NDVI: 0.72, data: 0.74) other regions lowest riparian thereby show new monitoring approaches grassland dynamics enable development sustainable management strategies, mitigation threats affecting Central Asia.