Abstract
Uzbekistan’s forest ecosystems, which cover approximately 6.7% of the national territory, face increasing threats from overgrazing, unsustainable land use, and climate-related adversities. Accurate mapping and monitoring are essential for the sustainable management of these vital resources, particularly in diverse terrains like the Dekhkanabad forestry region in the Gissar Range. This study employs an Object-Based Image Analysis (OBIA) approach to classify land cover by integrating multi-source remote sensing data, including Sentinel-2 and Kompsat-3 imagery, complemented by an ALOS PALSAR-derived Digital Terrain Model (DTM).
To ensure rigorous validation, ground truth data were collected from 1,980 plots using the FAO’s Collect Earth tool, categorized according to IPCC and Forest Resources Assessment (FRA) guidelines. The methodology utilized multi-resolution segmentation and supervised classification to overcome the limitations of traditional pixel-based methods, specifically addressing spectral overlaps in complex mountainous landscapes.
The results demonstrate high classification accuracy for dominant land cover types: pasture (91%), cropland (89%), and forest (80%). However, challenges remained for human settlements (53%) and wetlands due to spectral similarities and smaller sample sizes. The findings underscore the effectiveness of the OBIA framework and high-resolution satellite fusion in providing precise thematic maps. These spatial datasets offer critical insights for policymakers and forestry managers in Uzbekistan to support biodiversity conservation, resource allocation, and long-term environmental planning.