Abstract

Soil degradation threatens global food security and environmental sustainability, affecting approximately 33% of Earth's land surface. This study presents an innovative framework for digital mapping of soil degradation hotspots by integrating multi-temporal satellite imagery, machine learning algorithms, and field-based degradation indicators. We analyzed a 2,500 km² semi-arid region in Sub-Saharan Africa experiencing severe degradation pressures from 2015-2023. The methodology combined Sentinel-2 and Landsat-8 time series with terrain attributes, climate data, and land use information to map five degradation types: erosion, salinization, nutrient depletion, compaction, and organic matter decline. A Random Forest classifier achieved overall accuracy of 87.3% for degradation type identification, while regression models predicted degradation severity with R² values ranging from 0.82-0.91. Hotspot analysis revealed that 34% of the study area experienced moderate to severe degradation, with erosion affecting 18% of agricultural lands. Multi-temporal analysis identified degradation trajectories, showing 12% of areas transitioning from slight to moderate degradation over the study period. The framework provides actionable information for targeted intervention strategies, supporting land degradation neutrality goals. Integration of uncertainty mapping enabled confidence-based prioritization of restoration efforts, optimizing resource allocation for maximum impact.