Israelzemedkun/gidabo-basin-cldi-analysis

Machine learning-based land degradation assessment for the Gidabo Basin, Ethiopia. # Gidabo Basin Land Degradation Monitor ## Summary This project investigates spectral characterization of land degradation in Ethiopia's Gidabo River Basin using Landsat 5 (2000) and Landsat 8 (2024) surface reflectance data. A Combined Land Degradation Index (CLDI) classifies 500 sampled pixels; a Random Forest classifier achieves 95% cross-validated accuracy. Label validity is partially supported by 94.9% consistency with ESA WorldCover 2021. ## Live Dashboard Interactive app: gidabo-basin-cldi-analysis-… Features: spatial degradation map, zone filters, CLDI distribution, degradation risk predictor, and CSV export. ## Interactive Tools - **Streamlit Dashboard** (analysis results): gidabo-basin-cldi-analysis-… - **Gidabo Monitor** (interactive RUSLE risk assessment for Ethiopia): gidabo-monitor-ko6e.vercel.… ## Research Question This project investigates whether land degradation status in the Gidabo River Basin, Ethiopia can be reliably characterized using multi-temporal Landsat surface reflectance indices (NDVI, BSI, SI) and classified using a machine learning approach, without requiring field-collected ground truth data. ## Background The Gidabo River Basin, located in the Ethiopian Rift Valley south of Lake Abijata-Shalla, drains approximately 3,500 sq km of highland and rift-floor terrain. It supports a dense smallholder farming population relying primarily on rain-fed cereal cultivation, coffee agroforestry, and irrigated horticulture along the valley floor. Over the past two decades, expanding cropland, woodland clearance, and population-driven overgrazing have accelerated soil exposure and surface salinisation in the lower rift, while upper catchment forests have come under increasing charcoal and timber pressure. Reliable, spatially explicit monitoring of land degradation is essential for targeting soil and water conservation investments, advising f …