Land Use/Cover Dynamics in the Ruhuhu River Basin, Tanzania (1990–2023): A Multi-temporal Satellite Imagery and Machine Learning Analysis

Abstract

This study presents the first multi-decadal analysis of Land Use and Land Cover (LULC) dynamics in the Ruhuhu River Basin, Tanzania, over a 34-year period (1990–2023). Multi-temporal satellite imagery from Landsat 5 TM, Landsat 8 OLI, and Sentinel-2 MSI was classified using the Random Forest algorithm within Google Earth Engine into six LULC categories: water, barren land, forest, grassland, farmland, and shrubland. The classification accuracies were robust (overall accuracy: 81.60–98.70%; Kappa: 80.02–87.40%). The results revealed a sharp decline in forest cover from 56.92% to 28.60%, coupled with significant expansions in farmland (2.99% to 21.40%) and barren land (5.29% to 14.00%), driven by agricultural intensification and coal mining. These transformations have far-reaching implications for the ecological integrity of Lake Nyasa, including disrupted hydrological cycles, biodiversity loss, and reduced ecosystem services. This study provides spatially explicit evidence to support integrated land management and conservation strategies in the Ruhuhu catchment and similar sub-Saharan African river basins.