Spatiotemporal analysis of land use and land cover using random forest in Google Earth engine: A case study of the Grand Ethiopian Renaissance Dam basin and reservoir, Upper Blue Nile, Ethiopia

Abstract

Land use and land cover (LULC) changes have significant positive and negative impacts on the environment. The increasing water demand, the limited water resources, high soil erosion, and dynamic LULC and climate changes of the Grand Ethiopian Renaissance Dam (GERD) basin prompted this study to analyse LULC changes as an initial step toward further research. This study aims to analyse LULC changes in the GERD basin from 1986 to 2024 using Landsat images, as well as the GERD and Aswan High Dam (AHD) reservoir areas during GERD reservoir water filling with Sentinel-2 images (2019–2024). The classification was initially performed at 15 sub-basins using Random Forest machine learning in GEE and then mosaicked using ArcMap GIS. Approximately 39,000 points have been used for training and validation in 1986, 2000, 2014, and 2024 for the entire basin. The overall accuracy and Kappa coefficient for all sub-basins considering the auxiliary data sets were greater than 85% and 0.81, respectively. Eucalyptus trees, built-up areas, water, agricultural land, and bare land expanded by 4505%, 1537%, 56%, 36%, and 9%, while shrubs, forests, grassland, and wetlands shrank by 38%, 27%, 19%, and 10%, respectively, from 1986 to 2024. The GERD reservoir annual water filling based on the ALOS PALSAR 12.5 m Digital Elevation Model (DEM) indicated that the first filling started in 2020 rainy season (560 m, 300 km², 4.82 billion m³), 2021 (580 m, 460 km², 10.5 billion m³), 2022 (600 m, 840 km², 22 billion m³), 2023 (620 m, 1350 km², 43 billion m³), 2024 (637 m, 1765 km², 69.3 billion m³). On 9 September 2025, the full supply level was reached and inaugurated. When the GERD reservoir filled with water, the AHD reservoir's excess water overflowed into the Toshka lakes due to the high flow of the Nile River since 2020. The forest and shrubs in the GERD reservoir area were cleared before filling to prevent greenhouse gas emissions, decayed deposition on dead storage, and water quality issues. Ensemble, coupling, and comparisons of tools, methods, and factors should be considered for better classification accuracy. Effective LULC management that enhances ecosystem services and alleviates water stress on the GERD basin requires cooperation from all responsible stakeholders.