This study demonstrates that integrating cloud-based remote sensing and machine learning provides a robust and comprehensive framework for monitoring surface water resources in Ethiopia across diverse hydro-climatic and land use and cover (LULC) conditions. Spectral water indices, particularly WI and AWEIsh, prove to be effective for large-scale surface water monitoring, while machine learning approaches, particularly GTB offer high accuracy and valuable long-term spatiotemporal insights when supported by quality training data. The observed increase in surface water extent from 1986 to 2023, coupled with pronounced inter-annual variability, highlights the combined influence of climatic drivers, notably precipitation and temperature, and anthropogenic factors such as LULC change, and dam construction. XGBoost-SHAP-based interpretability further highlights the predominant role of climate alongside the significant, albeit secondary, contribution of LULC transformations to hydrological dynamics. Together, these findings underscore the need for adaptive data-driven water resource management strategies in Ethiopia. These strategies should consider development goals alongside ecosystem sustainability and support climate adaptation. Furthermore, future climate and land use scenarios need to be incorporated to ensure the sustainability and resilience of water resource management.
Verknüpfung zu Publikationen oder weiteren Datensätzen
