Assessing the spatiotemporal dynamics of seasonal and perennial surface water resources across Lesotho’s agroecological zones

Surface water resources are crucial for agricultural productivity and rural livelihoods, particularly in water-scarce regions such as Sub-Saharan Africa. In Lesotho, understanding the dynamics of seasonal and perennial water bodies is vital for informed water resource management and policy development. This study evaluates spectral indices for mapping and analyzing the spatiotemporal dynamics of surface water across different agroecological zones (AEZs) in Lesotho from 2016 to 2024 water years. Using harmonized Sentinel imagery integrated into a Random Forest machine-learning framework, we applied a range of water, vegetation, and soil indices to map surface water monthly and distinguish between seasonal and perennial water surfaces. Our findings reveal that the water ratio index was the most effective for mapping surface water across AEZs, outperforming others in distinguishing water from rangeland, cropland, and bare soil. Additional indices further improved water delineation in specific AEZs. Although no significant differences in classification accuracy were observed across AEZs (p > 0.05), visual inspection revealed misclassifications, mainly false positives, which could lead to overestimates of water area. Surface water trends vary regionally, with a significant increase in perennial water in the Foothills and Mountains, while seasonal water shows a non-significant decline, indicating divergent hydrological trajectories. These findings underscore the need for region-specific assessments and management strategies to address the evolving hydrological regimes. Our study provides a scalable framework for water resource assessment applicable beyond Lesotho, with significant implications for addressing water scarcity and guiding policies on water storage, climate-smart agriculture, and community-based governance in Sub-Saharan Africa.