REMOTE SENSING-BASED APPROACH TO EVALUATE ACTUAL SURFACE WATER USE FOR IRRIGATION IN BILATE WATERSHED, RIFT VALLEY LAKES BASIN, ETHIOPIA

<br> ABSTRACT<br> The Bilate watershed, situated within Ethiopia's Rift Valley lakes basin, confronts<br> escalating water scarcity driven by rapid irrigation expansion aimed at bolstering food<br> production. This expansion has significantly altered river flows and compromised ecosystem<br> integrity, particularly during irrigation seasons. Accurate estimation of irrigation water<br> abstraction is paramount for sustainable water resource management, enhanced agricultural<br> productivity, environmental preservation, and conflict mitigation. This study addresses critical<br> data gaps by leveraging high-resolution remote sensing (RS) imagery to quantify irrigation<br> water use, while acknowledging the inherent challenges in validating RS-derived products<br> across diverse climatic and landscape conditions. This research focused on evaluating the<br> accuracy of land use/land cover (LULC) classification, irrigated area mapping, and actual<br> evapotranspiration (AET) estimation, all crucial components of irrigation water use<br> assessment. Landsat 8's blend of spectral capabilities, spatial and temporal resolution, extensive<br> historical archive, free access, and reliable imagery make it an outstanding option for LULC<br> classification and monitoring. Yielding nine dominant classes and demonstrating seasonal<br> variations in accuracy, with overall accuracies ranging from 80% to 90% and Kappa<br> coefficients from 0.8 to 0.9. Irrigated areas were mapped using Sentinel-1, uses SAR<br> technology operate in all weather conditions. It provides high-resolution imagery, allowing<br> detailed and frequent monitoring of irrigated areas. Its dual polarization (VV and VH) enhances<br> the accuracy of irrigation mapping. Advanced classification algorithms, Random Forest, were<br> applied, resulting in accuracies of 88% for the Bilate watershed and 87% for the Gumara<br> watershed, with Kappa coefficients of 0.74 and 0.73. The two watersheds were selected due to<br> their similar emergence of small-scale irrigation and contrasting rainfall and land cover<br> characteristics. These maps revealed significant discrepancies with global irrigation datasets,<br> underscoring the necessity for localized validation. In this study, a thorough evaluation of five<br> satellite-derived AET products—including national (FAO WaPOR), continental<br> (modisSSEBop), and global (Climate Engine, modisSSEBop, and Synthesis) datasets—was<br> conducted. The comparison against ground-based reference evapotranspiration showed weak<br> correlations (&lt; 0.35) and significant differences in magnitude. Notably, FAO WaPOR<br> displayed the most consistent agreement with the other products. A spatiotemporal analysis of<br> irrigation water withdrawal and availability from 2017 to 2022 was conducted utilizing Google<br> Earth Engine (GEE), HEC-HMS, and FAO-CROPWAT. The high-resolution imagery from<br> Sentinel-2 within GEE allows for efficient monitoring of actual irrigated areas. With frequent<br> xxi<br> revisits and multispectral data, along with its free availability, it is ideal for accurately mapping<br> and tracking changes in small-scale irrigation systems. FAO-CROPWAT was used to<br> determine net and gross irrigation water requirements (IWR), considering climatic data, crop<br> parameters, and soil properties. The HEC-HMS model, calibrated with hydro-meteorological<br> data from the Ethiopian Meteorology Institute and Ministry of Water and Energy, simulated<br> streamflow with high accuracy (Nash-Sutcliffe Efficiency ranging from 0.91 to 0.96). This<br> study provides valuable insights for optimizing irrigation water management in the Bilate<br> watershed. The observed seasonal variability in LULC classification accuracy highlights the<br> importance of selecting appropriate temporal data for analysis. The identified spatial patterns<br> of IWR across sub-watersheds underscore the need for targeted water management strategies<br> and the promotion of water conservation practices, such as rainwater harvesting. By providing<br> an accurate assessment of irrigation water withdrawal through a robust RS-based approach,<br> this research aims to enhance water use efficiency, minimize water losses, and support<br> sustainable water resource management initiatives in the region.<br>