An object-based image analysis approach to assess irrigation-water consumption from MODIS products in Ethiopia

•A remote sensing-based method for per-field irrigation estimates is presented.•GEOBIA combined with the MODIS MOD16A2 allows near-monthly assessments.•Our method enables water-consumption assessment by smallholder irrigation practices.•Method has high potential in other regions using operational available remote sensing products. Efficient water-resource management is essential with regard to food security, growing populations and climate change. This is especially important for low- and middle-income (LMC) countries where food is often locally produced by traditional smallholder farming. Detailed knowledge of the spatio-temporal distribution of irrigation-water consumption provides valuable information to anticipate local food shortages and water scarcity as a result of climate variability. Yet, adequate techniques to quantify irrigation-water consumption at field level over large areas are lacking. Irrigation estimates generally have a coarse resolution making them inadequate for field-level assessments. This study developed a remote-sensing-based approach to quantify spatio-temporal patterns of irrigation-water consumption at field level using the MODIS evapotranspiration product (MOD16A2) and existing land-use maps on the spatio-temporal distribution of irrigated agriculture. Object-based image analysis was used to establish local evapotranspiration differences between irrigated and rainfed fields on a monthly basis, which are the irrigation-water consumption rates of the irrigated fields. This novel method was applied to a study area in the Central Rift Valley in Ethiopia where smallholder farming is dominant and only a few large-scale farms are present. Comparison with irrigation-water-consumption values of a local irrigation scheme showed that the monthly temporal dynamics were captured quite well, but lower values were calculated compared to the scheme's field data. Comparison with two validated remote-sensing based studies in Africa showed good agreement as irrigation-water-consumption estimates were in the same order of magnitude. Irrigation-water consumption follows the temporal rainfall pattern, i.e. irrigation practices intensify with increased water availability. Surface water is commonly used for irrigation in the study area. Our study shows that smallholder practices have a lower irrigation-water consumption compared to modern large-scale farms by approximately a factor 3. Irrigation-water consumption in the area is considerable, especiall