Fields of opportunity: satellite imaging uncovers water stress and potential for water savings from irrigation

In-situ observations provide accurate data on irrigation water use, but their spatiotemporal coverage is limited. This study employed a remote-sensing and machine-learning framework, combined with extensive field data on cropping patterns, to assess water stress and to highlight fields where irrigation water use could be reduced. To this end, actual evapotranspiration (ETa) and its uncertainty were computed from 30-m Landsat imagery using the METRIC method. Additionally, 10-m resolution multi-crop maps were extracted from Sentinel-1 and Sentinel-2 imagery by applying a random forest classifier at different intra-annual periods, based on the crop calendar. Results from the Lake Urmia Basin in northwest Iran demonstrated high crop classification accuracy of 83% to 91% across classification periods. Comparing farm-scale ETa against crop evapotranspiration revealed notable spatiotemporal patterns of water stress during crop development and mid-season, whereas high potential for water savings was observed during the initial and late growth stages.