Satellite‐Based Estimates of Irrigation Water Use Versus Modeling Reconstructions: How Much do They Differ?

Abstract

Knowing water amounts destined for agricultural practices, essentially consisting of irrigation water use (IWU), is essential for water resources management. Although explicit and consistent information on IWU is generally lacking worldwide, remote sensing and modeling platforms offer valuable solutions for estimating it. The objective of this work is to perform an intercomparison between satellite-based IWU products and modeling reconstructions over four main irrigation hotspots worldwide: the Contiguous United States (CONUS), the Ebro basin (Spain), India, and the Murray-Darling basin (Australia). Remote-sensing-based IWU estimated through Soil-Moisture-Inversion (SM-Inversion) and SM-Delta approaches are considered. As modeled data, five Global Hydrological Models are evaluated: H08, PCR-GLOBWB, WaterGAP, LPJmL, and QUAlloc. Reference irrigation volumes are exploited to assess which source is more reliable. Over CONUS and Murray-Darling basin, lower discrepancies between data sources (satellite or modeling) are detected. Referring to spatially-averaged IWU magnitudes, the comparison over CONUS provides RMSE ranging from 9.52 to 12.21 mm/month, BIAS between 3.24 and 7.06 mm/month, and Pearson correlation (𝑟 ) from 0.77 to 0.90. The same intervals for the Murray-Darling basin are 8.04–17.70 mm/month for RMSE , 4.81–12.63 mm/month for BIAS , and 0.42–0.85 for 𝑟 . Consequently, over such areas, similar performances in reproducing actual irrigation amounts by satellite estimates and modeling reconstructions, when available, are found. Over the Ebro basin, where the strongest discrepancies are found, satellite-based data sets outperform model outputs in reproducing reference data. The lack of in situ observations for India prevents a clear assessment of the results, but insights on differences introduced by parametrizations are established.