Uncertainty and performance of remote sensing-based irrigation water use estimates

Agriculture accounts for nearly 70% of global freshwater consumption; however, reliable regional-scale data on irrigation water use (IWU) remain scarce, limiting the design of effective agricultural water management strategies. This study quantifies regional IWU by developing a soil water balance model to simulate volumetric soil moisture in the root zone (RZSM) and actual evapotranspiration (ET) under non-irrigated conditions. To account for the simultaneous influence of irrigation on both RZSM and ET, residuals between irrigation-free model simulations and corresponding satellite-based observations of RZSM and ET were leveraged to retrieve irrigation signals. The analysis further evaluates the influence of satellite data spatial resolution on IWU estimation accuracy by using 30-m OpenET, 500-m MOD16, and 500-m PML ET data. The framework was applied to croplands in Arizona for the year 2017, when statewide irrigation data were available. The results showed strong agreement between modeled and remotely sensed RZSM during non-growing periods (RMSE = 0.011 m3/m3; R2 = 0.68). Annual IWU estimates derived from OpenET closely match benchmark values, exhibiting a moderate negative bias ranging from −16% to −23%, depending on irrigation efficiency assumptions. In contrast, the coarser-resolution MOD16 and PML products substantially underestimate IWU, with biases of −93% to −94% and −55% to −60%, respectively. More than 97% of IWU was attributed to ET residuals rather than soil moisture signals, highlighting the dominant role of high-resolution ET data. Uncertainty analysis indicated a coefficient of variation of 0.25 in IWU estimates, primarily driven by errors in meteorological inputs and remote sensing observations.