Spatial and temporal sensitivity of water footprint assessment in crop production to modelling inputs and parameters

Abstract

Owing to increasingly limited water resources, evaluating large-scale regional crop water footprints (WFs) within acceptable errors is the basis for efficiency and productivity improvements for sustainable agriculture. However, quantitative sensitivity analysis of outputs to inputs and parameters of advanced and widely used crop WF accounting models are still lacking. In this study, we evaluated the spatial and temporal sensitivity of the blue and green WFs of crops to key input variables based on year-by-year simulations for wheat, rice, maize, and soybean production in China from 2010 to 2019. We employed one of the most widely used AquaCrop models to test different water supply and irrigation scenarios. Nine input variables or model parameters were selected, including reference evapotranspiration (ETo), crop transpiration coefficient (KcTr), soil evaporation coefficient (KE), maximum canopy cover (MCC), precipitation (PR), canopy decline coefficient (CDC), planting density of the crop (DC), reference harvest index (HI0), and normalised water productivity (WP*). The results showed that crop WFs are generally more sensitive to ETo among the input variables and KcTr and KE among the model parameters. We also found that the blue WF was more sensitive than the green WF. In most cases, WFs in rainfed scenarios were more sensitive to changes in input variables compared to WFs of irrigated crops. Of the four studied crops, the WFs of maize and wheat were more sensitive than the WFs of rice and soybean. Wheat, maize, and soybean also showed a visible spatial variation in the sensitivity of the WF to each input variable. Overall, the WF sensitivity was influenced by crop type, water supply, irrigation method, and regional climate features. The current analysis provides a feasible approach to comprehensively identify the key inputs and parameters to be calibrated for accurate crop WF accounting.