Utility of Downscaled Climate Model Outputs to Support Crop Modeling at the Irrigation District Scale

Climate change impacts crop production through modifications of mean and variability of precipitation (P) and of the climatic variables controlling potential evapotranspiration (ET₀). Depending on the region, these effects may result in higher or lower irrigation needs to support the same level of productivity. Quantifying these impacts at the irrigation district scale is a challenging task that requires high-resolution climate projections to apply crop models. Here, we extensively compare a set of techniques to simultaneously bias correct and downscale P and the variables controlling ET₀ from the grids of general circulation models (GCMs) to the irrigation district scale. We focus on 12 irrigation districts covering an area of ~1,900 km² in central Arizona, which heavily depends on irrigation from limited water resources. We use an ensemble of 17 GCMs from CMIP5 and of 10 GCMs from CMIP6, and climate records from five weather stations as reference. We find that the best technique to reproduce daily P is based on quantile mapping with a two-component parametric distribution combining the generalized Pareto and exponential distributions; empirical quantile mapping is suitable for relative humidity, solar radiation, and wind speed; and variance scaling is appropriate for temperature. After bias correcting the GCMs to the station locations, we generate maps at 500-m resolution-required by the irrigation districts-through a spatial interpolation routine that uses terrain information. We validate the approach via leave-one-out cross validation at the station locations. Finally, we use the downscaled climate model outputs to apply a crop model at the irrigation districts and investigate the impacts of future climate on crop yield and irrigation requirements under the RCP4.5 and RCP8.5 scenarios for CMIP5, and SSP2-4.5 and SSP5-8.5 scenarios for CMIP6.