Simulation of regional irrigation requirement with SWAT in different agro-climatic zones driven by observed climate and two reanalysis datasets.

Irrigation water is one of the most substantial water uses worldwide. Thus, global simulation studies about water availability and demand typically include irrigation. Nowadays, regional scale is of major interest for water resources management but irrigation lacks attention in many catchment modelling studies. This study evaluated the performance of the agro-hydrological model SWAT (Soil and Water Assessment Tool) for simulating streamflow, evapotranspiration and irrigation in four catchments of different agro-climatic zones at meso-scale (Baitarani/India: Subtropical monsoon; Ilmenau/Germany: Humid; Itata/Chile: Mediterranean; Thubon/Vietnam: Tropical). The models were calibrated well with Kling-Gupta Efficiency (KGE) varying from 0.74-0.89 and percentage bias (PBIAS) from 5.66-6.43%. The simulated irrigation is higher when irrigation is triggered by soil-water deficit compared to plant-water stress. The simulated irrigation scheduling scenarios showed that a significant amount of water can be saved by applying deficit irrigation (25-48%) with a small reduction in annual average crop yield (0-3.3%) in all climatic zones. Many catchments with a high share of irrigated agriculture are located in developing countries with a low availability of input data. For that reason, the application of uncorrected and bias-corrected National Centers for Environmental Prediction (NCEP) and ERA-interim (ERA) reanalysis data was evaluated for all model scenarios. The simulated streamflow under bias-corrected climate variables is close to the observed streamflow with ERA performing better than NCEP. However, the deviation in simulated irrigation between observed and reanalysis climate varies from -25.5-45.3%, whereas the relative irrigation water savings by deficit irrigation could be shown by all climate input data. The overall variability in simulated irrigation requirement depends mainly on the climate input data. Studies about irrigation requirement in data scarce areas must address this in particular when using reanalysis data.