Improved hydrological modelling and ET estimation in watershed with irrigation interference

Abstract

Evapotranspiration (ET) is the true water consumption of the watershed and is key to achieving water conservation. The accurate estimation of ET by distributed hydrological model can provide an important basis for watershed water resources management. At present, the traditional approach calibrating distributed hydrological model parameters by runoff has low accuracy in watersheds under human interference. The reason is that the measured runoff is no longer natural due to various human activities such as river water withdrawals, reservoir regulation and inter-basin water transfer. While using ET for model calibration can effectively overcome this interference, making it a more accurate approach for watersheds under human interference. Therefore, we calibrated the Soil and Water Assessment Tool (SWAT) model using PMLV_2 ET in Ziya River under frequent human activities. The results indicated that calibrating the model with remote sensing ET could improve the model’s simulation performance, increasing the NS from 0.65 to 0.83. However, the annual ET calculated by the model was 21.7% lower than PMLV_2 ET. The largest underestimation was for the annual ET of cropland (agricultural ET), which was underestimated by 24.4%. This is partly because the model did not account for irrigation evaporation from cropland. Therefore we used Landsat 8 data to extract the planting areas of winter wheat, which consumes the highest irrigation water in the Ziya River Basin, from 2013 to 2018. The planting areas and irrigation regimes of winter wheat were incorporated into the SWAT model to calculate the agricultural ET. The results indicate that the agricultural ET calculated by the SWAT model that considers irrigation evaporation is 6.9% lower than that of PMLV_2 ET. The simulation accuracy of ET is significantly improved. This study provides a reference for accurate estimation of agricultural ET and a basis for ET management.