The Effects of Soil Water on Accuracy of Different Methods for Calculating Evapotranspiration from Winter Wheat Field

Abstract

【Objective】 Evapotranspiration from farmlands is an important component in the hydrological cycle, and it varies with many factors. The aim of this paper is to present an experimental study on the impact of soil water on reliability of different methods commonly used in the literature to calculate evapotranspiration from cropped fields. 【Method】 The experiment was conducted using large lysimeters with winter wheat used as the model plant. It consisted of three soil water treatments, achieved by criteria for irrigation: Resuming irrigation whenever the soil water content in the root zone dropped to 70% (T70), 60% (T60) and 50% (T50) of the field water capacity, respectively. The evapotranspiration in each treatment was calculated by methods using a single crop coefficient, dual crop coefficient, and the BP artificial neural network, respectively. 【Result】 With an increase in water stress due to the reduced irrigation, the total evapotranspiration and seasonable evapotranspiration at different growth stages decreased. Comparison with measured results from the lysimeters showed that the accuracy of the three models varied with soil water content. Under moderate water stress (T50), only did the method using dual crop coefficient reproduce the measured evapotranspiration reasonably well with NSE=0.646 and RSR=0.599. Under mild water stress (T60), the BP artificial neural network model worked better with NSE=0.872 and RSR=0.360, followed by the dual crop coefficient model with NSE=0.729, RSR=0.523. When there was a limited or without water stress (T70), all three methods accurately reproduced the measured evapotranspiration. 【Conclusion】 On average, the model using dual crop coefficient is more suitable for estimating evapotranspiration of winter wheat grown in soils under different water stress.

Keywords

• winter wheat
• large lysimeters
• water stress
• evapotranspiration
• crop coefficient