Evaluation and verification of two evapotranspiration models based on precision screening and partitioning of field temperature data

Field surface temperature are essential input of an evapotranspiration (ET) model. Crop canopy and soil surface temperatures (Tc and Ts, respectively) are often mixed in the field surface temperature at early stages, due to changes in crop growth, and row and plant spacing. To ensure the accuracy of ET estimation, the Tc and Ts at actual positions included in the field temperatures should be verified and optimized. In this study, the underlying field surface temperature data were determined from thermal infrared sensors through an automatic monitoring system. These were screened and partitioned into actual Tc and Ts, with an algorithm based on maize growth conditions. The Tc replaced the average temperature of field surface in the Seguin-Itier (S-I) model, and the unified air temperature in Shuttleworth-Wallace (S-W) model was substituted by the air temperature (Ta), Tc, and Ts at actual positions. Field ET estimated by the Penman-Monteith model and the Soil Water Balance model were also analyzed, based on synchronous monitoring data of maize for comparison. Experimental data was collected over two years for three record periods, within three areas in China. These included ZhongWei in Ningxia Autonomous Region (2019–2020), ChangChun in Jilin province (2018–2019), and JieFangZha in the Inner Mongolia Autonomous Region (2015–2016). Results showed that the accuracy of estimating ET in the S-I model improved, subsequent screening the temperature data, with the average coefficient of determination and index of agreement increased by 0.13 and 0.08 respectively, and the average of the root mean square error and relative error decreased by 0.10 mm d- 1 and 2.5% respectively. The key parameters a and b in S-I model calibrated in early growth stage can be an alternative over application in whole season, which can simplify calculations while maintaining accuracy. The average values of a and b in the early stage and whole season were a= 0.408/b= −0.455 and a= 0.856/b= −0.498 in ZhongWei, a= 0.898/b= −0.913 and a= 1.039/b= −0.936 in ChangChun, and a= 0.736/b= −0.328 and a= 0.672/b= −0.310 in JieFangZha, respectively. Soil evaporation and canopy transpiration were estimated more precisely by the Revised S-W model through field temperature data partitioning. The ET estimated by the Revised S-W model is in line with the results from Penman-Monteith model and Soil Water Balance model. Moreover, this study demonstrated that estimating ET (by S-W or RS-W model) is easily influenced by weather conditions. This research provides valuable insight into determination and application of the estimation of ET using different models in different regions, and can be used in precision irrigation management.