Application of six canopy resistance models for estimating winter wheat evapotranspiration

Abstract: Evapotranspiration (ET) is the main way of water loss in the farmland water cycle in precision agriculture. An accurate ET estimation is of great significance for water-saving irrigation. Extensive studies have been carried out on ET measurement and simulation in recent years. Among them, the single source Penman-Monteith (P-M) model has been one of the most commonly-used models. But, a major challenge still remained on the accurate parameterization of canopy resistance in the P-M. In this study, six commonly-used canopy resistance models were selected to test the simulated canopy resistance with the P-M for the direct estimation of winter wheat ET. The P-M simulated ET was compared with the measured values of Bowen ratio over the wheat canopy in Shunyi District, Beijing for two years (2020 and 2021). The results showed that: 1) The six models underestimated the canopy resistance of winter wheat, but overestimated ET with the original parameters. Among them, the Todorovic model (TD) performed the best, where the R2 between the measured and simulated for the canopy resistance and ET were above 0.605, the Mean Bias Error (MBE) of -82.8 s/m and 10.4 W/m2, respectively, with the Root Mean Square Error (RMSE) of 254.4 s/m and 33.5 W/m2. The coupled canopy resistance model (CO model) performed the worst, where the R2 for the canopy resistance and ET were 0.113 and 0.046, respectively, with the MBE of -236.4 s/m and 97.4 W/m2, as well as the corresponding RMSE of 373.8 s/m and 147.9 W/m2, respectively. The RMSE was ranked as a decreased order by TD, FAO56-PM, Katerji Perrier (KP), Garc?á - Santos (GA), Jarvis (JA), and CO. 2) Further parameter calibration with the data of 2021 and the verification with data of 2020 showed that there was greatly improved performance in the simulated wheat canopy resistance and ET by JA, CO, GA, KP, and FAO56-PM. Except the JA model underestimated the wheat canopy resistance, the others overestimated the canopy resistance and underestimated the wheat ET. Specifically, the KP obtained the best prediction, where the R2 values for the canopy resistance and ET were larger than 0.907, with the respective MBE of 41.1 s/m and -14.7 W/m2, and the corresponding RMSE of 94.0 s/m and 21.5 W/m2, respectively. The performance of the other five models was also feasible, where the R2 values for the simulated wheat canopy resistance and ET were larger than 0.641, the MBE ranging -25.4-24.0 s/m and -11.7-7.2 W/m2, respectively, with the RMSE of 76.8-265.2 s/m and 22.2-26.4 W/m2. The performance was then ranked as a decreased order by KP, GA, TD, FAO56-PM, CO and JA. 3) All six models can be used to predict the rs that are needed by the P-M model for the estimation of the winter wheat evapotranspiration. Nevertheless, the TD model obtained the best wheat rs and ET estimates, even without the local calibration. Thus the TD model can be the first choice in the data-lacking environment. The KP model also needed only a few parameters for the highest accuracy after calibration, particularly for the data-sufficient environment. The findings can provide a better guide application of the P-M's one-step approach to estimate the winter wheat ET in the region of North China.