Optimizing Parameter Estimation to Improve Evapotranspiration Calculation for Maize Fields

• Published in: Guan'gai paishui xuebao
• Main Authors: QIU Zhongqi, ZHOU Linlin, LIU Hongjuan, TIAN Qianglong, ZHAO Zijing, ZHANG Xiaomei, WEI Guoxiao
• Format: Article
• Language: Chinese
• Published: Science Press 2022-01-01
• Subjects: parameter optimization; evapotranspiration model; markov chain monte carlo
• Online Access: https://www.ggpsxb.com/jgpxxben/ch/reader/view_abstract.aspx?file_no=20220105&flag=1

【Objective】 Evapotranspiration is an important factor in agricultural water management, but its calculation is not trial, especially in areas lacking weather stations where measured meteorological data are incomplete or unavailable. The purpose of this paper is to propose an optimal method to estimate parameters which cannot be measured directly but required for estimating evapotranspiration. 【Method】 The analysis was based meteorological data measured from weather stations at Daman in the basin of Hei River We took corn fields in the basin as an example and assumed latent heat flux and sensible heat flux were the parameters. Differential evolution adaptive algorithms were compared to optimize the parameters in the evapotranspiration model by introducing an energy-unclosed-factor to the multi-objective function in the parameter estimation. The model was built on the Bayesian inference with the values of the parameters calculated by the Markov chain Markov chain Monte Carlo method. Based on traditional indexes including coefficient of determination (R2), linear regression slope, root mean square error (RMSE), consistency index (IA) and Nash coefficient (NSE), we evaluated the model against the original Shuttleworth-Wallace model. 【Result】 We separated the comparison into two phases: a calibration phase and a prediction phase. Comparing with the original Shuttleworth Wallace model showed the optimized model reduced the root mean square error by 52.46% and increased the consistency index by 17.3% in the calibration phase; the optimized model also improved the Nash coefficient of the latent heat flux to 0.82, though it did not show significant improvement over the original Shuttleworth Wallace model for estimating the sensible heat flux. For prediction, the optimized model reduced the root mean square error by 50.51% and increased the consistency index by 14.46%, compared with the original Shuttleworth Wallace model. The proposed model improved the Nash coefficient of the latent flux to 0.80, but it did not show significant difference from the original Shuttleworth Wallace model in other evaluation indexes. 【Conclusion】 We proposed a model to estimate parameters which are required for calculating evapotranspiration but cannot be measured or are missing. Tests against measured latent heat flux and sensible heat flux showed that the proposed model was superior to existing models for estimating the two parameters.