Introduce canopy temperature to evaluate actual evapotranspiration of green peppers using optimized ENN models

Abstract Canopy temperature (Tc) was introduced using the Elman Neural Network (ENN) optimized by Mind Evolutionary Algorithm (MEA) and Genetic Algorithm (GA) methods to improve the predicting accuracy of actual evapotranspiration (ET) of green peppers. The improving planting methods, rainwater harvesting partial plastic film mulching and regulated deficit irrigation (MFR-RDI) with straw mulling, had been employed to save irrigation water. The results indicated that MEA-Elman1 (RMSE = 0.349 mm/d, MAE = 0.303 mm/d, NS = 0.951) and GA-Elman1 (RMSE = 0.487 mm/d, MAE = 0.404 mm/d, NS = 0.913) models can accurately estimate the ET of green peppers, indicating that the Tc can well improve the accuracy of both MEA-Elman and GA-Elman models in predicting ET under the condition of MFR-RDI. Furthermore, the results showed that the performance of MEA-Elman model was better than GA-Elman model under the same factors. Based on the above results, different input factors in different growth periods of crops has been adopted using the MEA-Elman4 model to further improve the predicting accuracy of ET. Indeed, the MEA-Elman4 model exerted the best performance (RMSE = 0.339 mm/d, MAE = 0.285 mm/d, NS = 0.954), which should be given priority to predict the ET of green peppers.