Prediction Model for Reference Crop Evapotranspiration Based on the Back-propagation Algorithm with Limited Factors

Abstract

The accurate estimation of reference crop evapotranspiration (ET O ) is vital for regional water and irrigation water resource management and is bene cial to the rational allocation of regional water resources, and alleviates the disparity between water supply and demand. This study accurately estimates the ET O of 14 meteorological stations in southern China. Five neural network models (backpropagation neural network [BP], extreme learning machine [ELM], ant colony optimization [ACO]-BP, bird swarm algorithm [BSA]-BP, cat swarm optimization [CSO]-BP) were introduced to predict ET O with limited factors using different methods. The results demonstrated that models inputting T (maximum, minimum and average air temperature), sunshine duration (n) and relative humidity (RH) exhibited the highest accuracy of all studied combinations; the role of T, n, RH, wind speed (U 2 ) and average atmospheric pressure (AP) in relation to ET O gradually decreased. All the three biological heuristic algorithms effectively improved the performance of the BP model. The accuracy and computational cost of the CSO-BP model are better than those built by other algorithms. Therefore, it is strongly recommended to use the CSO-BP model for ET O prediction in southern China. This result provided a reference for the more accurate prediction of ET O for future irrigation decision-making and water resource management in southern China.