The Use of Artificial Neural Networks to Estimate Reference Evapotranspiration

Abstract

Evapotranspiration is defined as the loss of water from soil and vegetation to the atmosphere, driven by weather conditions. It reduces the availability of water for agricultural purposes, which affects the amount of irrigation water, particularly during the dry season. The objective of this paper is to present a comparative analysis of the estimated reference evapotranspiration value based on artificial neural networks (ANN) with backpropagation bias 1 (BP-1) and backpropagation bias 0 (BP-0) architectures. The model was fed with data of air temperature, relative humidity, and solar radiation. The model is utilized to calculate the evapotranspiration using the Hargreaves method as the training data. The performance of ANN model was evaluated using the mean square error (MSE), root mean square error (RMSE), and coefficient determination (R²). Our results showed that both ANN models performed well as indicated by low error (MSE < 0.01) and high R² (>0.99). Also, we found that air temperature and relative humidity determine the optimal prediction. Further, this proposed model can serve as a reference for other models seeking to determine the most appropriate computational model for evapotranspiration value estimation.