Document Type : Applied Article
Author
University of Torbat-e Jam, khorasan Razavi, Iran
Abstract
Evapotranspiration is one of the most important components in the energy and water budgets of watersheds. Accurate estimation of Evapotranspiration is necessary for water budget studies, irrigation scheduling, water management, and environmental studies. In recent years, the application of machine learning (ML) techniques has provided a promising approach to improving ET prediction accuracy. In this study, two artificial neural network models, multilayer perceptron (MLP) and radial basis function (RBF), were used to estimate reference crop ET at Mashhad, Sabzevar, and Torbat-e Jam stations in Razavi Khorasan province. Meteorological data, including temperature, relative humidity, wind speed, and sunshine hours, were collected and normalized from 1992 to 2023. The ET values calculated by the FAO Penman-Monteith method were used as target values for training the models. The results showed that the neural network models, especially MLP, could provide accurate estimates of ET and performed better than the Linacre and Hargreaves-Samani methods. Excluding wind speed at Torbat-e Jam and Mashhad resulted in a 5.1% decrease in the coefficient of determination (R²), while at Sabzevar, this decrease was 11.3%. In contrast, excluding sunshine hours had a negligible impact on model accuracy. The findings further revealed that even with limited data, using only the minimum and maximum temperatures could yield reasonably accurate evapotranspiration estimates. These findings suggest that the use of neural networks can be a fast and accurate approach for estimating ET in data-scarce conditions and can contribute to the management of agricultural water resources in the studied regions of this province.
Keywords
Main Subjects
)
Send comment about this article