نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشیار، گروه علوم و مهندسی خاک، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران.
2 کارشناسی ارشد، گروه علوم و مهندسی خاک، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران.
3 دانشیار، گروه مهندسی آب، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران
4 دانشجوی دکتری، گروه مهندسی آب، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Reference evapotranspiration (ET₀) is a crucial parameter in agricultural water management. In this study, two modeling approaches were investigated for predicting daily ET₀ at the Kish, Gorgan, and Shiraz meteorological stations: (1) a standard multilayer perceptron (MLP) model, and (2) an MLP model optimized using the stochastic gradient descent (SGD) algorithm. A dataset comprising 23 years (2000–2023; 1379–1402 in the Iranian calendar) of daily meteorological data was used to develop and validate the models under five different input scenarios, selected based on Pearson correlation analysis. Model performance was evaluated using multiple statistical metrics, including the correlation coefficient (R), root mean square error (RMSE), Nash–Sutcliffe efficiency (NSE), and the Willmott agreement index (d). Among the tested configurations, the fifth scenario of the combined MLP-SGD model showed the highest accuracy, achieving RMSE values of 1.04 mm/day, 1.20 mm/day, and 1.94 mm/day for the Shiraz, Gorgan, and Kish stations, respectively. The results indicate that the combined MLP-SGD model outperformed the standalone MLP model, demonstrating its reliability and effectiveness in predicting reference evapotranspiration. These findings underscore the potential of hybrid artificial intelligence models in improving water resource management and promoting sustainable agricultural practices.
کلیدواژهها [English]
DOI: 10.22034/ws.2021.45876.2415
DOI:10.1109/TAC.2013.2254619
DOI: 10.1016/j.agrformet.2007.04.012
DOI: 10.1016/j.agrformet.2020.108034