نوع مقاله : مقاله پژوهشی
1 استادیار /بخش مهندسی آب، دانشکده کشاورزی، دانشگاه شهید باهنر کرمان، کرمان، ایران
2 دانشآموخته کارشناسی ارشد مهندسی منابع آب/ دانشکده کشاورزی، دانشگاه شهید باهنر کرمان (عضو انجمن پژوهشگران جوان دانشگاه شهید باهنر کرمان)
عنوان مقاله [English]
Accurate estimation of evapotranspiration has a great influence on water resources management and planning, especially in arid and semi-arid areas. Different methods have been presented by researchers for evapotranspiration estimation. These include a variety of empirical equations and data-driven methods. In this study to estimate the daily reference evapotranspiration at eight semi-arid climates in Iran, three methods based on the adaptive network-based fuzzy inference system (ANFIS), support vector machines (SVM), and model tree (M5) as well as five empirical equation were used. Meteorological data including maximum and minimum temperatures, relative humidity, wind speed, and the sunshine hours were used. Eleven different combinations of these variables have been used as input variables in data-driven methods for evapotranspiration modeling for the period of 1980 to 2009. Eighty percent of the data were used for the training and twenty percent were used to test the models. The results were compared with those of the standard Penman-Monteith FAO-56 equation. Performance of the methods was evaluated using statistical indices of mean square of error (RMSE), coefficient of determination (R2), and index of agreement (d). Support vector machines and adaptive networks based on fuzzy inference system methods presented best performance with RMSE between 0.24~1.55 (mm.day-1) in nine combination of meteorological variables. RMSE of empirical equations varied between 0.71~5.96 (mm day-1). Blaney-Criddle and McGunness-Bordne equation presented the highest accuracy in most stations. M5 model has a lower performance compared to ANFIS and SVM methods in the studied climates.