نوع مقاله : مقاله پژوهشی
1 دانشجوی دکترای مهندسی منابع آب دانشگاه شهید چمران، اهوز، ایران
2 استادیار گروه مهندسی منابع آب دانشگاه شهید چمران اهواز، اهواز، ایران
3 استادیار گروه مهندسی آب دانشگاه شهرکرد، شهر کرد، ایران
4 دانشجوی کارشناسی ارشد مهندسی منابع آب دانشگاه ارومیه، ارومیه، ایران
عنوان مقاله [English]
In this study, for estimating grass reference crop evapotranspiration using support vector machine (SVM) and genetic programming (GP) techniques, monthly meteorological data of 6 synoptic stations in northwest of Iran during a 38 year period (1973-2010) were collected. At first, the monthly grass reference crop evapotranspiration values for selected station were calculated using the FAO-56 Penman-Monteith equation and considered as the output of SVM and GP models. In the next step, a regression equation was made based on effective climatic variables on evapotranspiration and different input patterns were defined for developing SVM and GP models. Since the relative humidity had the lowest effect on evapotranspiration, it was removed from the model’s input variables. In order to consider the effect of memory of time series in the accuracy of evapotranspiration prediction, the evapotranspiration values with one, two, three, four and five months lag time were used in model’s input patterns. Ten input patterns were made for every model. However, the obtained results showed the high accuracy of both SVM and GP models in predicting monthly grass reference crop evapotranspiration in northwest of Iran, but the SVM method had better performance than GP method. Also, the results showed that if there is no enough meteorological data, the memory of time series can be used for predicting monthly grass reference crop evapotranspiration.