%0 Journal Article %T Evaluation of Data Mining Methods and Experimental Temperature-Radiation-Based Models in Estimating Evaporation from the Pan (Case Study: East of Urmia Lake) %J Iranian Journal of Soil and Water Research %I University of Tehran %Z 2008-479X %A Mouneskhah, vahid %A samadianfard, saeid %A HADI, MOEIN %D 2020 %\ 11/21/2020 %V 51 %N 9 %P 2337-2348 %! Evaluation of Data Mining Methods and Experimental Temperature-Radiation-Based Models in Estimating Evaporation from the Pan (Case Study: East of Urmia Lake) %K Data Mining %K Jensen-Haise %K modeling %K pan evaporation %K temperature %R 10.22059/ijswr.2020.302179.668603 %X Evaporation from the pan has an effective role in water resources management. But due to the interaction of meteorological variables in the calculation of evaporation, several nonlinear relationships have been presented that their efficiency is arguable according to the climatic conditions of each region. Therefore, in the present study, the capabilities of temperature-radiation-based empirical equations and data mining methods of support vector regression (SVR), Gaussian process regression (GPR) and nearest neighborhood (IBK) were investigated under 10 different scenarios resulting from the combination of meteorological factors in estimating and predicting the evaporation amounts in 5 selected stations in the east of Urmia Lake basin. NRMSE and MAPE statistical indicators were used to evaluate the results. In order to model the effective parameters on pan evaporation, the effect of each parameter was calculated using the principal component analysis through the correlation values of parameters with the pan evaporation rate. The results showed that among the implemented meteorological parameters, temperature have the maximum impact and wind speed and precipitation have the minimum impacts on modeling process. Also, among the empirical methods, the Jensen-Haise method had the highest accuracy. Moreover, among the data mining methods, the SVR in Tabriz, Sarab, and Harris stations and GPR in Bostanabad and Maragheh stations had higher accuracies as compared to the others. In general, in all the studied stations, the accuracy of the best data mining scenario was higher than the best empirical method. Also, in terms of data limitation, the Jensen-Haise method had suitable accuracy. Also, despite the low accuracy of the IBK method compared to other data mining methods, this method reachs to its highest accuracy rates with the lowest input variable. %U https://ijswr.ut.ac.ir/article_77045_660869d2c0368a0a646f550091700c36.pdf