An Evaluation of Accuracy of Intelligent Methods and Sensitivity Analysis of Reference Crop Evapotranspiration to Meteorological Parameters in Two Different Climates

In this study, climatic parameters are used to evaluate the sensitivity of reference evapotranspiration in Isfahan and Rasht with two semi-arid and moderate climates using two intelligent methods during 1970-2010. In this research, multi-layer precpetron e (MLP) and gene expression programming (GEP) intelligent models are used in order to analyze the sensitivity of reference evapotranspiration to meteorological parameters. Results of correlation of the parameters showed, in Isfahan and Rasht, maximum and minimum temperatures have the highest correlation with correlation coefficients of 0.968 and 0.926 with other parameters, and in Isfahan wind speed and maximum temperature with a coefficient of 0.067 and in Rasht, minimum humidity and wind speed with a coefficient of 0.001 have lowest correlations. The results of the survey indicate that in Isfahan using MLP model and the maximum and minimum temperature, humidity, wind speed and number of sunshine hours as inputs, with the root mean square error (RMSE) equal to 0.418 mm/day was the best results for estimating reference evapotranspiration. The application of MLP model in Rasht, with the inputs of maximum and minimum temperature, the maximum humidity and the number of sunshine hours has less error and higher correlation coefficient and the values of reference evapotranspiration are much more similar to reality. Also, about the evapotranspiration of reference plant in relation to parameters that are used in this study the results showed that in two study places, the correlation coefficient between wind speed and reference evapotranspiration are 0.3 and 0.061 in Isfahan and Rasht, respectively which are the lowest coefficients.