Analysis of Reference Crop Evapotranspiration Sensitivity to Climatic Variables in Homogeneous Regions Using Computational Intelligence Techniques and Multivariate Statistical Methods

Evapotranspiration is one of the most important factors in basin water balance and water cycle in the nature. Several variables affect evapotranspiration rate. Therefore, all these variables cannot be considered for deriving the equations governing the evapotranspiration process, even if they are available. Determining the most important of these variables, and reducing the number of inputs in modeling process, can lead to increasing the model validity. There are several approaches for sensitivity analysis and identifying the relative influence of different climatic parameters on evapotranspiration. In this study, using computational intelligence techniques and multivariate statistical methods, sensitivity analysis and relative importance of different parameters affecting reference evapotranspiration were investigated in the western and northwestern Iran, based on the FAO Penman-Monteith model. In order to increase the accuracy of the models output results, considering the type of climate in the process of evapotranspiration, using the Ward linkage and Euclidean similarity measure in the cluster analysis technique, based on data from synoptic meteorological stations, the study area was divided into several climatically homogeneous regions. Then, in each of the homogeneous regions, the most important parameters affecting the reference evapotranspiration were identified using three methods of factor analysis, neural networks, and decision tree. Finally, verification and validation of these methods were performed for different homogeneous regions. The results showed that the artificial neural network method had better performance for determining the most important parameters affecting evapotranspiration than decision tree and factor analysis techniques. Finally, using the results of artificial neural network, regional equations were extracted and tested for each homogeneous region.