مقايسه تبخير وتعرق گياه مرجع (ET0) برآوردي از روش هاي تجربي و شبكه هاي عصبي مصنوعي با داده هاي لايسيمتري

Comparison of Reference Evapotranspiration (ETO) results of Empirical methods and Artificial Neural Networks with Lysimetric data

The purpose of this study is selection of the best method to estimation Reference Evapotranspiration (ETO). To this purpose, the ETO of the region was estimated using meteorological data and different empirical methods as well as Artificial Neural Network (ANNs). Than, the each estimated ETO was compared with a 2 year lysimetric data. For ANN various layouts with hidden layer, threshold function and learning rule were performed. The results showed that Penman-FAO method and ANN were useful for this purpose. The results (from empirical methods and ANN) assessment were performed with statistical parameters such as ERMS, Ema, R2 and SDema- According to this results Penman-FAO method had better than the other methods with the second level of R , the First level Erms, the second level Ema, second level SDema- Also, ANN with 6-6-1 layout had been (for the statistical parameters) 0.057, 0.069, 0.86 and 0.068 (mmd1), respectively. Comparison between Penman-FAO, ANN and the lysimeter data showed that high accuracy of ANN rather than Penman-FAO method. Thus, ANN because of low input information and high speed and little time was suggested as the best method. The regression results of ANN and empirical method showed that polynomial model had the high R2 value than linear model. So, polynomial model because of simplicity and having the high R2 value is an exact mean of ETO estimation