Assessment of Artificial Neural Network (ANN) in prediction of garlic evapotranspiration (ETC) with lysimeter in Hamedan

Evapotranspiration as one of the important elements in agriculture has a considerable role in water resource management. Therefore, using a more exact estimation method is an essential step of agricultural development, especially in arid semi-arid area. In this research, in order to exact estimate of garlic evapotranspiration using lysimeteric data, an artificial neural network (ANN) model was developed. Maximum and minimum air temperatures, maximum and minimum relative humidity values, wind speed and sunshine hours were used as the input layer data. The crop evapotranspiration was measured using 4 lysimetres of 2×2×2m of the Bu-Ali Sina agriculture collage’s meteorology station during 2006-2008. Statistic indicators RMSE, MAE, STDMAE R2 were used for performance evaluation of the models. The results showed the more exact method concerned to the multilayer perceptron (MLP) model with the back propagation algorithm. The 6-6-1 layout with Levenberg-Marquat rule and sigmoid function had the best topology of the model. The evaluation criteria were 0.088, 0.07 and 0.061 mm/day as well as 0.88, respectively. The results also showed that the average daily garlic evapotranspiration were 8.3 and 6.5 mm based on the lysimeter ANN methods, respectively. Overall, evaluation of ANN results showed that the errors of ANN were negligible. The ANN showed high and low sensitivity to maximum air temperature and minimum relative humidity, respectively.