Forecast of daily mean, maximum and minimum temperature time series by three artificial neural network methods

Temperature forecasting has been one of the most important factors considered in climate impact studies on sectors of agriculture, vegetation, water resources and tourism. The main purpose of this study is to forecast daily mean, maximum and minimum temperature time series employing three different artificial neural network (ANN) methods and provide the best-fit prediction with the observed actual data using ANN algorithms. The geographical location considered is one of Turkey's most important areas of agricultural production, the Geyve and Sakarya basin, located in the south-east of the Marmara region (40°N and 30°E). The methods chosen in this study are: (1) feed-forward back propagation (FFBP), (2) radial basis function (RBF) and, (3) generalized regression neural network (GRNN). Additionally, predictions with a multiple linear regression (MLR) model were compared to those of the ANN methods. All three different ANN methods provide satisfactory predictions in terms of the selected performance criteria; correlation coefficient (R), root mean square error (RMSE), index of agreement (IA) and the results compared well with the conventional MLR method.