Daily Streamflow Forecasting of Nooranchay River Using the Hybrid Model of Artificial Neural Networks- Principal Component Analysis

Accurate forecasting of the daily discharge plays a significant role in the efficient management of water resources. For this purpose in order to model more accurately the process of forecasting the daily discharge of Nooranchay river in Atashgah basin, the artificial neural networks model (ANN) was used. In addition, in order to increase the accuracy of ANN, the principal component analysis (PCA) was used for preprocessing of input data. Finally, the results of multivariate linear regression (MLR) model were compared with the obtained results in the mentioned hydrological simulation. The results indicated that the hybrid model of ANN-PCA in comparison with ANN and MLR, had the highest precision. So that the results of goodness-of-fit tests criteria, such as the correlation coefficient (CC), the efficiency coefficient (EC) and the root mean square error (RMSE) for the hybrid model of ANN-PCA (at the verification stage) were CC=0.9959, EC=0.9905 and RMSE=0.0071, and for the ANN (at the verification stage) were CC=0.9093, EC=0.8269 and RMSE=0.0405 and the results for the MLR were obtained as CC=0.8866, EC=0.7860 and RMSE=0.0926. Also the use of PCA as an effective method for pre-processing of data, created independent components which eliminated the multicollinearity. Therefore, the PCA increased the efficiency of the ANN.