Mohammad Azmi - Phd Student of Water Resources Engineering, Agricultural Technology and Engineering Faculty, University of Tehran
Shahab ARAGHINEJAD - Professor Assistant of Water Resources Engineering, Agricultural Technology and Engineering Faculty, University of Tehran
Behzad Moshiri - Professor of Control & Intelligent Processing, Center of Excellence, School of ECE, University of Tehran

In the process of data fusion in hydrological forecasting models, in addition to highimpact of the accuracy of individual forecasting models and data fusion methods, two other sections that include selection of the best predictor variables as the input of individual forecasting models and selection of the best individual forecasting models as the input of data fusion models should also be carefully considered. The aim of thepresent study is to evaluate the degree of impact of using the two aforementioned sections on overall accuracy of data fusion in a hydrological forecasting. The method applied for use in both steps is based on entropy concepts. Here, results of the case study of forecasting maximum annual flood of Red River in Canada will be evaluatedand compared in different conditions both before and after applying filtering method. The results reveal considerable improvement of forecasting accuracy after the application of the above mentioned method in both stages.

Entropy Criterion, Data Fusion, Predictor Variable, Individual Forecasting Model