Plenary Lecture
Rainfall-Runoff Modeling Based on Self Organizing Map and Wavelet Pre-Processing Techniques
Associate Professor Vahid Nourani
Faculty of Civil Eng.
University of Tabriz
Iran
formerly, Assoc. Prof. St. Anthony Falls Lab
and NCED, Dept. Civil Eng.
University of Minnesota, USA
E-mail: vnourani@umn.edu
Abstract: The precise rainfall-runoff model provides vital information for water resources planning and watershed management. Since the spatial and temporal selection of input data is a key element for accurate modeling, data screening and preprocessing become the main concern in any data driven modeling. To this end, this paper combined multi techniques to introduce an optimal Feed Forward Neural Network (FFNN) rainfall-runoff model. In this way, self-organizing map (SOM) clustering technique was applied to identify spatially homogeneous clusters of precipitation obtained from satellite data. Besides, the wavelet transform (WT) was utilized in order to extract dominant temporal features from runoff non-stationary time series. The comparison of proposed hybrid model (SOM-WT-FFNN) with single SOM-FFNN model and conventional auto regressive integrated moving average with exogenous input (ARIMAX) model, proved the efficiency of the developed methodology. Satellite data as a source of daily precipitation and runoff data from Gilgel Abay watershed in Ethiopia were applied for calibration and validation purposes. The results of the study provide promising evidence for combining spatial and temporal data pre-processing methods in forecasting runoff values specifically peak records using the FFNN method.
Brief Biography of the Speaker: Vahid Nourani was born in Tabriz, Iran in 1975 and received his B.Sc. and M.S. degrees in Civil Engineering from University of Tabriz, Iran in 1998 and 2000, respectively. He then continued his graduate study in Civil and Environmental Engineering in the field of Hydrology at Shiraz University, Iran and Tohoku University, Japan and was graduated in 2005. Nourani was with the Faculty of Civil Engineering, University of Tabriz as an Assistant Professor from 2005-2009; as Associate Professor from 2009-2012 and with Dept. of Civil Eng., University of Minnesota, USA at 2011 as visiting associate professor. In this period, 45 Ph.D. and M.Sc. students were graduated under his technical supervision. His research interests include rainfall-runoff modeling, Artificial Intelligence applications to water resources engineering, Hydroinformatics and computational hydraulics. His researches outcomes have been published as 51 Journal articles, 1 book, 2 book chapters and more than 55 papers presented in international and national conferences. He is currently researching on hydrological models generated via satellite data as well as pre-processing methods such as Self Organizing Maps and wavelets for developing optimum models.