FINAL REPORT FHWA/IN/JTRP-2002/2 REGIONALIZATION OF INDIANA WATERSHEDS FOR FLOOD FLOW PREDICTIONS (PHASE I) Studies in Regionalization of Watersheds by

The contents of this report reflect the views of the authors who are responsible for the facts and the accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the Federal Highway Administration and the Indiana Department of Transportation. The report does not constitute a standard specification, or regulation. The following five methods of regionalization of watersheds were tested with Indiana watershed and annual maximum flood data: (1) the L-moment based method, (2) the method based on hybrid cluster analysis, (3) the hybrid cluster method using rainfall data, (4) the fuzzy cluster method, and (5) the method based on artificial neural networks. The results of the L-moment based method and the hybrid cluster method with rainfall data were unacceptable because of the subjectivity involved with the former and the heterogeneity of the of the results obtained by the latter. The remaining three methods gave very similar results. The fuzzy cluster and artificial neural network based methods are much easier to use and hence are recommended. The results from any of these methods will not give homogeneous regions. The results from the clustering methods must be tested and revised to get homogeneous watersheds. The data from each of the regions were investigated by using tests based on simple scaling. The results from these tests confirm all the regions, except one, to be homogeneous.. 17. Introduction Several studies have claimed that regionalization of watersheds is essential to develop regional flood flow equations. These flood flow equations would be used to estimate flood magnitudes at locations where actual flood data are not available. Although several regionalization methods have been proposed, there is no agreement about the method or methods which are to be used. In this study of regionalization of Indiana watersheds, a two-step procedure was adopted. In the first step, regionalization methods in use were reviewed and the most promising of these were selected for testing. In the second step, the selected methods were tested by using the watershed and flow data. The following regionalization methods were tested: • The L-moment based method • The method based on hybrid cluster analysis • The hybrid cluster method using rainfall data • The method based on fuzzy cluster analysis • The method based on artificial neural networks. Findings The L-moment based method requires subjective judgment in regionalizing watersheds. Consequently, the results would not be unique and …