A Modified Region of Influence Approach for Flood Regionalization

The present investigation modifies the "Region of Influence" (RI) approach based on statistical and physical characteristics of the stream gauging stations in south eastern New South Wales (NSW), Australia. Statistical properties of flood data are first included as the attributes in similarity distance algorithm instead of geographical location. The RI approach is modified by the determination of the criteria for the choice of a set of six statistical attributes of catchments. The choice of the threshold value of the similarity distances and the effect of attribute interaction are also discussed. The modified method was then applied to determine the RI of some streamflow gauging stations in NSW using three independent statistical attributes. In regional flood frequency analysis, the defined regions must be homogeneous. In this study, the physiographic characteristics of catchments are also combined with flood statistics in terms of defining the RI for the selected sites in NSW, Australia, in order to develop optimally the RI approach. This approach transfers the relevant information from the dispersed catchments in the region to their reference site. Three independent physical attributes, Median Annual Rainfall, Coefficient of Compactness and Stream Length, and three independent statistical attributes, Coefficient of Variation, Q50/A and a new coefficient based on Pearson distribution as the statistical attributes are also applied, for defining the RI of each site. When physical attributes are included in the attributes set, sites in the resulting regions are more concentrated around the reference site. The results of this study indicate that a physically-statistically based RI approach can provide a more effective information transfer than a RI approach which only includes flood statistics. Thus, the physically-statistically based RI approach is recommended for flood regionalization.