Regional Flood Frequency with Hierarchical Region of Influence

Abstract

Regional flood frequency analysis involves combining extreme-flow information from numerous stations to enhance the reliablity of the temporal characterization of extreme-flow data at sites with short periods of record. The region of influence (ROI) approach used herein, defines regions such that each site has a potentially unique combination of stations. The similarity of the selected stations is ensured by incorporating a homogeneity test into the regionalization process. A hierarchical feature is added to the ROI approach to further enhance the efficiency of the spatial information transfer. This is done by taking advantage of the different spatial similarity scales that have been observed for different orders of moments for a flood frequency distribution. The incorporation of this concept into the ROI framework results in a set of ROIs for a site as opposed to a single ROI. The hierarchical ROI approach is evaluated through the use of a Monte Carlo experiment applied to data representative of a collection of unregulated catchments in the midwest portion of Canada. The simulation experiment shows that an improvement in flood quantile estimation is achieved.