Inter-comparison of regional flood frequency procedures for Canadian rivers

Abstract

This paper presents the results of a comparison of regional flood frequency procedures. The accuracy of regional quantile estimators as predictors of population quantiles at a number of key sites is examined. Reference sites have been selected with the basic requirement of at least 35 years of local flood records. Accurate at-site estimates of population quantiles can therefore be obtained at these locations. The regional estimators are used ignoring the local data. This is equivalent to flood estimation at an ungauged site. The case of some local information (10 years of records) is also considered. The bias and variance of the regional estimators are evaluated by comparing at-site and regional estimates. Two methodologies have been devised for the comparison of regional models: a regional bootstrap approach and an empirical Bayes approach. For each approach, a detailed description of the results of the various methods considered in the study is provided. It was shown that, in general, the two methodologies lead to similar results.