Robust simple scaling analysis of flood peaks series

Abstract

Simple scaling theory explains the spatial variability of peak flow processes by their indexation on a set of scale parameters, such as the size of drainage basins. On this basis, a regional analysis can be conducted to construct a unique growth curve. However, a drawback of the simple scaling approach is that it assumes that the coefficients of variation of peak flow distributions are identical throughout a region. Although empirical data might display statistical properties which correspond to simple scaling assumptions, the coefficient of variation of flood peaks depends on drainage area. We discuss this fact here and we suggest using the empirical Bayes method to take into account this regional variation of the coefficient of variation. The advantage of using an empirical Bayes analysis subsequent to simple scaling modeling is illustrated using a set of 109 flood series from the Province of Ontario, Canada. Key words: simple scaling, empirical Bayes, index flood, regional flood frequency analysis.