Regional hierarchical approach to flood frequency analysis

Usually flood frequency analysis copes with the flood quantile estimation related to a return period much more greater than the observation period of the historical data. Regional flood frequency analysis (RFFA) overcomes this deficiency by increasing the number of the involved data by means of the use of all the gauging sites in the region of interest. RFFA explores the spatial links of hydrological variables, providing at-site/regional or purely regional flood quantile estimates which can be more precise than at-site estimates, even in the presence of moderate heterogeneity. In the last decades, since the Flood Studies Report (NERC, 1975), many procedures of RFFA have been proposed, as reviewed, among others, by Potter (1987) and Cunnane (1987; 1988). The procedures differ by many features, as they combine in different ways at-site and regional information or use different kind of data (annual maximum flood peaks series, peaks over a threshold series). Moreover some RFFA methods consider both rainfall and flood data, in order to transfer information from the former to the latter within appropriate homogeneous region. Two parameter distributions are often adopted. Such distributions lead to quantile estimates with relatively short standard error but large bias. On the other hand, distributions characterized by three or more parameters have larger standard errors but are sufficiently flexible to be relatively unbiased, especially in the case of homogeneous regions and mildly heterogeneous regions. Anyway the use of these distributions within a regional scheme improves quantile estimations. In the following paragraphs some RFFA for gauged and ungauged sites are briefly discussed, with particular attention to the hierarchical regional approach.