The concept of credible duration for the Regional Frequency Analysis including historical data: application of the FAB method to a skew storm surge database

Abstract

Abstract. The design of effective coastal protections requires an adequate estimation of the annual probability of occurrence of rare events associated to a return period up to 10 3 years. The characterisation of extreme events is carried out by statistical methods based on the extreme value theory. Usually, these methods apply to very short 15 observations series and the uncertainties associated to high return periods are too wide to be used in a design approach. Nowadays, large and spatially distributed dataset are available and historical data approaches are providing new insight on past events. Combining local information with regional information and historical information will increase the precision and the reliability of low probability extrapolation. However, the past events data provided by single events analysis are not continuous and exhaustive as systematic records. Can we guess 20 how exhaustive the historical records are and how many historical events have been missed by the observations? This is a very important information for the estimation of the correct empirical frequency of the extreme events.To answer these questions, the FAB method is introduced here. It is based on a new definition of the "credible duration", representing the estimate duration of the support of the data.