Statistical simulation of flood variables: incorporating short‐term sequencing

Abstract

Abstract The pluvial and fluvial flooding in the United Kingdom over the summer of 2007 arose as a result of anomalous climatic conditions that persisted for over a month. Gaining an understanding of the sequencing of storm events and representing their characteristics within flood risk analysis is therefore of importance. This paper provides a general method for simulating univariate time series data, with a given marginal extreme value distribution and required autocorrelation structure, together with a demonstration of the method with synthetic data. The method is then extended to the multivariate case, where cross‐variable correlations are also represented. The multivariate method is shown to work well for a two‐variable simulation of wave heights and sea surges at Lerwick. This work was prompted by an engineering need for long time series data for use in continuous simulation studies where gradual deterioration is a contributory factor to flood risk and potential structural failure.