Revising Flood Return Periods by Accounting for the Co-Occurrence Between Floods and Their Potential Drivers

The return period of floods can be influenced by the extreme values of their potential drivers, which may vary among catchments. Understanding the risk and the associated changes in return periods due to these extreme values of drivers is therefore of interest in flood hydrology. In this study, floods are considered as compound events resulting from a combination of non-independent factors. The return periods of these events are estimated using joint distribution functions, accounting for the dependence of flood peaks and their potential drivers in two distinct catchments: (i) an inland catchment-Warunji Catchment, Krishna basin, India, and (ii) a coastal catchment-Usk catchment, United Kingdom (UK). The annual maximum (AM) rainfall, soil moisture and storm surge are considered as potential drivers of floods and their variations in time of occurrence are calculated to understand co-occurrence patterns. The pairwise co-occurrence frequency and dependence are estimated, and joint distribution is calculated with the survival copula distribution function. The results indicate that AM values of the variables tend to co-occur within a short time window, signifying that the flood risk changes with the extreme values of drivers. The maximum values of the AM series of drivers are observed in the same year as the largest flood in the series. The joint return periods of flood events show significant variations from their univariate estimates in both catchments, which have different flood-generating mechanisms. This work re-emphasises the findings in recent literature that the traditional univariate risk assessment methods based only on flood peak information may substantially underestimate/overestimate the risk of floods by neglecting the effects of their potential drivers and that a multivariate viewpoint is imperative for assessing the risk of floods.