Heavy tail controls along the flood process cascade

Abstract

Many observed precipitation and streamflow time series show heavy tail behaviour. This means that the occurrence probability of extreme events is higher than for distributions with an exponentially receding tail. Neglecting heavy tail behaviour can therefore lead to an underestimation of rarely observed, high-impact events. Using long time series and a better understanding of the relevant process controls can help with more robust estimation of upper tail behaviour. Here, a conceptual rainfall-runoff model is used to analyse how precipitation and runoff generation characteristics affect the upper tail of flood peak distributions. Long, synthetic precipitation time series with different tail behaviour are produced by a stochastic weather generator and used as input for a rainfall-runoff model. In addition, catchment characteristics linked to a threshold process in the runoff generation are varied between model runs. The upper tail behaviour of the simulated discharge times series is characterized with the shape parameter of the generalized extreme value distribution (GEV).Our analysis shows that the rainfall distributions asymptotically govern the flood peak distributions above a certain, catchment-specific return period. Below this return period, threshold processes in the runoff generation lead to heavier tails of flood peak distributions. We conclude that, for return periods that are mostly of interest to flood risk management, runoff generation is often a more pronounced control of flood heavy tails than precipitation.