How extreme are flood peak distributions? a quasi-global analysis of daily discharge records

A higher chance of extreme floods is often revealed by a probability distribution with a slower decay and a heavier tail. There are numerous and conflicting findings in previous studies that describe flood behavior in different regimes. Our research aims to reconcile these discrepancies by examining the influence of record length and flow regime on the upper tail behavior of floods which holds implications for engineering design and operational purposes related to flood prevention and management. Our investigation focuses on a quasi-global peak-over-threshold (POT) analysis using the Generalized Pareto (GP) distribution. While we employ a single distribution model due to its practicability and parametric parsimony, our primary objective is to understand how the variation in the distribution parameters, here GP shape parameter, reflects heaviness in the flood tails. We investigate 4,482 streamflow series across six distinct regime types with record lengths ranging from 30 to 213 years. Our findings reveal that as the record length increases, the variance of the GP shape parameter decreases, and the mean value converges to a fixed threshold, depending on the flow regime. We show that snow-dominated streams exhibit the lowest shape parameter (γ=0.074) and intermittent streams exhibit the highest (γ=0.384). We further investigate how the hydroclimatic and catchment controls affect flood tails and find that the influential controls of tail heaviness vary depending on the flow regime. Identifying catchments with a higher likelihood of experiencing extreme floods can help prioritize flood risk management practices. We recommend further research to identify the interactions between climatic and catchment controls of floods to improve the prediction of heavy tails. Our research provides insights into the behavior of flood tails across different streamflow regimes and their underlying drivers.

DDC Class

600: Technology

Options

How extreme are flood peak distributions? a quasi-global analysis of daily discharge records