Fallacy of the 500-year Flood

A Cautionary Note on Flood Frequency Analysis

Abstract

The flood of 1999 on the Tar River in eastern North Carolina was the largest in nearly 100 years of stream flow records, where recurrence interval estimates at several gaging stations exceeded 500 years. Nevertheless, the estimation of recurrence intervals for low frequency, large magnitude floods involves considerable uncertainty. This paper uses annual flood records from four gaging stations in the Tar River basin to demonstrate the level of inaccuracy associated with flood frequency analysis (FFA). The margin of error (90% confidence interval) for recurrence interval estimates of large floods on the Tar River are suggestive of the inaccuracy of flood frequency curves, which show that the 100-ycar flood may be under or overestimated by as much as 1.5-2 times. Although FFA is necessary for the effective management of floodplains, estimates of discharge for various recurrence
intervals should be evaluated in the context of several significant limitations: they arc often based on short records, the underlying assumptions arc routinely violated, and the margins of error arc usually large.