Statistical properties of hurricane surge along a coast

Abstract

[1] The validity and accuracy of approaches used to determine hurricane surge hazard risk received much attention following the hurricane seasons in mid- to late-2000, which caused record surge-related damage along the Gulf of Mexico coastline. Following Hurricane Katrina in 2005, research showed that most extreme-value statistics approaches underestimated the risk associated with this surge event. In this paper, two of the most popular methods for determining hurricane surge extreme-value statistics are reviewed: the historical surge population approach and the joint probability method. Here, it is demonstrated that both limited historical record length and random along-coast variability in hurricane landfall location can introduce significant errors into surge estimates. For example, the historical surge population approach gives errors of 9% to 17% for return periods between 50 and 1000 years when a surge record of 100 years is considered. In contrast, it is shown that the joint probability method yields significantly more reliable surge estimates, with errors of 2% to 3% for return periods between 50 and 1000 years when a storm record of 100 years is considered. Finally, we show that both methods remain robust when decadal-scale climate variability in the storm rate of occurrence is considered, so long as the hurricane history is long enough to capture the full decadal cycle. When used in conjunction with continuous surge response information, it can be concluded that the joint probability method is a practical and reliable approach for determining extreme-value hurricane surge statistics.