Abstract
The past 12 years have seen significant steps forward in the science and practice of coastal flood analysis. This paper aims to recount and critically assess these advances, while helping identify next steps for the field. This paper then focuses on a key problem, connecting the probabilistic characterization of flood hazards to their physical mechanisms. Our investigation into the effects of natural structure on the probabilities of storm surges shows that several different types of spatial-, temporal-, and process-related organizations affect key assumptions made in many of the methods used to estimate these probabilities. Following a brief introduction to general historical methods, we analyze the two joint probability methods used in most tropical cyclone hazard and risk studies today: the surface response function and Bayesian quadrature. A major difference between these two methods is that the response function creates continuous surfaces, which can be interpolated or extrapolated on a fine scale if necessary, and the Bayesian quadrature optimizes a set of probability masses, which cannot be directly interpolated or extrapolated. Several examples are given here showing significant impacts related to natural structure that should not be neglected in hazard and risk assessment for tropical cyclones including: (1) differences between omnidirectional sampling and directional-dependent sampling of storms in near coastal areas; (2) the impact of surge probability discontinuities on the treatment of epistemic uncertainty; (3) the ability to reduce aleatory uncertainty when sampling over larger spatial domains; and (4) the need to quantify trade-offs between aleatory and epistemic uncertainties in long-term stochastic sampling.
Highlights
The estimation of storm surges, from both tropical and extratropical cyclones, is critical to quantifying hazards, risks and resilience in coastal areas
Several examples are given here showing significant impacts related to natural structure that should not be neglected in hazard and risk assessment for tropical cyclones including: (1) differences between omnidirectional sampling and directional-dependent sampling of storms in near coastal areas; (2) the impact of surge probability discontinuities on the treatment of epistemic uncertainty; (3) the ability to reduce aleatory uncertainty when sampling over larger spatial domains; and (4) the need to quantify trade-offs between aleatory and epistemic uncertainties in long-term stochastic sampling
Analogous to methods used in hydrological analyses for historical datasets (Langbein 1949; Beard 1962; Chow 1964; US Water Resources Council 1967; Beard 1974), initial historical storm method (HSM) used peak water levels from annual records or hindcasts combined with parametric fitting techniques, typically based on rank–order plotting position, to estimate annual exceedance probability (AEP) or return period (TR), AEPðxÞ 1⁄4 1 À FðxÞ; or TRðxÞ 1⁄4 AEPðxÞ 1⁄4 1 À FðxÞ
Summary
The estimation of storm surges, from both tropical and extratropical cyclones, is critical to quantifying hazards, risks and resilience in coastal areas. In the aftermath of Hurricane Katrina, extensive efforts were funded and major progress was made in this area (Westerink et al 2008; Resio et al 2008; Irish et al 2009; Niedoroda et al 2010; Toro et al 2010; Irish and Resio 2010) These new methods represented significant advances over the previous approaches which were based only on hindcasts of historical or hypothetical ‘‘design storms’’ to estimate storm surges in an area. Typical estimates of inundation hazards assume that storm characteristics in a given region are drawn from a single stationary, homogeneous population In this context, known physical constraints on the energy balance in obliquely landfalling storms and the effects of extended episodes of low storm activity and high storm activity are often neglected. This paper investigates potential effects of such natural structure in estimates of statistical storm surge hazards used today for setting coastal insurance rates and for community planning decisions. We treat estimates of characteristics that depend on the number of samples, such as number of historical storms in an area, best-fit statistical parameters and the period of record, as aleatory
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
