Abstract: Populated coastlines influenced by tropical cyclone (TC) prone areas call for flood risk hazard assessments, including knowledge on the probability of occurrence of major TC-induced significant wave heights. Due to the scarcity of TC historical records, extreme value analyses often rely on fitting generalized extreme value distribution functions to extrapolate longer return periods. This paper describes a methodology that allows to obtain deterministic estimations of the tail probability distribution using long collections of high-fidelity tracks that reproduce similar historical diversity and frequency trends. Given the large dimensionality of the problem (spatiotemporal variability of track geometry and intensity), we implement a track parameterization to easily identify storms in a parametric space. A hybrid approach significantly reduces computational resources by enabling to narrow the number of non-stationary numerically simulated cases forced with vortex-type wind fields parameterized using the Holland Dynamic Model. The proposed surrogate model, HyTCWaves, is trained with a selected subset of maximum significant wave height (MSWH) spatial fields to which a Principal Component Analysis and interpolation functions are performed. Results show a useful approximation of spatialbased regional extreme value distribution of MSWH induced by TCs. The proposed model is applied to the target location of Majuro atoll.

Authorship

Download reference