Metocean Design Criteria in Tropical Cyclone Prone Areas: the Mozambique Channel

Abstract

Tropical cyclones represent one of the most catastrophic examples of Natural Hazard. The effect of such events on the environment and human activities is a foregone conclusion. Metocean design criteria are usually defined using databases of historical tropical cyclone tracks characterized by wind and central pressure values. From these historical databases wind and wave maxima are subjected to extrapolation in logarithmic space. In order to correctly evaluate the required return period (up to 1 on 10,000 years) extrapolating the tail of the distribution from historical database provides highly uncertain estimates, therefore, a synthetic tropical cyclone database is needed. The present work describes step-by-step the setup of a synthetic tropical cyclone database valid for the Mozambique Channel area and finalized at the definition of metocean design criteria in area characterized by such events. The synthetic tropical cyclone database, statistically consistent, has been build-up using an autoregressive model for increments in track speed, direction and maxima cyclone winds. Moreover the used methodology involves application of a physically based wave model in order to catch a correct representation of the main processes: wave shoaling/refraction, trapped fetch effects, wave breaking, correct representation of offshore islands, etc. In order to reduce the computational time without reducing the benefit of the wave modeling phase, the proposed methodology considers a track pre-selection step allowing wave simulation only for tropical cyclone tracks characterized by the higher waves at point of interest. In such way the tail of the distribution, obtained from the synthetic database and characterized by higher return periods (i.e. 1,000, 5,000 and 10,000 years), is effectively simulated by the wave numerical model allowing a correct definition of the corresponding metocean design criteria. Furthermore, fitting modeling results with an extreme distribution function (e.g. Weibull, Pareto, etc.) allows definition of extreme values at lower return periods (i.e. 100, 200 and 500 years).