Regional long-term extreme wave analysis using hindcast data from the South Atlantic Ocean

Abstract

Long-term extreme wave analysis is performed using a spatial approach applied to hindcast data, to study the sensitivity of extrapolated quantiles of significant wave height to different hindcast calibration methods and distribution functions. The probabilistic moments and L-moments are analyzed, as well as several extreme value distribution functions. The estimation of extreme values is addressed in two parts: (1) numerical modeling of wind and wave hindcasts with improved accuracy under extreme conditions; (2) investigation and tests of fit of extreme value distributions using the Regional Frequency Analysis. Therefore, the main goal of this paper is to show how hindcast data calibrated to account for extreme events can influence and improve return value estimates. Results show that simplistic calibrations, such as a linear regression applied to input winds and quick modifications on the numerical wave model, do not significantly improve the representation of the peak of storms – which has a small impact on the reliability of shape parameters of the quantile functions. The construction of a new wave hindcast using satellite winds inside extreme cyclones led to a reduction of the bias at higher percentiles and additional improvement of the distribution tail and its shape. These elements, combined with a robust extreme analysis methodology, such as the Regional Frequency Analysis, and thorough choice of the distribution function, produced reliable spatial return values of significant wave height.