CoAuthors

Jiuke Wang (Sun Yat-Sen University, Chine)

Dangerous sea forecasting remains a crucial topic for several sectors of activity such as maritime navigation and the protection of goods and people in the global ocean. The work of Le Merle et al (2021) has shown that the SWIM instrument is capable of providing good directional wave spectra from which we can calculate parameters that are sensitive to the detection of rogue waves, such as frequency spreading, Crest to Trough Correlation (CTC) and the two-dimensional Benjamin-Feir Index (BFI2D). The aim of this work is to implement a rogue wave indicator based on an estimation of the maximum wave height using a deep learning method. The first part of this work consists in developing a deep learning model that can be used to retrieve the maximum wave height along track of SWIM nadir. The model training is carried out using buoy wave data from the North Atlantic (Brittany and gulf of Biscay) and Campbell island off shore of southern New Zealand. The input parameters for the training model are Significant wave height, maximum wave height, mean wave period and peak period. The Validation of the training model with independent wave data from field campaign SUMOS shows a good estimate of maximum wave height along the track of CFOSAT.
In general, the occurrence of rogue waves is defined when the ratio of maximum wave height to significant wave height exceeds 2. The second part of this work consists in analysing the occurrence of rogue waves as a function of the parameters sensitive to the generation of these waves such as BFI2D, CTR, frequency spreading) derived from SWIM directional wave spectra. By using wave data from the SUMOS campaign in the bay of Biscay, we are able to estimate critical threshold values for parameters describing non-linear wave instabilities as a function of maximum height along the SWIM track. The results of dangerous seas in other ocean regions, such as the Agulhas region and the East Coast of Australia, are also examined. Further discussions and conclusions will be presented.