Though the directional wave spectrum can show multi-modal features under confused sea conditions with windsea and swell, there is not enough research about the windsea and swell on the multi-modal spectrum because of the difficulty in processing the multi-modal spectrum. This study proposed the advanced methods to identify windsea or swell on the multi-modal directional spectrum. We improved the inversion method for the directional spectrum and proposed a spectral partitioning method based on a statistical approach to identify each wave system from the multi-modal directional spectrum. We validated the proposed methods by the numerical test for the multi-modal spectrum.
We applied the proposed methods to identify each wave system from the in-situ multi-modal directional spectrum and evaluated their wave heights, periods, and directions. We classified each wave system into windsea or swell with wave age and evaluated wave forcing corresponding to each wave system. By classifying the data into the unimodal sea or multi-modal sea, we could more clearly interpret the results for the directional narrowness dependency on wave steepness. The proposed method is available for routine analysis on wave observation because it requires no trials by operators to determine the data processing criteria.

Key Words: non-negative least squares, directional spectrum inversion, information criteria, spectral partitioning, windsea, swell