Application of Maximum Entropy Method to the Real Sea Data

Two different versions of maximum entropy methods(MEM) were compared with two conventional methods for analyzing directional ocean wave spectra. The two MEMs were originally proposed by Lygre and Krogstad (1986) and Kobune and Hashimoto (1986), respectively, and the two conventional methods are the truncated Fourier series method(TFS) and Longuet-Higgins parametric model(LHM). The comparisons included hypothetical idealized cases and actual measured data. For the hypothetical cases, the MEM by Kobune and Hashimoto clearly performed better, particularly for dual-peaked spectra. As for the comparisons from measured data, the MEM generally yielded narrower directional spreading than the two conventional methods but all methods gave nearly identical main direction information. However, this MEM does have occasional convergence problem in real sea data analysis. The problem is removed with the aid of an approximation scheme. This modified scheme is employed in the automated directional spectral analysis of measured sea data.