An eigenfunction parameterization of a time sequence of wave spectra

Abstract A variant of principal component analysis termed eigenvector analysis is shown to provide an effective parameterization of a time sequence of one-dimensional (frequency) spectra of wind-generated water waves. The analysis indicates that as few as five parameters are required to reproduce 85 percent of the variation in a climatological record of wave spectra. Reconstructed spectra reproduce spectral peaks and periods for both uni- and multimodal spectra well, particularly during high energy conditions. The potential of the functions as a universal set of functions for representation of spectra is pointed out as are applications to statistical estimation of a spectral wave climate.