In offshore engineering, JONSWAP shapes are commonly used to fit wave spectra and characterise the peakedness of the sea state. The latter is defined as γ, the so-called peak enhancement factor, which is as large as the peak is acute, with values generally ranging from ∼1 to 7 (1 and 3.3 being usually adopted for fully-developed and average wind seas respectively). The determination of γ values permit to simulate realistic sea conditions later on, in particular for determining extreme loads on floating structures in severe sea states.
The consistency of the fitting with respect to the actual sea state is subject to — at least — two key, connected aspects, which are: the frequency resolution (i.e., the spacing between two consecutive spectrum estimates) and the sampling error (i.e., the statistical variability of estimates due to both the finiteness of the wave record and the spectral averaging).
Based on realistic sea state spectral simulations, this work addresses the sensitivity of the JONSWAP fitting to these aspects and shows that it systematically results in a bias in the peakedness estimate — the bias being aggravated when the sampling error is large for a given signal length (from 10min to 1h). It is shown, in particular, that the error committed with respect to the true γ value is generally higher than 5% for most types of peaks and may often exceed 30%.
The aim of this work is to emphasize this issue for practical offshore and oceanographic applications for it may, in turn, potentially introduce a bias in the dynamic response of floating structures in numerical simulations and tank tests.