The effects of uncertainties in the modeling of ocean waves is considered both with respect to fatigue and predicted extremes. A tension leg platform is adopted as the example structure and the platform is assumed to be located at Haltenbanken. A statistical model for the wave climate is presented and some uncertainties related to this model are identified. Various models for the wave spectrum are also considered. The adequacy of these models with respect to the actual structural quantity is indicated by also calculating the fatigue damage using mean measured spectra.
The stochastic long term method, Battjes (1970, 1979) , , Moan et al. , Inglis et al. , is a most convenient method for estimating the wave induced response of linear/linearized structural systems. The marginal (long term) distribution function for a given response quantity is then estimated as a weighted sum of the short term distributions for all possible sea states. Thus the probabilities of exceedance from the various sea states are properly accounted for and the problem of selecting a consistent set of design storms is avoided. From the marginal distribution function, the fatigue life IS easily estimated using the Miner Palmgren hypothesis. Extremes corresponding to a given return period are also conveniently estimated from the distribution function assuming statistical independence between the crest heights of adjacent zero-crossing cycles. An adequate formulation of the long term method requires that the wave conditions are properly described both in a short - and long term sense. Concerning the short term description, the waves are in a statistical sense completely characterized by the spectral density function, provided the sea surface elevation can be modeled as a Gaussian process. This hypothesis IS assumed to be of a reasonable accuracy as far as deep water conditions are considered.