The present paper is concerned with the accurate estimation of wave induced design responses in model test campaigns where only a relatively low number of seastates can be simulated.
Three response types are considered, namely an essentially linear response type such as the linear elevation of wave crests, a weakly nonlinear response type such as the second order crest elevation, and finally a discontinuous badly behaved response type such as wave in deck loading.
For each of these response types, the 100 year and 10 000 year response levels are calculated for a Norwegian Sea environment. The fractiles at which these response levels occur at the governing seastate along the 100 year and 10 000 year contours, are also calculated.
Having established these quantities, the efficiency of three methods is evaluated. Firstly an extreme value approach with a Gumbel assumption is considered and it is evaluated how many three hour simulations are required to obtain a good estimate of the relevant fractile levels in the governing seastate. Similarly, a peak over threshold analysis of maxima is considered and the number of seastates required is compared with the Gumbel results.
Finally, the 100 year and 10 000 year response levels are estimated directly by employing an iterative procedure on the probability integral of the long term distribution of the responses together with a simple interpolation procedure. It is found that the latter procedure is efficient, particularly for strongly nonlinear response types.