Hindcasts of wave conditions can be subject to large uncertainties, especially in storms. Even if estimates of extremes are unbiased on average, the variance of the errors can lead to a bias in estimates of extremes derived from hindcast data. The convolution of the error distribution and wave height distribution causes a stretching of the measured distribution. This can lead to substantial positive biases in estimates of return values. An iterative deconvolution procedure is proposed to estimate the size of the bias, based on the measured distribution and knowledge of the error distribution. The effectiveness of the procedure is illustrated in several case studies using Monte Carlo simulation.

This content is only available via PDF.
You do not currently have access to this content.