Optimizing a new de-clustering approach for relatively small samples of wind speed with an application to offshore design conditions

The effect of extreme wind speeds in applications for design is of great interest in a variety of fields such as meteorology and coastal engineering. In these fields a common problem is the scarcity of long datasets. To overcome this limitation, a common approach is to utilize the entire available dataset using the Peak-Over-Threshold (POT) approach. In small samples there may be a limited number of extremes and so re-sampling is often beneficial. However, the re-samples are often affected by dependency and the independence limitations are usually disregarded. To alleviate this effect, the DeCA Uncorrelated (DeCAUn) model is proposed taking into account the correlation effect when re-sampling. This model provides an improvement to the current physical De-Clustering Algorithm (DeCA), by re-sampling the samples of DeCA irregularly in time. The methodology proposed in this assessment is illustrated using wind speed data from a high resolution database over the North Sea, the Atlantic Ocean and the Mediterranean Sea. From this evaluation, the DeCAUn model is proposed as an alternative re-sampling strategy for observations irregularly spaced in time.