Historical methods are typically used to determine the 100-year seastates used in traditional design practice. In these methods, one fits a probability distribution to the historical record of Hs and wind speed. With the development of more rational design standards, interest has shifted to rare events-with annual exceedence probabilities of 10−4 or lower. If one uses historical methods to determine the seastates associated with these probabilities, one typically encounters high uncertainties as a result of limited sample size and modeling uncertainty associated with the choice of distribution shape. This paper presents the results from Phase-I of an industry-sponsored project aimed at the development of more reliable methods for the calculation of seastates (particularly Hs) associated with rare probabilities. These methods construct deductive (or physically based) probabilistic models of the storm characteristics and their effects, and then use these models (together with the laws of probability theory) to calculate the probabilities of rare seastates. The objective of this first (exploratory) phase was to develop and apply a deductive model for hurricanes affecting platform sites in the western Gulf of Mexico, compare the predictions of this model to those of the historical approach, investigate and compare the uncertainties in both approaches, and explore other benefits of using the deductive method. For the sake of efficiency and internal consistency, a simple parametric-hindcast method was used for both the deductive and historical approaches. Results from this project indicate that the deductive and historical methods predict identical values of Hs for annual exceedence probabilities of 10−2 per year. For lower probabilities (particularly for 10−4), the predictions by the deductive method are within the range of predictions by the historical method using different assumptions about distribution shape. The statistical uncertainty of the deductive method is comparable to that of the historical method (if the latter uses pooled data coming from the entire static region). If no pooling is performed with the historical method, the uncertainty in the deductive method is lower. In addition, the deductive method shows less sensitivity to distribution-shape assumptions. This paper also summarizes ongoing efforts to adapt the deductive approach to extra-tropical storms in the North Sea — which have more complicated wind fields than hurricanes — and to the entire Gulf of Mexico — with the associated geographical variation in storm rate and characteristics.

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