Probabilistic Models for Texas Gulf Coast Hurricane Occurrences

Publisher Website
Google Scholar
Abstract
Various statistical methods are used here to analyze the occurrence of Texas Gulf Coast hurricanes. Simple Poisson, periodic Poisson, and Markov chain models are fitted to the occurrence data for a site offshore of Mustang Island, near Corpus Christi, Tex. The periodic Poisson model provides the best fit to the data and permits a fairly Poisson model provides the best fit to the data and permits a fairly simple description of cyclical occurrence phenomena. Introduction The Gulf Coast of the U. S. is frequently visited by hurricanes, resulting in very large property damages. Since hurricanes are the most severe type of storm generally affecting the region, the high winds and waves associated with such storms produce critical conditions for which many offshore and nearshore structures must be designed. Designers require some method of predicting hurricane occurrence likelihoods over time spans of several years. Such prediction models may then be used to develop the prediction models may then be used to develop the suitable design criteria for a particular structure. Study of the Gulf Coast storm records reveals the rather irregular nature of hurricane arrivals at coastal sites. This irregularity suggests the treatment of hurricane occurrences as a stochastic process; that is, a random process in time. Hurricane occurrence probabilities may be obtained from the models of such probabilities may be obtained from the models of such a random process. This investigation is concerned with the development of probabilistic models for predicting hurricanes at the site offshore of Mustang Island, near Corpus Christi, Tex. (Fig. 1). The forecasting methods used in this study are based solely upon the available historical storm occurrence data. The methods are valid only for long-term forecasts made in ignorance of future weather conditions. Thus, the probabilistic treatment of long-term hurricane probabilistic treatment of long-term hurricane prediction serves as a complement to synoptic weather prediction serves as a complement to synoptic weather forecasting procedures, which are suitable only for short-term predictions. Various stochastic models for hurricane predictions, with parameters derived from the storm data for the site, are investigated and compared. Data The assembling, categorizing, and interpreting of available hurricane occurrence data are very important steps in predicting the likelihood of future storms. The past record of storm occurrences is the information upon which the stochastic models of this study are based. For illustration, consider an "average" hurricane with maximum winds of about 85 mph. This hypothetical hurricane is moving on a northwest heading in the vicinity of the coastline. The winds, which circulate counterclockwise around the eye of the storm, usually have both a higher speed and a longer fetch on the right-hand side of the moving storm. As a result, the winds, as well as the associated extreme wave action, are generally most severe in the region preceding the eye of the storm and in the quadrant to the front and light of the storms path. As the eye of the storm approaches the coast, the over-water fetch of the winds in the left front quadrant is limited, so that wave action on the left side of a storm proceeding inland is less severe. JPT P. 279

This publication has 0 references indexed in Scilit: