A statistical method for modeling tropical cyclone activity

Tropical cyclones are natural hazards that occur in many parts of the world. The North Atlantic, Northeast Pacific, Northwest Pacific, Indian Oceans and the Southwest Pacific around Australia are all basins where tropical cyclones form. The potential for extensive property damage and loss of life is of great concern in many countries. An important tool for evaluating the risk associated with these natural catastrophes is to develop statistical models for Monte Carlo simulation of potential losses. Historical records of tropical cyclone activity in different regions provide useful information on the frequency and intensity of these events; however, reliance on the data alone is insufficient for risk assessment since the data is often sparse, and events can occur in places that have previously seen no activity. Furthermore, historical loss statistics are unreliable due to population growth and changes of property values. Probabilistic models can be developed to overcome these difficulties. We discuss a statistical method for modeling tropical cyclone events. We develop probability distributions for annual storm frequency, storm genesis and storm termination. We also discuss appropriate time series models for storm characteristics, such as intensity, radius of maximum winds, and forward speed along the storm track. Illustrations are based on models for the North Atlantic and Northwest Pacific Basins.