Synthetic Tropical Cyclone Generation for Risk Analysis

Hurricanes and other tropical cyclones (TCs) are a major natural hazard that pose significant threats to countries and communities along coastlines. Understanding the spatiotemporal patterns of TCs is essential to predict the potential hazards and risk. Most of the previous TC genesis models were based on parametric estimation methods that estimated the probability density of TC occurrence and genesis locations by assuming a parametric probability function. However, historical TC data may not satisfy the assumed function. To alleviate this challenge, a nonparametric estimation method, KDE, is applied. Accordingly, there is no need to assume a specific probability function. This study developed the TC genesis locations and time based on both historical data and global climate modeling data. First, kernel density estimation was applied to obtain the probability density of annual occurrence number and genesis location and time of TCs in the Atlantic Ocean using historical TC data. A standard normal density function was applied as a kernel function. Then, Monte Carlo simulation was applied to generate 10,000 years of TC events. This genesis model has been validated by comparing the probability density obtained from historical TC data, in the dimension of genesis time and location. This TC genesis model lays a solid foundation for future full-track model development of TCs in the North Atlantic basin that will incorporate climate change.