Using Catastrophe Models to Analyze the Probabilities of Recent Tropical Cyclone Records

The 2017 Atlantic hurricane season brought many records for the basin and for the US and Caribbean islands, not the least of which was the end to the drought for major hurricane landfalls in the US. Other records that were set in 2017 such as two consecutive category 4 US landfalls within a fifteen day period (Harvey and Irma), three consecutive major hurricanes in the basin (Harvey, Irma, Jose) and three simultaneous hurricanes in the basin since 2010 (Irma, Jose, Katia), all of which may seem esoteric from the standpoint of landfall frequency and intensity, are nonetheless important to evaluate probability-wise from an insured and economic loss perspective because such behavior can impact annual aggregate loss, demand surge, and business interruption in a region.

Given that the historical record may be too brief to accurately capture all the activity that is possible, catastrophe risk models provide a convenient tool to provide a more complete picture. These models, specifically the hazard component, are built using most if not all of the historical data and can therefore in some way indicate what is possible even though it has not yet occurred at all or with appropriate probability. A typical catalog will contain ten thousand or more years of tropical cyclone activity – any one of which could play out in the near future. Each year has information on the number of storms, their location, and their intensity by hour.

In this talk, some of the recent record and near-record tropical cyclone activity will be evaluated in terms of likelihood of occurrence for the Atlantic and Pacific (west, central, and east) basins in the northern hemisphere using hazard models developed at AIR-Worldwide. Comparison to the historical record will also be done to provide some perspective as to what extent climate change may be playing a role in generating the recent record activity.