Assessment of the Impact of Climate Variability on Modeled Atlantic Hurricane Losses

Tropical cyclone activity over the Atlantic Ocean has shown considerable variability in the last 110 years. Hurricanes occur more frequently in certain decades than others as a result of interannual variability, such as El Nino/Southern Oscillation (ENSO); and multidecadal variability, such as the Atlantic Multidecadal Oscillation (AMO). Hurricane frequencies in catastrophe loss models have traditionally been based on a long-term view, e.g. uniform consideration of all historical data since 1900. However, many financial contracts tied to hurricane risk (e.g. reinsurance, catastrophe bonds, etc.) have lengths ranging from one to three years. Further, eras of active seasons put a strain on the financial resources of companies bearing hurricane risk, particularly due to the tendency of multiple active seasons to occur within decadal time scales. For these reasons, there is a need to better quantify the frequencies on these time scales. To meet this need, EQECAT has developed a near-term Atlantic hurricane frequency model, representing the best estimate of the risk over the next few years.

ENSO occurs on a time scale of two to seven years and has considerable impact on hurricane activity due to increased vertical wind shear in the main development region of the Atlantic Basin during El Nino events. Because of the short time scale of the ENSO cycle and the possibility of onset of an El Nino event occurring midseason, use of ENSO in the near-term model is impractical. The AMO is manifested in the variation of sea surface temperatures in the Atlantic Ocean. The warm phases of the AMO are believed to be associated with eras of active seasons such as the period 1926-1969 and the most recent period that began in 1995. EQECAT's current near-term hurricane frequencies are based on the historical events that occurred during the warm phases of the AMO. The near-term risk corresponding to the warm phases of the AMO produces an average annual loss (based on modeled losses from historical events) that is 29% higher than the long-term risk.