Insurance and reinsurance rates for property on the gulf and east coasts of the United States are strongly influenced by predictions of the possible number and intensity of future hurricanes making landfall in these regions. The insurance industry is interested in such predictions on a wide range of time-scales. In particular, there is interest in predictions of the number of hurricanes in the next one to five years. Such predictions are used for the pricing of insurance and reinsurance contracts, and for the allocation of capital to businesses and business units.
Following the high levels of Atlantic basin hurricane activity over the past decade, a considerable amount of research regarding the influences of natural and human-induced variability on hurricane behaviour has been published. Much of this work is relevant to our question of how to predict hurricane numbers and intensities on five year time-scales, although there is still some controversy regarding the mechanisms which have caused the recent increase in hurricane numbers as well as uncertainty regarding the likely nature of future changes. Work by Webster et al. (2005), Elsner (2006), Mann and Emanuel (2006) and Hoyos et al. (2006) suggests that, in the Atlantic, there is an increasing trend in frequency and intensity of hurricanes, in part due to the rise in SSTs brought on by global warming. Trenberth (2005), however, suggests that although increasing SSTs are likely to increase intensity and rainfall from hurricanes, the effect of increasing SSTs on hurricane numbers is unclear. Dynamical prediction models have been used to estimate the impact of global warming on hurricane intensity and precipitation (Knutson and Tuleya, 1999, 2004, Vecchi and Soden, 2007) and these models show a CO2-induced increase in storm intensity and rainfall and a small decrease in storm frequency over the next century. Further studies by Klotzbach (2006) and Goldenberg et al. (2001) suggest that long-term Atlantic hurricane variability is more influenced by natural variability, such as the Atlantic Multidecadal Oscillation (AMO), than by anthropogenic changes. A number of other authors have also looked into aspects of natural and human-induced impacts of hurricane variability (for example: Kerr, 2005, and Bengtsson, 2001). In summary, there is agreement that we are currently in a period of increased hurricane activity, and there is perhaps a weak consensus forming that the intensity of hurricanes may increase in the future, but there is no consensus as to what will happen to the future frequency of hurricanes. Based on this range of evidence and opinions we develop a variety of prediction models that are consistent with different underlying physical mechanisms and interpretations of the data.
In addition to the internal variability of the Atlantic Multidecadal Oscillation and the global warmingtrend in SST, other climatic features such as the North Atlantic Oscillation (NAO) (Elsner et al., 2000) and the El Nino Southern Oscillation (ENSO) (Bove et al., 1998; Camargo and Sobel, 2005) are known to influence hurricane activity. Is this relevant for our one to five year time-scale? Lyons (2004) has shown that ENSO has predictability on up to annual time-scales but for five year time-scales there is at present no effective way to predict ENSO (and there very possibly never will be). Similarly, the NAO, is not currently predictable out to five years and therefore lacks skill for our prediction purposes. The lack of predictability of ENSO on one to five year time-scales means that the prediction methods we describe below are very different from those used in seasonal predictions (see for example Saunders and Lea, 2005 or Vitart and Anderson, 2001).
Was this article helpful?