 By Karen Clark, President and CEO, Karen Clark & Company.
Superstorm Sandy has fuelled the debate about how climate change may be impacting hurricanes and hurricane losses. Are recent storms, such as Sandy, evidence we’re seeing more destructive hurricanes due to long term climate change or are short term climate factors influencing hurricane frequency and severity? While these questions cannot be answered definitively, this article sheds light on what the actual data suggest.The bulk of evidence suggests that current hurricane experience is consistent with past experience, and climate change has not yet influenced US hurricane activity.
While Superstorm Sandy caused large losses, these losses are not unprecedented when adjusted for demographic changes, such as increasing population and property values. If the Great New England Hurricane of 1938 occurred again today, the economic and insured losses would be twice as much as the Sandy losses. Historical hurricanes such as Carol (1954) and Donna (1960) would cause
Northeast losses comparable to Sandy.
Because major hurricanes happen infrequently but property values in coastal areas grow every year, hurricane losses increase over time in nominal terms, but not necessarily in real terms. When historical hurricanes since 1900are adjusted to today’s population and property values there is no increasing trend in losses. The most damaging hurricane would be a repeat of the 1926 Great Miami Hurricane which would likely cause insured losses in excess of $100 billion and economic losses of nearly $200 billion today.
Nor do the data on hurricanefrequency, as measured by landfalling storms, show an increasing trend. The historical annual rate of hurricane landfall in the US was 1.6 for the last century. Since 2000 the average annual frequency has been 1.5. In fact, the most recent scientific consensus report published by the Intergovernmental Panel on Climate Change (IPCC), the organization that won the Nobel Peace Prize in 2007 and is at the center of climate change research, states that most global climate models project a decrease in tropical cyclone frequency due to global warming.
The consensus report does conclude that tropical cyclones are likely to become more intense over the next 20 years with peak wind speeds ranging from two to five percent higher than currently. This is likely to be a gradual change and has not been evidenced as of yet. To date, the most intense US landfalling hurricanes have occurred in 1935, 1969, and 1992. Superstorm Sandy did not have extreme wind speeds and was not the most intense storm to impact the Midatlantic region. The high losses were the result of the unusually large size and the storm track.
Near, short, and medium term hurricane models
Annual hurricane activity is influenced by factors such as El Nino Southern Oscillation (ENSO) and the Atlantic Multidecadal Oscillation (AMO), but scientists will readily admit annual predictions of tropical cyclone activity have little skill. Likewise, the catastrophe model vendor attempts to project “near”, “short”, or “medium” term hurricane activity have missed the mark entirely.
A popular fiction in hurricane loss estimationis the near, short, and medium term models are prospective views of hurricane risk not based on extrapolations of historical data.In fact, all catastrophe models are based on historical data, not unlike other models used by actuaries. The medium term modelstypically start with the long term historical view as average and then project if the upcoming period is going to be above or below this average.
These projections were initially fueled by the observation that since 1900, there have been decadal and multidecadal periods of higher than average frequency followed by below average periods. For example, between 1940 and 1959 forty hurricanes made landfall in the US for an average of two per year. From 1960 to 1979 there were only 27 landfalling storms for an average annual frequency of 1.35. But observing high and low frequency periods in the historical data is a lot easier than predicting hurricane frequency for the upcoming years.
All three modeling companies introduced near term hurricane models in 2006, and all three initially projected Atlantic tropical cyclone loss levels at least 35% above the long term average for the five year period 2006 to 2010.Whether based on studying the historical relationships between sea surface temperatures (SSTs) and hurricane activity or simply asking different scientists for their subjective opinions, all the projections were wrong. The period 2006 to2010 turned out to be well below average with respect to both hurricane landfalls and insured losses.
Conclusions
The near, short, and medium term hurricane models are based on sparse information and therefore have relatively low credibility and skill. Hurricanes are random events that cannot be predicted for one, two, or even five year periods.
While not yet evident in the data, the most sophisticated global climate models indicate global warming will lead not to more storms, but to more intense storms in the future. Rather than roughup and down frequency adjustments, this trend can be more appropriately factored into loss estimates by judiciously adjusting peak wind speed assumptions over time.
Less than 200 hurricanes have made landfall in the US since 1900. By examining the evidence and actual data, actuaries can enhance theirunderstanding of historical as well as likely future hurricane activity.
|
