This week marks the third anniversary of Hurricane Katrina, butthat's not the only fact that prompted the lead article of ourAugust E&S/Specialty Lines e-newsletter about hurricaneforecasts, "Global Warming Not Linked To Increased HurricaneActivity."

The idea that specialty insurers and brokers might be interestedin the science of global warming and the possible link to hurricaneactivity didn't occur to me until I attended an AAMGA UniversityWeekend earlier this year.

During a break period of one particularly well-attended class,"Climate Change & Its Implications for the Property-CasualtyInsurance Industry," a local wholesaler told me his firm had juststarted placing coastal property coverage. While understandingintricacies of weather science isn't necessary to place insurance,he thought there was something to be gained from understanding whatclimate studies might reveal about future needs for windinsurance.

As the year progressed, catastrophe modeling firms sentannouncements to NU highlighting the fact that their models werebeing tailored and used by wholesale brokers--allowing them toanalyze specific catastrophic exposures on behalf of retail brokersand their insureds; to access real-time, natural hazard riskreports; and to use a variety of platforms to make sure theircustomers were getting the best quotes they could fromcarriers.

So when I happened to read a widely published Associated Pressarticle in May about a scientist from the Geophysical FluidDynamics Laboratory of the National Oceanic and AtmosphericAdministration who had changed his view on the link between globalwarming and hurricane frequency, it seemed like something worthexploring further.

As reported in Ara Trembly's lead article for this edition ofE&S/Specialty Lines Extra, the new NOAA view argues against thenotion that any recent tropical cyclone can be traced to globalwarming.

Further, and perhaps of more interest to participants in theproperty-casualty insurance market, the view suggests--based oncomputer simulations through the end of the century--that warmingmight actually decrease hurricane frequencies globally over thelong term.

Personally, one element of the AP story that had actually piquedmy interest was the identity of the researcher, Tom Knutson ofGeophysical Fluid Dynamics Laboratory of the National Oceanic andAtmospheric Administration, and another scientist, Kerry Emanuel ofthe Massachusetts Institute of Technology, who argued against hisnew view. Both had participated on a panel of experts convened byone of the cat modeling firms--RMS--in 2006 to help the Newark,Calif.-based firm develop a new five-year view of hurricaneactivity.

RMS, of course, was not alone. In the wake of criticisms thatcat models had underestimated hurricane activity in 2004 and 2005,the modeling firms put out these shorter-term views to replace100-year models and reflect a higher observed frequency ofhurricanes in recent years.

"The increased frequency and severity of hurricane activity inthe Atlantic Ocean Basin, as observed since 1995, are driven byhigher sea surface temperatures in the tropical North Atlantic andby associated changes in atmospheric circulation," RMS wrote in atypical statement, suggesting a high degree of certainty aboutcause of observed activity.

Given the new level of uncertainty in the scientific community,it seems reasonable to ask about possible insurance implications ofthe current thinking. Should cat models now change to reflect a newview of the world? Will insurance prices go down?

As reported in Ara's article, representatives of modeling firmsare now voicing just as much as uncertainty as GFDL scientistsabout the link between global warming and hurricane activity--andabout the likely long-term frequency of hurricanes.

Peter Dailey, director of Atmospheric Science at AIR Worldwidein Boston, told Ara: "There is now a near consensus that global airtemperatures are increasing, [but] no consensus on how this hasaffected the temperature of the world's oceans"--the Atlantic, inparticular.

Christine Ziehmann of RMS added: "It is not clear what effectglobal warming is having, and will have, on the frequency,intensity and geographical distribution of hurricanes in theAtlantic basin."

Computer models suggest that global warming should, in the longterm, lead to less frequent tropical cyclones globally, but they're"less clear about hurricane activity in individual ocean basins,"she said, noting that some indicate increased long-term frequenciesand some decreased long-term frequencies of Atlantic storms.

I haven't juxtaposed the 2006 statements with current commentsto highlight discrepancies between them. In fact, as Ara pointedout when I pressed him to investigate insurance implications ofwhat seems like a changed view of hurricane frequency, a carefulreader may conclude that nothing has changed to alter modelassumptions.

After all, the latest statements about potentially lowerfrequencies refer to global long-term trends--over multipledecades--not five years in the Atlantic Basin, he noted. Inaddition, RMS's 2006 press release disclosed that the expertsconvened by the firm held "different climatological perspectives onthe underlying causes of elevated hurricane activity."

But more casual readers are bound to ask questions. Without oldpress releases at their fingertips to compare exact phrases wordfor word, what most of us recall is that modelers told us warmingwas producing more hurricanes--which hiked their model results, andproperty rates went up along with them.

Despite the fact that hurricane levels were at record-highlevels, with no major storm making landfall, critics focusing onconsumer issues rather than insurer strength and solvency arelikely to start putting modeling firms and insurers on the hot seatagain--this time not asking why models are underestimating losses,but why they are overestimating them.

Such questions are already being raised. The July 1 edition ofThe Wall Street Journal, for example, ran a page one article,"Insurers Criticized For New Rate Models."

I asked Ara to give RMS a chance to address potentialquestions--focusing first on the question of whether the firm'sview has changed since 2006, when an official said its expert panel"agreed unanimously that a forward-looking view of risk shouldreflect a higher probability of landfalling hurricanes thanrepresented by long-term historical averages."

RMS responded: "The views of RMS remain based on the mostup-to-date scientific evidence and reflect the current range ofexpert opinion within the hurricane research community.

"It is clear there has been an increase in the annual frequencyof hurricanes in the Atlantic basin since 1995, and that estimatesof hurricane numbers over the next five years based on a simpleaverage of frequency since 1900 are likely to be too low.Similarly, the proportion of strong hurricanes has increased sincethe 1970s such that a simple average since 1900 will also tend tounderestimate the proportion of Category 3-to-5 hurricanes over thenext five years.

"There is currently no consensus among researchers about whetherthese changes in frequency and intensity are caused by globalwarming or natural cycles in climate.

"Computer models of the effect of global warming on tropicalcyclones project many decades into the future and so do notdirectly apply to hurricane activity in the recent past or nearfuture."

Another relevant question to ask may be how short-term modelscan even hope to be reliable when there's now little consensus overthe longer term?

Related questions about the appropriate uses of short-termmodels have been asked before--by Karen Clark, founder of AIR, andnow president and chief executive officer of Karen Clark &Company.

Even back in the days when she was president and CEO of AIR, shetold NU that she worried about wild swings in insurance premiumsthat would occur if insurers relied on short-term models forpricing purposes, suggesting that short-term views be used only forshort-term decisions, like how much reinsurance to purchase nextyear. (See related article, NU, March 13, 2006.)

Going forward, only one thing seems certain--the models willnever give a perfect view of storm activity, leaving modelers toblame users, and users to blame models for inaccuracies in ratesand shortfalls in capital that emerge as a result.