NU Online News Service, Jan. 20, 12:52 p.m.EST

Catastrophe modeling firms' hurricane damage predictionsoverestimated insured losses for a second year, according to acatastrophe prediction consulting firm.

Karen Clark & Company in a report said models designed toproject U.S. Atlantic hurricane insured losses for the five-yearperiod ending in 2010 "have significantly overestimated losses forthe cumulative 2006 through 2009 seasons."

"Hurricane activity is very difficult to project because theEarth's atmosphere is very complex and has many feedbackmechanisms," said the report. "Given all of the uncertainties,near-term projections do not have sufficient credibility to be usedfor important insurance applications such as product pricing andestablishing solvency standards."

Modeling firms said the reports conclusions were flawed, basedon wrong assumptions and misrepresentations.

In a report last year, the Boston-based firm found modelingoverestimates for the 2006 through 2008 seasons.

This year's Karen Clark report said catastrophe modelers AIRWorldwide, Eqecat and Risk Management Solutions initially projectedinsured loss levels at least 35 percent above the long-term averagefor the period 2006 through 2010. AIR lowered its figure toapproximately 16 percent in 2007, and Eqecat has made only minoradjustments to its original estimate of loss increases of between35- and 37 percent. RMS introduced modifications to its model in2009, but still predicted losses at 25 percent above the long-termaverage.

Assuming long-term average annual insured hurricane losses of$10 billion per year, these figures translate into cumulativeinsured losses for 2006 through 2009 of $48.8 billion, $54.5billion and $54.6 billion, respectively, for the AIR, EQECAT andRMS models, according to the report's calculations.

Karen Clark said the actual cumulative losses were $13.3billion, far lower than the model predictions, and only one-thirdthe long-term cumulative average of $40 billion. The 2009 Atlantichurricane season was below average in the number of named storms,hurricanes and major hurricanes, and was the lowest frequency yearsince 1997, it was noted.

The report noted that hurricane activity is influenced by manyclimatological factors, many of which are known, but some unknown,by scientists. It stated that there are complicated feedbackmechanisms in the atmosphere that cannot be quantified preciselyeven by the most sophisticated and powerful climate models.

The report recommended that insurers, reinsurers and regulatorsevaluate the efficacy of the near-term hurricane models in light ofthis uncertainty.

"Catastrophe models are powerful, broad-based tools that arevery good for particular applications," said Karen Clark, presidentof her self-named firm. "However, model users must recognize thatthere can be a very wide range of estimates associated with a givenmodel metric, such as average annual loss."

The report also addressed what has been termed the HurricaneFrequency Paradox.

It said some scientists suggest there has been an increase inAtlantic tropical activity, based on growth in the tropical cyclonecounts since data was first compiled in the late 19th century.Paradoxically, this apparent increase has not resulted in anincrease in hurricane landfalls in the United States.

It said researchers at the National Oceanic and AtmosphericAdministration (NOAA) have now concluded that the increase inannual storm frequency is in large part attributable toimprovements in observational technology leading to the increaseddetection of tropical storms and hurricanes.

"If one estimates the number of storms prior to 1970 that werenot detected, there appears to actually be a slight decreasingtrend in storm frequency. In addition, we have not seen anincreasing trend in hurricane losses when historical losses arenormalized to current exposure values," said Ms. Clark.

RMS said her report demonstrated "a fundamental misunderstandingof the purpose of Cat models by asking whether they can be used topredict actual catastrophe experience in a particular one, two, orfive-year period. It's important to understand that catastrophemodels deliver probabilistic forecasts not deterministicpredictions. A probabilistic activity forecast means that, onaverage, over many five-year periods, this is the number ofhurricanes to be expected. The actual number experienced in aparticular five-year period will be just one sample from a broaddistribution of possible outcomes.

Last year, in response to the report, Peter Dailey, AIR'sdirector of atmospheric science, said his firm rates hurricane riskbased on years in which sea surface temperatures (SSTs) were abovethe long-term climatological mean. The Warm SST catalog, he said,"is not intended to predict insured losses in a particular year.Instead, it estimates the sensitivity of landfall risk to a typicalwarm ocean condition," and AIR offers its Warm SST catalog as asupplement to, rather than a replacement for, its standard catalogof Atlantic hurricane activity.

Eqecat last year said that the firm had misrepresented "thebasis of natural catastrophe models for the infrequent risks whichundercuts its conclusion."

The Karen Clark report, "Near Term Hurricane Models -Performance Update," is available at www.karenclarkandco.com.

Karen Clark & Company provides software products andconsulting services for insurance companies to evaluate and improveexposure data processes, to understand catastrophe risk, and toutilize models and model results.