CSU Hurricane Forecasts: Where’s the value?

The annual April seasonal hurricane forecast is out today from the folks at Colorado State University (CSU) led by Phil Klotzbach and Dr. William Gray. After the past few years of iffy forecast results, some are asking; where's the value in producing seasonal hurricane forecasts?

The 2009 forecast was a huge bust for the CSU team. On April 7, 2009, the CSU team issued the spring updated forecast for the 2009 season. They called for near-average activity in 2009 of 12 named storms and 6 hurricanes.

Instead, 2009 produced just 9 named Atlantic storms and only 3 hurricanes. That's the fewest number of Atlantic hurricanes since 1997. No hurricanes made landfall in the United States in 2009.

Tracks of named storms in the 2009 Atlantic hurricane season.

Looking back at the past 5 years of April CSU hurricane forecasts, the accuracy of the CSU seasonal hurricane forecasts is at best mixed. They had a very good year in 2008, but in every other year the forecasts have show an error rate of 33% or higher in either the number of named storms or hurricanes. In addition to badly missing the lowest number of hurricanes since 1997 in the 2009 forecast, CSU's April 2005 forecast of 13 named storms and 7 hurricanes completely missed the notion of the most active Atlantic hurricane season on record. In addition to the devastating Hurricanes Katrina and Rita, 2005 produced a record 28 named storms and 15 hurricanes in the Atlantic. The CSU team under forecast both the number of named storms and hurricanes in 2005 by over 50%.

There are several groups in addition to the renowned CSU forecasters that issue seasonal hurricane forecasts. NOAA, and private forecasters such as Britian's Tropical Storm Risk also produce seasonal hurricane forecasts. Many in the insurance industry are now questioning the accuracy and usefulness of seasonal hurricane forecasts.

A recent post from The National Association of Professional Insurance Agents asks the following question:

"What's the Insurance Value of Pre-season Hurricane Predictions?"

At the start of December a pair of forecasters at Colorado State University kicked off the 2010 hurricane prognostication season with a call for an above-average hurricane season in the Atlantic. A year prior, the same team predicted 14 named storms, seven hurricanes and three major hurricanes for 2009. The 2009 hurricane season ended November 30 with nine named storms, only three of them hurricanes, in the Atlantic. These were the lowest totals since 1997. No hurricanes reached the U.S. mainland.

At the start of the 2009 hurricane season in June, the researchers at Colorado State lowered projections to 11 named storms, five of which were to become hurricanes. This caused some in the insurance industry to question the value of advance predictions for hurricane season. A.M. Best's Best Week talked with industry experts. Bottom line: forecasts are useful for seeing trends, but not useful for setting insurance rates.

Karen Clark, a pioneer in the catastrophe-modeling field and president of Karen Clark & Co. recently released an analysis of near-term hurricane models and found they "do not have sufficient credibility to be used for important insurance applications such as product pricing and establishing solvency standards." "There needs to be a deeper understanding of the inherent uncertainty in modeling," Clark said. "This means real money to real people."

February 2, 2010

The bottom line for those living in hurricane country is that you have to prepare for each hurricane season in the same way. The great variability in the number and intensity of storms from one year to the next does not seem to allow us to extract great value from seasonal hurricane forecasts.

Short term hurricane forecasts have more value:

The real, demonstrable value lies in short term hurricane forecasts such as those issued by the National Hurricane Center. A 2004 paper in The Journal of Applied Meteorology estimates the value of existing 48-hour hurricane forecast information to oil and gas producers averaged roughly $8 million per year during the 1990s, which substantially exceeds the operating budget of the National Hurricane Center.

It appears clear that the real return on investment value in improving hurricane forecasts lies in the continued improvement in short term hurricane landfall and intensity forecasts. The seasonal hurricane forecasts make great headlines, but are they really worth the time, effort and investment in research dollars?

PH