pleased he was wrong about the number of hurricanes which he forecast would occur last year.

The number of hurricanes in the Atlantic region last year was below normal, and below what was predicted by Mr. Gray, a professor of atmospheric science at Colorado State University.

On his Internet site Mr. Gray attributes the failure in his predictions to the early onset of El Nino, which was more intense than normal. El Nino refers to the warming up of southern Pacific waters every two to seven years and its effect on global weather patterns.

For insurers an El Nino year usually means a below normal level of hurricane activity in the Atlantic regions. Mr. Gray factored in this historical data when making his prediction for last year's hurricane season, which occurs between June to November.

Still El Nino took Mr. Gray and other hurricane forecasters by surprise. There were seven named storms -- about two below average -- and three hurricanes -- about four below average -- during the season. The storms also lasted for a shorter time than average. On average they persisted for about ten days. The average is 24.

"Our early 1997 forecasts called for a slightly above average hurricane season while our 6 August forecast was for an average season,'' Mr. Gray and other researchers stated on their Internet site. "This overforecast of activity was a result of our inability to anticipate that we would experience the most extreme El Nino event ever to be recorded during the period of June through October. This unprecedented El Nino dominated other normally hurricane-enhancing forecast signals, leading to a larger reduction in this year's hurricane activity than would have occurred if this El Nino had not been so intense.'' The below normal hurricane activity contributed to the drop by about 50 percent in insured damage from natural catastrophes worldwide in 1997, compared to 1996.

The reinsurers may have saved money by paying out less for claims due to hurricane damage, but ironically they want forecasters like Mr. Gray to get it right the next time. In an unpredictable world, reinsurers and insurers want to minimise their risks as much as possible through accurate forecasting. Such forecasting helps them set their rates at levels which will allow them to reap profits.

That reinsurers and other companies in the catastrophe business have continued funding the Bermuda Biological Station for Research as part of its Risk Prediction Initiative (RPI).

Now in its second year the project has already helped at least one reinsurer adjust its catastrophe prediction model. Bermuda-based Renaissance Reinsurance Ltd. invested $8 million in computer technology to help the company forecast the occurance of natural disasters.

Jayant Khadilkar, head of Renaissance's catastrophe modelling department, said the Biological Station's research has helped the company adjust its model regarding windstorms.

"We have gotten some really good information that either validates our catastrophe modelling or has helped us refine the assumptions in the modelling,'' he said.

He described current catastrophe modelling as still "primitive'' in terms of the quantification and dating of historical data used in predicting future storm occurance.

The project has also contributed to the industry by opening up a dialogue with the scientists involved in the research.

"We are extremely happy with the project,'' Mr. Khadilkar said.

Renaissance and ten other funders therefore have renewed or are in the process of renewing their contributions to the project. Contributions in the first year totalled about $1 million.

The money was used to fund 16 research projects. Through the research the scientists hope to extend the historical record of hurricanes as far back as possible.

Actuaries like Douglas Collins, who is with insurance consultancy firm Tillinghast-Towers Perrin, said extending the record will help them calculate hurricane probability more accurately. Tillinghast-Towers is one of RPI's founders.

"The research is important from a pricing standpoint in determining how often a certain size hurricane will occur,'' Mr. Collins said. "We only have historical records back for 100 years or so. It's important to know if the biggest storms have a frequency of once in 50 years, once in 100 years, or once in 300 years. It would be better to have 1000 years of experience...you would have comfort if you had the probability right.'' For example, RPI funded researcher Kam Biu-Liu of Louisiana State University has already found evidence that an intense hurricane hits Alabama every 600 years. RPI money helped fund his study of coastal pond sediments in the region.

RPI money is now being used to fund a multi-year study by meterologist Chris Landsea of the National Hurricane Centre in the US. Mr. Landsea will try to link more accurately the effects of El Nino on tropical cyclone activity around the world. The information could help reinsurers and insurers determine how to spread their risks.