Season Review: Fourth Consecutive Year of Accurate Hurricane Predictions from Colorado State University Forecast Team

Note to Editors: Forecast verification totals are in the attached chart. The complete 2002 season verification report, as well as related research, detailed descriptions of forecast factors, previous forecasts and explanations of individual storms, are available on the Web at or

William Gray and the renowned Colorado State University hurricane forecast team were on target with their later updated seasonal forecasts for a suppressed 2002 hurricane season. This was the fourth consecutive year that Gray and his colleagues were able to provide skillful predictions of hurricanes, hurricane days, intense hurricanes and intense hurricane days.

In early September, the team’s updated forecast predicted eight named storms, three hurricanes and one intense hurricane for the season, down slightly from their early August update of nine named storms, four hurricanes and one intense hurricane. As of Nov. 21, observed totals were 12 named storms, four hurricanes and two intense hurricanes.

"We consider this a successful forecast year with regard to hurricane activity but not for the large number of weaker sub-tropical storms that developed," said Gray. "Overall, we are making very good progress on improving statistical hurricane forecasting techniques and in better understanding why there is such variability in month-to-month and year-to-year Atlantic basin hurricane activity."

Other early August and early September forecasts for the season’s hurricane activity also came close to the mark. The Colorado State team predicted 10 hurricane days while 11 were observed, and two intense hurricane days were forecast while 2.5 occurred. Hurricane Destruction Potential, a measure of a hurricane’s potential for wind and storm surge destruction, was also on target. The team also predicted an early end to the 2002 hurricane season which witnessed no tropical cyclone activity after early October.

However, this year’s forecast was not as accurate regarding the number of named storms and named storm days. The team’s early August forecast called for nine named storms while 12 occurred, and predicted 35 named storm days and witnessed 54; the early September forecast was somewhat less accurate.

"We underestimated and do not well understand why such a large number of weak tropical storms formed at high latitudes and never gained hurricane strength," said Gray. "However, this is a continual learning process and we have become a little better educated as to why so many weaker systems formed in September."

This year’s unusual storm climate and ocean activity encouraged Gray and his team to significantly reduce their December 2001 and early April predictions that had anticipated an above-active storm season. In December, the team’s extended-range forecast called for 13 named storms, eight hurricanes and four intense hurricanes. However, according to Gray, an unanticipated massive rearrangement of global ocean and atmospheric conditions caused the forecast team to lower their later updated predictions to reflect this change to more inhibiting conditions.

"Although our December and April forecasts did not anticipate such low hurricane numbers, our late May, early August and early September updates captured much of the decrease in hurricane activity for this year," said Gray. "Our forecasts do not generally change significantly throughout the course of the season. However, the type of unpredictable change in hurricane signals witnessed this year is why we issue seasonal updates, to allow adjustment for changing conditions."

According to Gray, the forecasts were consistently reduced during scheduled updates for two primary reasons: a stronger than anticipated El Nino and an uncharacteristic weakening of the Atlantic thermohaline circulation that rapidly cooled the tropical Atlantic and brought higher Atlantic surface pressures.

Since 1995, Gray and his colleagues had witnessed an increase in the Atlantic thermohaline circulation. This increase warmed North Atlantic sea surface temperatures and lowered surface pressure which enhanced Atlantic basin hurricane activity. Additionally, annual hurricane activity since 1995 has averaged the highest on record. This year’s weakening of the Atlantic thermohaline circulation rapidly cooled the tropical Atlantic during the early summer months, inhibiting hurricane activity. This trend reversed in September, helping to again increase storm activity.

"As the season progressed, changes in Atlantic conditions and the development of a moderate El Nino event were the primary factors that led us to lower our winter and early spring forecasts," said Gray.

The forecast team originally expected a weak El Nino in the Pacific. However, as a moderate El Nino formed and was present throughout most of the hurricane season, the team correspondingly lowered the later forecast updates accordingly.

Some other unique characteristics of the 2002 hurricane season include the following.

  • The first hurricane, Gustav, did not form until Sept. 11, the latest date for the first hurricane of the season to form since 1941.
  • Although there were 12 named storms, only four developed into hurricanes and two of those were marginal (Gustav and Kyle). Isidore and Lili were the only significant hurricanes of this season. No year since 1950 has had as high a ratio of named storms to hurricanes.
  • Also since 1950, only three other seasons have had as many as 12 named storms with as few as only two intense hurricanes.
  • This season witnessed the first U.S. hurricane landfall since Irene in October 1999. Before Lili made landfall on Oct. 3, 2002, a record 21 consecutive Atlantic basin hurricanes did not make shore along the U.S. coast.
  • 2002 was a very active year for landfalling weak tropical storms with six making U.S. landfall. Since 1900, no more than five sub-hurricane intensity storms had made landfall in any one year.
  • Hurricane Kyle, which lasted from Sept. 20 to Oct. 12, was the third-longest tropical cyclone on record behind only Ginger, which lasted 21 days in 1971, and Inga, which lasted 15 days in 1969.
  • Eight named storms formed during September, a new record for the month. However, there was relatively little hurricane activity during the month.
  • 2002 was only the third year in the past 30 with no Atlantic basin tropical cyclone activity in October and November.

The Colorado State team measures their forecast skill by their ability to predict the variation of forecast parameters from climatological averages. Using this analysis, the team’s forecast skill has shown significant improvement over the past four years. The average forecast skill of the last four years at the early August period for their Net Tropical Cyclone activity was close to 80 percent.

William Gray and his Colorado State University team have provided seasonal Atlantic basin hurricane forecasts for the past 19 years. Until Gray began developing his forecast model in the early 1980s, there were no objective methods for predicting whether forthcoming hurricane seasons were likely to be active, inactive or near average. Ongoing research by Gray and his colleagues has since indicated that there are a number of atmospheric and oceanic precursor circulation signals that can be used for skillful Atlantic basin seasonal hurricane activity as early as December the prior year.

Extended-range forecasts are issued in early December with updates in early April, early June, early August and early September. The early August forecast now includes separate forecasts of August-only and September-only hurricane activity. The development of each new forecast and update emphasizes the current trends in oceanic and atmospheric conditions and their past years of association with seasonal and monthly hurricane activity.

The team is already working on their first seasonal forecast of 2003 Atlantic basin hurricane activity. This forecast report will be issued on Dec. 6, 2002. In addition to Gray, team members include Christopher Landsea, Philip Klotzbach, John Sheaffer and others.


Tropical Cyclone Parameters and 1950-2000 Climatology (in parentheses) December 7 Forecast April 5 Forecast May 31 Forecast August 7 Forecast September 2 Forecast Observed 2002 Totals
Named Storms (9.6)* 13 12 11 9 8 12
Named Storm Days (49.1) 70 65 55 35 25 54
Hurricanes (5.9) 8 7 6 4 3 4
Hurricane Days (24.5) 35 30 25 12 10 11
Intense Hurricanes (2.3) 4 3 2 1 1 2
Intense Hurricane Days (5.0) 7 6 5 2 2 2.5
Hurricane Destruction Potential (72.7) 90 85 75 35 25 31
Net Tropical Cyclone Activity (100%) 140 125 100 60 45 80

* Number in ( ) represents average year totals based on 1950-2000 data.