Hurricane Guesswork

by Conroy

This past week tropical storm Alberto churned off the coast of Georgia and the Carolinas, ushering in an early start to the Atlantic hurricane season [1]. For residents along the U.S. Atlantic and Gulf coasts, the hurricane season is a period of low-grade angst. There’s a relatively remote chance that coastal areas will be hammered by a truly devastating storm, but it happens. Since 1950 there have occurred, on average, 11 named storms [2], 6 hurricanes [3], and 3 major hurricanes (Category 3 or higher [4]), per year. A fraction of these, one or two hurricanes each year on average, will hit the U.S. mainland. And rarely, maybe once or twice a decade, a major hurricane makes landfall with catastrophic results, as witnessed by the deaths and damage caused by Camille, Hugo, Andrew, and Katrina, to name only the most prominent.

Recognizing the danger of hurricanes, it’s without question important to identify and follow every tropical depression as it forms, predict its storm track, warn people living in potentially affected areas as far in advance as possible, and prepare for emergency response as warranted. The news media and various local, state, and federal public agencies do the first three very well. Sadly, as vividly demonstrated by Katrina, emergency response can sometimes be hindered by bureaucratic obstacles and lack of initiative.

Hurricane Forecasting
The interest in hurricanes also leads to far more dubious types of planning and forecasting, namely the annual Atlantic Seasonal Hurricane Forecast. The most prominent of these is published as part of the Tropical Meteorology Project of Colorado State University [5], which was previously led by Professor William Gray and is now led by Professor Philip Klotzbach.

Gray and Klotzbach will explain that they use Atlantic and Pacific water temperatures, global weather patterns like El Nino and La Nina, and other climate factors to predict hurricane activity and probabilities for strikes in the Caribbean and North America. For example, this year’s forecast (from April) predicts a relatively quiet season with a total of ten named storms (tropical storms or hurricanes), four hurricanes, and two major hurricanes.  In addition, it places the probability of a hurricane making landfall somewhere along the U.S. coast at 42%.

Call me a skeptic, but I doubt the ability of anyone to accurately predict the number and intensity of Atlantic hurricanes, let alone the location of where they might develop and track, for any given season. Meteorologists can’t predict the weather for any one location with much accuracy even a week in advance. How can they predict the development of major weather systems over huge expanses of the equatorial and North Atlantic many months ahead of time (the Atlantic hurricane season usually peaks in early- to mid-September)? Add to that the fact that the dynamics of hurricane (or any tropical cyclone) formation are not entirely understood and the various factors that lead to hurricanes are specific to local weather conditions and cannot be predicted in advance.

Yet predictions are made each year, and each year they are given mention in the news. For that reason, Gray’s and Klotzbach’s forecasts deserve to be tested for their accuracy.

Bad Predictions
One way to check whether forecasts have any predictive value is to examine how previous forecasts have aligned with actual hurricane activity. This table summarizes predicted versus actual Atlantic hurricane activity from 1999 (the first year of Gray and Klotzbach forecast data available on-line) through 2011.

Number of hurricanes: named storms / hurricanes / major hurricanes
Year	CSU Predicted Number of Storms	Actual Number of Storms	Difference	% Difference
2011	16 / 9 / 5	19 / 7 / 4	3 / -2 / -1	19 / 29 / 20
2010	15 / 8 / 4	19 / 12 / 5	4 / 4 / 1	27 / 50 / 29
2009	12 / 6 / 2	9 / 3 / 2	-3 / -3 / 0	33 / 100 / 0
2008	15 / 8 / 4	16 / 8 / 5	1 / 0 / 1	7 / 0 / 20
2007	17 / 9 / 5	15 / 6 / 2	-2 / -3 / -3	13 / 50 / 150
2006	17 / 9 / 5	10 / 5 / 2	-7 / -4 / -3	70 / 80 / 150
2005	13 / 7 / 3	28 / 15 / 7	15 / 8 / 4	115 / 114 / 133
2004	14 / 8 / 3	16 / 9 / 6	2 / 1 / 3	14 / 13 / 100
2003	12 / 8 / 3	16 / 7 / 3	4 / -1 / 0	33 / 14 / 0
2002	12 / 7 / 3	12 / 4 / 2	0 / -3 / -1	0 / 75 / 50
2001	10 / 6 / 2	15 / 9 / 4	5 / 3 / 2	50 / 50 / 100
2000	11 / 7 / 3	15 / 8 / 3	-4 / 1 / 3	36 / 14 / 0
1999	14 / 9 / 4	12 / 8 / 5	-2 / -1 / 1	17 / 13 / 20
13 Year Average for Actual Atlantic Storms: 16/8/4

A review of the predicted versus actual hurricane activity shows that the forecasts are wildly inaccurate. Averaging the 13-year period from 1999-2011, Gray’s and Klotzbach’s forecasts have been off on a yearly basis by four named storms, three hurricanes, and two major hurricanes. That’s a huge discrepancy given the relatively modest number of named storms and hurricanes that actually occur each year.

No one could have predicted Katrina or the terrible 2005 hurricane season

2005 was the most active Atlantic hurricane season on record. That was the year Katrina inundated New Orleans and the even more powerful hurricanes Rita and Wilma pounded Texas and Florida, respectively. Did Gray and Klotzbach forecast 2005 to be a record season for total storms and storm intensity? No. Did they at least predict an above average year? Again no, in fact it was quite to the contrary, they predicted that the 2005 Atlantic hurricane season would be below average both in the number of storms and the number of powerful hurricanes. In fact their 2005 prediction was lower than their 2004, 2006, and 2007 forecasts, all years where the actual hurricane season was far less intense than 2005.

Gray’s and Klotzbach’s most accurate forecast was for 2008 when their prediction very nearly matched the actual hurricane activity. The thing to note about 2008 is that it was an average year in all respects, with the total number of named storms, hurricanes, and major hurricanes aligning with the 1999-2011 averages. In other words, Gray and Klotzbach were most accurate when they forecasted an average year.

In 1999 they predicted an average year, it was below average; in 2000, 2001, and 2003 they predicted below average years, they were average; in 2004 they predicted a below average year, it was above average; in 2006 and 2007 they predicted above average years, they were below average; and in 2010 and 2011 they predicted average years, they were above average. That’s 10 of 13 years where the predictions proved to be just plain wrong. I could do better than Gray’s and Klotzbach’s sophisticated models and complex methodology just by guessing that each year will be about average. It begs the question: What good is a bad prediction?

So take this year’s forecast of a quiet season with a grain of salt…we’re already off to a loud start with the first storm forming before June. Let’s just hope it’s not anywhere near as intense as 2005.

---

NOTES:

[1] The Atlantic Hurricane Season officially runs from June 1 through November 30.

[2] Classified as an organized storm with sustained winds of 39 to 74 mph.

[3] An organized storm with sustained winds of 75 mph or higher.

[4] A hurricane with sustained winds of 111 mph or higher.

[5] I certainly can’t be the only one who finds it strange that the most prominent center for tropical cyclone studies is located in landlocked, mountainous Colorado. Shouldn’t this be a project being led out of the University of Miami or some other southeast coastal school?