• World News
• | US News
• | Money
• | Science & Technology
• | Lifestyle
• | Conservatism 2.0

Forecasters have gotten it wrong more often than right in recent years — but they're not giving up.

On the list of things that fill any blogger’s heart with joy, puncturing media hype has to be right near the top. Thus, for me, it is something of an annual ritual to point out that the “official first day of hurricane season,” June 1, is a largely arbitrary and meaningless milestone, which serves mostly as an excuse for MSM types to trumpet seasonal forecasts and pay lip service to storm preparedness for a day, before forgetting about the tropics for the next three months.

Admittedly, it hasn’t seemed so arbitrary the last two years. Last year, the marginal Tropical Storm Barry formed on June 1, and this year, Tropical Storm Arthur — the gender-confused reincarnation of a Pacific storm previously known as Alma — formed late on May 31 and lasted all day on June 1.

Nevertheless, as the MSM hype machine recognizes, the “official” start of hurricane season is newsworthy mostly because it provides a convenient opportunity to look ahead at the season to come. Well, June 1 has come and gone, and hurricane season is “officially” upon us (although it doesn’t typically get active until late August and September). So, what should we expect between now and the “end” of the season on November 30?

The consensus among the seasonal forecasters is that it will be a somewhat above-average season, with between 11 and 16 named storms (the long-term climatological average is 9.6), between 6 and 9 hurricanes (the average is 5.9), and between 2 and 5 major hurricanes (the average is 2.3). Dr. William Gray, the dean (or groundhog) of season hurricane forecasters, is near the high end of those ranges, predicting 15 named storms, 8 hurricanes and 4 major hurricanes. He also forecasts an Accumulated Cyclone Energy index, or ACE, of 150%. (The seasonal ACE “takes into account the number, strength, and duration of all the tropical storms in the season.” Last year’s ACE was 72%, and in 2006 the ACE was 79%. Contrast that to 225% in 2004 and a whopping 248% in the record-shattering 2005 season.)

The forecast by Dr. Gray and his co-author, Dr. Philip Klotzbach, also “calls for an above normal chance of a major hurricane hitting the U.S., both along the East Coast (45% chance, 31% chance is normal) and the Gulf Coast (44% chance, 30% chance is average).” These are, it should be emphasized, very generalized predictions based on broad climatic patterns. It is impossible to predict individual storm landfalls more than a few days in advance.

Meteorologist and weatherblogger Dr. Jeff Masters summarizes the thinking behind the Klotzbach/Gray prediction:

“The forecasters cite the continuation of the above-normal hurricane activity period that began in 1995, the expected lack of an El Niño event, the continuation of above average sea surface temperatures and below average pressure in the eastern Atlantic, and slower trade winds (which result in reduced evaporative cooling of the ocean), as the justification for their forecast of an above average hurricane season.”

But wait, you’re thinking to yourself. Haven’t forecasters predicted above-average hurricane seasons the last two years? And haven’t they been wrong? Well, yes. Seasonal forecasters have taken a beating on the public-relations front recently, not so much because they’ve suddenly lost their predictive mojo (frankly, they never had much to begin with), but because the media scrutiny has become much more intense due to the historic 2004 and 2005 seasons and the subsequent politicization of weather as a result of the climate-change debate.

It is utterly foolish, as I explained last year, to suggest that some vast, agenda-driven, Al-Gore-fueled conspiracy is causing inflated predictions. For one thing, Dr. Gray himself is an outspoken skeptic of global warming. For another thing, forecasts in recent years have underestimated the storm totals more often than they’ve overestimated them!

That said, it is perfectly reasonable to take these seasonal forecasts with a grain of salt — indeed, I’d argue that it would be unreasonable not to do so — given their spotty track record. And I continue to think it’s worth asking, as I also did last year, whether they really serve any useful purpose. The forecasters themselves, perhaps stung by all the criticism, have taken a stab at answering the latter question. Indeed, the very first issue discussed in the Klotzbach/Gray forecast (PDF) is the question, “Why do you issue seasonal hurricane forecasts?” The answer reads:

There is an inherent curiosity amongst the general public about how active or inactive the coming season is likely to be. Using historical data, there is considerable hindcast (using the past to predict the future) skill available for predicting the upcoming season.

However, one must realize that these are statistical forecasts which will fail in some years. We find that we learn a lot from our forecast errors. Our end-of-the-season verifications give much information on explaining what the factors were that dictated the number and frequency of storms. Some of these factors may not have been considered in our forecasts for that particular year, and we often add new predictors in a quantitative or qualitative manner based on our end-of-the-season verifications.

There is also an educational component to these forecasts. For example, it was discovered about 25 years ago that El Niño reduced hurricane activity in the Atlantic. Through the issuing of these seasonal forecasts, this relationship has become well-known amongst the general public. Also, these seasonal hurricane forecasts have taught us many new relationships between climate features and Atlantic basin hurricanes such as sea surface temperatures, sea level pressures and levels of vertical wind shear in the tropical Atlantic. These relationships may have not been so readily elucidated had we not publicly made forecasts for which we are held accountable.

For its part, NOAA’s Hurricane Season Outlook — which pointedly notes that it “is not a seasonal hurricane landfall forecast” and “does not imply levels of activity for any particular region” — states that its purpose is to “provide the public with a general guide to the expected overall nature of the upcoming hurricane season.” It further states that it “may serve as a guide for large businesses, corporations, financial markets, governments, and industries to better prepare for and manage the potentially significant risk posed by hurricanes.”

Personally, I’m not sure whether these rationales are strong enough to outweigh the downsides of widely publicizing these seasonal forecasts. The downsides include: creating a “boy who cried wolf” effect when a season “underperforms” relative to the forecast; confusing people such that they begin to mistrust individual-storm forecasts, which are much more accurate than seasonal forecasts; and distracting coastal residents from the importance of storm preparedness, which is essential regardless of the seasonal forecast. (In the words of the new NHC director, Bill Read: “The seasonal forecast is meaningless to preparedness.”)

In fairness, the seasonal forecasters themselves are quick to make the latter point. In NOAA’s words: “Hurricane disasters can occur whether the season is active or quiet. Residents, businesses, and government agencies of coastal and near-coastal regions should prepare for every hurricane season regardless of the seasonal outlook.” Indeed they should. (You can get information on how to prepare for a hurricane from the Red Cross, Homeland Security, FEMA, NOAA, HHS, and the Florida Division of Emergency Management.)

There also seems to be a new focus among the forecasters on explaining the uncertainties inherent in their task. NOAA, for instance, now includes percentage probabilities along with its predictions of storm activity, somewhat like the margin of error in a public opinion poll. And the margin is quite high: “an above-normal season is most likely (65% chance), [but] there is a significant 25% chance of a near-normal season and a 10% chance of a below-normal season.” (Definitions here.) “This outlook is probabilistic, not deterministic,” NOAA’s introduction states. It is “based on predictions of large-scale climate factors known to be strong indicators of upcoming seasonal Atlantic hurricane activity,” but there are “uncertainties inherent in such climate outlooks,” which the percentage probabilities are designed to take into account.

Specifically, NOAA states, there are “three major sources” of forecast uncertainty:

• One is the timing of La Niña, which is essentially the opposite of El Niño. Just as El Niño conditions in the Pacific depress Atlantic hurricane activity, La Niña conditions enhance it. La Niña is currently ongoing, but much depends on precisely when it ends — and that is notoriously difficult to predict at this time of year. NOAA writes: “The period between March – July is referred to as the ’springtime forecast barrier,’ a period when predicting these phenomena can be difficult because the atmosphere is in a state of transition.”

• Another major source of uncertainty is the potential development of “weather patterns that are unpredictable on seasonal time scales [which] can sometimes develop and last for weeks or months.” For example, the Sarahan dust pattern that inhibited tropical development in 2006 could not have been accurately forecasted in May or June. Such relatively short-period events are a major reason why it is reasonable to expect that annual hurricane predictions may be less accurate than, for example, predictions of the Earth’s average temperature over the course of a decade or a century. For the latter type of forecast, such relatively minor, short-term events make little difference. However, when one is attempting to predict the number and intensity of tropical cyclones in a given year, they can make an enormous difference.

• The third and final major source of uncertainty — and another reason why this type of medium-term forecasting can actually be more difficult than long-term climate modeling — is the fact that “many combinations of named storms and hurricanes can occur for the same set of climate conditions.” As NOAA puts it, “one cannot know with certainty whether a given climate signal will be associated with several short-lived storms or fewer longer-lived storms with greater intensity.” Nor, I would add, can one know with certainty whether favorable atmospheric conditions will be “wasted” because a storm happens to immediately smack into a land mass, as happened with Arthur. (Last year’s entire season was a great example of this, too. As Dr. Masters wrote, “wind shear [was often] strategically high at the right time and right place, and storms…tended to form too close to land to develop.” The result was, in Sullivan’s words, “an exceptional number of duds.”)

Again, this sort of thing does not matter for truly long-term climate forecasts, but such essentially random variables can seriously skew the accuracy of these annual hurricane predictions, which occupy a strange hybrid area where “weather” meets “climate.”

Still, despite these acknowledged uncertainties, and despite the recent failures, forecasters have soldiered on and tried their best to accurately predict the 2008 season. In fact, the Klotzbach/Gray team has based its forecast on a newly tweaked model, designed to correct some of the errors of previous years. Cynics might compare this to college football’s BCS, which has repeatedly changed its formula to compensate for previous years’ problems — the sports equivalent of “hindcasting” — only to see brand new problems develop in subsequent seasons.

On the other hand, this is how the science evolves, and Klotzbach and Gray are forthright in admitting that it is a work in progress. In any event, “hindcasts” based on the new model come much closer to the mark than the real-time forecasts did in all of the last four years, which is significant, since 2004 and 2005 were both well above average (and were under-forecasted), while 2006 and 2007 were below average (and were over-forecasted). “The new hindcast model improves upon our real-time forecasts by approximately 60%…over the period from 2004-2007,” Klotzbach and Gray write. The model is also 40 percent more accurate than “dummy” predictions based purely on climatology, i.e., what you’d expect in an “average” year.

If those explanations and improvements aren’t enough to make you believe that Klotzbach, Gray & co. might be right this time — that there really is a wolf — you may want to consider what the aforementioned weatherblogger Alan Sullivan has to say. Sullivan (who, like me, isn’t a meteorologist, but plays one on the Internet) was a lone voice in the wilderness last year, arguing as early as May and June that an average or below-average season was possible due to lower-than-expected water temperatures and unfavorable upper winds. That prediction — ultimately vindicated — went against the conventional wisdom at the time, which was still calling for an above-average season.

This year, however, Sullivan is singing a different tune. In an excellent and detailed mid-March post, Sullivan wrote that he “expects the 2008 season to resemble 2004 or 2005 much more than 2006 or 2007. But I will make no specific guesses about numbers. I’ll leave that to the people with supercomputers.”

In early May he elaborated, “I suspect the 2008 season will get off to a slow start, but I am not inclined to back off my forecast of an active climax in late summer and fall.” Then on June 1, he wrote:

[The current atmospheric pattern] means tropical trouble for US coasts, unless it shifts before the peak of hurricane season. I am not optimistic about [any such] change in the pattern. La Niña conditions are weakening in the Pacific, and warmth has spread in the Atlantic; but the overall distribution of global sea temperature anomalies is holding steady. This does not bode well for hurricane-risk zones of the US or Caribbean basin.

Finally, on June 3, Sullivan — himself a resident of South Florida — reacted to the Klotzbach and Gray forecast by writing: “Conditions still seem right for a busy season. The pool of warm water in the Atlantic has actually grown, and tropical waves are already so active that I suspect the [Klotzbach/Gray] number [i.e., 15 named storms] may prove a little low.

For residents of South Florida, it’s time restock batteries, lantern fuel, and canned goods. Also we should get back in the habit of keeping the car’s gas topped off, no matter what it costs to fill the tank.” That’s good advice for anyone along the Atlantic or Gulf coast, no matter how many storms form in 2008. After all, as has so often been said, “An active year is the year when you get hit.”

Brendan Loy blogs at Hopefully Considered .

Podcasts

PJM Home