Is Global Warming Leading to an Increase in the Total Number of Atlantic Storms? (Part I: The Debate)

i-26a8c4c69bb6e36898b270ed42aaa626-3 storms.jpg It may seem a strange question to be asking in a season that, so far, hasn't yet seen an Atlantic hurricane. But while the weather in any given year can be tricky and unpredictable, there's no doubt that we're currently in an active period for Atlantic storms in general--and now, a new paper (PDF) from Greg Holland of the National Center for Atmospheric Research and Peter Webster of Georgia Tech explicitly ties this period of heightened activity to global warming by asserting that the total number of Atlantic storms has increased markedly over the past century, in correlation with rising sea surface temperatures.

At the outset, it's worth noting that National Hurricane Center specialist Chris Landsea's rebuttal (PDF) to the Holland/Webster paper was published in EOS back in May, well before the Holland/Webster paper itself officially came out (in the Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences). This is simply a quirk of the scientific publication process: Although they haven't appeared in a journal until now, the arguments about storm numbers made by Holland and Webster have been kicking around for some time. Indeed, back in January on this blog I reported on a public back-and-forth between Holland and Landsea over Atlantic storm numbers that occurred at the annual American Meteorological Society meeting in San Antonio.

Still, as the Holland-Webster paper is now garnering major media coverage, it seems an appropriate time to provide some commentary on it. Allow me to start out by providing the abstract of the paper so that you can see what is being argued (you'll find that abstract as well as my further analysis after the jump):

We find that long-period variations tropical cyclone and hurricane frequency over the past century in the North Atlantic Ocean have occurred in the form of three, relatively stable regimes separated by sharp transitions. Each regime has seen 50% more cyclones and hurricanes than the previous regime and is associated with a distinct range of sea surface temperatures (SSTs) in the eastern Atlantic Ocean. Overall, there appears to have been a substantial 100-year trend leading to related increases of over 0.7 C in SST and over 100% in tropical cyclone and hurricane numbers. It is concluded that the overall trend in SSTs and tropical cyclone and hurricane numbers is substantially influenced by greenhouse warming. Superimposed on the evolving tropical cyclone and hurricane climatology is a completely independent oscillation manifested in the proportions tropical cyclones that become major and minor hurricanes. This characteristic has no distinguishable net trend and appears to be associated with concomitant variations in the proportion of equatorial and higher-latitude hurricane developments, perhaps arising from internal oscillations of the climate system. The period of enhanced major hurricane activity during 1945-1964 is consistent with a peak period in major hurricane proportions.

Let's translate this into some numbers. Looking at the HURDAT or "best track" database maintained by the National Hurricane Center, Holland and Webster find that between 1855 and 1900 we had about 7-9 storms (tropical storms or stronger) per year; between 1905 and 1930 there was a decrease down to about 6 storms per year; and then between 1931 and 1994 there were around 6 to 14 storms per year with an average of 10. Since 1995, however, we're seeing dramatically more activity--so far, there have been 15 storms per year on average. All of this has occurred as sea surface temperatures, particularly in the eastern Atlantic, have warmed: "There has been an average of one additional tropical cyclone for each 0.1 C increase in SST and one hurricane for each 0.2 C increase," write Holland and Webster.

These scientists know very well that the hurricane database they're using to derive these numbers isn't perfect, and indeed, that it surely gets worse the further you go back in time (because more storms would have been missed in earlier eras). So they define the uncertainties thusly: "Our conclusion is that the number of earlier missed storms most likely lies between 1-3 per year prior to 1900, less than 2 in the early [20th] century and dropping off to essentially zero by 1960." Using these assumptions, Holland and Webster show a strong and very troubling trend towards increasing Atlantic storm numbers--and we're getting more intense storms, they say, largely because this increase in numbers gives more storms the opportunity to achieve their full potential. As they put it: "We are of the strong and considered opinion that data errors cannot explain the sharp, high amplitude transitions between the climatic regimes in the NATL, with an increase of around 50% in cyclone and hurricane numbers, and the close relationship of these regime changes with SSTs."

But, here's where the National Hurricane Center's Chris Landsea--quoted in recent news stories calling the latest work "sloppy science"--comes in with a counterpoint (PDF). Landsea says data errors can explain the apparent trend. As his paper puts it, referring both to the Holland/Webster study and another piece of work: "both analyses, with no indication of uncertainty or error bars, presumed that tropical cyclone counts are complete or nearly complete for the entire basin going back in time for at least a century...this presumption is not reasonable...improved monitoring in recent years is responsible for most, if not all, of the observed trend in increasing frequency of tropical cyclones." In fact, Landsea suggests that we should probably assume something more like 3.2 missed storms per year from 1900-1965 and 1 per year from 1966 to 2002--which, he explains, would be enough to essentially nullify the apparent trend.

Landsea further notes that the percentage of total storms that make landfall has been decreasing in recent years, and from this he deduces that in earlier eras, storms remaining at sea weren't detected nearly so frequently as they are today, when we have a much denser observing network. To illustrate this point, in his paper Landsea includes the graphic below, which compares the tracks of storms during the 1933 hurricane season (which saw 21 officially recorded storms) and the 2005 season (which set the record with 28). As you can see, landfall proportions were far higher in 1933 and far fewer storms out at sea were detected:

i-2076a6cbccdc3b13d7cf83ad780ff747-Landsea-small.jpg

Let me be completely frank: I have no idea who's right in the current argument between Holland/Webster and Landsea. Indeed, in some sense it's probably unknowable--we're talking about missed storms, after all, and now that they've been missed of course we don't know how many of them there were.

From a policy perspective, though, we don't have to remain completely agnostic regarding this debate. There are several important points to take away from this latest dustup, and in a follow-up post, I will tease out those implications. But for now, if I've left everyone scratching their heads about who to trust in the current argument, all I can say is, I'm scratching my head too....

More like this

Hm, one problem would be that the storms that do not make landfall are also those most likely to be missed. A storm on the open sea will only be noted if its path and that of a ship happen to cross (AND the folks in the ship live to tell the tale).

Who's right?

I think we have to consider that both sides have a pretty good point, and that there is likely some combination of problems with the data , inter-decadal climate variability and AGW all going on simultaneously, creating a noisy time series. The trick is to determine the proportions of these different factors. I suspect the reality of anthropogenic climate change will become clearer in the next 3-4 years.

You are also right that the uncertainty is no reason 'to remain completely agnostic'. It has been fairly well demonstrated that the necessary preparations for hurricane management aren't good enough for the current climate, and it might just get worse.

->CL

Landsea's remarks about Holland and Webster's work having "no indication of uncertainty or error bars" is criticism that can be applied to much of the underlying "science" of climate change.

When error estimates are included they are often anecdotal, novel or ill-defined. The main tools of the AGW doomsayers are climate models which of course are not based on actual measurements so have no real "error" calculations associated with their predictions.

Chris's wise statement about the prospect that this dispute will never be resolved is an important one to keep in mind when reading (or writing) a news story on this contentious topic.

When studies by serious scientists diverge sharply in their findings, that is a good sign that the research is at an early stage of understanding, meaning very tentative.

It's important for news stories to reflect that instead of perpetuating what I've called "whiplash journalism" -- giving each development equal play without characterizing the overall state of some field of research. It's not good enough to be "right for a day." That will just cause consumers of news to lose confidence, fuzz out, retreat.

More on this here: http://www.environmentwriter.org/resources/articles/010807_cspan.htm)

While the climate basics are now clear (more co2=warmer world) the details (how strong storms may get, how fast seas may rise, how fast and far temps will climb) remain laden with uncertainty (which should offer little comfort to anyone).

AFAIK there's no public access version of the new paper, but most of the information in it is available in this slideshow.

Note that Kerry Emanuel has endorsed the Holland/Webster findings.

One odd thing is Landsea's claim that post-1960 storms would have been missed. I would think the number would drop to zero by 1970 (the start of the satellite era), but for some years prior to that the density of cross-Atlantic flights would have been such that none would be missed even if ships weren't spotting them all. As soon as daily flights were inaugurated from London to Sao Paolo, Rio and Miami, e.g., one would think that any storm crossing those flight paths would be spotted.

A related point I wonder about is whether the density of military ship and aircraft traffic in the Central Atlantic during WWII war would have been enough to catch all the storms then. Anti-submarine aircraft flew far out into the Atlantic on a constant basis, and ship-based ASW operations (which included a number of small aircraft carriers) were ongoing farther out. Just looking at commercial shipping and aircraft flight paths would miss all of that activity.

By Steve Bloom (not verified) on 30 Jul 2007 #permalink

One odd thing is Landsea's claim that post-1960 storms would have been missed. I would think the number would drop to zero by 1970 (the start of the satellite era), but for some years prior to that the density of cross-Atlantic flights would have been such that none would be missed even if ships weren't spotting them all. As soon as daily flights were inaugurated from London to Sao Paolo, Rio and Miami, e.g., one would think that any storm crossing those flight paths would be spotted.

Landsea's claim isn't odd at all. He provides examples from 2003-2006 of storms that almost certainly wouldn't have been identified prior to the AMSU and QuickScat data platforms. It's quite possible that, 20 years ago, procedures at NHC wouldn't have named either of the storms in the Atlantic this year.

A related point I wonder about is whether the density of military ship and aircraft traffic in the Central Atlantic during WWII war would have been enough to catch all the storms then. Anti-submarine aircraft flew far out into the Atlantic on a constant basis, and ship-based ASW operations (which included a number of small aircraft carriers) were ongoing farther out. Just looking at commercial shipping and aircraft flight paths would miss all of that activity.

There is the question of whether the military traffic would have put the reports into a location where they would have gotten into the database. Lots of other extreme weather reports drop dramatically during wars given the attention given to higher priorities at the time.

By Harold Brooks (not verified) on 30 Jul 2007 #permalink

Steve,
The storms would have been spotted, passenger planes would assuredly have avoided them....but how would they have been *classified*? As tropical cyclones? Would they have been given names?

Landsea argues that more weak/subtropical storms are now being added in post-season classification due to better instrumentation, and this is a trend unfolding just over the past 5 years or so.

Thanks to everyone for the great comments.

Oops, missed the H/W paper pdf link at the start of the post. The slideshow is still worth looking at, though. It seems to have a couple of extra graphics plus a bit more plain-language explication.

Good point on the aircraft, Chris, but wouldn't the same issue apply to ships? And isn't the issue in part whether storms were spotted at all? Of course there's a classification problem generally back in those times, so I'll be curious to see how the paper touches on that point.

By Steve Bloom (not verified) on 30 Jul 2007 #permalink

Harold, I was trying to draw a distinction between being missed entirely (which I thought was the main point of the graphics reproduced above) and misclassified. Wouldn't misclassification also have been a problem with 1933 storms that were spotted but stayed mostly out to sea?

Regarding possible WWII records, the Navy tends not to throw that sort of thing away (remembering that they are *extremely* interested in major storms), but finding them and getting the data out could well be a nightmare. OTOH, isn't that why grad students were invented? :) Wasn't it the Navy that began the hurricane hunter flights toward the end of the war, BTW?

By Steve Bloom (not verified) on 30 Jul 2007 #permalink

I've wondered if there is a change in over all planetary wind speed that would cause steering winds to push the hurricane track to the east and cause less landfall for Atlantic hurricanes. I haven't been able to find info on wind speeds world wide, so it's just a thought.

Steve, it's possible that storms could have been misclassified in 1933, but that would just add to the undercount from 1933. All of the Landsea examples are of storms that wouldn't have made it into the database in the 1930s. I can't see how the 2006 storm would have been identified at all if it happened in 1933-the identification depended on AMSU, QuikScat and diagnostics from numerical weather prediction models of a storm that peaked at 45 kts and was of tropical storm strength for 24 hours.

As far as the war is concerned, sure, the Navy was collecting data where it operated, but I'm not sure what impacts on ship tracks there were. Obviously, there were some areas, such as the northeast US to the UK, that got a lot of traffic, but I don't know what else. The recon flights started in '44 and their impact shows up in the data, although they don't go very far east for a a number of years later. If you break the data in Landsea's Fig. 2b into three periods (prior to '00-'44, '45-'64, '65-present), the middle period shows a decline in the fraction of landfalling cyclones, which could reasonably be associated with the increasing use of recon. The landfalling data aren't of good enough quality in the US, particularly Florida, to have confidence that all of them were being collected prior to 1900.

By Harold Brooks (not verified) on 31 Jul 2007 #permalink

The scientific literature on hurricane climate contains a number of papers that include proper methods of accounting for the uncertainty in the historical storm data base (see Hurricane Climate). For example in the paper "Prediction models for annual U.S. hurricane counts" the models include a component that estimates the probability that a storm was missed or miss classified assuming it actually occurred. Moreover, better than simple error bars, the models provide a complete posterior distribution on the storm numbers that account for epistemic and aleatory uncertainty.