28 November 2011

A New US Hurricane Record

On December 4, 2011 it will have been 2,232 days since Hurricane Wilma made landfall along the Gulf coast as a category 3 storm back in 2005. That number of days will break the existing record of days between major US hurricane landfalls, which previously was between 8 Sept 1900 (the great Galveston Hurricane) and 19 Oct 1906. Since there won't be any intense hurricanes before next summer, the record will be shattered, with the days between intense hurricane landfalls likely to exceed 2,500 days.

If you are in the insurance or reinsurance business and want to stir up a little constructive mischief, you should ask your favorite catastrophe modeling firm or ratings agency to show you the mathematics behind their estimate of the probability of zero intense hurricane landfalls from 2006 to present (both made at the time and what they'd say today). (Hint: Zero. Zip. Nada.).

This remarkable streak has to end sometime, and likely won't be repeated anytime soon.

36 comments:

  1. Another proof that we're all gonna die.

    ReplyDelete
  2. The heat ( nee Energy ) in the pipeline is being saved up in massive hurricanes that'll zap us at some future time. It'll be much worse that we thought: catastrophic, even.

    ReplyDelete
  3. 100 years ago the oceans were cooler, Arctic ice coverage more extensive, sunspot cycles low, atmosphere carbon dioxide content 100 ppm lower, 5 billion people less on the planet and virtually no asphalt roads anywhere sucking up heat...

    Where the hell did the energy for all those hurricanes come from back then?

    ReplyDelete
  4. So is it time for a climate change reporting first - "It's NOT Worse Than We Thought?"

    ReplyDelete
  5. @Pirate - is that a serious question?

    I'm no hurricane expert but one suspects that ocean temperatures then were frequently in the range that spawn hurricanes now. It doesn't have to be the hottest part of the year, or the hottest part of the ocean, for hurricanes to form; many other factors are involved.

    Someone with more expertise got the details here?

    ReplyDelete
  6. @Models - I don't think the heat is being "saved" for later; it's not like hurricanes actually transfer a great deal of heat from the ocean, in relative terms.

    ReplyDelete
  7. Hi Roger - how does the "no US Cat 3 landfalls" figure correlate with other activity data over the century? I know data is much sparser as you go back more than even a few decades...

    ReplyDelete
  8. Cherry-picking. The U.S. Southeast is not the world. Plenty of major Asian typhoons this year.

    ReplyDelete
  9. What is more striking is that there are no major hurricanes during the current strong La Nina event.

    ReplyDelete
  10. Gadfly -
    Isn't the more important point the trendlessness in US landfalls?

    It's a bit dated, but consider this 2007 study indicating no trend in typhoon landfalls. http://www.swissre.com/media/news_releases/new_study_on_pacific_windstorms_reveals_regional_patterns_but_no_rising_trend_of_typhoon_landfalls.html

    ReplyDelete
  11. Somebody somewhere is already working on a paper about how some species depends on periodic hurricanes to effect some needed change in habitat, nutrients, land cover or whatever and that this landfall lull is troubling evidence of AGW-induced damage to the biospohere.

    ReplyDelete
  12. Gadfly

    Even when you add in the rest of the Globe, there is not that many Tropical Cyclones worldwide compared to just 5 years ago. This you can see plainly here:
    http://policlimate.com/tropical/

    In both Global Accumulated Cyclone Energy and in 4 out of the 6 areas the YTD ACE is below the NORMAL YTD ACE (There is no Southern Hemisphere YTD ACE listed). The only area that has an increased ACE is the North Atlantic and that is due to all the storms that formed (including the less than a day TC's) and never hit the US, they curved off.

    So bringing up the Pacific to try and prove your point actually hurts your position since there has been less than normal TC Energy across the Pacific.

    ReplyDelete
  13. The heat is being saved for fulfillment of the Mayan Prophesy. AlGore is a Mayan agent.

    ReplyDelete
  14. Roger - care to comment?:

    [Link:


    Here's an interesting excerpt from an interview with Nathan Urban on his new paper: “Climate Sensitivity Estimated from Temperature Reconstructions of the Last Glacial Maximum”:

    --snip--

    Q: Any other thoughts on the skeptics’ reception of your paper?

    One blog did surprise me. World Climate Report doctored our paper’s main figure when reporting on our study. This manipulated version of our figure was copied widely on other blogs. They deleted the data and legends for the land and ocean estimates of climate sensitivity, and presented only our combined land+ocean curve:

    [THE ORIGINAL ARTICLE HAS A MANIPULATED GRAPH HERE]

    Upper: World Climate Report’s manipulated image removing the Land and Ocean data.

    Lower: The actual figure as it appears in Science, with the Land and Ocean curves included.

    They did note that their figure was “adapted from” ours, and linked to our paper containing the real figure. On the other hand, Pat Michaels duplicated this doctored version of our figure again in an article at Forbes, and didn’t mention at all that it had been altered. (A side note with respect to the Forbes article: Science didn’t “throw a tantrum” about posting our manuscript on the web. They never contacted us about that. I took it down myself as a precaution, due to the journal’s embargo policy.)

    I find this data manipulation problematic. When I created the real version of that figure, it occurred to me that it would be reproduced in articles, presentations, or blog posts. Because I find the difference between our land and ocean estimates to be such an important caveat to our work, I made sure to include all three curves in the figure, so that anyone reproducing it would have to acknowledge these caveats. I didn’t anticipate that anyone would simple edit the figure to remove our caveats. I can’t say why they deleted those curves. If you were to ask them, I’d guess they’d say it was to “clarify” the figure by focusing attention on the main result we reported.

    Regardless of their intent, I find the result of their figure manipulation to be very misleading, especially since their blog post strongly implies that our study eliminates the “fat right tail” of the climate sensitivity distribution, and has proven the IPCC’s climate sensitivity range to be incorrect. Our land temperature curve, which they deleted, undermines their implication. They intentionally took our figure out of the context in which it was originally presented, a form of “selective quotation” which hides data that does not support their interpretation.

    In summary, I find World Climate Report’s behavior very disappointing and hardly compatible with true skeptical inquiry. I can only imagine how they would respond if they found a climate scientist intentionally deleting data from a figure, especially if they deleted data that undermined the point of view they were presenting.

    --snip--

    I particularly like Urban's statement about "very disappointing behavior" that is "hardly compatible with true skeptical inquiry."

    ReplyDelete
  15. There has been some work on Pacific cyclones, specifically those making landfall in eastern Australia. As with other data sets, interannual variability especially on ENSO time scales is a big part of the landfall variability.

    see:

    Callaghan, J., and S. Power (2011), Variability and decline in the number of severe tropical cyclones making land-fall over eastern Australia since the late nineteenth century, Clim. Dyn., 37(3), 647-662, doi:10.1007/s00382-010-0883-2.

    Recent studies have raised concerns that tropical cyclones (TCs), particularly severe TCs, have become more frequent in many places in response to global warming. Other studies discuss errors in TC data that can cause large inaccuracies in some of the observed trends. Additional studies conclude that TCs are likely to become more intense in the future in response to global warming, while regional modelling studies for the south-west Pacific near north-eastern Australia project an intensification of TCs and either a decrease or no change in TC numbers. Here we describe and use a new data base of severe land-falling TCs for eastern Australia derived from numerous historical sources, that has taken over a decade to develop. It provides one of the world’s longest reliable records of tropical cyclone activity, and allows us to document changes over much longer periods than has been done previously for the Southern Hemisphere. Land-fall numbers are shown to vary a great deal on interannual, decadal and longer time-scales. The interannual variability is consistent with previous studies using much shorter data sets: land-fall numbers are well-simulated as a Poisson process and are modulated by the El Niño-Southern Oscillation (ENSO). Land-falls occurred almost twice as often in La Niña years as they did in El Niño years, and multiple land-falls only occurred during La Niña years. The statistical link between land-falls and pre-season values of the Southern Oscillation Index provides a modest predictive capability. Decadal variability in ENSO drives some of the decadal variability in land-fall numbers. The sign and magnitude of trends calculated over 30 years periods vary substantially, highlighting that caution needs to be taken in making inferences about trends based on e.g. satellite era data only. The linear trend in the number of severe TCs making land-fall over eastern Australia declined from about 0.45 TCs/year in the early 1870s to about 0.17 TCs/year in recent times—a 62% decline. This decline can be partially explained by a weakening of the Walker Circulation, and a natural shift towards a more El Niño-dominated era. The extent to which global warming might be also be partially responsible for the decline in land-falls—if it is at all—is unknown.

    ReplyDelete
  16. Roger,

    I believe that your graph is better represented by a "non-linear" trendline.

    In fact, post-1921, it seems clear to me that longer and longer stretches without landfalling storms are just about certain. ;)

    ReplyDelete
  17. News from the campaign trail, July 2012:
    In a speech in this important swing state, President Obama said that he was proud that his administration had "protected the great state of Florida from devastating hurricanes." A statement from the campaign of his Republican opponent noted that the first three-plus years with no hurricanes were due to the policies of former President Bush. ;)

    ReplyDelete
  18. Joshua,
    No one is surprised that you would seek to hijack the thread. After all it does not conform to your views and you are far too limited to actually discuss topical issues.
    So back tot he topic at hand:
    It is interesting how durable the AGW catrastrophist dogma is no matter what valid counter data is presented.

    ReplyDelete
  19. Um... Huricane Ike? Twenty-five percent of Galveston, Texas has been gone for good since it hit in October 2008. The winds were Category Three but it was a hundred-year flood, despite the seawall.

    ReplyDelete
  20. Roger,
    having lived through the non major tropical storm Irene, after telling my neighbors in Vt it would be no big deal because i knew the wind speeds would be exaggerated by the press (they were), I was shocked at the devastation caused by flooding. No one had ever seen a storm do this much damage, and it was at least as bad as the hurricane of 1927 that was in conjunction with another storm form the midwest, and the other major one in 1938. Maybe it wasn't category 3 at landfall, but it did a hell of a lot of damage.

    ReplyDelete
  21. -20-tonylearns and -19-Barza

    Yes, Cat 1 and 2 storms can be incredibly destructive. Ike was a Cat 2 at landfall, Irene a Cat 1

    ReplyDelete
  22. That graph (like every other) is clear proof of CAGW. As everyone knows the longer the pause before the storm, the bigger and nastier the storm. A windup pause of this length dictates that Gulf coast residents should consider moving to Alberta.

    ReplyDelete
  23. FoFS -

    Of all the counter-arguments that people throw my way, I have to say that the claims that I'm attempting to "hijack" a thread are the ones I find the funniest.

    I asked Roger for comment because I was interested in reading his comment.

    Although, apparently, you think that you're serving some deep and important purpose by posting comments on a blog, I don't share your perspective. I have no reason to "hijack" anything that you might want to talk about.

    So that's funny - but what's even funnier is that you would write a post not on the topic of the thread, but on whether or not I'm intending to "hijack" the thread, as a way of making it clear that you wouldn't want the thread to be "hijacked."

    Think about that for a bit.

    ReplyDelete
  24. Roger,

    In 2006, one of the catastrophe modelling companies you refer to made a forecast of major (cat 3-5) US hurricane landfalls of around 0.9 per year over the next 5-years with a Poisson distributed rate of occurance.

    The Poisson probability of having no landfall for 6 years, assuming a process with annual rate of 0.9, is 0.5%. This is certainly small, but it isn't "Zero. Zip. Nada."

    For comparison, the probability of having no landfall for 6 years using the average major landfall rate from 1900-2011 (0.63 per year as calculated from the HURDAT landfall summary table) is 2.4%. The quiet of recent years is clearly a rare but not impossible event regardless of whether one thinks insurance/reinsurance hurricane risk should be priced using an ‘active-period' or a historical mean view.

    ReplyDelete
  25. -26-Jessica Turner

    Thanks for the comment. I have a comment on your math.

    To calculate the probability of a year with no Cat 3s or greater, the relevant Poisson variable would be the probability of an intense storm making landfall per year, not the average rate.

    For instance, if the average rate is 1.0 per year over five years, you would get a different probability of years with no storms if you assume 1.0 per year versus all 5.0 in one year, even though the average is the same. In the former, the probability of a year with no intense storms (over the 5 year period) is 0.0 and in the latter, 0.80.

    So average rate is not a sufficient basis to calculate a streak of 6 years without landfalls.

    Here is an example of one study by a firm that produced a 5-year forecast back in 2006:

    https://www.rms.com/Publications/Research/Lonfat_2007.pdf

    Can you point in the paper to where it provides information that would allow such a calculation?

    I think that such information is not provided and one would have to dive into the actual expert elicitation methodology to discern the PDF of occurrence implied by the forecast.

    ReplyDelete
  26. Sorry but crowing about this seems asinine. It's like saying, Well I don't have any Stage 3 skin cancer from the neck up so I must be okay. ( Spare me ,I'm just reporting this this not making judgements. This has been reported in the right wing press ad nauseum.)

    But I guess only left wing journo's are bandwagoner's, right?

    ReplyDelete
  27. Roger,

    Thanks for your reply.

    The question that my calculation was answering was “What is the probability of 6 years without a US major hurricane landfall if we assume that the arrival time of these landfalls is Poisson distributed.” In that case, it is a straightforward application of the probability distribution function where:

    P{N=n} =(lambda^n * e^-lambda)/n! and N=0,1,2…

    Here N=0 for 0 occurrences and lambda is, if calculating using the 1900-2011 mean rate as an estimator for lambda, (0.63 * 6).

    By definition, the Poisson distribution is time-independent, which means the arrival time of each event does not depend on what happened before. It’s irrelevant to one’s estimation of lambda if over a 5-year period 5 hurricanes arrived 1 at a time or if 5 arrived one year and none the next four (in the later case you may very well wonder if your data is really Poisson distributed, but that’s a different issue!

    I’m not commenting in order to defend any particular firm but catastrophe modelling in general. My larger point was that I don’t believe a modeller would say a thing that has never happened before has "Zero. Zip. Nada.” probability of occurrence since they are in the business of predicting events that have never occurred.

    ReplyDelete
  28. -29-Jessica Turner

    Thanks! I am simply saying that your lambda is incorrect.

    If intense hurricanes occur in 1/3 of all years then your lambda would be 0.33, even if the average rate is 0.63. You cannot derive this lambda from the average rate.

    The unit of analysis here is years with an intense landfall.

    ReplyDelete
  29. Roger (#30): What you say doesn't make any sense. A Poisson distribution has only one parameter. Lambda is, by definition, the average rate (avg. # landfalling major hurricanes per year).

    If you were arguing that the distribution is not Poisson, that would be one thing; but if it's Poisson then you know everything about the probability distribution from the average rate.

    Here's what Jessica is doing: The average number of hurricanes per year is 0.63, so the Poisson distribution implies that the probability of no hurricanes in a single year is (0.63 ^ 0 * e^-0.63)/0! = e^-0.63 = 0.53

    Now, if the statistics are truly Poissonian, then we can rescale the time frame to 6 year intervals: the average number of hurricanes in a 6 year interval is just 6 times the average number per year (if it's not, then your statistics aren't Poissonian) so 0.63 * 5 = 3.78. The probability of zero hurricanes in 5 years is (3.78^0 * exp(-3.78)) / 0! = exp(-3.78) = 0.023, or 2.3 percent. Not, as Jessica notes, "Zero. Zip, Nada." (Alternatively, using conjunction of independent events, the probability of zero in six years is the sixth power of the probability of zero in a single year, which gives the same answer.)

    If you think this is wrong, please explain in detail how a Poisson distribution can have a lambda parameter that's anything other than the average rate.

    ReplyDelete
  30. @Roger: to add one more comment for clarification, because this detail seems to be a source of confusion that's getting in the way of your understanding Jessica: The probability of a major hurricane making landfall in a year is 1 minus the probability of zero making landfall in a year.

    If major hurricane landfalls are Poisson distributed and if the average rate of major hurricanes making landfall in a year is 0.63, then the probability of at least one major hurricane making landfall in a year is (1 - P(0)) = 1 - exp(-0.63) = 0.47, or 47%. From this we can derive that the probability of 6 years with no major hurricane landfalls is (1 - 0.47)^6 = 2.3%

    Thus, contrary to what you wrote in #30, if major hurricane landfalls are Poisson distributed, then the average rate uniquely determines the probability of a major hurricane making landfall in a year.

    Another interpretation of what you're writing would be that you're trying to tell Jessica that you can't tell from the average hurricane rate alone whether they are Poisson distributed, but this would be a puzzling interpretation as well because both you (http://sciencepolicy.colorado.edu/admin/publication_files/resource-2786-2009.47.pdf) and RMS (http://www.rms.com/Publications/Research/Jewson.pdf) seem to agree that hurricane statistics are Poissonian.

    ReplyDelete
  31. -32-Jonathan

    Thanks, you write:

    "If major hurricane landfalls are Poisson distributed and if the average rate of major hurricanes making landfall in a year is 0.63, then the probability of at least one major hurricane making landfall in a year is (1 - P(0)) = 1 - exp(-0.63) = 0.47, or 47%."

    I can comment more later, but this is not empirically correct -- major landfalls (memory) occur in about 1/3 of all years.

    ReplyDelete
  32. -32-Jonathan

    OK, a bit longer reply ...

    There are two issues here. One is how to compute the probability of a 6-year period with no IH landfalls. The second is what basis the cat modelers provided for doing so in 2006.

    On the first, I am not quibbling with your math. One could also calculate the probability using the binomial distribution and an annual probability. The issue we are debating is not math.

    On the second more interesting question, the number that Jessica presents is the output of an expert elicitation that resulted in a changed frequency and distribution (as compared to climatology). I do not know that this new distribution is Poisson (in fact I strongly doubt it, having participated as an elicitor). If you dig into the expert elicitation you will find that it was impossible for the elicitors to anticipate a 5-year period with no landfalls -- which I'd be happy to elaborate on.

    ReplyDelete
  33. This is in error. The longest stretch on record for no lanfalling major hurricanes was 3,316 days in 1860-1869 according to the official HURDAT database. Link here. http://www.aoml.noaa.gov/hrd/hurdat/UShurrs-detailed-2011.html

    Specifically, there was a major hurricane that made landfall on August 11, 1860 and the next major hurricane to make landfall in the USA was not until September 8, 1869. A period of 3,316 days.

    I thought you would want to know.

    ReplyDelete
  34. A reporter just asked me about Ike in 2008 -- it was a Cat 2 at landfall, here is the official NHC report:

    http://www.nhc.noaa.gov/pdf/TCR-AL092008_Ike_3May10.pdf

    ReplyDelete