New Holland and Webster Paper

A new Holland and Webster paper is here. Some discussion already started on this thread http://www.climateaudit.org/?p=900

Update: Judith Curry sent in the following data file http://data.climateaudit.org/data/hurricane/natldata.xls

295 Comments

  1. bender
    Posted Dec 27, 2006 at 11:52 AM | Permalink

    From that section 3:

    Standard statistical tests are used throughout. Unless otherwise indicated all variance and correlation coefficients are valid at the 95% level using the t-statistic with correction for serial correlation (Davis 1976).

    Davis (1976) is not an appropriate statistical authority. That doesn’t mean their correction method is wrong. It just means they’ve doubled the amount of work a reviewer has to do to make sure their statistics are correct. Also, I think it is more appropriate to state p-values for individual statistics than make a global claim that all p-values are below 0.05. There’s a BIG difference between p

  2. Judith Curry
    Posted Dec 27, 2006 at 11:57 AM | Permalink

    Reply to Dave Dardinger (on the Emanuel thread:)

    The issue of greenhouse warming and increasing hurricane activity is a hypothesis (see our BAMS article that was discussed on a previous thread). I have argued that the greenhouse warming argument is the best available explanation for the increase in global hurricane intensity and the increase in total number of North Atlantic tropical cyclones (the magnitude of the increase associated with the greenhouse warming is the main issue of contention, rather than the existence of the increase).

    For a hypothesis to be elevated to a theory, it must pass 3 tests (discussed in the BAMS article):
    1. Survive scrutiny and debate, including attacks by skeptics.
    2. Be the best existing explanation (physical and statistical) for the particular phenomenon.
    3. Demonstrate predictive capability.

    As concluded in the BAMS article, this hypothesis is not yet ready to be elevated to a theory owing to uncertainties in the data and the need for further observations with continued warming to assess whether this hypothesis has any predictive value. Statistical analysis is a key element, but not the only element. The short data records and uncertainties (which are very difficult to quantify) are a substantial challenge to our field. The data issues in combination with the complexity of the climate system make this an extremely challenging scientific problem.

    In the past year and a half, there have been many papers published on this topic (most of which are referenced by the Holland/Webster paper). However analyses made this far have not been sufficiently convincing to reject any of the subhypotheses in the causal chain that we propose that links an increase in hurricane intensity and global warming.

    Dave’s analogy re the hammer and nail is apt. Until the Emanuel and Webster et al. papers were published, the main hammer around was the natural variability one. Now we have a new hammer on the scene. As stated in my BAMS article:

    ” Investigative scientists bring prior experience, understanding and prejudices to their investigation. Amongst a group of investigators, if these priors are so different that the investigators genuinely reach divergent conclusions from the available sparse data, then none of the hypotheses can be raised to the status of a theory.”

    So we are talking about a hypothesis here, that has not yet been elevated to a theory.

  3. bender
    Posted Dec 27, 2006 at 12:06 PM | Permalink

    Dr. Curry says:

    The number of NATL tropical cyclones (when filtered for el nino and other short term variability) does show a correlation with SST. Webster has done the calculation for 5 year running means, and the correlation between NATL SST and TC counts from 1910-2005 gives a correlation of 0.78 that is significant using the Mann Kendall statistic (this is a curry/webster paper that is in press, to be published around June 07).
    If Steve M, Bender, or Willis can provide any further insights into the statistical significance of this relationship, I would certainly be interested in pondering it and would certainly forward to Holland/Webster.

    However I believe we explained, and even demonstrated, a few months back that the Mann-Kendall statistic is indeed susceptible to serial correlation.

    Furthermore, I remember showing that the SST-TC correlation was due to high decadal coherence, such that the significance of the trend coherence would be quite low after correcting for serial correlation.

    Furthermore, filtering does not merely remove subdecadal El Nino effects, as advertised; it also exaggerates coherence at the wavelength of filtering, introducing correlations that were not formerly there. Filtering is fine for graphical presentation, but not for statistical inference.

    Post your data script and let’s have a go at it.

  4. Steve McIntyre
    Posted Dec 27, 2006 at 12:06 PM | Permalink

    Davis 1976 is online here . bender, you’re right about this – it’s annoying to see these sorts of “trade” references rather than proper statistical authorities, although at least they didn’t cite themselves as statistical authorities a la Wahl and Ammann.

  5. bender
    Posted Dec 27, 2006 at 12:15 PM | Permalink

    Dr. Curry says:

    none of our arguments stand or fall solely based upon statistical analyses of imperfect data

    I’d be willing to discuss that. Remember that when your hypothesis is built from a chain of linked inferences that the error propagates among links. As logical structures these inference chains are surprisingly fragile – somewhat like a house of cards.

  6. bender
    Posted Dec 27, 2006 at 12:17 PM | Permalink

    To finish #1:
    There’s a BIG difference between p lt 0.05 and p lt 0.0001. Ask Martin Juckes.

    (Why does neither “&lt” nor the lt symbol work anymore?)

  7. bender
    Posted Dec 27, 2006 at 12:54 PM | Permalink

    Re #4
    What exactly is the correction method in Davis (1976) that Holland & Webster are using? To answer that question you need to sift through:
    -18 pages
    -more than 10 equations
    -2 appendices
    Happy hunting.

  8. Judith Curry
    Posted Dec 27, 2006 at 12:55 PM | Permalink

    Bender,

    I’ve emailed a data set to Steve M. No guarantees that this is exactly what Holland used, I am on travel and don’t have full access to my data sets. Also, a reminder that a few months ago there was a hurricane data compilation thread where these data were assembled.

    With regards to the SST/TC correlations. There is a negative correlation associated with El Nino; this needs to be filtered out in some way to examine the correlation associated with the longer time scales. The optimum way to do this while still maintaining statistical significance has perhaps not yet been accomplished, have a go at it.

  9. bender
    Posted Dec 27, 2006 at 1:06 PM | Permalink

    Re #4
    Read p. 252 paragraph of Davis (1976) carefully. If this is Holland & Webster’s correction method, then there is a very important caveat cited in that sub-section. It is not clear that the hurricane/SST data satisfy the restriction that the time-series record length be long relative to the time-scale of the autocorrelation.

    Re #8
    Great. Let’s have a look.

  10. Posted Dec 27, 2006 at 1:08 PM | Permalink

    Here’s Holland and Webster (2006) Figure 1:

    Figure 1. Tropical cyclone occurrence (blue points indicate annual totals and the black line is a 9-y running mean) in the North Atlantic together with East Atlantic SST anomalies for the hurricane season (red line) from 1855-2005. TC1, 2 and 3 refer to climate regimes discussed in the text.

    Forgive me for a naive interpretation of this, but if there’s an opportunity here for a spurious correlation then other scientists have not put that chance down.

  11. jae
    Posted Dec 27, 2006 at 1:20 PM | Permalink

    Judith, you say:

    I have argued that the greenhouse warming argument is the best available explanation for the increase in global hurricane intensity and the increase in total number of North Atlantic tropical cyclones (the magnitude of the increase associated with the greenhouse warming is the main issue of contention, rather than the existence of the increase).

    Why can’t the cause of the warming be something other than greenhouse gas warming? If hurricane parameters are affected by SST, why can’t the SST be due to natural variations, such as those caused by solar effects? This automatic attribution of higher temperatures to mankind bugs me.

  12. Dave B
    Posted Dec 27, 2006 at 1:47 PM | Permalink

    Dr Curry, you wrote:

    “There is a negative correlation associated with El Nino; this needs to be filtered out in some way to examine the correlation associated with the longer time scales. The optimum way to do this while still maintaining statistical significance has perhaps not yet been accomplished, have a go at it.”

    please forgive me as a statistical layperson, but shouldn’t one first “filter out” the negative correlation with el nino, THEN determine whether one still has statistical significance? rather than presume one will “still maintain statistical significance”?

    thank you for considering this.

  13. Tim Ball
    Posted Dec 27, 2006 at 2:43 PM | Permalink

    I have difficulty with the title. It poses the question, “Natural variability or climate trend?” Surely a climate trend is a natural variability, unless you define climate trend as human caused and this is not possible unless you can definitively identify natural variability. Any attempt to even infer that a portion is human is unjustified, especially in a scientific paper, as jae indicates.

    The variability is based on a 100 year sea surface temperature (SST) record, but I don’t know of a reliable instrumental or proxy record. Recent computer models of SSTs, presumably based on the same record, were comletely at odds with what was actually measured.

    I have difficulty with Judith Curry’s definition of hypothesis and theory (#2). One dictionary defines hypothesis as, “A supposition that appears to explain a group of phenomena and is advanced as a basis for further investigation, a proposition that is subject to proof or to an experimental or statistical test.” Wikipedia defines theory as follows. “In science, a theory is a proposed description, explanation, or model of the manner of interaction of a set of natural phenomena, capable of predicting future occurrences or observations of the same kind, and capable of being tested through experiment or otherwise falsified through empirical observation.” Is there a difference? I wait to be convinced.

    The key phrase is, ” capable of predicting future occurrences or observations of the same kind” and the computer models have failed miserably in that regard. There is also research indicating the trend is as much a function of changes in measuring, observing and analyzing as a real change. The complete failure of hurricane predictions for 2006 indicate inadequate data or understanding and provide no basis for even speculations, yet here we have it subtly inserted in the title. Why?
    Am I making semantic arguments? I don’t think so. The subtle use, misuse and inappropriate use of terminology are telling and disturbing. Why is the question even necesary in the title? It certainly has no scientific function, especially as used. I will not speculate on motive.

  14. Steve McIntyre
    Posted Dec 27, 2006 at 3:28 PM | Permalink

    #9. bender – I’ve browsed through Davis (1976). ON a first pass, it sure seems like a strange choice of statistical authority for what they are doing. IT would definitely be worthwhile reconciling what Davis 1976 actually recommended against what H&W actually did.

  15. Judith Curry
    Posted Dec 27, 2006 at 3:34 PM | Permalink

    Tim, re hypothesis vs theory: there is some confusion in the scientific vs lay use of these terms. Google theory hypothesis, or scientific method for clarification. Here are some words I pulled from the web that may clarify this for you:

    Hypothesis: A hypothesis is a rational explanation of a single event or phenomenon based upon what is observed, but which is not yet generally accepted. Hypotheses can be supported or refuted by experimentation or continued observation.

    Theory: A theory is an explanation of a set of related observations or events based upon proven hypotheses and verified multiple times by detached groups of researchers. One scientist cannot create a theory; he can only create a hypothesis. Some scientific theories include the theory of evolution, the theory of relativity, and the quantum theory. Scientists continue to tinker with the component hypotheses of each theory in an attempt to make them more elegant and concise, or to make them more all-encompassing. Theories can be tweaked, but are rarely entirely replaced.

    The way were are framing the hurricane-global warming issue is outlined in our paper, which was written to be accessible by lay readers:

    The link between greenhouse warming and increasing surface temperature is arguably a theory (one can argue about the magnitude of the warming, but no credible scientist disuputes the link), whereas the hurricane-global warming link is a hypothesis.

    Bender, yes the causal chain we have formulated depends on the robustness of each of its links. I don’t see any other way to deal with the problem given its complexity. The combination of complexity of the phenomena and issues with the data set present substantial challenges, but at the same time, these same issues make it difficult to refute the arguments as well (none of the hypotheses in the causal chain have been convincingly falsified or otherwise refuted).

  16. Gerald Machnee
    Posted Dec 27, 2006 at 3:34 PM | Permalink

    Re #2 & @11
    **I have argued that the greenhouse warming argument is the best available explanation for the increase in global hurricane intensity and the increase in total number of North Atlantic tropical cyclones (the magnitude of the increase associated with the greenhouse warming is the main issue of contention, rather than the existence of the increase).**
    I have 4 questions:
    1) Did the SST in the North Atlantic increase or decrease from 2005 to 2006?
    2) What was the change in temperature?
    3) Was the change caused by AGW?
    4) What is the proof?

  17. bender
    Posted Dec 27, 2006 at 3:40 PM | Permalink

    Re #13
    That comparison should be easy because I’m sure their code would be made available upon request. And I’m sure that the folks reviewing the paper have the time and interest and courage to make that request to the managing Editor. After all, we all know that (a) Peer Review is flawless, and (b) it’s the truth that everyone’s after, not LPUs (least publishable units).

  18. Judith Curry
    Posted Dec 27, 2006 at 3:44 PM | Permalink

    Jae, in the context of the causal chain that we pose, there are two separate links:

    1. increased SST causes increased global hurricane intensity and increased number of NATL tropical cyclones

    2. the increase in SST since 1970 is predominantly attributed to greenhouse warming (climate model simulations).

    Looking at the entire SST record since 1854 (which is in the xcel spreadsheet i sent to steve), the SST variations over that time period can be attributed to a combination of solar forcing, aerosol forcing, greenhouse forcing, and natural internal variability such as the AMO. So there is no automatic attribution of all warming to greenhouse gases. The proportion of warming since 1970 attributable to greenhouse gases can be debated, but no credible scientist suggests that there is NO temperature increase from the increases in greenhouse gases.

    The Holland/Webster paper addresses link #1. It does not explicitly address link #2, although they refer to other papers that address this link

  19. bender
    Posted Dec 27, 2006 at 3:50 PM | Permalink

    Re #14
    Ironic that we can agree on so much, yet disagree when it comes to the basic principle of cautious scientific interpretation given uncertain data vs. alarmist hand-waving on the remote chance we’re right.

    To say there is a SST-TC link is as meaningless as to argue black or white as to the value of A in AGW. The problem is that the scientists are talking about a generic link, while the politicians are thinking about a specific link, in terms of Katrina and the 2005 hurricane season. The fact is the vast uncertainty around the probability of X hurricanes occurring over Y days dwarfs any putative AGW link. That’s what, I think, gets RPJr’s nose out of joint. To focus on that thread of a link opens the door to suboptimal policy-making and extortionist profiteering by the insurance companies. The way to solve that problem is to stay quantitative wherever possible, and avoid categorical statements. Categorical statements have a way of distorting perspective to society’s detriment.

  20. jae
    Posted Dec 27, 2006 at 3:52 PM | Permalink

    OK, Dr. Curry; can’t disagree with that. Thanks for the clarification.

  21. Judith Curry
    Posted Dec 27, 2006 at 4:18 PM | Permalink

    Bender, I agree with your statement in #18. In the BAMS article, i discussed the values gap between the scientist and the policy maker. Alarmism doesn’t help the situation at all, but it isn’t as bad as completely ignoring the issue either. How to appropriately frame the scientific issues in a way that addresses what people really care about in a way that is scientifically credible is the challenge. The middlemen between the scientists and public (the media, bloggers, congressional committees) have not been helping. what is needed is clarification of values (economic and moral issues) surrounding this topic, and some precautionary planning to assess what the options are and the costs. Then, where opportunity meets value, we can develop some sensible policies.

    in case you didn’t catch it, integrityofscience posted some talks on this subject at AGU, addressing these issues from a scientists point of view (I liked these talks better than the talks on dealing with the media) see
    In

  22. Judith Curry
    Posted Dec 27, 2006 at 4:27 PM | Permalink

    Bender, I agree with your statement in #18. In the BAMS article, i discussed the values gap between the scientist and the policy maker. Alarmism doesn’t help the situation at all, but it isn’t as bad as completely ignoring the issue either. How to appropriately frame the scientific issues in a way that addresses what people really care about in a way that is scientifically credible is the challenge. The middlemen between the scientists and public (the media, bloggers, congressional committees) have not been helping. what is needed is clarification of values (economic and moral issues) surrounding this topic, and some precautionary planning to assess what the options are and the costs. Then, where opportunity meets value, we can develop some sensible policies.

    in case you didn’t catch it, integrityofscience posted some talks on this subject at AGU, addressing these issues from a scientists point of view (I liked these talks better than the talks on dealing with the media) see http://scienceblogs.com/integrityofscience/

  23. Dane
    Posted Dec 27, 2006 at 4:47 PM | Permalink

    Dr. Curry,

    This statement is a little insulting don’t you think “The proportion of warming since 1970 attributable to greenhouse gases can be debated, but no credible scientist suggests that there is NO temperature increase from the increases in greenhouse gases.”

    As geologist with a MS in Environmental Geology, myself and all of my geologist coworkers disagree with your statement. Firts of all, with only 150 yrs of data, you really can’t say one thing about natural variability. All you can say anything about is the variability within the dataset you have, which is very small relative to the amount of time since the end of the last Ice Age, and hence the assumed warming of SST’s. Proxies for tempurature are good tools, but not exact measuring devices.

    I find your Ivory tower attitude slightly annoying. Just becasue scientists like myself are not in Academia or publish on a regular basis, does not mean we do not understand the debate taking place and are not “Credible Scientists”. From my vantage point it appears you assume that AGW is true, and are looking for any link to justify that assumption, no matter how weak the link may be. It also appears a class in basic geologic time is needed in the entire climate field.

  24. Steve Bloom
    Posted Dec 27, 2006 at 5:13 PM | Permalink

    Re #20: Sorry, Dane, but professional geologists working for the fossil fuel industry aren’t credible scientists. One, they are not qualified (fuds and pubs are rare), two, on the rare occasions any of them do publish (outside of trade pubs) it seems to be under questionable circumstances (and with often ludicrous “science”) such as with those two Russian characters from USC, and three, they made a collective laughingstock of themselves when they gave that award to Crichton. But by all means try to prove me wrong: Do some research and publish on the SST-global warming connection. Heck, just do an “audit” of the papers that demonstrate the relationship and post it here. BTW, nobody is under the illusion that the fossil fuel industry and its employees don’t understand the debate.

  25. Dave Dardinger
    Posted Dec 27, 2006 at 5:15 PM | Permalink

    re: #2, #20

    From my vantage point it appears you assume that AGW is true, and are looking for any link to justify that assumption, no matter how weak the link may be.

    Which brings us back to what I was saying to Dr. Curry that she responded to in #2. It’s true that preconceptions can cut both ways, but what bothers me is that the AGW crowd seems bent on painting the anti-AGW position as unacceptable as you (Dane) pointed out in phrases like “no credible scientist.” Of course when called up on it the tendency is to say something like, “Surely you don’t link yourself with those non-credible ones who deny any warming whatsoever, do you? That no specific scientists with such positions are brought forward for branding as non-credible, at least by the AGW proponent (various trolls and misc. bloggers who will brand all and any skeptic as non-credible are another story), tends to be overlooked.

  26. bender
    Posted Dec 27, 2006 at 5:24 PM | Permalink

    You might find it annoying; however it is not Dr Curry’s personal verdict, but prevailing professional/expert opinion. Like it or not it is a fact that proponents of A=0 have little or no credibility among the scientific community at large. Agnosticism, ok, somewhat acceptable. A LT 50% Ok, maybe. But A=0, untenable. If you dissent with expert-opinion-at-large, spell out your case.

    I’m tempted to admonish: don’t shoot the messenger. Except, of course, that’s sort of how alarmism works: the most direct way to stop the spread of the alarming message is to terminate the transmission process at the source. (Alternatively, you could damp the impact of the message at the receiving end by making sure the receiver is able to put the alarmist message in proper context. Pretty challenging though.)

  27. TAC
    Posted Dec 27, 2006 at 5:24 PM | Permalink

    #13 I just took a quick glance at the manuscript, and was surprised to encounter:

    All data are therefore smoothed by a simple 9-y running mean. Standard statistical tests are used throughout. Unless otherwise indicated all variance and correlation coefficients are valid at the 95% level using the t-statistic with correction for serial correlation (Davis 1976).

    First point: This approach is absolutely not a “standard statistical test”. In fact, I find the claim reminiscent of the late and beloved HS; each of the components has a legitimate place somewhere in the world of statistics, but the combination is something else altogether.

    Having said that, I have a hard time understanding why one would begin by taking a lag-9 moving average — which would typically introduce correlation out to lag 9 — and then try to fix things by using a correction for serial correlation (AR(1), or lag 1). Without the correction, the standard trend tests will be hugely biased in the direction of finding significance when it isn’t there. So the question is: Does the “correction for serial correlation” take care of this problem? It might, I suppose. Without seeing any equations, it is hard to say.

    In summary, I have no idea what these results mean. Has anyone had any luck working out the properties of this bizarre test?

  28. bender
    Posted Dec 27, 2006 at 5:26 PM | Permalink

    Re #21
    Again, a subsample of 20 CA skeptics pegs A at 20%, not 0%, as many warmers would like to assert.

  29. J. Curry
    Posted Dec 27, 2006 at 5:29 PM | Permalink

    speaking of shooting the messenger, i had a post blocked (“waiting moderation”) on my regular email account

  30. bender
    Posted Dec 27, 2006 at 5:30 PM | Permalink

    Has anyone had any luck working out the properties of this bizarre test?

    First things first. Has anyone had any luck identifying the actual test extracted from Davis (1976)? We have to identify it before we can characterize it.

  31. bender
    Posted Dec 27, 2006 at 5:33 PM | Permalink

    Re #25
    Sudden increase in commenting frequency after a long lapse since previous correspondence?

  32. TAC
    Posted Dec 27, 2006 at 5:45 PM | Permalink

    Following up on #23 and the statement that “the standard trend tests will be hugely biased in the direction of finding significance” (unless corrected), it seems that, without the correction for serial correlation, the probability of rejecting H0 when we have 156 observations of white noise is about 50 percent.

  33. J. Curry
    Posted Dec 27, 2006 at 5:46 PM | Permalink

    repeating the text of my “censored” message:

    Bender, I agree with your statement in #18. In the BAMS article, i discussed the values gap between the scientist and the policy maker. Alarmism doesn’t help the situation at all, but it isn’t as bad as completely ignoring the issue either. How to appropriately frame the scientific issues in a way that addresses what people really care about in a way that is scientifically credible is the challenge. The middlemen between the scientists and public (the media, bloggers, congressional committees) have not been helping. what is needed is clarification of values (economic and moral issues) surrounding this topic, and some precautionary planning to assess what the options are and the costs. Then, where opportunity meets value, we can develop some sensible policies.

    in case you didn’t catch it, integrityofscience posted some talks on this subject at AGU, addressing these issues from a scientists point of view (I liked these talks better than the talks on dealing with the media) see http://scienceblogs.com/integrityofscience/

    p.s. to the skeptics: it is pretty difficult to argue ZERO increase in surface temperature for 30% or 100% increase in a major greenhouse gas like CO2. You may not agree with climate model attribution studies for the 20th century, or their projections for the 21st century, but to say ZERO increase is pretty indefensible. Among the most extreme AGW skeptics (that actually publish on these topics), Bill Gray anticipates 0.3C increase for doubling CO2. A more moderate skeptic (Lindzen) anticipates a 1C increase (which is just shy of the 1.5C increase that comprises the IPCC lower limit on their range of expected increase). The real debate (scientific and policy) is 1C (do we care?) vs 5C (yeah we definitely care). Do the 20% CA skeptics say ZERO increase from doubling CO2, or do they say yes some increase but we have no idea how much and we aren’t convinced by the allegedly “alarmist” values of order 3C or greater?

  34. J. Curry
    Posted Dec 27, 2006 at 5:48 PM | Permalink

    to the skeptics: it is pretty difficult to argue ZERO increase in surface temperature for 30% or 100% increase in a major greenhouse gas like CO2. You may not agree with climate model attribution studies for the 20th century, or their projections for the 21st century, but to say ZERO increase is pretty indefensible. Among the most extreme AGW skeptics (that actually publish on these topics), Bill Gray anticipates 0.3C increase for doubling CO2. A more moderate skeptic (Lindzen) anticipates a 1C increase (which is just shy of the 1.5C increase that comprises the IPCC lower limit on their range of expected increase). The real debate (scientific and policy) is 1C (do we care?) vs 5C (yeah we definitely care). Do the 20% CA skeptics say ZERO increase from doubling CO2, or do they say yes some increase but we have no idea how much and we aren’t convinced by the allegedly “alarmist” values of order 3C or greater?

  35. John S
    Posted Dec 27, 2006 at 5:58 PM | Permalink

    Dr. Curry,

    I note with interest your comment:
    “…these same issues make it difficult to refute the arguments as well (none of the hypotheses in the causal chain have been convincingly falsified or otherwise refuted).”

    and link it with the requirement for falsifiability in the scientific method. This suggests the data is not capable of sustaining robust scientific methodological exploration. The answer depends on the assumptions not the data. That is, the current hypotheses are not falsifiable given the current data and this means that it is less a matter of science and more a matter of faith – at least until better data, capable of falsifying the hypotheses, is available.

    Comment?

  36. J. Curry
    Posted Dec 27, 2006 at 6:02 PM | Permalink

    re #18 Bender, I agree. (here is another try to get my “flagged” message through). In my BAMS article, we discussed the values gap between scientists and policy makers. which is worse, alarmism or ignoring the issue all together? hard to decide which is worse, the are both wrong, but those seem to be the choices we are presented with. How to appropriately frame the scientific issues in a way that addresses what people really care about in a way that is scientifically credible is the challenge. The middlemen (the media, bloggers, aadvocay groups) have not been helping.

    What we need (instead of alarmism vs ignoring) is clarification of values (economic and moral issues) surrounding this topic, and some precautionary planning to assess the options and their costs and benefits. We can then develop some sensible policies where opportunity meets values. Instead, we fight this silly war over the science that is admittedly uncertain.

    in case you didn’t catch it, integrityofscience (won’t try to post the link, that may be why this message has been censored?) posted some talks on this subject at AGU, addressing these issues from the perspective of scientists

  37. Welikerocks
    Posted Dec 27, 2006 at 6:12 PM | Permalink

    Hey messangers, at this point in time I don’t even believe there is “warming”. I find the various temperature reconstructions unconvincing. You guys argue fractions of C over small timescales like “since 1970″ and simulate the living breathing changing earth on computers as if you know “it” well. Besides the uncertainty, and the bad data, and the models, I can’t swallow it because of the attitude!

  38. J. Curry
    Posted Dec 27, 2006 at 6:12 PM | Permalink

    John S: not faith, but a hypothesis. Better observations are needed (reanalysis of existing data, a few more years in the time series), continued examination of our theories, and continued improvement to our models will eventually resolve this issue. Much work to be done. But the hypothesis of greenhouse warming contributing to an increase in hurricane activity is presently supported by the observations, the theory, and models. They all agree on the sign of the relationship. They disagree on the magnitude of the relationship, and whether any increase can actually be detected in a significant way with the data that we do have. So again, the issue is not so much whether we would expect global hurricane intensity to increase with greenhouse warming, but how large the increase might be. The intriguing thing about the North Atlantic data set is that the quality and quantity of the data may be sufficient to quantify this relationship in a credible way, that is the focus of activity right now. The correlation of SST and tropical cyclone counts is a key link, the statistics of this needs to be sorted out. I’m sure Holland’s analysis (and Emanuel’s) is not the last word on this. I look forward to CA’s take on this (i’ve also asked Carlos Hoyos for his opinion on how to address this)

  39. J. Curry
    Posted Dec 27, 2006 at 6:38 PM | Permalink

    welikerocks: the physics of radiative transfer indicate the direct forcing for doubling CO2 is 0.4C increase in surface temperature. Pretty hard to argue against these physics, which go beyond theory to the status of scientific law. The issue is of course the feedbacks in determining the actual magnitude of the temperature increase (no conceivable feedback scenario suggests an actual global cooling associated with increasing CO2). So ZERO increase of surface temperature with a 30% (so far) and 100% (next century) increase of CO2 is scientificially indefensible.

  40. jae
    Posted Dec 27, 2006 at 6:44 PM | Permalink

    Dr. Curry: I hope you have read (and studied) this paper. It appears to me that anyone who posits highly significant amounts of warming from GHGs has to be able to “tackle” the basic logic and physical explanations presented therein. Modelers should especially pay attention to it, since it offers a much better approach than computer modeling for estimating the effects of GHGs (IMO, of course).

  41. jae
    Posted Dec 27, 2006 at 6:45 PM | Permalink

    Darn it, I linked the wrong paper. HERE.

  42. Dane
    Posted Dec 27, 2006 at 7:03 PM | Permalink

    Bender and Dr. Curry,

    Most people ie scientific skeptics in geologic and environmental consulting don’t have time to post on CA, or even know CA exists for that matter. Just because benders little survey says 0%, means nothing to me. It is also irrelevent that we do not publish in the field for the purpose of this discussion. What is important is our detailed knowlege of earth processes, not just climate science, that allow us to understand the current data and compare it to older examples from the geologic record. When one does this, one finds Co2 levels were much higher in the past, yet climate continued to fluctuate radically from warm to cold. Lag times in ice cores is another problem I have with correlating Co2 content in the atmosphere and tempurature averages on earth. They don’t match. Apparently this doesn’t bother bender or Dr Curry, but as an environmentalist and concerned scientist, it sure bothers me.

    I also am bothered by the lack of time and precision in the limited real world data we do have. As a consultant who deals with government regulatory agencies on a regualr basis, as well as a former regulator myself, I am also fully aware that data can be “Massaged”, smoothed, whatever you want to call it, statistically manipulated, to give one the result one desires.

    Bender please find me one article showing proof of a link between Co2 concentrations in earths atmosphere and earths temperature? No models please, real data. The ice cores have all been shot down, what else do you have?

    I understand the earth is a complex adaptive system, NOT a greenhouse. I also see potential benifits in earth becoming warmer, many more benifits than from a colder earth. I think we all can agree that the climate WILL NOT stay the same, regardless of what we do. Dr. Curry, am I to believe that you would like the earths climate to just stay the same? You know perfectly well that will never happen. That is the feeling I get from proponents of AGW.

  43. Paul Linsay
    Posted Dec 27, 2006 at 7:11 PM | Permalink

    #10, Is the raw data that was used to construct Fig. 1 of Holland and Webster available somewhere in convenient form, eg, ascii?

  44. welikerocks
    Posted Dec 27, 2006 at 7:29 PM | Permalink

    #34 Dr. Curry,
    Thanks for your reply.

    the physics of radiative transfer indicate the direct forcing for doubling CO2 is 0.4C increase in surface temperature.

    I understand all the models. And I understand that is still one. It is Earth simulated in a lab. That is, physics with a lid on it and not a real world.

    And I believe most people everywhere are willing and ready any time to change engines and fuels for reasons besides the climate- so your “What we need (instead of alarmism vs ignoring) is clarification of values (economic and moral issues) surrounding this topic” isnt what I think we need here from a scientist. Truth, fact-and politics out of the lab and heads of scientists who run in circles of acadamia and “the media” is what we need.

    Can you get sued for being wrong as working scientists do in the real world?

  45. J. Curry
    Posted Dec 27, 2006 at 7:38 PM | Permalink

    Dane, climate varies on many time scales, associated with external forcing and natural internal variability. AGW is one of many causes of climate variability. Whether or not AGW in the next century will be swamped by highly active volcanoes or unanticipated solar variability (or some crazy internal oscillation in the North Atlantic), who knows. The existence of other forcing mechanisms (like solar variability and volcanic activity) that we can’t predict doesn’t imply that AGW will not be a significant factor in the climate variability in the next century or two. The physics of relating atmospheric composition and radiative transfer is very well understood. You can’t make major changes to the atmospheric greenhouse gases and expect climate to remain unchanged. The geologic record is part of the puzzle, but not the only part. There is a very real risk that AGW will be a significant factor in the climate variability in the next century or two. Should we really ignore this risk because you don’t think that the geologic record alone provides convincing evidence that a rapid increase in atmospheric CO2 will result in warmer surface temperatures?

  46. bender
    Posted Dec 27, 2006 at 8:01 PM | Permalink

    Re #37
    “bender’s little survey” (n=20) said A ~ 20% ± 10% (standard error), much like Lindzen. The purpose of this survey was not to establish any kind of consensus, dear friend. It was merely to point out that A=0 is not the “skeptic position” (if there is such a thing). It was a counterpoint to some turkey who was suggesting that was the skeptic position.

    No models please

    Why not? Do you dismiss all models as a matter of principle? It’s one thing to argue that models X,Y,Z are flawed. Quite another to dismiss them all as a matter of principle. You aren’t that twisted, are you?

  47. J. Curry
    Posted Dec 27, 2006 at 8:03 PM | Permalink

    Welikerocks: 99.99% of climate scientists do not get involved in any prescription of policies. They do their job which is to investigate nature, and publish papers. Many scientists feel a responsibility to communicate their research and its social and ethical implcations (surely this is better than keeping scientific research hidden and inaccessible to the public). Scientists are often accused of having an “agenda” simply by virtue of publishing a relevant paper. Then policy makers and the media push a scientist into uncomfortable positiona (like asking them to do IPCC assessments, asking impossible questions in inteviews and then misquoting them, etc.). Then certain elements of the media, bloggers and certain advocacy groups (and geologists) attack them for destroying the U.S. economy, etc, while others on the opposite side of the political spectrum tout each new paper (and the scientists) that supports their agenda. Scientific research does not translate into policy (which is clear since we have had alot of science but no policy to deal with the issue, at least at the federal issue). Policy is a complex interplay of uncertain science, technologies, economics, politics, and values. Whether there are policy decisions to ignore AGW or to do something about it, surely it is best that at least some scientists engage with the decision makers so the risks of the different policy options are undersood.

    You may not like the “attitude” of the 0.01% of scientists that are out there publicly on this issue, talking about tipping points, 10 years to act, etc. But dislike of this “attitude” of a few scientists is not an objective reason to reject all of the scientific arguments being made. Skeptics need to be skeptical of their skepticism, to insure that you are not overreacting emotionally to attitudes of a few people

  48. Ken Fritsch
    Posted Dec 27, 2006 at 8:10 PM | Permalink

    Jim Kossin presented the linked presentation at the AMS Environmental Science Seminar Series October 20, 2006 where as I recall Holland also reviewed his paper under discussion here. I was curious whether Kossin’s paper is in publication/review at this point and how it is being received.

    It would apparently put much of the previously uncorrected tropical cyclone data outside the NATL and Eastern Pacific into question as it shows the mean intensity of storms diverging from that reported the further one goes back into time.

    Summary:

    The trends in the Atlantic and East Pacific intensity records are well supported by our new record. Upward in the Atlantic. Downward in the East Pacific. No upward trends were found in the West Pacific, Northern and Southern Indian, and South Pacific oceans. The vast majority (85%) of global hurricane activity takes place in these basins. Similar warming trends are found everywhere in the tropics. Why is the Atlantic behaving so differently? If the data are not good enough to accurately measure long-term hurricane behavior, then our path to understanding how hurricanes will change in a warming world must be through better physical understanding. This is our present research challenge.

    http://www.ametsoc.org/atmospolicy/documents/Kossin102006.pdf

    Landsea papers on the quality of tropical cyclone data were reviewed on CA in the thread below.

    http://www.climateaudit.org/?p=864

  49. David Smith
    Posted Dec 27, 2006 at 8:10 PM | Permalink

    I’ve been reading the Holland/Webster (HW) report and will try to summarize some of its points. Judith or others, correct me where I err:

    * First, the paper is a self-described “broad brush” review. It is not a rigorous, tightly-argued paper but rather a weaving of ideas.

    * The subject is trends in North Atlantic (NATL) storms (hurricanes + tropical storms). The data examined covers a wide time range of admittedly varying quality. Nevertheless, the attempt is to make sense out of the available data.

    * Storm count data since 1900 is seen as good enough to be useable and intensity data since 1944 (start of aircraft reconnaissance) is seen as good enough to be useable

    * The search for trends is confounded by the overlap of various natural oscillations (El Nino and some interesting stuff at 2-3,5,9 and decadal intervals). To try to remove this short-term variation, they use a 9-year moving average in places.

    * The paper also uses SST data for various parts of the Atlantic. They believe it is accurate enough for the purposes of the paper.

    * The HW paper tweaks the noses of Landsea and Gray in places, especially Landsea

    * Figure 1 uses smoothed (9-yr average) data to show that Atlantic storm count (hurricanes + tropical storms) has increased in steps, with a step increase around 1930 and another step increase around 1995. Each step-up increased storm count by about 50%.

    * Figure 1 also plots SST of part of the Atlantic, including a distinct rise beginning in the mid-1970s (DS note: seems like I’ve seen “mid-1970s change” mentioned before, somewhere)

    * The paper says there’s a statistically significant relationship between the storm count and SST.

    * Figure 2 is a “phase diagram” which shows SST/storm count relationships. There are “regimes” of squiggles. The most recent regime (1995+) has not stabilized.

    * Figure 3 shows no trend in the % of Atlantic storms which become hurricanes and no trend in the % which become major hurricanes. However, it does show an absolute increase in the number of hurricanes, because the overall number of Atlantic storms has increased.

    * Figure 4 shows that the % of storms which become major hurricanes oscillates over time, independent of SST.

    * Figure 5 shows that the major storms form south of 25 degrees latitude (Miami)

    * Figures 6 and 7 are to show in a simple way that Atlantic SST are part of a global warming of SST

    * The causal chain is:

    1. global SST has increased due to increased greenhouse gases
    2. there is a statistically significant relationship between SST and storm count in the Atlantic
    3. the SST/storm count relationships are due to shifts between climate regimes

    * hurricanes are forming deeper in the tropical Atlantic, which leads to more-intense hurricanes

  50. jae
    Posted Dec 27, 2006 at 8:19 PM | Permalink

    Dane: You have friends here, you know….You seem to misunderstand this site. Read some of the threads.

  51. jae
    Posted Dec 27, 2006 at 8:29 PM | Permalink

    There is a very real risk that AGW will be a significant factor in the climate variability in the next century or two. Should we really ignore this risk because you don’t think that the geologic record alone provides convincing evidence that a rapid increase in atmospheric CO2 will result in warmer surface temperatures?

    Noted. But what does “ignore” mean, here. Should we alter the whole economic structure of the world and screw the poor nations, “just in case?” Here we go with the overblown “precautionary principle” again. Just like DDT: kill millions and millions of people because of the precautionary principle. Disgusting!

  52. Dane
    Posted Dec 27, 2006 at 8:30 PM | Permalink

    Bender,

    Why attack and call me twisted? What do you gain from that? In my scientific experience models. although useful tools in many situations, are usually flawed and unreliable at best. They are by no means the standard in good science. As a geoscientist I want to see REAL data, why do you have a problem with that? If Co2 was such a major player in global climate (as say Malinkavitch cycles are), you would think that it would leave its signature somewhere on earth, where we can find it and study it. In 1991, models predicted we would loose 10,000 troops in Gulf War 1, they were WAY off. Those were excellent models I might add that were perfected in WW2 and again in ground combat operations in Vietnam. Am I foolish enough to belive that something even more complex than modern human ground combat, like Global climate for example, is better understood?

    A better question is why do you have so much faith in the models?

    Dr. Curry,

    I quote “The physics of relating atmospheric composition and radiative transfer is very well understood”. I know this. My point is that those physics are for a model, not the real working earth, which you must agree is far more complex than any model humans have yet developed.

    I quote again “You can’t make major changes to the atmospheric greenhouse gases and expect climate to remain unchanged”. actually you can if those gases aren’t as important as you think they are on a real working planet. Again you assume the earth is a greenhouse, I don’t.

    You think there is a real risk in the next century or two from GHG, I don’t. I look at all the data, and think people who live in areas that may be affected by future warming will have time to adapt (which will create jobs), and that the risk is exaggerated greatly by both acedemia and especially the mainstream media. It is literally in the news EVERY DAY! article, news peices on local TV, TV shows, jokes, sitcoms etc, the propoganda is endless and the reality is we don’t even know if its happening, then we don’t even know if it will be a bad thing. Again, picture the Littte Ice Age, did Co2 conc. drop? Or during the Medieval Warm period did Co2 increase? What was better for humanity? I would argue warm is far better than cold. Panic about 100 to 200 years from now does nobody any good at all unless your trading futures or in carbon credits. And last, please admit you hurricane data is sparse at best in intensity, distribution, and numbers over the past 150 years. You must realize you are looking at a snapshot of time as far as the real earth goes, geologically speaking. To think you can make policy based on the weak evidence you have is scary, to the point of extremism. Do I think we should continue to study the issue? Of course, it should remain well funded, if for nothing else than to gain the knowledge of how the earth works. Do I want off Oil as our main energy source? Yes, but not due to some fallacy about Co2 and fossil fuels warming the planet. The science so far is weak in my opinion, almost criminal when it comes to the media propaganda and blitz we have seen over the last 10 years ( I began to seriously study the issue in the early to mid 1990′s).

  53. bender
    Posted Dec 27, 2006 at 8:34 PM | Permalink

    Lag times in ice cores is another problem I have with correlating Co2 content in the atmosphere and temperature averages on earth. They don’t match. Apparently this doesn’t bother bender

    That’s it, keep it up, friend. Keep putting words in my mouth. Keep impugning motive where there is none. Shoot the messenger. You’re a gem.

  54. Jim Edwards
    Posted Dec 27, 2006 at 8:34 PM | Permalink

    #40, 42

    Dr. Curry:

    I agree wholeheartedly that we shouldn’t be forced into choosing between two antithetical positions for acting on GW – “ignoring” and “alarming.”

    I hope that you would also agree that passing zero laws / enacting zero regulation Re: CO2 and other GHGs while funding vigorous research is not the same as “ignoring” GW as a potential problem – when thinking people can agree that there is no compelling evidence demonstrating A in AGW is not just likely more than zero, but actually a demonstrably significant and disturbing factor in climate change.

    As long as you and your colleagues are receiving grants of taxpayer dollars, and continue to investigate, the problem is not being “ignored.”

  55. bender
    Posted Dec 27, 2006 at 8:37 PM | Permalink

    why do you have so much faith in the models

    Err, because my faith is strong? Do I win? Keep it up. Don’t bother reading the blog. Best just to assume you’re right and I’m wrong.

  56. J. Curry
    Posted Dec 27, 2006 at 8:38 PM | Permalink

    Ken, re Jim Kossin’s work, apparently this has not yet been published (although I believe that papers have been submitted). Hence his work hasn’t quite received the scrutiny yet that it will receive once it has been published. There are two reasons i like his study:
    1) he uses an objective analysis technique that does subjectively depend on the analyst (in the way that the Dvorak technique does)
    2) he seems personally objective and unbiased on the subject of hurricane trends

    His paper seems to demonstrate that the NATL and EPAC intensity data sets that have been used in studies such as Webster et al. are reliable (i.e. his analysis agrees with the standard data sets). However, in the other ocean basins, his analysis disagrees with the conventional analyses. Since his algorithm was trained on the NATL dataset, it is unclear how accurate this algorithm will be in different regions characterized by storms that are on average larger and more intense. The community is currently focused on sorting out the discrepancies in the WPAC data set; WPAC has 40% of the global TCs, and played an important role in both the Emanuel and Webster et al. papers. In addtion to the JTWC dataset (used by Emanuel and Webster), and Kossin’s analysis, several other WPAC data sets are also out there (from Hong Kong and Taiwan), all of them disagreeing with each other (there was a recent EOS article on this). There is some hope that the aircraft reconnaissance flights in the WPAC during 1983-1987 can be used to sort out these discrepancies (note Kossin and JTWC show large disagreements during 1983-1987). Also, Kossin did not explicitly calculate PDI or NCAT45, so exactly what the discrepancy in his dataset with that used by Webster and Emanuel remains unclear.

    Bottom line: there have been numerous claims (e.g. landsea and gray) that the global TC data base is no good. I view Kossin’s work to be the first important word on this subject. However, it is far from the last word. The discrepancies among the different data sets need to be sorted out, the accuracy of Kossin’s method needs to be assessed in the different regions, and Kossin’s analysis needs to be extended to the period prior to 1983. Until this is sorted out, the community is focusing on the NATL data (and people should also start looking at EPAC more closely as well.)

    p.s. one statistical issue. it seems that he calculated the trends using the 5 year averages, I doubt that the trends from old and new analyses calculated this way are significantly different from each other?

  57. bender
    Posted Dec 27, 2006 at 8:46 PM | Permalink

    Why attack and call me twisted?

    I didn’t. I asked if you dismissed models as a matter of principle. You haven’t answered that question, although I’m starting to get a hint of your reply. In which case, you definitely have mine.

    It’s not about what I gain from attacking you, but what jae and the solar gang gain. jae needs those models that you so eagerly dismsiss, if he’s going to have a home for cosmic ray flux as a forcing process on equal footing with GHGs.

  58. nanny_govt_sucks
    Posted Dec 27, 2006 at 8:50 PM | Permalink

    …the reality is we don’t even know if its happening, then we don’t even know if it will be a bad thing. … I would argue warm is far better than cold.

    Absolutely.

    There is much evidence that the planet is benefiting from recent warming. Greening, and poleward-expanded habitats for animals is certainly a good thing when compared to the opposite. Isn’t it possible that just maybe this planet is currently too cold and dry?

    And has anyone ever stopped to consider what the concentration of CO2 SHOULD be? It is said that plants virtually starve at 200ppm, and studies show that the vast majority thrive best at around 1000-1200ppm. Shouldn’t we consider that what is best for plants is best for the animals (including humans) that eat those plants?

  59. J. Curry
    Posted Dec 27, 2006 at 8:55 PM | Permalink

    Jae, why do you assume that if AGW is a real risk, then the only policy option is to destroy economies? 50 years from now, do you really expect that we will have a fossil fuel based economy (even if AGW wasn’t an issue?) Economies aren’t static, they will change independently of AGW issues, in response to technological innovation, politics, natural resource availability, etc. There are many AGW policy options to consider, from adaptation to mitigation (which includes not just emissions reductions, but carbon sequestration), some of which can be regarded as “no regrets”. Too much energy is being spent to sow confusion about the science (more so than actually exists) and hence we haven’t spent sufficient time exploring the various policy options. Precautionary planning is different from applying the precautionary principle to policy. Its time to start putting energy into developing policy options that are technically feasible, cost effective, and politically viable. Why should we automatically assume that such policy options don’t exist?

  60. David Smith
    Posted Dec 27, 2006 at 8:59 PM | Permalink

    Re #44 One problem I have with HW is the quality of the older data. This is an old problem and covered in detail in the past at CA. It’s hard for me to build up much enthusiasm for the HW analysis when I’m bothered by the data on which it is founded.

    To beat this horse once again, I want to add a new ocular exercise, with regards to holes in the data. This will take a minute because I have not consolidated these:

    Glance at 1931 , 1932 , 1933 , 1934 , 1935 , 1936 , 1937 , 1938 , 1939 , 1940 , and on to 1945. I think the visual impression is that there are few storms east of 55W. What was happening between 55W and Africa? It’s like a void existed, with few exceptions.

    Now, glance at 1995 , 1996 , 1997 , 1998 , 1999 , 2000 , 2001 , 2002 , 2003 , 2004 and 2005 .

    The activity east of 55W is widespread.

    I would say what has changed (today vs 1930s/40s) is data collection in these remote parts of the ocean, thanks to reconnaissance and especially satellite. In my opinion, some weaker storms in the east Atlantic were missed and some peak strengths were missed. HW may say “regime change” with hotter SST and more activity in the east Atlantic. I think it’s a data collection problem.

  61. Brooks Hurd
    Posted Dec 27, 2006 at 9:05 PM | Permalink

    David,

    I perused the HW paper and came away with the same impressions which you outlined above. When I looked at the step changes, what came to my mind as an explanation is changes in tropical cyclone observation methods. No one would argue that our observations today identify essentially all TCs. Our methods have improved significantly in steps over the past 150 years. We certainly missed a high percentage of TCs in the second half of the 1800s. In addition to counting TCs, the categorization methodologies have also changed.

    Judith

    In a previous thread you had said that you believed that all the historical TC data needed to be looked at and adjusted for changing methodologies. How well do you believe this paper has been adjusted for changing observational methods and categorization methodologies?

  62. welikerocks
    Posted Dec 27, 2006 at 9:10 PM | Permalink

    J. Curry,

    Then certain elements of the media, bloggers and certain advocacy groups (and geologists) attack them for destroying the U.S. economy

    And geolgoists? I knew it. I just said in another thread how since the get go, the climatolgists had to discredit the field of geology for their A in AGW.

    Scientific research does not translate into policy (which is clear since we have had alot of science but no policy to deal with the issue, at least at the federal issue). Policy is a complex interplay of uncertain science, technologies, economics, politics, and values. Whether there are policy decisions to ignore AGW or to do something about it, surely it is best that at least some scientists engage with the decision makers so the risks of the different policy options are undersood.

    I live in California. Need I say more? I am being taxed and policied, fined and regulated, and my children are taught AGW as fact in their school books.

    Skeptics need to be skeptical of their skepticism, to insure that you are not overreacting emotionally to attitudes of a few people

    No. I am making a opinion, like everyone else here, on the data, the temperature reconstructions and the statisically flawed models THEN not liking the attitude of the scientists who promote these things. I am not as shallow as you assume. I don’t care who the scientist is or what the science is about-only the data matters.

    Bender, I think you should lay off Dane and visa versa. You both are misunderstanding each other. Dane said models are useful tools and of course jae needs his models, but the difference is solar scientists aren’t running around proclaiming they know everything by what they see in models. I see a major difference in wording and presentation of ideas here. Climatologists and the powers that be are taking the climate, which does bother, affect, change and run human lives, and they are using these play by play snap shots, or models to control things politically. It’s brilliant propaganda, because every single senerio fits somebody’s fear or grudge or concern doesn’t it? In the meantime, we’ve had record low temperatures where I live and nary a peep in the media about it as unusal. My plants I have had for years and years are freezing. Doesn’t fit the agenda I suppose.

  63. J. Curry
    Posted Dec 27, 2006 at 9:21 PM | Permalink

    David and Brooks,

    Since 1960 (when we have had satellite data in the NATL, i am virtually 100% certain that we have caught all of the TCs. Since 1944 (and probably since 1941, with all the cross atl transport assoc with WWII), we probably caught virtually all of the storms

    Prior to 1944, we can try to do consistency checks (e.g. HW looking at the ratio of hurricanes to total TCs) and David’s checking of the storms E of 55 degrees). David, one issue that may explain the lack of tracks east of 55W is that the eastermost portions of the tracks may have been frequently missed, with the tracks picked up once the storm moved west. A statistical analysis of the post 1970 data for the number of TCs that stayed wholly in the eastern portion for its entire lifetime would help identify the fraction that might have been missed. One can imagine other consistency tests, such as looking at the El Nino/TC relationships in the early period vs the period since 1970. Also, the ratio of U.S. landfalls (arguably more reliable than total NATL TCS) to total NATL TCs would be another consistency check. It is probably a useful exercise to see if we can come up with some sort of adjustment for early undercounting. But i suspect that in the 19th century, the undercounting doesn’t exceed 25% (and is likely to be more like 10-20%).

    Emanuel has argued that the strong correlation of SST and TC counts on multidecadal time scales, even back to about 1870, lends credence to the older TC counts. I for one do think that the AMO does influence NATL TCs, and see what looks like an AMO feature reflected in the TC counts with minima around 1850, maxima around 1885, and then minima around 1915.

    one way to test whether such an adjustment would matter is to adjust all annual TC counts in the 19th century upward by 25%, and then do the correlations with SST.

  64. Ken Fritsch
    Posted Dec 27, 2006 at 9:30 PM | Permalink

    David that was a great review and now I have some puzzlements to post.

    Holland and Webster indicate that they used July to September SSTs in the Eastern Atlantic (5N-25N,55W-20W) for correlation with NATL cyclones (that is what the labeled graphs indicated as they reference the Eastern Atlantic). They also give Goldenberg and Shapiro (1996) as a reference for the MDR, the main development region for tropical storms, who in turn would appear to indicate that as bounded by (10N-20N,80W-20W) in the link below and as shown in the graphic below.

    Emanuel 2005 used SSTs for (6N-18N,60W-20W) for the month of September only for his correlations of PDI with SST.

    My question to you, David, or anyone else willing to take it on: What is the sensitivity of the cyclone activity/intensity to SST to the SST bounds selected and the months of SST data used?

    http://iri.columbia.edu/climate/ENSO/globalimpact/TC/Atlantic/windshear.html

  65. Brooks Hurd
    Posted Dec 27, 2006 at 9:37 PM | Permalink

    Re: 42

    Judith,

    Then certain elements of the media, bloggers and certain advocacy groups (and geologists) attack them for destroying the U.S. economy, etc, while others on the opposite side of the political spectrum tout each new paper (and the scientists) that supports their agenda.

    I agree with you on this issue. Personally, I believe that one of the major problems that we have in describing scientific issues to the press (worldwide) is that the press is so scientifically challenged. I feel that the reason that the press missquotes scientists is that the press has no clue what scientists are talking about. The press reports what they think the scientists said rather than what the scientists meant. All too often this means that there is huge gap between what ends up in the press and the scientists’ statements upon which the story was supposedly based.

    It is possible to get the press to comprehend scientific points, but it takes time and effort. As scientists and engineers we have not done a very good job of providing resources to the press so that they can get the story right. We need to do a better job of this I find it unlikely that the press corps is going to solve its problem any time soon.

    There are agenda driven groups who are happy to share their opinions with the press, however this tends to give the press an opinion that there is only one side to the story and the presence of a hypothesis (theirs) means that all the science is settled. I really believe that the public deserves to hear more than just one side of a story.

  66. bender
    Posted Dec 27, 2006 at 9:44 PM | Permalink

    I want to see REAL data, why do you have a problem with that?

    “Why?” Because, unfortunately, there’s a limit as to what you can learn from statistical analysis of the limited available data. All theings being equal, I agree: data are always preferable to models. Unfortunately good data do not come cheaply or quickly enough. Hence the value of modeling. It’s not either/or. Both need to be pursued simultaneously in proportion to their benefit:cost.

    Rocks says “lay off” so I will. I like rocks.

  67. Ron Cram
    Posted Dec 27, 2006 at 9:52 PM | Permalink

    re:14

    Dr. Curry,

    You wrote:

    The link between greenhouse warming and increasing surface temperature is arguably a theory (one can argue about the magnitude of the warming, but no credible scientist disuputes the link), whereas the hurricane-global warming link is a hypothesis.

    This is not accurate. You might want to read the wikipedia article found
    here.

    One scientist does not believe the increase in CO2 is mainly from mankind. Another thinks the current warming is purely natural, a cycle that will be reversed beginning somewhere between 2012-2015.

    I direct you especially to Khabibullo Abdusamatov and Zbigniew Jaworowski.

  68. Tim Ball
    Posted Dec 27, 2006 at 9:53 PM | Permalink

    #55
    Thanks David for drawing attention to the data and the changes in coverage, measurement devices and techniques, as well as analysis changes. I raised these issue in #12, as well as the problems of complete prediction failure for 2006 based on the same data sets, but all this was avoided by focussing on definitions of theory and hypothesis. The definitions provided did not resolve the problems I identified with the original definition in #14, but they roiled the water sufficiently to appear like an answer. I am not impressed by attempts to qualify the reliability of the data, especially when there is any subjectivity. Something about silk purses and sows ears.

  69. Dane
    Posted Dec 27, 2006 at 10:08 PM | Permalink

    Bender,

    One last point, Why does the data have to be statistically manipulated? In Geologic consulting we avoid that hole since it leaves us open in court to being easily manipulated.

  70. Ken Fritsch
    Posted Dec 27, 2006 at 10:11 PM | Permalink

    It is possible to get the press to comprehend scientific points, but it takes time and effort. As scientists and engineers we have not done a very good job of providing resources to the press so that they can get the story right. We need to do a better job of this I find it unlikely that the press corps is going to solve its problem any time soon.

    The problem with the press, and the media in general, is that they judge in order to sell their product they have to jolt the public. This is evident from the differences between what one takes away from story headings and story content. It is not confined to scientific articles either. The process has been dumbed down because that is what sells to the current public. There is also the cases of the biases of the writers getting involved. The journalist and/or from whomever they get their orders, I think, know very well what they are doing.

    If a scientist judges that the they have been misquoted or misinterpreted, the paper would normally be more than happy to afford them space for doing so and particularly so a competitor in the media. The split personality of a scientist/policy advocate may not be sufficiently disturbed to set an embellishment straight.

  71. Brooks Hurd
    Posted Dec 27, 2006 at 10:16 PM | Permalink

    Re: 63

    Judith

    one way to test whether such an adjustment would matter is to adjust all annual TC counts in the 19th century upward by 25%, and then do the correlations with SST.

    How confident can we be that historic SST data is significantly better quality than historic TC data? My concern is that the errors in these data are sufficiently large that a correlation would be meaningless.

  72. David Smith
    Posted Dec 27, 2006 at 10:19 PM | Permalink

    Judith, first I want to thank you for spending time here today. You help the conversation greatly.

    Let me touch on another data problem: here are Otto (2004) , Nicole (2004) and Gamma . The green portion of the tracks are where they were of minimal tropical storm (40 MPH) strength.

    (I’d also post the recent 10′th tropical storm of 2006, the one found while reviewing satellite photos, but Unisys has not posted it yet. It lasted all of 18 hours.)

    How dinky can a storm get? What are the chances of a ship encountering one like these in olden days and recognizing that it’s a “storm” and not a depression or active wave or squall and reporting it?

    These are small, weak storms, mostly at sea. I call them “junk storms” because they junk up the records and are of no significance. Yet they are part of the modern record, especially in the last 10 years. I’ve never catalogued them but my impression is that junk storms like these are not part of the pre-reconaissance record and aren’t common in the early satellite era.

    I’ve counted about two junk storms a year since the late 1990s. It’s difficult for me to understand their use in long-term data comparisons.

  73. Dane
    Posted Dec 27, 2006 at 10:29 PM | Permalink

    Post #53

    Bender,
    Quote “That’s it, keep it up, friend. Keep putting words in my mouth. Keep impugning motive where there is none. Shoot the messenger. You’re a gem.” I didn’t shoot the messenger. I read the blog daily, and know you do too. You didn’t address my question and assumed I put words in your mouth. Never assume, you know why. Address the question and settle the issue.

    Post #55,

    Quote “Err, because my faith is strong? Do I win? Keep it up. Don’t bother reading the blog. Best just to assume you’re right and I’m wrong.” Whats that supposed to mean? I read the blog, have done so for about a year. I know your style. You attacked me, not the other way around as you want others to believe.

    Post #57,
    Quote again, I asked if you dismissed models as a matter of principle. You haven’t answered that question. I did answer the question, did you even bother to read the post? Here it is again, Models are useful tools, but are often wrong, see post about GW1.

    Post #66,

    I already answered above.

  74. Willis Eschenbach
    Posted Dec 27, 2006 at 10:33 PM | Permalink

    Dear Dr. Curry:

    First, I applaud your willingness to explain, clarify, and defend your ideas on this site. It is very valuable, and I read your posts quite carefully.

    Next, some comments. You say:

    The number of NATL tropical cyclones (when filtered for el nino and other short term variability) does show a correlation with SST. Webster has done the calculation for 5 year running means, and the correlation between NATL SST and TC counts from 1910-2005 gives a correlation of 0.78 that is significant using the Mann Kendall statistic (this is a curry/webster paper that is in press, to be published around June 07).

    If Steve M, Bender, or Willis can provide any further insights into the statistical significance of this relationship, I would certainly be interested in pondering it and would certainly forward to Holland/Webster.

    As has been shown numerous times on this site, and as is widely known, many standard statistical tests (including the Mann-Kendall statistic) are worse than useless for auto-correlated datasets. I say worse than useless, because they show statistical significance where none exists.

    Even on white noise, a 5-year running average introduces strong autocorrelation, with a corresponding reduction in degrees of freedom. For a 5 year running average of white noise, N = 101 as in H/W, lag 1 corr = ~ 0.8, effective N = ~ 8.

    But using an average of already autocorrelated datasets such as SST is much worse. The 1910-2005 ACR region has inherent lag 1 autocorrelation = 0.56. The lag-1 autocorrelation of a 5-year running average of the ACR temperature is 0.92, with an effective N of only 2 … in other words, any statistical tests such as the Mann-Kendall statistic on such averages are useless.

    Second, you ask:

    With regards to the SST/TC correlations. There is a negative correlation associated with El Nino; this needs to be filtered out in some way to examine the correlation associated with the longer time scales. The optimum way to do this while still maintaining statistical significance has perhaps not yet been accomplished, have a go at it.

    I must admit to being totally astounded by this statement. The methods for adjusting for confounding variables are a well known sub-discipline of statistics, and are taught at the undergraduate level. Perhaps I have misunderstood your question?

    If not, the most obvious one is to establish some sort of statistical relationship (linear, power, log-linear, log-log, etc.) between the El Nino and the hurricane datasets. Use the Durbin-Watson statistic of the residuals to distinguish between possible models and select the best one. Then regress the El Nino data on the hurricane dataset using the selected relationship, and subtract out the regression from the hurricane dataset.

    w.

  75. Brooks Hurd
    Posted Dec 27, 2006 at 10:48 PM | Permalink

    Re: 70

    Ken

    I agree that “if it bleeds, it leads” is the motto used by some in the press. However, I beleive that you are being overly cynical to paint everyone in the press with the same broad brush. Personally I think that most people in the press want to present factual stories. The problem is how can they check their facts?

    With crime and civil action stories the press can always call the police, DA, or attornies to check their facts. It normally takes a quick phone call or two. Where is a reporter to turn to check facts in a science story?

    Scientific issues tend to be complex and require explanations from people who:
    1. Understand the issues and,
    2. Can explain complex issues to press people with almost no knowledge of science.

  76. J. Curry
    Posted Dec 27, 2006 at 11:12 PM | Permalink

    Willis, I have no involvement in the H/W paper. I am prepared to accept that it is possible that the statistical analysis in the paper was not optimal. I would be interested in seeing an alternative analysis, and assessing whether what H/W did adversely effects the conclusions drawn from the data set. I will be very surprised if there is no correlation between these two time series. I look forward to any analysis that you, bender, or steve m might do. I will consider your arguments and ask Carlos Hoyos for his opinion since I am not a statistician. I will make sure that I provide Holland with any useful input. The insights that I can provide are on the data and the physics.

  77. J. Curry
    Posted Dec 27, 2006 at 11:27 PM | Permalink

    In a recent NPR interview of Lindzen (i think i posted a reference to it on the Gray thread), he said something like “the data is what it is”. Sad, but true. Climate research is NOT like controlled laboratory experiments where you can take detailed measurements and carefully determine and constrain the accuracy and precision of these measurements. As a result, the science proceeds as an iterative process with analysis of imperfect data, theory, and use of imperfect models. Re the historical SST dataset, Kaplan describes the issues here http://www.cgd.ucar.edu/cas/guide/Data/kaplan_sst.html. Its hard to tell what kind of error bars to assign to the historical SST data. So there is a tradeoff: if you want really accurate data, stick with the data since 1944. If you need more data for statistical significance, then the real significance issue becomes a convolution of data quality and sample size. So i would appreciate any input on the optimal selection of the time series for analysis.

  78. J. Curry
    Posted Dec 27, 2006 at 11:35 PM | Permalink

    Brooks and Ken, the whole issue with the media makes me want to tear my hair out. Re the hurricane and global warming story, one science journalist has taken it on himself to delve deeply into this, and has written a book (due to be published summer 07) called “storm world” , here is the book description from amazon:

    One of the leading science journalists and commentators working today, Chris Mooney delves into a red-hot debate in meteorology: whether the increasing ferocity of hurricanes is connected to global warming. In the wake of Katrina, Mooney follows the careers of leading scientists on either side of the argument through the 2006 hurricane season, tracing how the media, special interests, politics, and the weather itself have skewed and amplified what was already a fraught scientific debate. As Mooney puts it: “Scientists, like hurricanes, do extraordinary things at high wind speeds.” Mooney”¢’‚¬?a native of New Orleans”¢’‚¬?has written a fascinating and urgently compelling book that calls into question the great inconvenient truth of our day: Are we responsible for making hurricanes even bigger monsters than they already are?

    I have no idea what to expect (although the statement about extraordinary things at high wind speeds makes me a bit nervous). But with all the work that Mooney, a very serious science journalist, put into this, this has to be the best shot at making sense of the implications of this issue. We’ll see.

  79. Dane
    Posted Dec 28, 2006 at 12:06 AM | Permalink

    Dr. Curry,

    I am still wondering what your thoughts are on the goeologic data from ice cores and the reliability of other geologic data that doesn’t support the AGW theory?

    Also, does it bother you that minus the models, most of the AGW debate falls apart?

    You seem to be an authority in your field, as my Proff for my MS was in his, your thoughts will be considered as such. Please elaborate.

  80. bender
    Posted Dec 28, 2006 at 12:19 AM | Permalink

    Re #73
    This thread is about the HW manuscript. Let’s try to focus on that. Recapitulations are silly, as the course of discussion is a matter of record.

    Re #76

    I will be very surprised if there is no correlation between these two time series.

    It’s not the magnitude of the correlation coefficient that is the real concern, but its significance level. A correlation can be high but non-significant if there are not many degrees of freedom available for the hypothesis test. AS Wilis points out, this appears to be the case. A more precise analysis will follow, I’m sure.

    If I may, Dr. Curry … I think this is the complaint people have about climate science: it’s not about a mere lack of statistical rigor. It’s more fundamental than that. It’s about a perceived generalized failure to be self-critial, in the Feynmanian sense of how science ought best to proceed.

    It’s the drive toward self-criticism that should motivate the sceintist to seek robust statistics, not what a reviewer or an Editor demands. Some get the sense that this is lacking, and frankly I’m inclined to agree. I think there is a reason climatologists tend to hold the statisticians off at arms length: they don’t want to hear the bad news. I’m prepared to change my view on this, but it would take a pretty solid rebuttal.

  81. bender
    Posted Dec 28, 2006 at 12:28 AM | Permalink

    Re #76

    I will consider your arguments and ask Carlos Hoyos for his opinion since I am not a statistician.

    Dr Curry, you don’t need to be a statistican to understand the damning argument Willis presents in #74:

    The 1910-2005 ACR region has inherent lag 1 autocorrelation = 0.56. The lag-1 autocorrelation of a 5-year running average of the ACR temperature is 0.92, with an effective N of only 2

    How on earth can one expect a significant result with only two independent observations? I expect Dr. Hoyos will basically agree. Though he may have some interesting supplementary commentary.

  82. bender
    Posted Dec 28, 2006 at 12:36 AM | Permalink

    Re #77

    Sad, but true. Climate research is NOT like controlled laboratory experiments where you can take detailed measurements and carefully determine and constrain the accuracy and precision of these measurements.

    And that is precisely why cautious interpretation is necessary … if you want to maintain scientific credibility.

    [I am not a climatologist, but I am sympathetic, because the very same problems plague the field of ecology. I wish it were not so difficult to determine the probability of species extinctions under climate change. Alas, it is nigh impossible except for the very, very best-studied systems. (Polar bears not being one of them.)]

  83. Posted Dec 28, 2006 at 12:36 AM | Permalink

    Re: #78

    Dr. Curry wrote:

    I have no idea what to expect (although the statement about extraordinary things at high wind speeds makes me a bit nervous). But with all the work that Mooney, a very serious science journalist, put into this, this has to be the best shot at making sense of the implications of this issue. We’ll see.

    This might be a clue as to what we can expect, given this recent book by Mooney: The Republican War on Science

    My guess is we will see another attack on the current administration, laying the blame for more and stronger hurricanes.

    This “very serious science journalists” has written with an agenda in the past, using half truths as evidence. We will have to see if he has changed his stripes.

  84. bender
    Posted Dec 28, 2006 at 12:44 AM | Permalink

    Re #83
    This is your answer as to why Gore gets a standing O at AGU. Not because the room was full of believers, but because scientists are generally tired of being marginalized by an elected leadership that doesn’t seem to want to listen.

  85. fFreddy
    Posted Dec 28, 2006 at 2:33 AM | Permalink

    Re #84, bender

    …scientists are generally tired of being marginalized by an elected leadership that doesn’t seem to want to listen.

    Is this referring to the GW enthusiasts ? Or something broader – and, if so, what ?

  86. Posted Dec 28, 2006 at 6:19 AM | Permalink

    Judith Curry should be commended for being involved in the climate debate – far too few working climate scientists are involved in communicating their informed views to the public and wider scientific community. That said, CO2 driven warming/cooling is not the only theory available to explain climate change accross all time scales, including the past few decades. Published/working scientist Nir Shaviv, not funded by big oil or big tofu, gives an objective overview here:

    http://www.sciencebits.com/CO2orSolar

    Climate sensitivity:

    http://www.sciencebits.com/OnClimateSensitivity

    http://www.phys.huji.ac.il/~shaviv/articles/2004JA010866.pdf

    http://www.sciencebits.com/IPCCbias

    Much published/working climate scientist Roger Pielke Sr is a good source for papers on land use change being at least as big a human climate influence as CO2, and poor siting of temperature stations which contribute to the flawed ‘global average surface temperature’ metric.

  87. welikerocks
    Posted Dec 28, 2006 at 7:21 AM | Permalink

    re; 85
    Yeah, I’d like to know the answer to that one as well.

    elected leadership that doesn’t seem to want to listen.

    because that sounds like complete BS to me. Now you know what elected leadership doesn’t want to listen to? Is there a computer model for this too or did you read that in the newspaper? Sheesh. The elected leadership in my state is in control of this issue from the governor’s office to the EPA to the classroom. In my experience, those who cry foul so often are the folks really doing the fouling! GW is real, haven’t you heard?

    Everything negative and wrong I’ve come to understand about this issue is all represented right here in this thread. I suppose the elected officials who ban french fries are more open to scientific knowledge then?

    Judith Curry should be commended for being involved in the climate debate

    Yes that is true and I do thank her. But I’d appreciate being commended for not being a starry eyed little guy who doesn’t question the elite media and scientists who adore the guilt fest and politics of this issue just as much as the science, and who give a speech like that given by a politician like Al Gore a standing O at the AGU. And to use Bender’s observational skills I’d say: They seem to insist they are smarter and know so much more about the world, my mind, what I need, and what is best for me and my family then I do, and they do not honor my vote on any issue because they believe this about “me” with all their hearts-all based on a computer model of the climate.

    SteveM, I apologize for my offshoot of a rant here. But I do not believe these new papers are about real science anymore, from Mann to Juckes to this one. I hope it gets back to it- and goes more on topic from Willis’ post here.

  88. TAC
    Posted Dec 28, 2006 at 7:59 AM | Permalink

    Judith, I would be interested in hearing Hoyos’s response to Willis’s comment (#74). What concerns me here, and elsewhere, is that the statistical methods often employed in climate science seem inappropriate given the widely recognized characteristics of climate processes. For example, no one argues with your statement (#45) that “climate varies on many time scales, associated with external forcing and natural internal variability,” yet you glance at H&W and see use of a statistical test — which is a proof by contradiction — assuming that climate data (adjusted in peculiar ways) correspond to an AR(1) process. Needless to say, statistical significance is meaningless if the null hypothesis fails to represent the “untreated” system.

    To be absolutely clear, here’s a silly example: Let H0 be: “All cars are blue and there’s no trend in hurricanes.” Now we go out and observe a red car, which means we reject H0 and conclude ( ;-) ) that there is a trend in hurricanes. QED.

    Is this even remotely relevant? I think it is. Modify H0 by replacing “All cars are blue” with an equally bogus statement like “Natural variability is white noise (or even AR(1)).” QED.

    Does it matter? As I noted in #27 and #32, the type 1 error rate — the true probability of incorrectly concluding that there is a significant trend based on a nominal 5% test — can easily go above 50 percent.

  89. David Smith
    Posted Dec 28, 2006 at 8:20 AM | Permalink

    Re #64

    Ken, you raise an interesting question. In addition to wondering about the choice of SST “box” and duration for HW Figure 1, I also wonder about the impact of the 9-year running average.

    Here are some possible SST boxes:

    Here is the SST box used by HW (5-25N; 55-20W). It looks reasonably like the plot in Figure 1.

    Here is the SST box you mentioned from Goldberg.

    Here is Emanuel’s box.

    I used July/Aug/Sept in all three.

    To my eyes, HW’s box shows a slight decline until the mid-70s, when an uptrend began.

    Emanuel’s box shows Atlantic tropical SST in a slow decline until about 1995, then an uptrend.

    Goldberg’s box shows SST in a slow decline until the mid-1990s, when temperatures returned to 1950s level, then a spike in the last 5 years. Alternately, Goldberg’s box might show a step-drop circa 1970 with a step-up circa 1995. It all depends on how one squints.

    As you probably suspected, the choice of boxes affects the SST pattern and presumably any correlation with storm count. However, the more smoothing I use, the more the plots of SST resemble one another.

    One consequence of using the Goldberg or Emanuel box is that the uniqueness of the SST rise in the last several decades fades. In Goldberg and Emanuel, the circa 2000 SST looks about the same as circa 1960. Note that circa 1940 was probably higher than circa 1960.

    The paper states that, “The eastern Atlantic region includes both the main development region defined by goldberg and Shapiro and the area of tropical SST fluctuations associated with the Atlantic multi-decadal oscillation”, so i suppose that the rationale for the expanded box is as part of their search for an AMO signal. However, my recollection is that the clearest AMO SST signature is farther north, closer to Spain, which is 35 to 40N. So, I’m not clear on their objective.

    Maybe Willis or someone else can plot the pre-1950 SST data for the various boxes and see how they compare with each other and with storm count.

  90. J. Curry
    Posted Dec 28, 2006 at 8:42 AM | Permalink

    Willis, Bender, TAC,

    I absolutely agree that rigor in statistical analysis is desired. There are two factors that contribute to a perceived lack of rigor in statistical analysis used by climate scientists. The first one (which you will accept as obvious) is that many climate researchers have at best rudimentary training in statistical analysis. The second one (which you need to ponder) is that statistical analysis alone (and obsessing about the details of confidence intervals) doesn’t get our field very far. It is the combination of statistics and physics that move the field forward. When we a have a physical reason for expecting a relationship to be significant (theories and laws, even), then details of statistical significance are less relevant. Simply because we have a data record of insufficient length or accuracy for statistical significance, this does not mean that we should not proceed forward with exploring or using the relationship. Then there is an interplay between the two factors, where climate scientists don’t take the issue of statistical significance as seriously as they should in situations where it is truly important.

    For the issue at hand, e.g. the correlation of SST and NATL TC counts, we don’t have a physical explanation and the issue of statistical significance is relatively important. Hence, I look forward to the results of your analysis. Even if the correlation turns out not to be statistically significant, scientists will use the relationship to further explore the scientific issues and to test to see if the relationship has any predictive capability.

  91. J. Curry
    Posted Dec 28, 2006 at 8:52 AM | Permalink

    Not to belabor the point, but the so called republican war on science is an important issue. If you haven’t read mooney’s book, it is really about the govt war on science which is largely a bipartisan issue. Mooney argues that the last good science president was JFK. Since then, Bush 41 was a relatively good science president (Bush 43 is the worst by a long shot). It is a serious piece of scholarship that has garnered him numerous awards and nominations for even bigger awards. (it remains to be seen what he does with the hurricane book). But the relevant issue that needs to be considered is the extent to which this war on science has provoked the behavior of some scientists that you object to. I believe that this is a real issue. Back to statistics.

  92. welikerocks
    Posted Dec 28, 2006 at 8:58 AM | Permalink

    There are more papers on Atalntic hurricanes listed here on a page of the AGU website.

    First on the list is a paper by a Holland also, but I don’t think it’s the same person.

    “A decomposition of the North Atlantic hurricane energy dissipation index and the mechanisms contributing to its increase since 1970″

    * Holland, C L (cholland@utig.ig.utexas.edu) , University of Texas, Institute for Geophysics, 4412 Spicewood Springs Rd, #600, Austin, TX 78759, United States
    Scott, R B (rscott@utig.ig.utexas.edu) , University of Texas, Institute for Geophysics, 4412 Spicewood Springs Rd, #600, Austin, TX 78759, United States

  93. welikerocks
    Posted Dec 28, 2006 at 9:08 AM | Permalink

    J Curry, I personally wrote an email to my local election office before GWB was even elected and asked them to please look into this Hockey Stick business carefully because I was concerned about junk science. And I am an independent voter. I believe the “War on Science” and more propaganda just like your comment about geologists.

  94. J. Curry
    Posted Dec 28, 2006 at 9:17 AM | Permalink

    David and Ken, the issue of the SST averaging region is an important one. Peter Webster made a presentation at AGU (I think Steve M mentioned it somewhere) about the changing character of the tropical warm pools (not just temperature increase but area increase/change). It may also be that the MDR region changes with time. For the longer time periods, given the low resolution of the observations, it may be best to look at the SST of the entire tropical atlantic.

  95. TAC
    Posted Dec 28, 2006 at 9:40 AM | Permalink

    Judith,
    There is much to mull over in your comment:

    But the relevant issue that needs to be considered is the extent to which this war on science has provoked the behavior of some scientists that you object to. I believe that this is a real issue.

    I accept your explanation that some scientists have felt “provoked” by a perceived government “war on science.” But does this justify a scientist exaggerating claims or publishing invalid results? I hope we can agree that the answer is no.

    I also hope we can agree that climate processes are complex, that it is not a trivial task to identify and apply appropriate statistical methods, and that, for now, uncertainty is a large and important part of the climate story. This last point is obvious. As you and I observed at AGU, there are literally thousands of scientists presenting new and surprising findings every year. Under the circumstances, it seems counter-productive, for both science and society, not to represent the situation as accurately as possible.

    However: “Back to statistics” sounds good to me!

  96. Posted Dec 28, 2006 at 9:44 AM | Permalink

    #90

    The second one (which you need to ponder) is that statistical analysis alone (and obsessing about the details of confidence intervals) doesn’t get our field very far.

    You are right, need to ponder this. Specially that obsessing about the details of confidence intervals part. Something related to MBH99? We won’t get far in climate science if we do it right? How far the field must go?

  97. John Norris
    Posted Dec 28, 2006 at 9:57 AM | Permalink

    re #90
    ” … It is the combination of statistics and physics that move the field forward. …”

    “… Simply because we have a data record of insufficient length or accuracy for statistical significance, this does not mean that we should not proceed forward with exploring or using the relationship. …”

    No problem with that here, just don’t expect skeptics/deniers/gadflies/big oil/war on science mongers/etc … to buy the results. HW stated that there is more work to do to reach more conclusions, but apparently they don’t think it is of insufficient length to reach the conclusions that they did.

  98. Jeff Weffer
    Posted Dec 28, 2006 at 10:05 AM | Permalink

    Judith Curry brought up the negative correlation with the ENSO.

    I think if a person looks closely into this, you will find a strong correlation of the ENSO with TSs in the Atlantic which overrides the Atlantic SSTs. You might also find that average global temperature changes are closely correlated with the ENSO as well. You might also find that the ENSO shows no correlation whatsoever with CO2 levels in the atmosphere.

    1997-98, near-record El Nino, high TS counts, highest global temperature.

    2005, weak La Nina, record TS counts, lower global temperature (taking into account the slanting of the data by Hansen).

    2006, stronger El Nino (still going strong), near record low TSs, lower global temperature (World Meteorological Association beats Hansen to the punch this year and declares 2006 as the sixth highest year – Hansen would have made it the highest if he would have got to the media sooner.)

    The problem is that the ENSO varies throughout the year while the Atlantic TSs are limited to a specific season which makes the statistical analysis complicated. Since most climate researchers are concerned with proving/slanting the AGW case and are not particularly good statisticians, they have botched up the link.

    http://www.cdc.noaa.gov/people/klaus.wolter/MEI/

    http://en.wikipedia.org/wiki/Image:NOAA_ACE_index_1950-2004_RGB.svg

  99. David Smith
    Posted Dec 28, 2006 at 10:18 AM | Permalink

    Re #94 Agreed. If one is comparing Atlantic SST versus Atlantic storm count, then one should use Atlantic tropical SST, not a subsection, unless there’s a compelling reason.

    The plot for my definition of Tropical Atlantic (25 to 10N, 90 to 20W) is given here .

    Using the GISS plotter , it looks to me like 1990-2000 SST in the tropical Atlantic was slightly less than 1935-1945 SST. Adding in the sharp rise in SST after 2001 makes the recent decadal SST warmer than 1935-1945, but that increase happened after the step-up in storm count.

    Judith, one of the things I’ve wondered is whether the regime changes, like 1995, are due to shifts in the summer ITCZ and the path of tropical waves (700mb level), bringing the two into closer interaction, so that the wave benefits from the moisture/heat content of the ITCZ. It’s an indirect effect of higher NATL SST, which “draws” the ITCZ northward. Has that thought been kicked around anywhere? Thanks.

  100. bender
    Posted Dec 28, 2006 at 10:32 AM | Permalink

    Rocks et al, the “war on science” is real, although I agree the label is hyperbolous. Fact is a large number of government science programs were terminated prematurely and/or downsized in order to pay for the “war on terror”. The negative correlation is direct, and that’s where the hyperbolous phrase comes from. Ask any scientist about their grant writing experiences in the 2001-2002 period; they will tell you the same story. And since then, look at the (relative) stagnation in American universities, compared to the demographic renewal going in in every other developed country in the world. Broad brush at the national scale, and every thing looks fine … on the surface. But go talk to the profs on campus and they will generally agree with Dr Curry: the universities have been neglected and are overdue. It’s not just AGW climate scientists we’re talking about, but all departments, all parts of the spectrum. It’s sad. We used to be the best of the best. But it’s been slipping away.

    Back to stats …

  101. fFreddy
    Posted Dec 28, 2006 at 10:37 AM | Permalink

    Re #91, J. Curry

    the so called republican war on science is an important issue … I believe that this is a real issue.

    OK, so what is your experience of GWB’s war on science ?

  102. fFreddy
    Posted Dec 28, 2006 at 10:38 AM | Permalink

    Re #91, J. Curry
    And please don’t refer me to this Mooney fellow. His reaction to the Wegman report concluded :

    And make no mistake: Despite the new fireworks, that big picture remains unaltered. Whether the “hockey stick” is right, wrong, or irrelevant, the underlying message on global warming is that we’re causing it. Period. End of story.

    So he either doesn’t understand the issues, or else doesn’t care. So I doubt that he is worth taking seriously.

  103. fFreddy
    Posted Dec 28, 2006 at 10:49 AM | Permalink

    Re #100, bender
    This is the first time I have seen this “war on science” being referred to in the context of a general decrease in funding for science at universities. Every other time I have come across it, it has been in the context of preaching hatred for GWB on the subjects of creationism, embryonic stem cells, global warming, etc. Particularly from that Mooney fellow.
    Given what we have learned here about the global warming stuff, I have been curious if there was anything to any of the other stuff – so thank you for a frothless response.
    Regarding “the demographic renewal going in in every other developed country in the world” – not here in the UK, I’m afraid…

  104. jae
    Posted Dec 28, 2006 at 11:00 AM | Permalink

    59: Dr. Curry, I agree with you, in general. I guess I was reacting to some of the plans for high taxes on fossil fuels, which I disagree with. That’s a ploy to make alternate forms of energy SEEM to be economically feasible. Such taxes are inflationary and can wreak havoc on the economy. And they put much more burden on the poor than on the wealthy. And I doubt that they would not have any measureable effect on CO2 levels, especially if China and India did not participate. And what will all this tax money be used for? I’m with Bjorn Lomborg’s views in the Skeptical Environmentalist.

  105. John Norris
    Posted Dec 28, 2006 at 11:12 AM | Permalink

    Starving, neglected scientists?

    HW gives funding credit to NSF. Annual NSF funding level looks pretty consistent back to 1997.

    http://dellweb.bfa.nsf.gov/awdfr3/default.asp

    There is a substantial increase in the number proposals going after that government funding though …

  106. EP
    Posted Dec 28, 2006 at 11:30 AM | Permalink

    There are two wars on terror: the one against Islamists and the second against spurious scientific results that try to alarm people.

    The US seems quite capable of winning Noble prizes in science; contrast with the UK’s decline. And this is a nation whose main parties are vying for the crown of “Greenest of all.”

  107. welikerocks
    Posted Dec 28, 2006 at 11:32 AM | Permalink

    Fact is a large number of government science programs were terminated prematurely and/or downsized in order to pay for the “war on terror”.

    Fair enough opinion Bender, but some of us think that might be just a sacrifice at the moment and not a war on science. Some of us understand sacrafices and some of us have experienced abuse by acadamians for serving in the military. The disdain for this President I do not share at all, and I don’t agree with him on everything, and that so this kind of comment needs some specifics for my understanding of how “bad” it is at universities. I am also not so far removed from acadamia and government programs to know alot of money pumped in from these taxes I pay, that go to universtities for things I may feel are stupid or not worthy scientifically and may only benefit small groups and trendy ideas. On the other hand too, universities are payed enormous amounts of money by ‘regular” folks to educate their children so they can go out into the real ugly messy world and function. Did the tuition fees get so high because of the war on science or a president? I don’t think so but I go off on another tangent here…

  108. Posted Dec 28, 2006 at 11:41 AM | Permalink

    Dr. Curry is quite game in holding up her end of the discussion on climate modeling and the physics of climate. She risks losing all credibility when she wanders onto the field of political debate and Bush-bashing.

    Please stick to the topic at hand, Dr. Curry, and maintain the respect you deserve.

  109. welikerocks
    Posted Dec 28, 2006 at 11:41 AM | Permalink

    I also believe I read just recently in my local paper the USC employs more people then any other entity or company in the city of Los Angeles.

  110. EP
    Posted Dec 28, 2006 at 11:42 AM | Permalink

    Regardless of scientists’ political views on Bush, funding or the colour of Janet Jackson’s nipple where does this paper leave the debate on AGW? Do relatively short term trends in warming an cyclone frequency mean we should change our economic policies?

  111. bender
    Posted Dec 28, 2006 at 11:51 AM | Permalink

    It is interesting how people at risk of losing a debate will often try to ridicule their opponent by exaggerating the extremeness of their view. Thus you have Dane in #42 arguing that the earth is not a greenhouse, now John Norris in #104 talking about starving, neglected scientists. What other ridiculous statements do you wish you could attribute to a real person?

    Annual NSF funding level looks pretty consistent back to 1997.

    Go ahead and cherry-pick your statistics to promote your POV (NSF is only one program); I will not argue with you. University scientists have it good, no doubt. My point is that competition in the university system for seemingly limited dollars is unprecedented cut-throat, which is one of the reasons why the race to connect the AGW dots is so intense. Disagree with me if you like. Trot out any stats you like, I’m telling you about people’s perceptions on campus. I’m telling you why Gore gets an ovation, and why books like Mooney’s sell. Would you care to argue with those facts?

    There is a substantial increase in the number proposals going after that government funding though

    And that’s what competition is, not the number of dollars available but the number of dollars per applicant. In your funding trend analysis please be sure to factor in (1) the increasing cost of science and (2) the tendency toward BIG programs (e.g. FACE). There may be lots for the big players on big teams; but there’s not a lot for the little players without a team.

  112. bender
    Posted Dec 28, 2006 at 11:53 AM | Permalink

    Re #108 And does USC not create great technologies that fuel the capitalist enterprise?

  113. Brooks Hurd
    Posted Dec 28, 2006 at 11:58 AM | Permalink

    Re: 90

    Judith

    The second one (which you need to ponder) is that statistical analysis alone (and obsessing about the details of confidence intervals) doesn’t get our field very far. It is the combination of statistics and physics that move the field forward.

    In my opinion, the way that the climate science community can avoid this pitfall is to keep the statistics as simple as possible. Anytime that an author uses an arcane statistical procedure to support a paper’s conclusions, the author owes his/her readers a well thought out explanation as to why the statistical treatment was selected. Furthermore, an author who selects an arcane statistical method should explain their methodology in sufficient detail such that it is made clear to most readers what was done and more importantly why it was done.

    When an author uses an arcane statistical method without providing clear explanations of why and how, the author is inviting after the fact oversite by professional statisticians, obssessing over confidence intervals and other details. Why should an author expect wholesale acceptance of an arcane and unexplained methodology? Many of the posters on this blog are asked by our customers to justify why we did things and to explain the methodologies that we use. I can not speak for other posters, but this is a daily occurance for me. This is why many of us find it difficult to understand why some authors will not answer the sort of questions with which we deal every day.

  114. J. Curry
    Posted Dec 28, 2006 at 12:08 PM | Permalink

    Looks like I unleashed a monster with the war on science thing.

    One comment on a statement by TAC:

    “I accept your explanation that some scientists have felt “provoked” by a perceived government “war on science.” But does this justify a scientist exaggerating claims or publishing invalid results? I hope we can agree that the answer is no.”

    The answer of course is no. But when ideologues take over the skeptical arguments, using the strategies of the defense lawyer out to defend a guilty client, then an unfortunate chain of events (unfortunate for the science) may take place. Although I suspect that I will regret throwing this idea out there, I think that hockey stick war was inflated by ideologues, provoking a circle the wagon response by the scientists and necessitating unusual strategies by M&M (I regard Steve M as a bonafide scientific skeptic and not an ideologue). I.e. the hockey stick war is almost a predictable result of the war on science (I wonder what Chris Mooney would do with this idea). And I state this as someone who has no stake in the hockey stick wars either way and as someone who accepts the the assessment of the NRC committee.

    So this is the real danger of the war on science. The government should not suppress or alter scientific research (if you want some examples from the climate field, see Rick Piltz’s site http://www.climatesciencewatch.org/
    If the government wants to devalue or ignore the science (after considering the various assessments and uncertainties) in their policy making (i.e. for economic, political, or whatever reasons), just say so (and preferably why), without trying to distort or suppress the science to justify ignoring it. Distorting and suppressing the science has some very serious backlash effects on the scientific process.

  115. Ken Fritsch
    Posted Dec 28, 2006 at 12:19 PM | Permalink

    Re: #91

    From a Mooney interview linked below, I see, from my libertarian prospective, a political screed against the Republicans for what to me is simply a general and predictable problem with government getting into science. It is like the attempts to make the Katrina disaster and aftermath a Republican failure that I see as a government failure at all levels. Science with government involvement will become politicized from both Republicans and Democrats.

    The excerpt below from Mooney in the interview is most puzzling to me. I see quite the opposite trend in the media in that it does not present all the facts and let the reader/listener decide for themselves the merits of any arguments. I do not want my news predigested by some science writer or writers in general. The media I believe delivers to a mass audience pretty much what that audience wants. Most of that audience is simply not all that interested in a subject that they see having little or no day to day affect upon them. The truly interested reader/listener will seek out many avenues of information in an effort to find the truth and that person knows what that process is.

    http://www.grist.org/advice/books/2005/09/27/mooney

    The press doesn’t generally help these matters. This is an argument I have made twice now in Columbia Journalism Review. Through their instinctive tendency to create a “balance” between two sides, journalists repeatedly allow science abusers to create phony “controversies,” even though the scientific merits of the issue may exclusively be with one side.

  116. J. Curry
    Posted Dec 28, 2006 at 12:23 PM | Permalink

    EP, here is our answer as to what we we should do about the North Atlantic hurricane threat. You can link to the entire statement at Kerry Emanuel’s web page http://wind.mit.edu/~emanuel/Hurricane_threat.htm. This statement (discussed in the New York Times), and involved 10 scientists from both sides of the debate (including Landsea, Mayfield, Emanuel, Webster, Holland).

    Here is the text of the statement:

    As the Atlantic hurricane season gets underway, the possible influence of climate change on hurricane activity is receiving renewed attention. While the debate on this issue is of considerable scientific and societal interest and concern, it should in no event detract from the main hurricane problem facing the United States: the ever-growing concentration of population and wealth in vulnerable coastal regions. These demographic trends are setting us up for rapidly increasing human and economic losses from hurricane disasters, especially in this era of heightened activity. Scores of scientists and engineers had warned of the threat to New Orleans long before climate change was seriously considered, and a Katrina-like storm or worse was (and is) inevitable even in a stable climate.

    Rapidly escalating hurricane damage in recent decades owes much to government policies that serve to subsidize risk. State regulation of insurance is captive to political pressures that hold down premiums in risky coastal areas at the expense of higher premiums in less risky places. Federal flood insurance programs likewise undercharge property owners in vulnerable areas. Federal disaster policies, while providing obvious humanitarian benefits, also serve to promote risky behavior in the long run.

    We are optimistic that continued research will eventually resolve much of the current controversy over the effect of climate change on hurricanes. But the more urgent problem of our lemming-like march to the sea requires immediate and sustained attention. We call upon leaders of government and industry to undertake a comprehensive evaluation of building practices, and insurance, land use, and disaster relief policies that currently serve to promote an ever-increasing vulnerability to hurricanes.

  117. jae
    Posted Dec 28, 2006 at 12:31 PM | Permalink

    144: Didn’t you post this before? It is right on target, IMO.

  118. Ken Fritsch
    Posted Dec 28, 2006 at 12:35 PM | Permalink

    Re: #114

    I had already copied that statement to an earlier CA comment –as I thought it was excellent advice. At the time I commented that

    Federal disaster policies, while providing obvious humanitarian benefits, also serve to promote risky behavior in the long run.

    was a pleasantly surprising statement to me as it goes to the nature of unintended consequences of the “good intentions” of government.

  119. Posted Dec 28, 2006 at 12:37 PM | Permalink

    Re David Smith (99)

    I rarely visit blogs but I am intrigued to see that our paper has drawn comments as far ranging as President Bush’s ability as an adminstrator to questions of statistics and the war on science. What next I wonder? Perhaps I should get out of science and become an agitator. I have to be honest and say that these were unintended consequences of this pre-publication. But in the spirit of give and take I would like to comment on Mr. Smith’s version of science. In fairness, it was posed as a question to Judy Curry but I think the following might help.

    In (99) he considers the ITCZ and African waves as separate entities thinking that perhaps the 2006 season was anomalous because one was closer to the other at that time. In other words, the 2006 season was the linear addition of two separate phenomena. Alas, the ITCZ is a statistical artifact and is made up from a combination of waves propagating through a particular zone. If one were to look at a satellite picture during the summer on any one day, one sees blobs of convection moving westward separated by a 1000-2000 km. Viewed the next day, the blobs will have continued to move westwards at roughly 6-10 m/s and some will be stronger, some will disappear and occasionally some will become tropical storms. Three questions come from this: What forms the wave and why do they move eastward; why do some dissipate and why do some become stronger?

    An attempt to answer the first question (why the waves) is given in a paper (Webster and Tomas 1997: and Tomas et al 2000) available and downloadable on my website (http://webster.eas.gatech.edu, downloadable papers). Essentially, the waves (and their statistical mean: the ITCZ) arises form instabilities that has a strong cross-equatorial pressure gradient. Out of this same instability comes waves at the intertial frequency of roughly 3-4 days: hence the waves, the time scale and the location of the ITCZ. The mean position of the latter is set by the large scale and slowly varying cross-equatorial pressure gradient which determines the location of the waves which (in the mean) appears as the ITCZ. If you want to know why cross-equatorial flow is unstable: read the papers above.

    But perhaps more interesting to you is the second set of questions: why do some waves (so produced) become stronger and others weaker. Enter, I think, the background flow. First though, what determines the background flow and why should it persist for such a long period effecting the entire season. Slowly varying flow is produced by slowly varying boundary conditions such as the SST and the integrated heat content of the ocean. It is not just the magnitude of the SST (although that is important because of Clausius-Clapeyron saturated vapor pressure considerations) but also the distribution of SST. Considering the latter for the moment, the distribution of SST changes the structure of the atmosphere allowing highly sheared waves to penetrate into the tropics. These act to limit convective growth and waves dissipate. At the same time, the gradient of SST along the equator causes changes in the east-west wind along the equator. Now in regions where the gradient is negative (dU/dx

  120. jae
    Posted Dec 28, 2006 at 12:37 PM | Permalink

    90: Dr Curry said:

    Simply because we have a data record of insufficient length or accuracy for statistical significance, this does not mean that we should not proceed forward with exploring or using the relationship.

    Of course not, but you certainly should not be misleading people by publishing definite conclusions with this information. The weaknesses should be clearly stated.

  121. J. Curry
    Posted Dec 28, 2006 at 12:39 PM | Permalink

    Brooks, re 112 I agree with your statement, lets keep the statistics “simple” unless the problem demands something more complex in which case the methodology should be clearly outlined and justified. However there is no standard definition of what is standard vs arcane. In the case of Holland’s selection of Davis (1976), this could easily be explained by the fact that he was educated in the Australian system (standard methodologies may not be considered as standard across various fields and geographic regions) and he may have gotten used to using this technique. I doubt this was selected to confound people from figuring out what he actually did

  122. Paul Penrose
    Posted Dec 28, 2006 at 12:45 PM | Permalink

    Dr. Curry,
    First let me thank you for the frank and continued discussion on this topic. I really appreciate you spending your limited and valuable time helping us to understand these issues.

    In message #90 you seem to be saying that if you have a really good physical understanding of something, that you can basically ignore failing stastical scores when analyzing the data and vice versa. Please correct me if I’ve missinterpreted what you’ve said. I, however, see these two items, the physical understanding and the data analysis, as two sides of the same coin. If you think you have good understanding of the underlying physical processes but the correlations in your data are not stastically significant, then I think you have to question whether you really have a good understanding of the physics. The converse is also true in that good correlation could be spurious and uncertainty is high if you don’t understand the underlying physics. In either case I don’t think a good scientist should be coming to any conclusions, let alone making them public.

    Only when both are in general agreement should any conclusions be drawn. Obsession in only one area or the other is counter-productive, and if one can’t do justice to both on one’s own, then a colaboration should be sought with an expert in the other area. Failure to do so is, in my opinion, one of the major weaknesses of climatology today.

  123. jae
    Posted Dec 28, 2006 at 12:58 PM | Permalink

    Paul: that’s why I line the solar theories so well: the physics and the correlations match.

  124. John Norris
    Posted Dec 28, 2006 at 12:58 PM | Permalink

    re #110

    Point taken, you used the term “neglect”, I colored it further with “starving”, my poor attempt at sarcasm.

    By the way I agree with you 100% on your standing O observation. They feel they are getting ignored and shorted on funding. I have a different opinion.

    Also by the way, I did not cherry pick my data, I just quickly looked for a science funding stat most relevant to this HW paper thread. NSF apparently funded the HW paper, see page 24. If I read it correctly NSF annually funds 10,000 awards with a median average of $100,000. Most I suspect not towards climate science, but obviously some. As you said NSF is just one program, but obviously not a small program. Do you think the US taxpayer is getting their moneys worth out of funding Mann and this HW paper?

  125. Posted Dec 28, 2006 at 1:03 PM | Permalink

    When we a have a physical reason for expecting a relationship to be significant (theories and laws, even), then details of statistical significance are less relevant. Simply because we have a data record of insufficient length or accuracy for statistical significance, this does not mean that we should not proceed forward with exploring or using the relationship. Then there is an interplay between the two factors, where climate scientists dont take the issue of statistical significance as seriously as they should in situations where it is truly important.

    This is simply backwards.

    If the physical relationship berween to factors is described by physical law, then the statistical significance should be very high. If its not, then either the physical law is misunderstood or more simply, the relationship between the two factors has been misapplied.

    For example, a key assumption of global warming theory is that there should be a relationship between rising temperature and the width of tree rings. Now that’s an assumption that is taken to be axiomatic as physical law. So when statistical evidence shows little significance between temperature and the width of tree rings, its ignored.

    Or when a statistically significant relationship between tree ring growth and global temperature in bristlecone pines in the Western US is described, there is no account taken that the BPs do not have a statisitically significant relationship with local temperatures. Instead the concept of “teleconnections” is blatently abused to suggest that the original relationship is not spurious.

    I do not follow your claim that if a physical law relates two factors then statistical significance is less important. As in the rest of this, the claim should be backed up with evidence.

  126. J Edwards
    Posted Dec 28, 2006 at 1:16 PM | Permalink

    Re: #110

    Excellent observations bender. Another point to note is the declining “share” of R&D under government funding. Like Ken, my libertarian tendencies are to keep government out of most things, but I realize that is a double-edged sword in some areas of R&D, and might be perceived to some as being antithetical toward science, as opposed to merely ambivalent. The IntegrityofScience website that Dr Curry linked to, has a presentation by Dr Gleick which includes a slide (#20) that shows the declining market share of Government-funded R&D compared to the waxing share represent by corporate R&D (URL: http://www.pacinst.org/topics/integrity_of_science/AGU_IntegrityofScience_Gleick.pdf).

  127. EP
    Posted Dec 28, 2006 at 1:17 PM | Permalink


    We are optimistic that continued research will eventually resolve much of the current controversy over the effect of climate change on hurricanes. But the more urgent problem of our lemming-like march to the sea requires immediate and sustained attention. We call upon leaders of government and industry to undertake a comprehensive evaluation of building practices, and insurance, land use, and disaster relief policies that currently serve to promote an ever-increasing vulnerability to hurricanes.

    Katrina may have happened last year even without any increase in cylone frequency. Without proper state funding you will always have what happened in NO when a hurricane develops. I suspect that any genuine increase in hurricane frequency will discourage people from living in the coastal areas most affected. I see from the H&W paper that (assuming we can rely on complete frequency observations prior to satellite surveys)cylones forming equatorward of up to 25 degrees seem no more frequent than in 1950.

    Those in charge of NO got away with it for 55 years…

  128. EP
    Posted Dec 28, 2006 at 1:20 PM | Permalink

    That should read “cyclone” above ;)

  129. Dave Dardinger
    Posted Dec 28, 2006 at 1:37 PM | Permalink

    re: #117

    Welcome, Dr. Webster! Always nice to have an author of a paper we’re discussing stop bye. Thanks also for your well stated and interesting comments. You mention:

    Perhaps I should get out of science and become an agitator.

    Let me warn you that that’s what got Dr. Curry in trouble here. Blogs will eat away at your time like nothing else. I realize I should go and read the papers your mention and perhaps I will, but there’s a real case of “paper overload” on this blog. The person who wanted to could spend all day, every day, just reading papers of interest. But that’s been the problem for decades in other ways. It’s probably a large part of the reason I’ve only got a MS rather than a PhD. I preferred spending time reading interesting papers in the library to working in the laboratory or studying for classes.

    Nonetheless, we’re glad you’ve stopped by and I’m sure people will be pestering you with questions, pointed and otherwise. I hope you mostly enjoy whatever time you choose to spend here and don’t take the occasional knock as anything personal. Science is a contact sport of a sort and you’re a star, so people will want to test your mettle.

  130. John Norris
    Posted Dec 28, 2006 at 1:39 PM | Permalink

    re #110, 124

    Dr Gleick’s slides are excellent.

    From this link it looks to me as the government share decrease in total R&D is mostly due to a large increase in private industry R&D, at least up to 2004.

    R&D Chart

  131. Jim Edwards
    Posted Dec 28, 2006 at 1:40 PM | Permalink

    90

    Dr. Curry:

    I agree w/ #120 on your comments, such as,

    “When we a have a physical reason for expecting a relationship to be significant (theories and laws, even), then details of statistical significance are less relevant.”

    If this is a straw man argument like, “all scientists believe A in AGW is more than zero..” then I suppose I agree with it, to the extent you’re talking about a highly constrained, well-understood experiment [e.g. - water should freeze in the freezer].

    But if you’re talking about the real world, you’re wrong. If statistical confidence doesn’t exist for what common sense and physical understanding predict in a complex environment, then it’s likely that there’s at least one missing factor at play [e.g. - solar effects...].

  132. Brooks Hurd
    Posted Dec 28, 2006 at 1:42 PM | Permalink

    Re: 113,

    Ken,

    Mooney, like so many other agenda driven people, wants to use the media as a club to beat the other side into submission. My reading of his statement is that he sees the media as providing balanced and thus unfair coverage of the views that he opposes.

    This is same sort of argument that one frequently hears that it is unfair to have Fox News and talk radio. These are the folks who believe that when the US Constitution mentions “Freedom of Speach” in the Bill of Rights, it only applies to them and to those who share their opinions.

    Consequently, I can imagine that the theme of Mooney’s book will be.

  133. Chris H
    Posted Dec 28, 2006 at 2:22 PM | Permalink

    #117 Dr Webster, Thanks for your interesting and informative post. Please could you could repost it without the less than symbol so that we can see the whole of it. HTML doesn’t like less than or greater than symbols because they form part of the syntax and the various ways to avoid this don’t tend to work very well in blogs, so writing ‘less than’ or ‘greater than’ is safer. This site also supports latex if you are familiar with it.

  134. bender
    Posted Dec 28, 2006 at 2:24 PM | Permalink

    Re #118
    Dr Webster was burned by the ol’ less-than-sign problem. Care to finish that comment, Dr Webster? Very, very interesting. (Thanks for commenting at CA.)

  135. Dave B
    Posted Dec 28, 2006 at 2:28 PM | Permalink

    dr curry, you said:

    (holland)”was educated in the Australian system (standard methodologies may not be considered as standard across various fields and geographic regions) and he may have gotten used to using this technique. I doubt this was selected to confound people from figuring out what he actually did.”

    i agree. i don’t think the deception is bred from a desire to deceive or obfuscate. i think it stems from an unwillingness to consult appropriate statistical authorities. in the case of the hockey team, dr wegman and NAS tend to agree.

  136. bender
    Posted Dec 28, 2006 at 2:37 PM | Permalink

    Dr. Curry said:

    I doubt this was selected to confound people from figuring out what he actually did.

    Dave B in #134 agrees … and so do I.

    That is not the point however. The point is in #9. Please, can someone tell me which equation in Davis (1976) HW are invoking as a correction method? Has #9 got it?

    You see why Steve M is a fan of transparency and accountability when it comes to statistical methods? Then it doesn’t matter if the method is “arcane” or “standard” – it is what it is. We can debate its suitability once we know what the heck it is.

  137. Ken Fritsch
    Posted Dec 28, 2006 at 2:39 PM | Permalink

    Re: #129

    That chart should be of much interest to the peripheral discussion here and I need the practice producing images –so here goes. The Mooney/Curry argument for the various administration’s being good for R&D seem to be less clear on looking at this chart.

    The chart is visible in the preview so I’m pulling the trigger.

  138. Ken Fritsch
    Posted Dec 28, 2006 at 2:40 PM | Permalink

    Re: #136

    Damn!!

  139. Posted Dec 28, 2006 at 2:41 PM | Permalink

    Here is a relevant discussion of NATL hurricane data, comments welcomed:

    http://sciencepolicy.colorado.edu/prometheus/archives/author_pielke_jr_r/001035draft_paper_for_comm.html

  140. Nordic
    Posted Dec 28, 2006 at 2:44 PM | Permalink

    RE#129: Looking at that chart one would conclude that the war on science (research funding) was waged and lost by the elder President Bush, the Clinton administration kicked the can on down the road, followed by capitulation by the latest Bush administration.

    One might guess that the decrease in research funding during the GHW Bush administration might have something to do with the cold war ending and subsequent “Peace Dividend” that came with the downsizing of the military.

    I think the growth in private-sector research is more interesting. If one can’t find NSF funding to explore the linkage between rising SSTs and cyclone intensity maybe Munich RE would be interested in funding a little research to help them understand why they made such a bundle this year.

  141. Posted Dec 28, 2006 at 2:45 PM | Permalink

    On trends in federal funding for research:

    http://www.cspo.org/ourlibrary/perspectives/Pielke_October04.htm

    On the “war on science”:

    http://www.cspo.org/ourlibrary/papers/scientizing%20politics2.pdf

  142. fFreddy
    Posted Dec 28, 2006 at 2:52 PM | Permalink

    try anything once …

  143. fFreddy
    Posted Dec 28, 2006 at 2:54 PM | Permalink

    Jolly good. Ken, that was :
    [less than sign] img src=”http://www.aaas.org/spp/rd/usr04sm.gif” [greater than sign]

  144. J. Curry
    Posted Dec 28, 2006 at 2:59 PM | Permalink

    The so called war on science is NOT about research funding. It is about suppressing and distorting science. For recent examples related to climate, see Rick Piltz’s site Climate Science Watch http://www.climatesciencewatch.org/, his earlier posts are most relevant.

  145. J. Curry
    Posted Dec 28, 2006 at 3:02 PM | Permalink

    Re Davis (1976), i will contact Holland to try to find out which equations he used.

  146. David Smith
    Posted Dec 28, 2006 at 3:05 PM | Permalink

    Re #118 Dr Webster, I am honored by your reply and I appreciate the article references.

    My concept has been that easterly waves have an axis which extends well northward into the tropical Atlantic, away from the “cloud puffs” of the ITCZ, and that the location of greatest wind-shift is sometimes, perhpas often, often located away from the “cloud puff”, sometimes in a rather dry region. That’s the crux of my speculation – perhaps there are times/seasons/regimes when the two (moisture and what I’d call (perhaps incorrectly) vorticity) are better aligned.

    My concept of the ITCZ has been that of a mean moisture axis, with surface convergence, through which easterly waves (and their associated storms) travel.

    I appreciate your response to the question, and will do homework!

  147. jae
    Posted Dec 28, 2006 at 3:07 PM | Permalink

    Thanks for the funding link, Roger. I sure don’t see any “war on science” by the current administration there.

  148. fFreddy
    Posted Dec 28, 2006 at 3:11 PM | Permalink

    Re #144, J. Curry
    Note to anyone following Dr Curry’s link : you will need to delete the comma from the end of the URL.

    Dr Curry, I tried putting Wegman in the search box at that site and got no results. A quick flick through the first page doesn’t seem to show any scientific stuff, just advocacy stuff.
    Why do you regard this as a credible source ?

  149. Dave Dardinger
    Posted Dec 28, 2006 at 3:15 PM | Permalink

    re: #139

    Cute, Roger [g].

    You are aware, aren’t you, that we (and/or you brought up this same point here a few months ago? Almost surely the first possibility you mention is the correct one; i.e. the TS count was low up until the 1940s.

  150. Paul Penrose
    Posted Dec 28, 2006 at 3:15 PM | Permalink

    Re #125:

    The fact that the Government and Coporate lines in that graph are almost mirror images should tell you something. The fact of the matter is that Government spending on science has remained relatively stable since 1964, once inflation is taken into account. The only reason for the apparent drop in Government spending in that graph is because Coporate R&D spending has increased considerably over the 20 or 30 years. This is a good thing. I work in an industry that did not even exist in 1964, but because of large R&D investments since about 1970 has saved many lives and continues to save lives every day.

    Now I won’t argue that the selection of what gets funded may have changed since 1964, or that now large and expensive projects tend to be funded in favor of larger numbers of smaller projects; I just don’t know. However we are not cutting funding to science in the US, it is actually increasing if you count Coporate R&D, and despite what some may think, Coporate science is not evil. It has produced many break-throughs beneficial to society, often in cooperation with various universities around the world.

  151. Ken Fritsch
    Posted Dec 28, 2006 at 3:23 PM | Permalink

    RPJ, thanks for the countervailing views/links on hurricanes and R&D spending. Aint free speech and the internet great.

    The excerpt from your linked paper below is to me the most important aspect of it. Have you made the same calculations/simulations with these data that Emanuel (2005) outlined in his paper and concluding that for PID landfall numbers could be realized by chance?

    Secondarily, and regardless of other considerations, the important concerns for most of the human population would seem to me to be with, not the total number of cyclones and their intensities, but the number making landfall and their intensities.

    Emanuel (2005) raises the possibility that simple randomness might explain the lack of a trend in landfalling storm intensities in the presence of such a trend in the entire NATL. This may very well be the case (Pielke 2005); however, Emanuel’s point was made with respect to an integrated index of power dissipation, for which the landfalling component was only about 1% of the total NATL data. In the case of total storms, about 20% of the total storms make landfall over the entire dataset. It is therefore quite unlikely that randomness alone explains these results.

  152. Jim Edwards
    Posted Dec 28, 2006 at 3:26 PM | Permalink

    Is the portion of Fed Science budget related to biomedical research broken out ?

    Pres. Nixon began the much ridiculed “War on Cancer” over 30 years ago – much federal funding was directed there, to the supposed detriment to physics and climatological research funding.

    Much of the recent private research has been to expand upon the fruits of Nixon’s War on Cancer programs.

    If Federal biomedical research is correspondingly decreasing, presumably that would provide relative increases for the physical and earth sciences.

  153. J Edwards
    Posted Dec 28, 2006 at 3:28 PM | Permalink

    Re #150
    Paul, I agree with you. Roger Pielke’s paper on the subject demonstrates that government funding as a percentage of all discretionary funding has in fact increased over the same period. The big factor is that the US GDP has far outstripped both government funding and inflation, which is not just good, but excellent.

    Also, as you point out, corporate funding is not necessarily “evil”. It is, however, generally biased. Companies are going to fund reasearch in areas that will allow them to recoup profits. Unmarketable research is generally a dead end. For this reason, I think there will always be a place for government funded research. But the days when the preponderance of research was government funded are long gone.

  154. Posted Dec 28, 2006 at 3:31 PM | Permalink

    Dave-

    Thanks. I’m not at all surprised that it has been raised here (it does seems like a point that would be obvious to check in the data, as Judy Curry notes above in this thread). The point comes up all of the time in discussions of our normalized loss data, which shows no long-term trends (see, e.g., my exchange with Kerry Emanuel Dec, 2005, Nature). I am nonetheless very surprised that this aspect of the NATL TC data hasn’t yet been explicitly discussed in the peer-reviewed literature even as papers are published using data relatively uncritically on historical storm counts.

    But this seems to be changing, and not just with my short and simple draft — I know of another paper in the works (completely independent of mine), with a different slant on the same sort of issue.

  155. Ken Fritsch
    Posted Dec 28, 2006 at 3:35 PM | Permalink

    Re: #143

    Jolly good. Ken, that was :
    [less than sign] img src=”http://www.aaas.org/spp/rd/usr04sm.gif” [greater than sign]

    Jolly good indeed, fFreddy and thanks. When I saw the man himself misfire on some images I thought it might have been the site and not me. Are you sure that there is no backslash immediately before the greater than sign?

  156. TAC
    Posted Dec 28, 2006 at 3:40 PM | Permalink

    #141 Roger, thanks for posting Sarewitz’s review of Mooney’s “war on science”. Sarewitz reports:

    As an unapologetic critic of the Bush administration, I was eager to read a penetrating political analysis of how the current regime has sought to wring partisan advantage from the complex and difficult relationship between politics and science. Alas, what I found was a tiresome polemic masquerading as a defense of scientific purity.

    Having some familiarity with Sarewitz’s writings (“Frontiers of Illusion” and “Prediction,” among others), I am confident that his assessment is accurate.

  157. J. Curry
    Posted Dec 28, 2006 at 3:54 PM | Permalink

    I’m losing interest in this discussion. The so called war on science is not about research funding, it is about suppression and distortion of scientific research and assessments. Rick Piltz provides many examples of this related to climate research and assessments at his web site Climate Science Watch.

  158. Paul Linsay
    Posted Dec 28, 2006 at 3:55 PM | Permalink

    #144, J. Curry,

    The so called war on science is NOT about research funding. It is about suppressing and distorting science.

    When you talk about suppressing science do you mean Jim Hansen refusing to appear at a Congressional hearing because “deniers” like Spencer and Christy will be there?

    When you talk about distorting science do you mean the paper under discussion here which uses some mediocre data and unjustified smoothing to make its point? How about the hockey stick? What about the refusal to accept the satellite temperature measurements and the insistence on using the ground stations?

    Think of the refusal to accept AGW and its imagined future catastrophes as a correct application of the precautionary principle. Since the science is so uncertain it would be foolish to make large social and economic changes just because some subset of the scientific community believes deep in its heart that disaster looms a century from now. A subset which happens to have good political and oratorical skills so that its viewpoint is broadcast, broadcast to the detriment of others who have sincere and deep doubts about the validity of the basic science.

  159. jae
    Posted Dec 28, 2006 at 4:05 PM | Permalink

    When you talk about distorting science, think about what Chicken Little Al Bore is saying. LOL.

  160. bender
    Posted Dec 28, 2006 at 4:08 PM | Permalink

    All these funding trend data are fine. However the facts remain: (1) the war on terror is costly, (2) universities and the natural and earth sciences are paying for that choice.

    On top of the move toward expensive “Big S” science programs, you have basic physical science programs being “watered down” to include more social and economic dimensions, thus leaving less for basic earth observation and experimentation.

    It’s somewhat irrelevant how the left chooses to define an alleged capital W “War on Science”. I’m talking about a little w war of attrition against scientific programs. It is a demonstrable fact that science programs were cancelled when the decision was made to fund the war on terror. You can view it the way welikerocks does: “it’s one of the sacrifices we need to make”. But you can’t deny it.

    Spin it any way you like; you might influence public opinion, but you are not going to change perceptions on campus with such superficial analyses. Sorry to have to bring you this unpalatable message.

  161. J. Curry
    Posted Dec 28, 2006 at 4:28 PM | Permalink

    I am rather astonished that so many skeptics on this site, which should be about weighing the evidence for yourselves and drawing your own conclusions (according to what I think is Steve M’s goal), are so quick to form conclusions about Chris Mooney and the Republican War on Science tbased upon a few posts that make accusations of bush bashing and RP Jrs cherry picking of a single book review from the hundreds that you can find on the web. Has anyone other than RP Jr actually read the book? Google “Republican War on Science” and see what pops up. Hundreds upon hundreds of reviews. Check out this one from the Christian Science Monitor (which is surely not a bush bashing journal): http://www.csmonitor.com/2005/0927/p11s02-bogn.html. Even more astonishing is the automatic assumption about my opinion on the book, which i didn’t even discuss (beyond a brief description of what it was about), and the automatic assumption that I think there should be more research funding.

    Such assumptions and stereotypical behaviour (by everyone except Bender who has responded on this) reduce the credibility of this site as a place for bonafide scientific skeptics.

    For some examples of suppression and distortion (which can be found on Piltz’s web site), consider the following

    http://www.climatesciencewatch.org/index.php/csw/details/censorship-and-secrecy/

  162. jae
    Posted Dec 28, 2006 at 4:43 PM | Permalink

    quick to form conclusions about Chris Mooney and the Republican War on Science tbased upon a few posts that make accusations of bush bashing and RP Jrs cherry picking of a single book review from the hundreds that you can find on the web.

    Huh? I sure haven’t formed any conclusions, yet. IMO, it’s good to see the various views that are presented here.

  163. Nordic
    Posted Dec 28, 2006 at 4:45 PM | Permalink

    I am just skeptical that this is anything new. Administrations and agency heads change and different people’s oxes are gored while others feast. News flash: Tax dollars are appropriated by politicians and allocated by agencies under the executive branch of Government – headed by elected officials who, unsuprisingly, are politicians.

    If you want to completely remove politics from science lobby to have all federal research dollars given to research Universities with no strings attached. Universities could then distribute the funds to their researchers as they see fit. This would solve the problem because, as all of us that have been graduate students in research institutions know, Universities are completely free of internal and external politics;)

  164. Tim Ball
    Posted Dec 28, 2006 at 4:52 PM | Permalink

    #146
    The factor not being mentioned here, but important in the transition from an easterly wave to an enclosed system, is coriolis force (CF) and angular momentum. Below 10° of latitude CF is insufficent to trigger the circular system associated with tropical depressions, tropical storms, and then hurricanes. The greater the latitudinal shift the more angular momentum changes and influences the pattern and force of the system. If the easterly waves are forming further north in the eastern North Atlantic because of a shift in the Intertropical Convergence Zone (ITCZ) then it is likely more closed systems will develop. It is also important to consider that there are two Hadley Cells with the point of contact between them being the ITCZ. This contact zone will vary in width depending on several dynamic factors and thus extend or narrow the zone. This is then important with regard to latitude of part or all of the ITCZ and formation and transition of easterly waves. It is also important because of the land/ocean pattern on the coast of Africa at the critical latitudes.

  165. Posted Dec 28, 2006 at 4:53 PM | Permalink

    Judy-

    I thought that you wanted to talk about hurricanes;-)

    Mooney’s book is flawed from its title. The Bush Administration is not conducting a “war on science”. Quite the opposite; they are pursuing an ideological agenda by exploiting the nature of science — as every presidential administration has done. Their ideological agenda depends upon science, it would make no sense to conduct a war. What they are conducting a war on is their political opponents. I personally despise much of what the Bush Administration has done in many areas, but to suggest that there is a “war on science” is simply wrong and misleading.

    I have debated Mooney on this issue, and it is my view that he is sincere but that his view of the history of science and politics starts in 2001. To call his book “scholarly” is to fail to understand what scholars in science studies research actually do. I can document many instances of what you call “suppression and distortion” going back to the early 1960s across presidents and parties. For a few examples from more recent administrations see:

    http://sciencepolicy.colorado.edu/admin/publication_files/resource-1935-2004.27.pdf

    I debated Mooney on his thesis in October, 2005, and I wrote at that time, “Ultimately, I worry that Mooney’s thesis will be so readily received by practicing scientists, many of whom share Mooney’s political predispositions and the widespread misperception that science should dictate certain political outcomes, that it will have the perverse effect of furthering the politicization of science by scientists.”

    I won’t clog up this site further discussing this issue (which discuss ad nausem at Prometheus), as I’d rather hear views on my draft hurricane paper;-)

    As far as Dan Sarewitz, he is one of the leading _scholars_ of science in politics, and while like any expert his views should be read critically, at the same time he probably ought to be listened to as he knows what he is talking about (just like you’d want people to listen to climate experts on issues related to climate):

    http://www.cspo.org/home/newatcspo/sarewitz_chronicle.htm

  166. J. Curry
    Posted Dec 28, 2006 at 4:56 PM | Permalink

    Here is my hypothesis regarding the lack of an increase in U.S. landfalling tropical cyclones:

    1. AGW is contributing to an increase in the number and intensity of NATL TCs

    2. The AMO also contributes to the number and intensity of NATL TCs, and it also plays a dominant influence on the tracks of NATL TCs on decadal time scales (this may arise from gradients in the SST that influence atmospheric circulation , or possibly some other mechanism). In an active AMO phase, more TCs make U.S. landfall.

    3. Lets combine AGW and AMO and see what happens to U.S. landfalling TCs.
    (at this point it may be useful to reference figures 1 and 4rin my congressional testimony

    http://reform.house.gov/UploadedFiles/GT%20-%20Curry%20Testimony.pdf

    The number of U.S. landfalling tropical storms in the past 10 years exceeds (slightly) the previous peaks in the AMO (ca 1950, 1880). Oceanographers tell us we are probably 15-20 years away from the current peak in the AMO. So already we have slightly exceeded the previous peak in U.S. landfalls, and we have say 15 years to go before reaching the peak of the current cycle.

    The point is, trying to fit a trend line through this very cyclical data for the U.S. landfalls doesn’t get you anywhere. However, when you interpret the time series in the context of physical understanding of the AMO, if you compare successive AMO cycles, it can be argued that we are already seeing a slight signal in the U.S. landfalls that is associated with the increased number of tropical cyclones (relative to where we are at in the AMO cycle). This argument was presented in my testimony. No, nothing statistically significant here. But this is a good example of using physical arguments to interpret what is going on to pose a hypothesis about what is going on, and predicting that we will soon be seeing a signficant signal.

    Several issues here. Some scientists such as Kerry Emanuel think that the AMO doesn’t influence the tropical atlantic and the tropical cyclones. I think it does. Actually, the scariest scenario for the north atlantic is if you combine AGW with the rising AMO.

  167. fFreddy
    Posted Dec 28, 2006 at 5:01 PM | Permalink

    Re #161, J Curry
    My first experience of this Mooney fellow was at the time of the Wegman report, to which he responded here. His comment concludes :

    And make no mistake: Despite the new fireworks, that big picture remains unaltered. Whether the “hockey stick” is right, wrong, or irrelevant, the underlying message on global warming is that we’re causing it. Period. End of story.

    You may regard him as a “very serious science journalist” (#78): I regard him as another sleazebag making a quick buck out of preaching hate instead of rational thought. I have no great desire to contribute to his book royalties.

    The CSM article to which you link says :

    And today, a few GOP mavericks like Sen. John McCain speak the truth on issues like global warming.

    I assume this McCain fellow is another climate alarmist ? If so, it implies a less than open-minded position on the part of the reviewer.

    Regarding your statement that :

    Even more astonishing is the automatic assumption about my opinion on the book, which i didn’t even discuss

    Well, in #91 you said :

    It is a serious piece of scholarship that has garnered him numerous awards and nominations for even bigger awards.

    Based on this, I assumed that you have a good opinion of this book. If I misunderstood you, then I apologise.

    Regarding the very turgid article at ClimateScienceWatch to which you link, I find it very light on hard data. Given that this site describes itself on its home page as an “advocacy project”, I suppose I shouldn’t be surprised.

  168. Ken Fritsch
    Posted Dec 28, 2006 at 5:07 PM | Permalink

    On top of the move toward expensive “Big S” science programs, you have basic physical science programs being “watered down” to include more social and economic dimensions, thus leaving less for basic earth observation and experimentation.

    I think that you are looking at a cultural issue here that might be more related to university policy than directly to politics. As for me personally, I think the government spends too much on wars and R&D.

    Such assumptions and stereotypical behaviour (by everyone except Bender who has responded on this) reduce the credibility of this site as a place for bonafide scientific skeptics.

    We all have our individual “politics” and policy advocacies and that includes you and me. We voice them peripherally here and without the degree of justification that we expect from scientific pronouncements and conclusions that are the main topic of discussion. I personally get tired of attempts to label those with whom we disagree and make it part of the science argument. I am fully capable of hearing all sides of the arguments, whether they be politically or scientifically based, considering potentially biasing of those doing the arguing and then deciding for myself what I find closest to the truth.

    I personally think that Mooney makes some good points about politics and science, I just think that he fails to see the general problem of the mixing of science and government. Even those that advocate for more government involvement in science or any other endeavor surely do not reject the reality of the price that ultimately needs to be paid for that involvement. Mooney was aghast at Representative Barton’s involvement with science and naively forgets that it if you want to be independent of such controls then do not put your hand out for taxpayer money. We are not yet to the point of having scientist kings.

  169. Dave Dardinger
    Posted Dec 28, 2006 at 5:12 PM | Permalink

    Judith, (I don’t use Dr. Curry here since you’re speaking from a political POV and not a scientific one)

    I want and looked at Piltz’ CV on line as soon as I went to his site. It’s clear he’s an environmentist partisan, which is, of course his right. But according to Roger Pielke (Sr. or Jr. I don’t know):

    The irony here is that Rick Piltz who drafted the reports has no formal training in climate science, and is a Democrat.

    which means that his complaining isn’t very important.

    And you don’t find out is someone is reporting things straight by listening to those who agree with him. Which is seem to be what you’ve done, regardless of what someone said in the Christian Science Monitor. … and actually reading that review, I see nothing which actually supports Mooney. He just lays out his position and says that it will appeal to the usual suspects and perhaps those on the right who distrust government in general.

  170. Posted Dec 28, 2006 at 5:14 PM | Permalink

    Hi Judy-

    You write, “The number of U.S. landfalling tropical storms in the past 10 years exceeds (slightly) the previous peaks in the AMO (ca 1950, 1880).” Here are landfalling hurricanes by decade:

    1880s – 23
    1950s – 19
    1997-2006 – 23

    Maybe you meant 1996-2005. Another interesting “decade”:

    1908-1917 – 26

    Near as I can tell 1917 is about the bottom of the supposed AMO cycle:

    http://www.aoml.noaa.gov/phod/amo_fig.php

    I’m no climate scientist, but I don’t think that the data on landfalls supports your physically-based explanation.

  171. jae
    Posted Dec 28, 2006 at 5:14 PM | Permalink

    Dr. Curry: what is your take on the lull in hurricane activity during 2006?

  172. jae
    Posted Dec 28, 2006 at 5:19 PM | Permalink

    The mood on campus that Bender mentions is very understandable. The universities have become bastions of liberalism, following the myth that liberal people are more intelligent and caring than conservative people. So naturally they find as much fault in the present administration as possible.

  173. fFreddy
    Posted Dec 28, 2006 at 5:23 PM | Permalink

    Re #155, Ken Fritsch
    No backslash – do it just as I quoted, with the square brackets and their adjacent spaces replaced by lt/gt signs.

  174. J. Curry
    Posted Dec 28, 2006 at 5:29 PM | Permalink

    Jae: re my take on 2006 NATL TC season, I would say I agree with Jeff Masters

  175. welikerocks
    Posted Dec 28, 2006 at 5:32 PM | Permalink

    #112 So? Actually Bender I was thinking of the Department of Paleoceanography and Paleoclimatology Program at USC that SteveM posted for the topic about the Medival Warm Pool. Sheesh.

  176. J. Curry
    Posted Dec 28, 2006 at 5:34 PM | Permalink

    Roger, you should ask your colleagues Landsea and Mayfield, as far as i can tell they think that the AMO controls U.S. lanfalling TCs.

  177. Roger Pielke Jr.
    Posted Dec 28, 2006 at 5:37 PM | Permalink

    Re: 169

    Dave- To put the excerpt of mine (Jr. not Sr.;-) into context. Piltz was complaining about a Republican who had no science training editing an interagency report that provided an overview of gov’t climate science research. The irony of course was that Piltz (before he resigned) helped write that same report and he is a Democrat with no science training.

    The more general point is that everyone has interests, values, history, biases, etc. etc. It is a loosing proposition to attack based on such characteristics, at some point we’ll have to turn our attention to actual content — like hurricane landfall data;-)

  178. peter Webster
    Posted Dec 28, 2006 at 5:39 PM | Permalink

    Tim,
    Thanks for the comment. Rotation has EVERYTHING to do with the ITCZ location, that there are waves at all and that some of them turn into hurricanes. My field (tropical meterology and oceanography) has languished for a long time because we have considered the of too simpl as observed e systems: systems that have no variation in the zonal direction (I mentioned the importance of this before) and systems that exist in strong cross-equatoprial pressure gradient forces (CEPG). This opens up a new class of dynamics and they are important. The CEPGs are set up by the warm water to the north of the equator (NH summer) and the colder water to the south. The CEPG is increased in the african region by the sahara and its very low surface pressure. This CEPG drives air from the southern hemipshere to the northern hemipshere. If there were no rotation one would finish up with a constant Hadley Circulation and really not much fun! Up to here we agree.
    But the problem is that if you look at data you will find that the Hadley Circulation is anything but steady. The are strong day-by-day variations and the Hadley circulation (like the ITCZ) is a statistical artifact. If you are really interested I can provide you with some diagrams from a paper we are about to submit in a couple of weeks (proprietary). But here is what happens. The CEPG drives air across the equator. Because the earth is rotating, it brings across air from the SH with the worng signed vorticity. To correct this “inertial instability” (sorry, you have to read Tomas and Webster 1997) the Hadely ciruclation speeds up generating the opposite signed vorticity. This produces waves that propagate westward. The location of the mean ITCZ is were this correction takes place.
    These are the reasons why the ITCZ, the ascending branch of the Hadley circulation is located where it is and niot at say, 30N. This is the reason why waves propagate across the atlantic in the 5-15 N band. So it is rotation , but with the caveat that the CEPG causes an instability in this rotating system. Hurricanes form 10-20N in the NATL becasue that is where the source of the waves is, where the SST is warmest and etc.
    I think all of this hangs together and I don’t think we are speaking at odds. Its just a matter of explaining how rotation works.
    One intersting observation is that the mean ITCZ in regions where the CEPG is substantial lies equatorward of the warmest SST. If there were no rotation or if it were not important then the ascending air should be over the warmest SST and the lowest pressure.This rarely occurs! To me this is intriguing and proof in the eating of the pudding. Tomas and webster (1997).

    Peter W

  179. Roger Pielke Jr.
    Posted Dec 28, 2006 at 5:40 PM | Permalink

    Re: 177 – Judy- Please don’t forget that they are _our_ colleagues. And I don’t believe that either Mayfield or Landsea has testified before Congress (or posted here) on the relationship of the AMO and landfall ;-) My response was to you . . .

  180. J. Curry
    Posted Dec 28, 2006 at 5:52 PM | Permalink

    Sorry to see the thread degrade so badly, starting with my simple mention that someone was writing a book about hurricanes and global warming. Too much noise, greatly amplified by RP Jr., and I am finished posting here (for the same reason i stopped posting a few months ago as a result of degradation of the Gray thread.) I’ll check back occasionally to see if bender or willis comes up with anything useful on the SST-TC correlation. Again, be skeptical of your skepticism.

  181. jae
    Posted Dec 28, 2006 at 6:04 PM | Permalink

    Aw shucks, Judy left. I wanted to ask her what she thought about Roger’s paper at Prometheus. TCphiles should read it.

  182. bender
    Posted Dec 28, 2006 at 6:37 PM | Permalink

    Dr Curry,
    Before you go, have a look at Willis’s comment(s) on Quenouille’s method (use search tool at top right). He asked if it was a legitimate method for correcting for autocorrelation. Looks good to me, but maybe Dr Hoyos would have an opinion.

  183. TAC
    Posted Dec 28, 2006 at 6:47 PM | Permalink

    If it isn’t rude to ask: Does anyone know which equation in Davis [1976] provided the basis for H&W’s statistical method?

  184. Posted Dec 28, 2006 at 6:53 PM | Permalink

    Is it just my impression, or is it actually the case that every time I try to engage Judy Curry on substance (with data, papers, evidence, here or on our site) she packs up and leaves, usually saying something critical about “noise” or somesuch?

  185. EP
    Posted Dec 28, 2006 at 6:57 PM | Permalink

    The idea that climate science is being suppressed beggars belief. I’m sure there have been cases of certain funding being denied but you’d have to be living in a dungeon somewhere in the highlands of Scotland to think there’s no climate science being done in the US.

  186. EP
    Posted Dec 28, 2006 at 7:00 PM | Permalink

    Does anyone have a model to explain why the cyclone frequency for the 25 degree band about the equator dipped between 1950 and 2000?

  187. David Smith
    Posted Dec 28, 2006 at 7:05 PM | Permalink

    Re #166, #170

    A caution: in recent decades there may be a tendency by the NHC to classify weak systems as “tropical storms”, whereas in earlier decades those weak systems may not have been named. The decadal data (windspeed at landfall) should be plotted to see if it is skewed towards weaker systems nowadays, indicating a change in classification practices.

    If that modern bias exists, it may be due to the greater amount of storm (satellite) data available nowadays and a greater risk-aversion (the course of least regret).

  188. Steve McIntyre
    Posted Dec 28, 2006 at 7:51 PM | Permalink

    #185. TAC, I agree 1000% with your question which was unanswered in an earlier post. I see nothing in Davis 1976 that supports the test of Holland and Webster. The following comment in Davis 1976 is almost opposing:

    Thus if a 6-month filter were applied to the data the correlation scale of SST could be expected to approach one year and any less than perfect correlation could not be regarded as significant.

  189. Steve McIntyre
    Posted Dec 28, 2006 at 8:23 PM | Permalink

    Children, children. Too much politics. I’ve deleted a couple of posts and will delete some more if this keeps up. Please – no more war on science discussion on this thread.

  190. jae
    Posted Dec 28, 2006 at 9:49 PM | Permalink

    Anyway, is there no feedback on Roger Pielke’s paper? I think it is very relevant to the TC discussion. Where is the AGW “signal” here? Where is the SST signal here? Does warming have anything to do with hurricanes? It should, IMO, but Roger’s paper makes me wonder. Does chaos rule?

  191. Steve McIntyre
    Posted Dec 28, 2006 at 9:59 PM | Permalink

    #190. jae and others – let’s stay on narrow and technical issues on this thread please. There are many other forums (and threads) for general discussions, but few for technical discussions. I do not wish to continue going over ground that is well trodden. For now, let’s keep to statistical issues.

  192. Ken Fritsch
    Posted Dec 28, 2006 at 10:01 PM | Permalink

    [snip - Steve:  Ken,  I don't disagree with your comments but please lay off on this for a while.]

  193. Ken Fritsch
    Posted Dec 28, 2006 at 10:12 PM | Permalink

    Steve M, I just saw your get-back-on the subject post after posting my OT one.

    I agree and would like to ask RPJ again about whether he contemplates doing a statistical analysis of his landfall cyclone data to determine the probably of it being the result of chance. The landfall data has important implications for the validity of past cyclone data and its use in forming conclusions as was done in the Holland and Webster paper under discussion here.

  194. Willis Eschenbach
    Posted Dec 28, 2006 at 11:03 PM | Permalink

    Dr. Curry, I’m sorry you feel that the discussion has degenerated. Me, I treat the trolls and such that are on every blog like I treat the spam in my email … I just ignore it.

    In any case, I’ve taken a look at the Hunter-Webster paper. Despite the autocorrelation, the relationship between the number of named storms and the SST is in fact significant (p=0.03, 5 year trailing average). However, claims of correlations on averaged values must be treated with some skepticism. As two series are averaged, the correlation increases. The only meaningful number is the correlation between the datasets themselves. Sea temperature only explains about a quarter of the variance in Atlantic hurricane numbers (r^2=0.27).

    However, what all of that ignores is the measurement and sampling errors. This is all too common in the climate science field, I fear.

    The HadSST2 sea surface temperature dataset provides these errors, so I have used it for the SST. Here are the data and errors for the July-September temperatures:

    As we would expect, the M&S errors in the SST database are quite large in the earlier years and smaller in the more recent time. Note that we can’t tell whether the 2005 temperature was warmer than the 1940s or not, since their 95% CI’s overlap. Also, note the marked cooling in 2006.

    I have assumed that the standard error in the “Named Storms” is ±0 post 1970, +1/-0 in 1945, and +3/-0 in 1910, with a linear transition between the values. When we include these errors, the phase diagram they use to show the 3 “regimes” (Hunter-Webster (Figure 1) looks like this:

    I have placed a half-oval (thin black lines) around the area in which there is 95% confidence that each point might be found. This is not a rectangular box, since the errors add in quadrature. It is a half-oval rather than a full oval since our hurricane negative error is 0. If two half-ovals overlap, we cannot say that the data points are statistically different.

    As you can see, because of the large size of the errors, the “regimes” claimed by Hudson-Webster are not statistically distinct. As bender has correctly stressed, data without error estimates (including M&S errors) is simply not science.

    Finally, it is worth looking at the “regimes” without the 5-year filtering done by H/W. Here is that plot:

    Regimes?

    Also, note that rather than continuing on an up and away increase, in fact 2006 has come back into the middle of the fold … go figure.

    w.

  195. paminator
    Posted Dec 28, 2006 at 11:31 PM | Permalink

    Re Willis #194-

    Great work, Willis. Those error bars pretty much indicate the paper’s findings are rubbish. When I was in grad school, we had a phrase for data that looked like your last graph- “no thesis under that rock…”
    I noted your HadSST2 graph that extends back before 1950, which leads me to a question for RPJ on his posted short paper that is currently up for discussion.

    RPJ, did you look at US hurricane landfall proportions prior to 1950 using this or similar SST data, to see if the null result persists? If not, are there reasons to avoid this portion of the dataset, other than the larger error bars?

  196. bender
    Posted Dec 28, 2006 at 11:51 PM | Permalink

    Now, with Willis’s figures as a backdrop, I would like to reiterate my initial remark from the previous thread:

    Figure 4 (of HW) is comical. The hypothesis here is that a step function fits the hurricane proportion data better than a smooth oscillatory function. Where is the hypothesis test? How were the parameters of the step function chosen? Where are the methods?

    The idea that the climate system is switching states according to this step function is the essence of the phase diagram. Willis’s enhanced phase diagram illustrates why the idea of switching among disrete state is pure fancy: the error is enormous. Cute (hey, who doesn’t like Lorenz’ butterfly attractor?), but fanciful.

    Moreover this underlines my statement in #80 about a lack of self-criticism in climate science. They systematically neglect error because they are more interested in hypothesis confirmation than hypothesis refutation. Seems statistics just get in the way.

  197. bender
    Posted Dec 28, 2006 at 11:57 PM | Permalink

    For those interested in statistically robust trends in landfalling hurricanes, recall this was discussed some months ago. Use search tool at top right to locate relevant comments. Graphics are currently missing from those threads, but maybe they’ll be back soon?

  198. bender
    Posted Dec 29, 2006 at 12:11 AM | Permalink

    Since HW try to make the connection between discrete state-switching & the multistable Lorenz attractor, I may as well ask my naive question about time-scale compatibility of the two works. I had always assumed Lorenz (1963) was interested in the problem of weather prediction, and that his attractor was supposed to be relevant to short-run thermodynamics. Whereas HW are looking at the possibility of superdecadal changes in climatic regimes. Am I wrong? Are HW inappropriately conflating time scales, or am I just overinterpreting the connection between their phase diagram and Lorenz’s? Apologies to the authors if I’ve misrepresented their work.

  199. Louis Hissink
    Posted Dec 29, 2006 at 3:04 AM | Permalink

    I think Willis’ latest effort points to a possibility that hurricanes might be caused by something else. My guess is Birkeland currents but means leaving a Victorian gas-light era physics model and looking instead at a 21st century one which includes ‘electricity” as the dominant force in nature, where gravity is a minor bit player.

  200. Bob K
    Posted Dec 29, 2006 at 5:18 AM | Permalink

    Here is a breakout of all wind speed readings from the unisys hurricane database
    for the Atlantic basin. The file I used is tracks.atl that ends with 2005.
    It appears that as time goes on, more low readings are being included.
    There appears to have been a shift in observational techniques around 1950
    and again around 1980.
    I don’t know how this affects the statistical reliability, but thought it might
    have a bearing on the lower percentage of US landfalling storms. Might it be
    appropriate to only use obsevations above a certain wind speed threshhold?
    Say 30-35kts. Those are certainly the ones most likely to be recorded over long
    periods of time.

    10kt, 15kt, 20kt, 25kt, 30kt, 35kt, 40kt, 45kt, 50kt, 55kt, more, total
    1851-60, 0, 0, 0, 0, 0, 0, 69, 0, 147, 0, 685, 901
    1861-70, 0, 0, 0, 0, 0, 0, 54, 0, 140, 0, 827, 1021
    1871-80, 0, 0, 0, 0, 52, 0, 367, 0, 396, 0, 1252, 2067
    1881-90, 0, 0, 0, 4, 56, 167, 291, 143, 349, 67, 1268, 2345
    1891-00, 0, 0, 2, 47, 94, 326, 332, 223, 405, 116, 1463, 3008
    1901-10, 0, 0, 4, 74, 288, 498, 269, 181, 190, 86, 873, 2463
    1911-20, 0, 3, 29, 44, 114, 225, 153, 142, 98, 80, 667, 1555
    1921-30, 0, 2, 13, 33, 55, 370, 191, 116, 97, 56, 861, 1794
    1931-40, 0, 18, 51, 72, 78, 732, 418, 173, 149, 113, 971, 2775
    1941-50, 0, 18, 26, 105, 80, 360, 295, 230, 188, 109, 1094, 2505
    1951-60, 0, 27, 10, 301, 178, 327, 180, 229, 191, 103, 1241, 2787
    1961-70, 0, 11, 27, 475, 502, 193, 171, 178, 170, 137, 1419, 3283
    1971-80, 0, 20, 90, 342, 478, 210, 240, 218, 160, 136, 886, 2780
    1981-90 21, 47, 81, 302, 418, 281, 247, 244, 188, 178, 816, 2823
    1991-00, 6, 21, 164, 251, 475, 329, 308, 315, 264, 171, 1334, 3638
    2001-05, 34, 29, 139, 273, 378, 293, 212, 250, 221, 166, 816, 2811
    total, 61, 196, 636, 2323, 3246, 4311, 3797, 2642, 3353, 1518, 16473, 38556

    Here’s mean wind speed for all readings and also for speeds of 30kt or more.
    All, 30kt+
    1851-60, 71.6, 71.6
    1861-70, 71.7, 71.7
    1871-80, 62.7, 62.7
    1881-90, 61.2, 61.3
    1891-00, 61.4, 62
    1901-10, 52.1, 53
    1911-20, 57.8, 59.6
    1921-30, 61.2, 62.2
    1931-40, 53.1, 54.8
    1941-50, 58.6, 60.9
    1951-60, 58.5, 63.2
    1961-70, 55.1, 60.8
    1971-80, 49.2, 54.1
    1981-90, 48.4, 53.3
    1991-00, 52.7, 56.9
    2001-05, 50.6, 56.5
    total, 56.3, 59.3

    Hmmm. Doesn’t seem to format well.
    The data could be saved as a .csv file and put in a spreadsheet for better viewing.

  201. TAC
    Posted Dec 29, 2006 at 5:25 AM | Permalink

    #194 Willis: Nice graphics! Also, your characterization of the pre-1970 record as “censored” data is absolutely right — we have a lower bound on the number of TCs, but we might have missed some (but not too many; that’s why we might have an upper bound).

    There is a rich literature on how to deal with censored data, and the corresponding statistical tools (aka Survival Analysis, Tobit regression, etc.) are readily available.

    One obvious point is that mistakenly treating the lower bound as a standard observation — apparently what was done in H&W — yields biased results, particularly where the censoring mechanism is structurally related to one of the predictor variables.

  202. David Smith
    Posted Dec 29, 2006 at 6:06 AM | Permalink

    RE #194 willis, one of the questions still in my mind has to do with the choice of SST region. If your Figure 1 used 25-5N and 90-20W, would that make a significant difference in your results?

    Beyond statistics, Dr Webster if you’re around, a question related to the climate regime concept: any general thoughts on the structure of the tropical Atlantic atmosphere in the current regime? Has something shifted, is something moving faster/slower, is something less stable, etc? I’m interested in what I’d call the “mechanics” of the atmosphere in the new new regime.

    One piece of (limited) data I noticed is this sea level pressure plot . This covers much of the Western tropical Atlantic. What I note is the apparent sharp drop circa 1995. It makes me wonder if Atlantic pressure patterns shifted at that time. I’m trying to develop a mental picture of what’s different.

    Thanks

  203. TAC
    Posted Dec 29, 2006 at 6:28 AM | Permalink

    #194, #202 David, Willis and others: Is there a physical explanation for why the number of hurricanes should not be strongly correlated with SST? Willis’s figures — particularly if you exclude 2005 (an outlier by many accounts) — provide little or no evidence of a connection between SST and TCs. However, my (limited) understanding of the physics leads me to believe there should be at least some connection, so I am wondering what I am missing. Does this make sense to you?

  204. Posted Dec 29, 2006 at 7:30 AM | Permalink

    All, a few replies-

    Re: #193 Ken- Two replies. In my short paper see Figure 1 which shows the number of landfalls over time. What needs to be explained is the remarkable consistency of landfalls, especially if there are in fact big things going on in the Atlantic. In the discussion I present some very simple statistical analysis on the probability of various periods seeing different landfall rates assuming that the rate in the earlier period holds in the second.

    Re: #194 Willis- Very interesting! A small name correction- It is Holland/Webster;-)

    Re: 195 Parminator- I did not look at pre-1950 SSTs. Certainly would be worth a look, for completeness if nothing else.

    Re: 196 Bender- I agree with your point here, but I’d phrase it a bit differently. Within the community there are those who offer criticism in the best of the scientific tradition, but it is often dismissed as “nose”, “denial”, or an “attack.” The community’s pathology lies is in the fact that such dismissals usually win the day (for a range of reasons, a few political, but many cultural and professional).

  205. Steve McIntyre
    Posted Dec 29, 2006 at 7:55 AM | Permalink

    Judith Curry sent in the following data file http://data.climateaudit.org/data/hurricane/natldata.xls with the following caveat:

    Note: I am on travel this is the only data set I have with me. I have added TC data for 2005, 2006 but do not have the 2005 and 2006 SST data. So no guarantees that this is exactly what Holland used.

  206. Posted Dec 29, 2006 at 7:59 AM | Permalink

    Steve- Thanks, link doesn’t work ..

  207. John Norris
    Posted Dec 29, 2006 at 8:01 AM | Permalink

    re #205

    I got this response to the link to Dr. Curry’s data:

    Forbidden

    You don’t have permission to access /data/hurricane/natldata.xls on this server.

  208. Ron Cram
    Posted Dec 29, 2006 at 8:05 AM | Permalink

    re: 177

    Dr. Pielke,

    Great post there on 177. Thanks for explaining the facts on his resignation.

  209. welikerocks
    Posted Dec 29, 2006 at 8:26 AM | Permalink

    Within the community there are those who offer criticism in the best of the scientific tradition, but it is often dismissed as “nose”, “denial”, or an “attack. The community’s pathology lies is in the fact that such dismissals usually win the day (for a range of reasons, a few political, but many cultural and professional).”

    Aint that the truth.
    You can see it right here. Many un-political concerns and questions asked to scientists by equally educated scientists went completely unanswered in this discussion. My father always says, “watch carefully what they do not say”. I am also completely disgusted with the level of data accuracy, even though I am a nobody, I can reason and follow all the angles that are covered in a problem or puzzle and Willis’ graphs help alot. In fact, after this exchange its made me think and look at other sources.

    On top of the SST and the “global average temp” I am starting to suspect at this point that the “pre-industrial CO2 levels” may have been fudged, cherry picked – exaggerated on the low side in this case (in part from the link you posted Ron Cram). All this fudging, even minor instances adds up- like karma dude. Luckily I have a real working scientist living with me that doesn’t dismiss my thoughts as childish, silly “emotional” or unworthy of his ear.

    Roger Pielke Jr. or anyone,
    For fun and laughs I suggest reading Crichton’s lastest book “Next”. I got the book for Christmas and started reading it yesterday. The parallels and format in the tale compared to elements in this discussion and the state of science are surreal! There are even graphics of “Scientific PR press releases ” ” news paper articles-by “science reporters” “commentary about peer review” and he even mentions the brain and ambgu-la la (however it is spelled) as Al Gore did at the AGU. LOL Cheers!

  210. David Smith
    Posted Dec 29, 2006 at 8:37 AM | Permalink

    Re #203 TAC, my (classical/quaint/antiquated/layman) view is that there should be some connection between SST and the number of storms in the Atlantic.

    There is something of a “magic temperature” (26C, give or take a bit) below which it is difficult for a tropical storm to form. For storms to form, the surface water needs to be 26C or higher. As the Atlantic warms over time, the area that is +26C expands, and lasts longer, so that seedlings have a greater opportunity to turn into storms.

    This is especially important in the middle tropical Atlantic, where the seedlings (easterly waves) progress westward. The sooner they encounter +26C water, the sooner they can strengthen. This is important, because they often encounter unfavorable conditions when they enter the eastern Caribbean: they need to “get healthy” before entering the Caribbean/ Central Atlantic so that they have a better chance of survival.

    So, the “classical” view is that warmer water should lead to some increase in the number of Atlantic storms, and probably has, but it may be hard to discern in the poor-quality data.

    However, there are mysteries (at least to me). The global number of storms is remarkably constant at about 90 per year : why is that? Sea surface temperatures have risen but the global number of storms has been more-or-less constant. Perhaps it’s related to some aspect of energy release, perhaps it is upper-atmospherics. perhaps there are offsetting factors. My classical view would expect some increase in global storms as SST increases, but that does not seem to have happened (based on the admittedly weak global historical data).

    On the question of intensity, the classical view is that there should be some (modest) increase in intensity as tropical SST increase. As SST rise the warmth goes deeper into the ocean (which reduces the cooling due to ocean mixing during a raging storm), the surrounding surface air should be more humid, evaporation is enhanced, the seedlings get an “earlier start”, there’s a greater delta-T between the surface and upper air, and so forth. But, for a 0.5C increase, I would expect the increase to be modest and probably hard to detect.

    Directionally, I think that the Curry/Websters and the Landsea/Grays agree with each other – the difference is over magnitude.

    (They may also differ on the importance of any natural oscillations, which this HW paper addresses but which has not been discussed here (so far at least).)

  211. TAC
    Posted Dec 29, 2006 at 8:54 AM | Permalink

    #210 David: Thanks. That is exactly what I was looking for.

  212. Jeff Weffer
    Posted Dec 29, 2006 at 9:04 AM | Permalink

    This is much like the tree ring problem.

    Tropical Storms are affected by a large number of variables – SSTs, ENSO wind shear, jet streams, dust from the Sahara, High and Low Pressure systems around it, Solar?, atmospheric water vapour – you name it.

    Much like tree rings, trying to tie tropical storms to global warming is going to be a difficult case because tropical storms themselves are multivariate. Besides that, the data sets we have are incomplete and not long enough.

    All of these studies are trying to tease out a very weak signal and, in so doing, the statistics used are always reaching and questionable.

    Go back to the original tropical storm models (Bill Gray’s for instance), which include all these variables.

  213. bender
    Posted Dec 29, 2006 at 9:11 AM | Permalink

    Re #203

    Is there a physical explanation for why the number of hurricanes should not be strongly correlated with SST?

    But it is: r^2=0.27, p=0.03. (Well, not strongly.)

  214. Dane
    Posted Dec 29, 2006 at 9:24 AM | Permalink

    [snip- Steve: Dane, take this to another thread please]

  215. bender
    Posted Dec 29, 2006 at 9:24 AM | Permalink

    Re #204

    The community’s pathology lies in the fact that such dismissals usually win the day

    Interesting. To AGW believers: if beating your head against a wall hurts, I have a suggestion …

    Re #212

    All of these studies are trying to tease out a very weak signal

    I have to agree. (Strong politics, weak signals.)

  216. Jean S
    Posted Dec 29, 2006 at 9:52 AM | Permalink

    Willis, Judith, Roger and others interested in these storm data:

    IMHO, there is a fundamental problem with the statistical approach of all of the storm papers (at least the ones I’ve seen referred here). That is, the underlying data set is count data, and furthermore, the level of measurement (if the storms are given in categories like “named strom” and hurricanes in Saffir-Simpson categories as they should IMO as the early wind speed measurements are inaccurate) is ordinal. Hence the standard linear statistics may not be approriate here.

    IMO, something like Generalized Linear Models (multivariate Poisson regression, Multinomial,…) should be used for analyzing the relationship between, e.g., SST and hurricane counts. However, I do not think standard GLM is appropriate due to apparent autocorrelation, but there exist extensions to time series. The standard references include

    Cameron & Trivedi: Regression Analysis of Count Data, Cambridge University Press, 1998

    Harry Joe: Multivariate Models and Dependence Concepts, Chapman & Hall, 1997.

    Also the following recent paper looks very interesting:

    R. Zhu & H. Joe: Modelling Count Data Time Series with Markov Processes Based on Binomial Thinning, Journal of Time Series Analysis Volume 27 Issue 5 Page 725 – September 2006.

    Abstract. We obtain new models and results for count data time series based on binomial thinning. Count data time series may have non-stationarity from trends or covariates, so we propose an extension of stationary time series based on binomial thinning such that the univariate marginal distributions are always in the same parametric family, such as negative binomial. We propose a recursive algorithm to calculate the probability mass functions for the innovation random variable associated with binomial thinning. This simplifies numerical calculations and estimation for the classes of time series models that we consider. An application with real data is used to illustrate the models.

  217. Steve McIntyre
    Posted Dec 29, 2006 at 10:59 AM | Permalink

    #207. I’ll try to fix it. After our changeover to the new server, a variety of things don’t seem to work as they did before. I’ll try to chase down John A.

  218. Ken Fritsch
    Posted Dec 29, 2006 at 12:54 PM | Permalink

    Re: #197

    For those interested in statistically robust trends in landfalling hurricanes, recall this was discussed some months ago. Use search tool at top right to locate relevant comments. Graphics are currently missing from those threads, but maybe they’ll be back soon?

    Thanks, Bender, for the reminder. The discussions appear to point to the same interpretation as RPJ used in his draft paper, but without definitive statistical analysis.

    Re: #204

    In the discussion I present some very simple statistical analysis on the probability of various periods seeing different landfall rates assuming that the rate in the earlier period holds in the second.

    Thanks RPJ for the reply. I want back to the paper for a closer read and did see the probability measures assuming a binomial distribution.

    Overall I think I agree with the “over 26 degree centigrade” theory of which David Smith reminded me and additionally that the effect of SST on TC activity and intensity could well be sufficiently small to make finding the signal difficult.

    I also recall a paper reviewed here earlier that used delayed or anticipatory SSTs on the track or as near as measurements were available to the TC track to relate SST to PDI or ACE and found no significant correlation. I need to search for that author’s name as I think I would recognize it if I saw it.

    In the Holland/Webster paper, the authors seemed to me to be in a hurry to make the policy statement of connecting AGW to significantly more TCs without deference to countervailing evidence that scientists would normally be expected to make. My view may be the result of a layperson not understanding all the nuances involved. It is also frustrating not to be able to hear their comments and replies to some of the seeming weaknesses in their paper noted here, Dr. Curry’s efforts notwithstanding. It is too easy without those replies or an opportunity to make them to talk amongst ourselves as though their results/conclusion have been refuted.

  219. David Smith
    Posted Dec 29, 2006 at 1:42 PM | Permalink

    RE #220 Ken, the paper you’re recalling may be this one by Patrick Michaels. It got lost in all the activity on CA last September.

  220. Ken Fritsch
    Posted Dec 29, 2006 at 2:16 PM | Permalink

    Re: #221

    Yes Patrick Michaels — I would have remembered the name. Thanks David. I think he was ranked way down on the Curry list for AGW advocacy by climatologists.

    TC win speed reaches a threshold at 28.25 degrees C for SST and platueas there with SST being defined in anticipation of the storm and as a maximum. The authors conclude that the relationship found would tend to contradict what Webster and Emanuel found and predict no significant increase on TC counts or category 4 and 5 hurricanes but that with 2 or 3 degree increases the portion of higher category storms could increase.

    I have noted that this paper was not referenced in the Holland/Webster paper under discussion here.

  221. Jeff Weffer
    Posted Dec 29, 2006 at 3:14 PM | Permalink

    Here are the estimates of the CO2 content of the atmosphere over the past 500 million years.

    We are at the historic low in the history of the planet in terms of CO2. Over the eons, plants and geological processes have taken the CO2 out of the atmosphere and buried it. Where do you think all that coal and oil came from in the first place.

    The climate does not seem to have varied much over time with these changes in C02 levels at all. For example, 580 million years ago, the last snowball earth took place in which the entire planet froze over. CO2 levels at the time? – 5000 ppm or 13 times higher than today.

  222. David Smith
    Posted Dec 29, 2006 at 3:21 PM | Permalink

    Re #222 If I remember correctly, Michaels wrote a book some years ago in which he said that Earth would indeed warm due to CO2, CH4, etc at about the no-feedback rate (1 to 1.5C for 2x). But, the consequences would not be severe. Needless to say, he’s been shunned.

    Sounds like Michaels is a lukewarmer, like me

  223. J. Curry
    Posted Dec 29, 2006 at 5:20 PM | Permalink

    A few quick replies:

    Steve M, thanks for trying to keep the noise down on this thread

    Re Pat Michaels, i don’t believe that I have ever been publicly critical of him. In fact i think his paper has some useful points (I referenced it in my BAMS article). In particular he talks about SST thresholding for hurricane intensity. Rather than Holland’s regime change, i prefer either the SST thresholding and/or a combination of AMO+global forcing (including AGW) as contributing to the explanation for the lower frequency (non el nino) variations in the TC time series

    Jae, re Roger’s paper, see my previous reply #166. You can’t use U.S. landfalling TCs to infer anything about the causal mechanisms of global TCs or even NATL TCs, but it is certainly legitimate to ask why the increase in landfalls is less than than the increase in total NATL TCs. I have put forward a hypothesis see #166. The landfalls are of obvious sociocoeconomic importance, but a landfall is rather a fluke of weather patterns, we don’t really understand much about what controls the tracks in a climatological sense

    Willis, thanks for your analyses. I have seen others report a 0.3 correlation (statistically significant) for the yearly values previously. There is a sense that the correlation is lower for the higher frequencies and higher for the lower frequencies in the time series, which was the point H/W were trying to address.

    Jean S, I am intrigued by your references, thanks, I will forward them to Carlos Hoyos.

    I haven’t heard back yet from Carlos Hoyos, and Greg Holland is on vacation (e-incommunicado)

  224. Willis Eschenbach
    Posted Dec 29, 2006 at 5:33 PM | Permalink

    Jean S., as always, a very interesting post. Your point is well taken, the data is not well described by a Gaussian distribution. Here is a comparison of three distributions and the hurricane data (1851-2006).

    As you can see, the data is best described by a Gamma distribution, shape = 5.1, scale = 1.6 … about which I have reams of ignorance. Perhaps you could comment on the statistics of Gamma distributions …

    w.

  225. TAC
    Posted Dec 29, 2006 at 6:11 PM | Permalink

    #224 Willis, something is not quite right about the figure. The Poisson is a discrete distribution — i.e. only defined for integer values — which means that its CDF should be jagged (saw-toothed). Perhaps your graphics package contributed some unwanted linear interpolation?

    In any case, my guess (see previous posts (e.g. #98) here) is that, given the sample sizes, the difference between a continuous appoximation (e.g. Gaussian or Gamma) and the Poisson is negligible.

    On the other hand, it is the case that the Poisson is correct (in theory) distribution to describe the number of arrivals for a Poisson Process — which seems to provide a pretty good description of hurricane landfalls.

  226. Posted Dec 29, 2006 at 7:23 PM | Permalink

    #225, #224

    The use of various distributions, especially Poisson, to describe various aspects of hurricane/TC frequencies has been well-explored in the literature, especially in the work of Jim Elsner. See, e.g.,

    http://ams.allenpress.com/perlserv/?request=get-pdf&doi=10.1175%2F1520-0442%281999%29012%3C0427%3AFINAHF%3E2.0.CO%3B2

  227. Ken Fritsch
    Posted Dec 29, 2006 at 7:52 PM | Permalink

    Re Pat Michaels, i don’t believe that I have ever been publicly critical of him. In fact i think his paper has some useful points (I referenced it in my BAMS article). In particular he talks about SST thresholding for hurricane intensity. Rather than Holland’s regime change, i prefer either the SST thresholding and/or a combination of AMO+global forcing (including AGW) as contributing to the explanation for the lower frequency (non el nino) variations in the TC time series

    Dr. Curry, here is your list to which I was referring. I should keep it posted at my computer for quick reference when reading papers. Pat Michael’s ranking in the list appears a bit conflicted.

    http://www.climateaudit.org/?p=851

    Re #12: David, I would characterize the papers in the context of a spectrum (no attempt to develop a pdf). Consider a scale of 1-10, and for simplicity i will refer to the authors rather than the papers in an attempt to integrate both their published (in some cases multiple papers) and their public statements on the subject. This is my take on this anyways.

    1 Emanuel (all of the increase is AGW, existence of AMO is questioned)
    1.5 Mann (ditto, but does not question existence of AMO, just its influence on hurricanes )
    2 Greg Holland (coauthor of WHCC and former Gray student, strong pro AGW statements)
    2.5 Trenberth (mostly AGW, a little AMO)
    3.5 Webster (clear AGW signal, but we cannot project forward without better understanding)
    3.5 Curry (recent presentations: about 2/3 AGW, 1/3 AMO + 20 yr cycle)
    4 Knutson (AGW, but the effect should be smaller than that inferred by Emanuel and WHCC)
    4.5 Pat Michaels GRL paper (the paper is in stark contrast to his public statements, which rate (a smiley that did not reproduce from Dr. Curry’s original list).
    4.9-5.1 RP Jr (trying to carve out a scientifically neutral position; Roger pls clarify)
    6 Chelliah and Bell (AMO, but something different has happened since 1995 that may be externally forced)
    8 Landsea (AGW will cause a small increase, but nothing detected yet)
    8 Pat Michaels (public statements, in contrast to his published paper)
    9 Klotzbach
    10 JJ Obrien (nothing published, but a presentation on his website has really been making the rounds; no AGW)
    (15) Bill Gray (arguably off the spectrum, he expects global cooling and diminishing hurricane activity in 3-8 years)

    To me, the more interesting thing is to look at 2.5-8, the people/papers that are saying both, but with varying emphases (rather than focusing on the extremes).

    in terms of statistical activities, can the position at 1 or 10 be falsified?

  228. Willis Eschenbach
    Posted Dec 29, 2006 at 9:37 PM | Permalink

    TAC, thanks for your comment. You say that the Poisson is a discrete distribution, which it is. However, the cumulative distribution changes slowly, and I have sampled it at 28 points (corresponding to the 28 maximum hurricane count), so it looks smooth, although you can see slight kinks in the tight turns.

    Also, the difference between the Gaussian and the data is one hurricane, and occurs at about the average number of hurricanes (~9). This may or may not be significant, but the difference in the statistics of their expected error, effects of autocorrelation, and the like may be.

    In any case, I’m not sure if the Poisson distribution is more likely than the Gamma distribution. Poisson fit improves if the data is detrended, but the Gamma distribution still beats it. Gamma is generally used for Mean Time Between Failure (MTBF) analysis, and hurricanes can certainly be considered as a “failure” of the normal heat redistribution modes of the tropical ocean … dunno.

    w.

  229. TAC
    Posted Dec 29, 2006 at 9:55 PM | Permalink

    #228 Willis, thanks for the reply.

    My real point is that, given the count sizes, from a practical perspective it makes almost no difference which distribution you use. However, for substantially smaller expected counts, one would be right to worry about whether it is appropriate to approximate the Poisson by the Normal or Gamma.

    Just to be clear about the graphical/interpolation issue: When you plot the CDF for a Poisson variate, the first derivative is “always” 0 — at the integers you have a step, everywhere else it is flat — and the result should look like a staircase. However, many graphics packages default to “connecting the dots” — I assume that’s what happened in the figure — which makes the Poisson look even more “Gaussian” ;-)

  230. Willis Eschenbach
    Posted Dec 29, 2006 at 10:38 PM | Permalink

    TAC, your point is taken, poisson is a discrete function, and I have connected the dots.

    w.

  231. Posted Dec 30, 2006 at 12:20 AM | Permalink

    Re # 83

    “My guess is we will see another attack on the current administration, laying the blame for more and stronger hurricanes”

    Wow I will sleep well tonight knowing my forthcoming book, Storm World, whatever its flaws, certainly won’t be susceptible to this rather caricatured attack. For those interested, the publisher website is here

    http://www.harcourtbooks.com/bookcatalogs/bookpages/9780151012879.asp

    I’m sorry I didn’t see this thread earlier but I’ve been kinda out of it. Thanks to everyone for all the lively discussion of my work, and happy new years to all

    cm

  232. Steve McIntyre
    Posted Dec 30, 2006 at 12:35 AM | Permalink

    #23 OK, Chris, you’ve plugged your book – if you wish to advertise here, I’d be happy to discuss. Everyone else – please keep this thread on statistical issues.

  233. McCall
    Posted Dec 30, 2006 at 2:05 AM | Permalink

    Statistically and journalistically, Mr. Mooney has a virtual NG-tube direct from RC. His lack of balance and objectivity on this issue rivals the worst of AGW spewmen.

    In fairness, I’m reviewing r^2 calculations of such statements — preliminary calculations near .90, surprised even me! But like Dr Lambert, he’s had many chances to show a professional level of objectivity — whiffing at each opportunity, that .90 may actually be low?

    Please Dr Curry, where has Mr Mooney written of AGW critical or skeptic position in a positive light? If RC, Quiggin, Huffington, Laurie David, among the many who cite this guy, well — the statistics of his biases aren’t that hard to analyse, are they?

  234. McCall
    Posted Dec 30, 2006 at 2:21 AM | Permalink

    Correction: spewpersons

  235. Willis Eschenbach
    Posted Dec 30, 2006 at 3:56 AM | Permalink

    Judith, thank you for your comment. You say:

    I have seen others report a 0.3 correlation (statistically significant) for the yearly values previously. There is a sense that the correlation is lower for the higher frequencies and higher for the lower frequencies in the time series, which was the point H/W were trying to address.

    This reveals nothing about the signal. The point that you seem to be missing is that this increase in correlation is not a consequence of the signal, it is a consequence of the filtering. I just ran a pair of red noise series, random walks. The raw data has an r^2 of 0.03, p=0.04 (significant). With a Gaussian filter of width 20 (FWHM), the same data has an r^2 of 0.16, p=0.045. Filtered at width 36, it has an r^2 of 0.50, p=0.05. This increase in correlation is because as the filter length increases, the resultant curves more closely resemble straight lines, and the correlation increases. It is an inevitable consequence of filtering. But I certainly would not say that one red noise series “explains” half of the variance in the other series …

    Thus, the claim that increasing correlation for lower frequencies in the hurricane data has a physical meaning would have to be supported in a statistical manner, by showing that the correlation was significantly greater than would be expected from merely the increasing filter length. As far as I can see, H/W have not done that. This was the reason that I said above that:

    However, claims of correlations on averaged values must be treated with some skepticism. As two series are averaged, the correlation increases.

    w.

  236. Willis Eschenbach
    Posted Dec 30, 2006 at 8:03 AM | Permalink

    Man, Davis 1976 is twisting my head. I can’t make out their paper. In equation (2) they say

    S = 1 – average(p-phat^2)/average(p^2)

    Shouldn’t that be

    S = 1 – average(p-phat)^2/average(p^2)

    ??

    I still haven’t figured out how they are deciding if their results are significant given autocorrelation …

    w.

  237. Dave Dardinger
    Posted Dec 30, 2006 at 8:22 AM | Permalink

    re: #235

    Willis, for those of us who are statistically challenged, could you or someone else put this in the form of a R script? I’d like to play with it and get a feel for how things work.

  238. J. Curry
    Posted Dec 30, 2006 at 9:38 AM | Permalink

    Willis, I haven’t had sufficient time to go through all this. The big issue in moving the science forward is how we can combine physics and statistics to make sense of what is going on. Stating that H/W’s attempt to consider the low frequency correlation between SST and TC is not statistically significant may very well be correct. The challenge is then to try to devise some sort of analysis that has some hope of being statistically significant that can tease out the hypothesized low frequency link between SST and TC. If the data quality and length simply won’t allow this (to clarify whether a significant link exists or not), then scientists will then explore theory and models to try to understand this link, and will also test to see if the relationship has any predictive value. Note in terms of predictive value, this low frequency relationship might be potentially useful in predicting averate TC activity 20 years from now with an SST increase of 0.5C (25 years from now we will be able to verify this prediction and we will also have 25 more years of data and potentially more accurate historical data base). So, in the context of how science works, demonstrating that one analysis is not statistically significant isn’t sufficient to reject the hypothesis that there is a low frequency correlation between NATL SST and TC counts.

    I am not defending the H/W paper, I am just trying to clarify how science works when you don’t have sufficient accuracy/length of data. The hypothesis of the low frequency TC-SST link is an interesting one, even if the statistical link is not sufficiently established by H/W. By the same token, Pat Micheal’s idea about thresholds of SST for increasing TC intensity was also very interesting, but would not meet your statistical significance test either. These hypotheses, supported by the data that we do have, provide fuel for further scientific progress, and should be part of the published scientific literature.

    (and we don’t need to diverge here to talk about what policy makers do or don’t do with the science, we’ve already covered that territory).

    p.s. on the status of the H/W paper. It is accepted for publication, but it is part of a special issue that is awaiting completion of other papers. I understand that some revisions are being contemplated. If this thread can provide constructive input on how better to analyze the data or otherwise interpret the data, I’m sure it will receive careful consideration.

  239. Posted Dec 30, 2006 at 10:03 AM | Permalink

    Hi Judy-

    This is great to hear: “I understand that some revisions are being contemplated. If this thread can provide constructive input on how better to analyze the data or otherwise interpret the data, I’m sure it will receive careful consideration.”

    1. I recommend that HW carefully consider the issues of uncertainty in the underlying datasets, as described by Willis here:

    http://www.climateaudit.org/?p=980#comment-75389

    Error bars should appear on all figures. The discussion should acknowledge the implications of these uncertainties for the analysis.

    There is good justification for quantifying uncertainties in the TC count data in these papers:

    http://www.climateaudit.org/?p=991#comment-75503

    2. And also a far more comprehensive treatment of statistical significance through the analysis, along the lines described by Willis here:

    http://www.climateaudit.org/?p=991#comment-75568

    Specific statistical test should be reported as well as statistical results (e.g., p values). Steve M.’s request for clarification on how the Davis paper was actually used would be good also.

    3. 2006 should be included.

    The above are relatively simple extensions of what is in the paper already, and would greatly increase its statistical rigor.

    As you know I have a paper under review for the same issue (you saw a very early version), an based on that I think there is indeed time for revisions ;-)

    Happy new year!

  240. bender
    Posted Dec 30, 2006 at 10:06 AM | Permalink

    Re #236
    -I hadn’t noticed that error. Thanks for catching it.
    -Agreed: Davis (1976) is a confusing paper.

    Re #238

    If this thread can provide constructive input on how better to analyze the data or otherwise interpret the data, I’m sure it will receive careful consideration.

    Find a proper citation from the statistics literature on correcting for autocorrelation. Or clarify what exactly was extracted from Davis (1976).

    It is accepted for publication

    Really?! I wonder who the reviewers were?

  241. bender
    Posted Dec 30, 2006 at 10:08 AM | Permalink

    #240 was cross-posted with #239.
    I agree with changes recommended in #239.

  242. Posted Dec 30, 2006 at 11:09 AM | Permalink

    #238

    So, in the context of how science works, demonstrating that one analysis is not statistically significant isn’t sufficient to reject the hypothesis that there is a low frequency correlation between NATL SST and TC counts.

    What would be needed to reject that hypothesis? I’m bit worried about ‘low frequency correlation’. In general, for given sample, correlation is a non-monotonic function of running-mean window length.

    I am just trying to clarify how science works when you don’t have sufficient accuracy/length of data

    I see, i.e. the effect is of a magnitude that remains close to the limit of detectability, or many measurements are necessary because of the very low statistical significance of the results.

  243. Posted Dec 30, 2006 at 1:07 PM | Permalink

    #242 cont.

    Here’s the figure. Generated 2 i.i.d Gaussian vectors (N=200). Computed sample correlation coefficient (red in the figure). Filtered both vectors, running mean n=3,5,7..,63. At each n computed correlation coefficient again (used data points 50..150 to avoid truncation problems). Plotted correlation coeff wrt n. Repeated 4 times. 50..150 original data is here .

    (site administrators are free to copy the figure here, if necessary )

  244. Willis Eschenbach
    Posted Dec 30, 2006 at 8:11 PM | Permalink

    Judith, thank you for your comment, which contains many interesting insights. However, in it, you say:

    So, in the context of how science works, demonstrating that one analysis is not statistically significant isn’t sufficient to reject the hypothesis that there is a low frequency correlation between NATL SST and TC counts.

    Ummm … respectfully, that’s absolutely not how science works. It is not anyone’s responsibility to reject the hypothesis that there is a low frequency correlation between SST and TC counts.

    On the contrary, it is the responsibility of those who assert the low-frequency hypothesis (e.g. H/W) to show that we can reject the null hypothesis, which is that there is no low-frequency correlation between TC and SST.

    To do so, they need to show that the relationship is statistically significant at some level above random chance (typically p less than 0.05), which they have not done.

    Again, thank you for your participation in this discussion.

    w.

  245. Posted Dec 30, 2006 at 10:36 PM | Permalink

    Willis-

    Last year Kevin Trenberth wrote in Science in regards to hurricanes that perhaps we should be focused on the “inverse” null hypothesis:

    “In statistics, a null hypothesis–such as “there is no trend in hurricane activity”–may be formed, and it is common to reject the null hypothesis based on a 5% significance level. But accepting the null hypothesis does not mean that there is no trend, only that it cannot be proven from the particular sample and that more data may be required. This is frequently the case when the signal being sought is masked by large variability. If one instead formulates the inverse null hypothesis–”there is a trend in hurricane activity”–then the 5% significance level may bias results in favor of this hypothesis being accepted, given the variability. Acceptance of a false hypothesis (a “type II” error) is a common mistake. Rather than accept the hypothesis, one may be better off reserving judgment. Because of the weakness associated with statistical tests, it is vital to also gain a physical understanding of the changes in hurricane activity and their origins.”

    http://www.sciencemag.org/cgi/content/full/308/5729/1753

  246. Dave Dardinger
    Posted Dec 30, 2006 at 11:34 PM | Permalink

    re: #245,

    I can’t see how you could expect either the regular or inverse null hypothesis to be rejected in his example.

    If there is truly a trend, it may be too weak to be able to reject the null hypothesis that there is no trend. But in that case, clearly the inverse null, that there is a trend, could not be rejected.

    And if there is truely no trend, then obviously the null hypothesis that there is no trend couldn’t be rejected. But it’s not clear that the inverse, that there is a trend, must be rejected. Here the problem is that we don’t have a WFH (Well Formed Hypothesis). Surely a well formed hypothesis must have some sort of number attached. I.e. that there is a trend greater than 5% increase per year (or decade or whatever). Obviously if there is a trend but it’s .0005% per year, you couldn’t reject the hypothesis that it’s there without a great amount of data or a great precision of measurement.

  247. Willis Eschenbach
    Posted Dec 31, 2006 at 3:00 AM | Permalink

    Like many things in climate science, the problem with cyclones is that our data is very poor.

    1. The older cyclone data is known to be under-reported, giving a false trend from the beginning to the end of the data.

    2. The further back we go in the record, the larger the errors in the SSTs become.

    I keep trying to emphasize this question of errors, let me give it another try. Here, for example, is the data that is used to construct the “AMO”, the Atlantic Multidecadal Oscillation, including the measurement and sampling (M&S) errors.

    I have not shown the most probable value, because I want to emphasize that in fact, the real line could be in there anywhere.

    Note that because of the measurement and sampling errors alone, and totally ignoring statistical errors, we can’t even say that the Atlantic is warmer now than in 1870, or in 1950. Nor can we say much about the AMO, because a straight line trend crosses almost every error line.

    Now, to statistics. Ignoring the M&S errors, the most probable underlying trend is 0.3°/century, with a 95% CI between 0.1 and 0.5°/century. Here we are discussing a couple hundredths of a degree difference in trend per century … folks, I got bad news. We can’t measure the temperature of the North Atlantic Ocean to ±0.02° even today. The 95% CI in the 2006 data is ten times that amount.

    And even ignoring M&S errors, it is very difficult to find any significant correlations between SST and hurricanes, and when they are there, they are quite small. But the rude reality is, we can’t ignore the problems in the data, and when they are included, we cannot find significant correlations at all.

    What we can say is that if there are correlations between SST and hurricane numbers, they must be pretty weak. Why? Because we can’t find them. Yes, there’s flawed data … but even with flawed data, a strong signal rises above the noise. This is the dirty little secret of climate science, the data is so poor “¢’‚¬? the datasets are so short, and the signal we are looking for is so small, that we can’t say much about the climate at all.

    This is hidden because scientists don’t put error bars on their data. People keep saying the recent warming, say 1980 on, is “unprecedented”. That’s nonsense. If we put error bars on all of the 2006-1980 = 26 year long trends, we find that rises of this magnitude are found a number of times in the record. Which brings me back to the null hypothesis question.

    Before we spend millions of dollars on preventing “catastrophic climate change”, we need to determine of anything unusual is going on. I can find no data to reject the null hypothesis, which is that the current climate is well within the historical record regarding trends and temperatures. Until someone provides some evidence to the contrary, we can’t reject the null hypothesis that we are seeing natural variations in temperature.

    Me, I get nervous when people start making inflated claims based on shaky statistics …

    Trenberth has the right idea when he says:

    Rather than accept the hypothesis, one may be better off reserving judgment

    w.

  248. John S
    Posted Dec 31, 2006 at 4:52 AM | Permalink

    Re #245, #246 et al.

    See my comment at #35.

    The data are so noisy and the power of our tests sufficiently low (at least those tests so far used) that the assumption of H0 determines the outcome. When you have assumptions determining the result it is not really empirical science – faith may be too pejorative a word for some in this context – but whatever it is, it is not empirical science because it is not falsifiable. (I will allow for assumptions to determine results in the queen of sciences but otherwise you better have a really good explanation.)

  249. Ken Fritsch
    Posted Dec 31, 2006 at 1:35 PM | Permalink

    Re: #243

    Not a site administrator but a poster needing to determine whether I have image posting finally correct.

  250. Ken Fritsch
    Posted Dec 31, 2006 at 1:37 PM | Permalink

    I used the 800px/600px code???

  251. J. Curry
    Posted Dec 31, 2006 at 2:09 PM | Permalink

    Willis,

    We have put forward these ideas as hypotheses. I hope I made it clear in earlier posts that there was insufficient evidence to accept the hypotheses and subhypotheses surrounding AGW and TC activity. The hypotheses have not been rejected, they are still being used working hypotheses for further scientific investigation. In trenberth’s terminology, the scientists are reserving judgment on this one.

    The confounding factor is your concern about spending millions on catastrophic climate change. There is climate science on one hand, and economics on the other hand. Both areas are fraught with uncertainty. Risk management is about dealing with uncertain risks. I don’t want to distract this thread on that topic, but this issue seems to get in the way of trying to do science here on the topic of explaining TC variability on multidecadal time scales . We should be trying to provide a range of future projections of TC activity that can be used in risk management. If we simply say the data isn’t good enough to do anything with, we should ignore the whole subject, and close our eyes and cross our fingers and hope for the best, we might be in for some very unpleasant surprises in coming decades.

    So the challenge we should be addressing is to try to develop an understanding, or several different versions of understanding, that can help us project some future scenarios. the fact that advocacy groups use the scientific research to promote agendas (two diametrically opposite agendas are being pushed by different advocacy groups) is politics. We still need to do the best job we can on the science, since this is an issue that people clearly care about. Roger tried to get people to project TCs for 2100 (in one of his papers); I am trying to do something more modest and look at NATL out to 2025 (i felt the uncertainties going out to 2100 were just too large). An ensemble of projections could be used by risk managers to bound the future risk so that we can deal with the issues sensibly

    Scientists working with statisticians can figure out how to make the projections and the uncertainty to assign to these projections. I’m hoping that is what we can try to do here.

  252. Willis Eschenbach
    Posted Dec 31, 2006 at 6:18 PM | Permalink

    Judith, you say:

    We should be trying to provide a range of future projections of TC activity that can be used in risk management. If we simply say the data isn’t good enough to do anything with, we should ignore the whole subject, and close our eyes and cross our fingers and hope for the best, we might be in for some very unpleasant surprises in coming decades.

    I do not see the connection between poor data on the one hand and ignoring the subject, closing our eyes, and hoping for the best on the other hand. One does not imply the other. At times, the best and most useful sentence in the world is “we don’t know”. Let me see if I can summarize the situation as I see it, one of the beauties of this blog is that plenty of people will let me know if any of these are wrong.

    “‚⠠ Global climate is a tera-watt scale heat engine, with an energy source at about 345 W/m2 and an energy sink at about 0°K.

    “‚⠠ Thunderstorms, squall lines, frontal systems, tropical cyclones, and hurricanes are all local heat engines which move heat from the surface to the upper atmosphere. As such, they are driven by some form of àƒ⣃ ‹’€ ”‚¬⟔, the temperature difference between the surface and the upper atmosphere. They form a parasitic loss to the global heat engine.

    “‚⠠ In a warming world, this àƒ⣃ ‹’€ ”‚¬⟔ will increase. However, and this is a very important point, it will not increase as fast as àƒ⣃ ‹’€ ”‚¬⟓, the change in the surface temperature. This is because both the atmosphere and the surface are warming.

    “‚⠠ In a warming world, the sum total of parasitic heat loss through all possible mechanisms (e.g. local climate heat engines as above, ordinary convection, evapotranspiration, hydrometeors, wind driven evaporation, Hadley cell circulation, ENSO events, atmospheric turbulence, mobile polar highs, changes in vegetation, etc.) will increase.

    “‚⠠ We have very little information on how this increased parasitic heat loss will be apportioned among the various mechanisms. Nor is there any reason to assume that the various proportions will be the same year-to-year.

    “‚⠠ Near as we can tell, the world has been warming over the last century, yet globally the number of hurricanes hasn’t changed much. The number of hurricanes in the Atlantic has increased. We don’t know why either of those might be the case.

    “‚⠠ While theoretical calculations indicate stronger hurricanes in a warmer world, observations find less of an effect than predicted. We don’t know why.

    “‚⠠ Near as we can tell, there is only a very weak correlation between landfalling named storms and hurricanes and sea surface temperature in the Atlantic. (ACR temp vs landfalling named storms 1851-2006, r^2 = 0.07, p=0.03. ACR vs landfalling hurricanes, r^2 = 0.03, p=0.05. Maximum 50-year trailing correlation r^2 = 0.14)

    Now, should this confusing situation mean that we should “close our eyes and cross our fingers and hope for the best”? By no means. But it doesn’t mean that we should pretend to know more than we do. Given our current knowledge, we cannot “provide a range of future projections of TC activity that can be used in risk management” in anything more than the most general sense:

    1. If the world warms in the future, we will likely see a slight increase in Atlantic hurricanes. On average, they will be slightly stronger. On average, there will be very little change in landfalling cyclones or hurricanes.

    2. If the world cools in the future, we will likely see a slight decrease in Atlantic hurricanes. On average, they will be slightly weaker. On average, there will be very little change in landfalling cyclones or hurricanes.

    3. For a “range of future predictions”, see 2005-2006.

    But this doesn’t mean a lot, because In either case, ALL THE NORMAL PRECAUTIONS APPLY. Don’t build in flood-prone areas, or if you do, take the normal precautions. Don’t build in hurricane prone areas, or if you do, take the normal precautions. Don’t build in areas exposed to extreme winds, and if you do, take the normal precautions.

    The mathematics for occasional extreme events are curious. Suppose I live in an area where a hurricane strikes occasionally. Being a prudent person, I figure out the highest wind I expect to encounter over the next hundred years, add a safety margin, and build to those specifications. I figure the house may be hit by a hurricane say six times in a hundred years. I know that there is a statistically small chance, let’s call it 5%, that sometime in the next hundred years a bigger wind may come along and, like the wolf in the story of the Three Little Pigs, “blow my house down”.

    Now, there is only an extremely weak correlation between sea temperature and land-falling hurricanes or cyclones in the Atlantic, with less than 10% of the variation explained by SST. Suppose somehow we could see into the future and we knew for a fact that the world would warm by a degree over the next century. Now, instead of expecting six hurricanes over a hundred years, I expect 6.15 hurricanes, and instead of having a 5% chance that my house will blow down, it’s now a 7% chance …

    What difference will it make?

    This is a much wider question than just hurricanes, because all of the doomsday evils foretold by the climate alarmists are with us today. Droughts? We have extra, no shortage there. Floods? Check with Bangladesh, they might have some to spare. Hurricanes? See 2005. Rising see levels? Go talk to Holland. Disease? Chat with some Africans. Hot spells? Yep, aplenty.

    Will any given one of these increase in the future? Quite possibly, but the ugly truth is, we don’t know.

    Should we just “hope for the best”? Nonsense. Should we worry about the year 2100? We don’t have enough knowledge to do that, and besides, we have pressing problems and limited money.

    The best thing we can do about future climate problems, our best defense, is to deal with the current climate problems. The more we know about solving current droughts, the better we will be able to deal with future droughts … and the same is true about hurricanes, floods, and all of the vagaries of weather.

    w.

  253. Dave Dardinger
    Posted Dec 31, 2006 at 7:15 PM | Permalink

    Excellent post, Willis. I hope everyone here will make it a New Year’s resolution to take it to heart and work to make it the operational principle for the new year.

    The best thing we can do about future climate problems, our best defense, is to deal with the current climate problems.

  254. welikerocks
    Posted Jan 1, 2007 at 9:11 AM | Permalink

    Willis,

    Thank you for your excellent post. “you rock!” which means the same thing. You have a gift of diplomacy some of us lack, among other gifts. ;) It is nice to read something calm and reasonable for the New Year.

    Dane, my husband thanks you as well (he’s off on a Pacific Coast Hwy trip with daughter-ending in Santa Barbara, CA. There are epic winter waves going off and she got in the line up; and caught a wave at Rincon with an 8ft swell; just like her dad did in the olden days yesterday-Yay! a dose of joy provided by the same dynamic planet ) :)

    #253 Agreed.
    The past is history, the future is a mystery, all we have is right now. Happy New Year!

  255. Brooks Hurd
    Posted Jan 1, 2007 at 11:16 AM | Permalink

    Re: 121

    Judith,

    I agree with your statement, lets keep the statistics “simple” unless the problem demands something more complex in which case the methodology should be clearly outlined and justified. However there is no standard definition of what is standard vs arcane.

    I apologize about the delay in replying to you.

    Yes, I agree with you that there are no accepted definitions of standard and arcance statistics. My belief is that authors have a duty to their readers to either use terms and methods which are fimiliar to their target audience or provide sufficient information in their papers to explain what they did and why they did it.

    I have concerns along with many other posters on this Blog about how errors are handled by some authors. Empirical data is seldom of the quality that we would like. I believe that all authors owe their readers a discussion of the errors in their data. All data are based on measurements which have accuracy and precision errors.

  256. Ken Fritsch
    Posted Jan 1, 2007 at 12:17 PM | Permalink

    Re: #252

    While theoretical calculations indicate stronger hurricanes in a warmer world, observations find less of an effect than predicted. We don’t know why.

    Willis nice summary in this comment, however, I was of the view that the theoretical calculations and models would predict a sufficiently small change in wind speeds to make it virtually undetected at this point and that some researchers in the field, such as Emanuel and Mann, believe that they can show a significantly larger effect than the predicted one.

  257. J. Curry
    Posted Jan 1, 2007 at 1:03 PM | Permalink

    Re #253, 256: Yes, the debate is that the observed intensity increase for 0.5C SST increase (Emanuel, Webster) is a factor of 2-3 greater than expected from theory and model simulations.

    Willis, your summary is nice. However, there are some things that we DO know:

    “‚⠠ global temperature is warming
    “‚⠠ anthropgogenic CO2 is making some contribution to that warming (the magnitude of the contribution is debated)
    “‚⠠ everything that we know about hurricanes (observations, theory, model simulations) suggests an increase of hurricane intensity with increasing SST (the magnitude of the increase is debated)
    “‚⠠ While there is some observational and modelling evidence linking an increase in the number of NATL TCs with increasing SST, there is no evidence of such a link in other ocean basins

    So we actually “know” quite a lot about the science (but of course far from everything that one might like to know about this subject). How a projection of an additional 0.5C SST increase in the next 20 years, and likely greater increases after that, translates into risk management of U.S. coastal regions is a complex interplay between federal, state/local, insurance industry policies as well as the individuals values and tolerance for risk.

    Should government, insurance industry, and individuals ignore the risk of elevated hurricane activity in the coming decades? No. But they may decide for a variety of reasons not to do anything about the risk, other than to gamble and pay the costs if/when they accrue. Or they may implement various strategies to manage the risk. But denying that the risk exists would not be wise.

    Whether you are a proponent of AGW or AMO or a combination of the two, all agree that we are headed for several decades of elevated TC activity, at least as bad as the 1950′s (or possibly worse if AGW is a factor). Prior to the 1980′s, the gulf coast was relatively undeveloped, largely owing to the hurricane issue. The hurricane lull of the 1970′s and 1980′s encouraged the coastal development. The development really took off in the 1980s, after hurricane frederic (a big one), which is politically significant since it was the first hurricane that received federal disaster aid (it also occurred a few months after FEMA was formed). The combination of the disaster relief plus 15 prior years of low hurricane activity, gave the green light for big coastal development. At this point, the government (federal and state) is subsidizing coastal development in hurricane prone areas through flood and disaster insurance, regulation and subsidization of insurance, etc. We are paying the bill through our taxes and higher insurance rates. So the issue of elevated hurricane activity projected for the next few decades (and likely beyond) does have serious economic consequences.

    Florida has been clobbered in the last 11 years, with about half of the U.S. landfalls. in 2004, apparently 20% of all homes in florida were damaged from the landfalling hurricanes. The concentration of population and wealth in particularly in florida and the gulf coast, combined with increasing TC activity in the coming decades (AMO, AGW, or both), adds up to huge vulnerability.

  258. Steve McIntyre
    Posted Jan 1, 2007 at 1:33 PM | Permalink

    Florida has been clobbered in the last 11 years, with about half of the U.S. landfalls. in 2004, apparently 20% of all homes in florida were damaged from the landfalling hurricanes. The concentration of population and wealth in particularly in florida and the gulf coast, combined with increasing TC activity in the coming decades (AMO, AGW, or both), adds up to huge vulnerability.

    No one’s arguing this point. What’s this got to do with statistics?

  259. Posted Jan 1, 2007 at 3:27 PM | Permalink

    When the Holland and Webster paper was first discussed at Prometheus, I noticed some of the comments centered around the issue of whether one could base conclusions about changes in hurricane frequency using data prior to 1945. Landsea (2006) appears to believe that data prior to systematic air reconnaissance may have under counted. However, Holland and Webster believe the total number of storms must have been counted correctly. Moreover, Holland and Webster appear to base most of their on data comparisons that require one to compare pre-1945 and post-1945 storm counts as if there is no bias in the count.

    I performed a quick fiduciary check to compare the sst vs storm count data prior to 1995 and after 1995, and I believe it’s clear that storms were systematically undercounted prior to the advent of air reconnaissance. The bias is approximately equal to 15%-20% of hurricanes (or roughly 1.6 hurricanes/ year.) Because pre-1945 data appears in every single figure in Holland & Webster, and used to support every single argument in that paper, I believe one should view all the conclusions in Holland and Webster with some suspicious. (This, of course also affects the conclusion of any analysis that assumes there is no significant bias in storm counts prior to 1945. )

    In many ways the analysis is captured by the infomation in this figure:

    (Hopefully this shows!)

    The difference in the intercept of these two graphs is statistically significant, suggesting hurricanes were undercounted prior to 1945.

    More details, with text that should be relatively accessible to those who were required to take some statistics as undergraduates is available here.

  260. J. Curry
    Posted Jan 1, 2007 at 6:26 PM | Permalink

    Margo, your analysis is very interesting. This is a good example of one of those consistency tests I mentioned before (but one i hadn’t thought of). Several issues (perhaps fodder for further analysis)

    1. If TCs are undercounted, I would expect something like this
    1944-1960: perhaps some small undercounting, but maybe just random errors where some TCs were missed and some tropical depressions were misclassified as TCs (perhaps undercounting of 0.5 TC
    1910-1944: some undercounting, probably with more in 1910 than 1944 (say 1 TC average undercount)
    1870-1910: more undercounting (say 1-2)
    before 1870: undercounting by 2-3
    Maybe try chopping this into 3 periods, with a cutoff say 1910

    2. The average SST for your two periods is different (avg SST is higher 1944-2004). the “true” relationship between TC and SST may not be linear, there may be thresholds (or “regimes”). To test strictly for undercounting, it may be best to lop off some of the recent years, enough to make the two periods have the same mean SST. Then check to see if the difference in intercepts is significantly different

  261. Steve Bloom
    Posted Jan 1, 2007 at 7:22 PM | Permalink

    Re #252: Willis, don’t forget that rising SSTs alone will allow cat 4/5 hurricanes to range much farther up the East Coast. Long-term, this oft-forgot problem has (IMHO) the largest practical implications; e.g., worst-case scenarios such as a cat 5 strike on Chesapeake Bay or a cat 4 strike on New York Harbor. On the West Coast, major hurricane strikes into the L.A. area might become possible.

    Re #253: If I were an actuary at a reinsurance company, I’d have to say quite a bit!

  262. Posted Jan 1, 2007 at 8:07 PM | Permalink

    Judy,

    1. I am absolutely astonished someone (like you), who has been dealing with this storm data so long had not thought to do a t-test to compare the difference in the mean value of storms predicted at some constant temperature SST. Applying T-tests to determine confidence in the slope and intercept is the sort of thing undergraduates in engineering are expected to do in laboratory courses.

    2. I am aware the data for pre-1945 and post-1945 have different mean in storm counts. I set up the t-test with this in mind and performed what I believe is the appropriate comparison. (Had I made the mistake of comparing the two means, I would have come up with a confidence interval that is truly minuscule!)

    3. Your speculation that quadratic terms may be important is “interesting”. You could, test for the quadratic term. It is straight forward and can be done with EXCEL. If you do, I suggest you begin by simply fitting a quadratic to the full data set. Then remember to test the coefficients on the quadratic term to learn whether they are statistically significant. Then come back and tell me what you find. ;) (Note: I’ve already done this.)

    4. In light of the fact that the hypothesis that there is “no change” is categorically excluded by a simple t-test, I should think those who wish to insist there is no change (which seem to include you or Holland and Webster ) have an uphill battle here. However, if you wish, you may of course, chop the data into any size chunks you wish. (Afterwards, you can try to explain why it makes sense to chop up a very noisy signal into smaller and smaller bits– and in some cases a batch of with as few as 16-17 data pairs.)

    5. It is not a question of “if”. Unless there is a change in physics, storms were undercounted. If you wish to tease out precisely how many were undercounted in each of various eras, knock yourself out. Meanwhile, this impacts the conclusions in the Holland and Webster paper which include somethng like 7 figures with data going back as far as 1855.

  263. Armand MacMurray
    Posted Jan 1, 2007 at 9:46 PM | Permalink

    Re:#263
    Based on her posts, I would think that Dr. Curry is deserving of a more polite tone (and such would encourage her to continue her useful engagement here).

  264. Posted Jan 1, 2007 at 10:24 PM | Permalink

    Armand,

    You may be correct about tone. I didn’t mean to be snippy. If Judy is suggesting she will have her research team and frequent collaborators to consider these sorts checks, I would highly recommend that. I apologize for failing to interpret her comments in the manner she intended.

    However, if should they attempt to do these analyses, I would caution that trying to come up with “corrections” using the short 16 sample period between 1944 and 1960 as the benchmark would be dubious. I would also point out that trying to tease out any information about the quadratic term will result in the conclusion that the quadratic term is not significant either statistically or in magnitude. (Besides that, she’ll also find that if she really wants to keep the quadratic term, it suggests the bias is worse than suggested by a simple linear analysis.)

  265. Steve McIntyre
    Posted Jan 1, 2007 at 10:34 PM | Permalink

    #260. Margo, I think that your analysis here is quite convincing.

  266. Willis Eschenbach
    Posted Jan 1, 2007 at 11:40 PM | Permalink

    Judith, as always, an excellent post. As you point out, the main issue is not changes in cyclones/hurricanes (C/H), it is changes in landfalling cyclones/hurricanes, because of the damage that they cause.

    Whether you are a proponent of AGW or AMO or a combination of the two, all agree that we are headed for several decades of elevated TC activity, at least as bad as the 1950′s (or possibly worse if AGW is a factor). Prior to the 1980′s, the gulf coast was relatively undeveloped, largely owing to the hurricane issue. The hurricane lull of the 1970′s and 1980′s encouraged the coastal development. The development really took off in the 1980s, after hurricane frederic (a big one), which is politically significant since it was the first hurricane that received federal disaster aid (it also occurred a few months after FEMA was formed). The combination of the disaster relief plus 15 prior years of low hurricane activity, gave the green light for big coastal development. At this point, the government (federal and state) is subsidizing coastal development in hurricane prone areas through flood and disaster insurance, regulation and subsidization of insurance, etc. We are paying the bill through our taxes and higher insurance rates. So the issue of elevated hurricane activity projected for the next few decades (and likely beyond) does have serious economic consequences.

    What I don’t understand is your claim that landfalling C/H increase with temperature. While there is a clear relationship between C/H and temperature in the Atlantic, there is no such relationship between landfalling C/H and temperature. According to your data, the relationship is (all references in this post are to landfalling C/H.

    ACR Temp/Landfalling tropical cyclones

    1851-2006 r^2 = 0.07, trend = 0.05 TCs/°C, p = 0.05
    1851-1944 r^2 = 0.08, trend = 0.06 TCs/°C, p = 0.04
    1945-2006 r^2 = 0.07, trend = 0.05 TCs/°C, p = 0.13

    ACR Temp/Landfalling hurricanes

    1851-2006 r^2 = 0.03, trend = 0.05 TCs/°C, p = 0.05
    1851-1944 r^2 = 0.05, trend = 0.05 TCs/°C, p = 0.05
    1945-2006 r^2 = 0.06, trend = 0.06 TCs/°C, p = 0.14

    This is an increase of 1/20th of a landfalling cyclone/hurricane for each 1° rise in ACR temperature. Note also that the relationship is no longer statistically significant in the recent era when we have better data.

    Finally, while there is relationship between the AMO and landfalling C/H, it is not statistically significant (p greater than 0.1 1856-2006, greater than 0.5 1945-2006). However, since we need to make decisions based on uncertainty, let’s ignore that. The trend between the AMO and TCs is 4.6 TCs/°C, and the trend between the AMO and hurricanes is 2.6 hurricanes/°C. Currently, the AMO is at about +0.2, which translates to 0.9 extra TC/year and 0.5 extra hurricane per year.

    Thus, I am totally at a loss why this small increase is even an issue at all. You keep saying that if the temperature goes up, so will the number of landfalling C/H, but there is no such trend in the data. Also, the AMO is probably within a year or so of peaking, but even if it is not, the change is likely to be small.

    If I were asked about the possible C/H numbers in the coming years, I’d say “somewhere between the numbers for 2005 and 2006, with a most probable value about a quarter of the way up, call it 6 landfalling TCs and 3 landfalling hurricanes. I don’t think that the data supports a more detailed forecast than that. And if I were in charge of planning for Florida, I’d plan for another year like 2005 … might not happen, but you’d have to plan for it. Doesn’t seem like science can help us much in this case, history is a better guide.

    Many thanks, best of the coming year to you,

    w.

  267. Posted Jan 2, 2007 at 1:27 AM | Permalink

    #249 Thanks, Geocities seem to handle the traffic now!

    Bottom left is interesting example. Sample correlation changes from 0 to 0.3 just by increasing the filter length from 10 to 20. And the real correlation is zero in all those cases, of course. So, be careful with arbitrary filter lengths. (I’m not claiming that HW have chosen the filter length to maximize the correlation, just general note on filtered correlations)

    BTW, is the raw data for Figure 1 easily available somewhere?

  268. Willis Eschenbach
    Posted Jan 2, 2007 at 1:47 AM | Permalink

    Re 262, Steve B., thanks for the insightful comment. You say:

    Willis, don’t forget that rising SSTs alone will allow cat 4/5 hurricanes to range much farther up the East Coast. Long-term, this oft-forgot problem has (IMHO) the largest practical implications; e.g., worst-case scenarios such as a cat 5 strike on Chesapeake Bay or a cat 4 strike on New York Harbor. On the West Coast, major hurricane strikes into the L.A. area might become possible.

    An interesting question. I suspect that I can put some numbers on this. Let me see … back in a while …

    OK, thanks for waiting, it took three hours. Here’s a graph. I don’t have sea temps under the hurricanes, so I have compared wind speed with average sea temperature by latitude, by taking the latitude of each 6 hour observation and comparing it to the average sea temperature for that latitude. This represents all hurricanes, 1851-2006.

    Now, the question is, if the sea temperature at 40°N rises, will there be more cat 4/5 hurricanes? As you can see, at the average temperature of NYC, cat 4 hurricanes are rare, and cat 5 nonexistent. Remember, each blue dot represents only 6 hours.

    There’s another way to look at it. We can graph the winds by latitude, and then take the slope of the latitude/temperature line and adjust the latitude of New York to see what the equivalent latitude would be. At the latitude of New York, the ocean temperature is rising about 1.2°C for each degree of latitude you go south. Here’s the chart:

    Again, we find that there’s not much change. A 1° change in temperature only makes NYC the equivalent of about 1° further south, so there’s no real difference in terms of risk.

    So the conclusion is, although hurricanes do go north, the stronger ones require warmer water, and the slope of the temperature vs latitude line is steep enough to make even a 1° temperature rise not very significant in the northern latitudes. In other words, it is not true that “rising SSTs alone will allow cat 4/5 hurricanes to range much farther up the East Coast”. At the latitude of NYC it will only change the range by about one degree of latitude per degree of warming.

    My appreciation for raising an interesting issue,

    w.

  269. Tim Ball
    Posted Jan 2, 2007 at 1:49 AM | Permalink

    In the list of things we DO know we read about the human contribution of CO2 adding to the warming. I would like to see the evidence that human contribution of CO2 is contributing to the warming. I do not want to hear about any computer output because inputs are completely inadequate to reach any conclusions. A quick look at the range of estimates for all sources of inputs of CO2 show how it is impossible to determine whether human CO2 is contributing to global temperature. I suppose you can argue that theoretically the human addition should make a contribution, but proving it and saying we DO know is a completely different matter. I suppose if I light a match in my basement in Northern Canada in winter I might argue it is adding to the heat, but measuring it and determining its significance is an entirely different matter.

  270. Steve Bloom
    Posted Jan 2, 2007 at 3:40 AM | Permalink

    Re #269: Willis, how did you figure those SST numbers? I ask that because for these example we would be interested in SSTs running northeasterly from the Gulf Stream into NYC and Chesapeake Bay. Also, perhaps I’m not reading the graphs right but it looks as if you’ve lumped NYC and Chesapeake Bay together.

  271. Steve Bloom
    Posted Jan 2, 2007 at 3:44 AM | Permalink

    Re #271: “northerly”

  272. David Smith
    Posted Jan 2, 2007 at 11:33 AM | Permalink

    Data on SST versus storm strength, using the SST beneath each storm, is given in Figures 3 and 4 here .

    Figure 4 pretty well shows that, to achieve category 4, a SST of 28C is needed.

    Summertime ocean off of Virginia runs about 23C, I believe. Here is a global SST chart for September, 1996, which was the best I can find at the moment and it should be similar to today.

    So, about 5C of warming would be necessary for a credible threat of cat 4 storms to exist.

    The Gulf Stream is clearly shown offshore, with warmer water. It’s possible that a category 4 storm could stay strong until leaving the Gulf Stream, and if it moves quickly it may not spin down much before getting to Cheasapeake Bay, but it will definitely weaken.

    The key is how fast the storm moves.

  273. Judith Curry
    Posted Jan 2, 2007 at 3:59 PM | Permalink

    WIllis, again a very thoughtful post. A few comments. Total NATL TCs are important in trying to understand the causal mechanisms, while the landfalls are associated with socioeconomic impacts. So both are worthy of our attention.
    As per a previous post somewhere, I stated that no one thinks that we are within a year or so of the AMO peak. People estimate that we are 10-40 years away from the peak or a downturn (this includes everyone from hurricane forecasters to oceanographers). 1/20th, or 1 additional storm on average 20 years from now is a pretty big deal. This would bring the average landfalling TC count to 6.5 (which would be unprecedented and a big deal). My earlier 2025 “prediction” was for 8 landfalling TCs, which is what we saw in 2004/2005 (this assumes that we are not at the AMO peak, that the analogue for the last decade is the 1930s). I agree that people should be planning for what we saw in 2004/2005. They aren’t (other than the insurance companies). The Army Corps of Engineers are planning for a cat 3 in terms of the guidelines for rebuilding the NOLA levees.

  274. Judith Curry
    Posted Jan 2, 2007 at 4:02 PM | Permalink

    Willis and David, I agree that anything greater than cat 3 striking NYC or vicinity is pretty unlikely in the current or slightly warmer climate. At cat 3 would be pretty catastrophic tho.

  275. Steve Bloom
    Posted Jan 2, 2007 at 5:33 PM | Permalink

    Re #273: David, as you point out, and as we saw with the effect of the loop current on Katrina, it is the SSTs experienced by the storm that seems to be most important. Because of that, the monthly averages are a little misleading. (I should also note that it’s not strictly SSTs that storms like, but warm water at some depth; in that sense SSTs are really just a proxy, but probably a reasonable one for this discussion.) As a comparison, here’s the monthly for August 1998 (probably close to the record warm month, so probably a better choice since we’re talking about the effects of future warming). Generally there’s a lot more warm water (contrasted to the graphic you linked) up to Cape Hatteras, and then it falls off rather rapidly. By contrast, this is the August 4, 1998 anomaly map. (Also see this map for August 6, 2005, which is even more interesting; the northern Atlantic coast appears much warmer than 1998.) I know these anomaly maps aren’t strictly comparable, but the point is that the monthly can mask shorter-term conditions much more favorable to a hurricane maintaining strength into a more northerly landfall. As you say, speed is important for such landfalls now and that will remain true (although perhaps much less so for Chesapeake Bay than for New York Harbor).

    All of that said, I don’t know if there’s any better approach to this than the one Willis took. For the near future, all else being equal major hurricanes can be expected to be a little more common in more northerly locations. Long-term, OTOH, a 3C increase overlaid on existing Atlantic coast sea temperatures will be a different story (depending on storm frequency and landfalling locations, both of which may change in unpredictable ways).

  276. Steve Sadlov
    Posted Jan 2, 2007 at 6:09 PM | Permalink

    RE: “On the West Coast, major hurricane strikes into the L.A. area might become possible.”

    No way. Not possible with current ocean basin shapes and current tectonic plate configurations. Meaning, you can’t fight the ocean currents that result from large slow convection cells interacting with the Coriolis Force interacting with land masses. The California Current will remain a cold current, there will not be tropical waters off of any place north of mid Baja.

  277. george h.
    Posted Jan 2, 2007 at 7:46 PM | Permalink

    “we are lead to the confident conclusion that the recent upsurge in tropical cyclone frequency is due in part to greenhouse warming and this is most likely the dominant effect”

    What bothers me most about this paper is not the statistics, but its clear reliance on assumptions about greenhouse warming which are just taken for granted by the authors: 1) There is good accurate global temp data which shows unprecedented warming which can be compared to hurricane counts, 2) there is a direct relationship between concentrations CO2 and global temp which is resulting in greenhouse warming. 3) If warming is increasing hurricane frequency or intensity, it is the result of anthropogenic C02 emissions. These assumptions are at best unproven and at worst demonstrably false. How can one possibly compare TC activity over a long enough period to ascertain a trend without confidence in the GMT data over the same period. There is a huge UHI, land use and selection bias in the temp data between 1880 and the present. In city after city in the US where population and land use have remained constant, there has been no change in average temperatures in the last 120 years. The measurement error alone may exceed the entire average temp change over 100 years. Assuming that the IPCC 0.6 °C ± 0.2 °C Twentieth Century warming is accurate, there is nothing inconsistent with what is known about previous interglacial periods. This planet is still coming out of a 90,000 year glacial period and warming during the recent 10,000 to 15,000 years is a natural result. Second, there is no historic or geologic evidence to suggest that CO2 plays any significant role in forcing. In Miocene times the climate was about 10°F warmer and CO2 concentration was less than today. During the Pleistocene, CO2 increased and was accompanied by global cooling. Between 1940 and 1970 when CO2 concentrations were on a rapid upslope, surface temps were declining, not increasing. Ice core data shows carbon dioxide rises lag global temperature increases. The authors may have demonstrated a relationship between TC frequency and warming, I don’t know, but they did not show one to greenhouse warming.

  278. Steve Bloom
    Posted Jan 2, 2007 at 9:46 PM | Permalink

    Re #277: You’re likely right about the California Current, but bear in mind that LA was hit by a weakish storm (probable TS) in the 1930s and an out-and-out hurricane in the 1830s. Tropical storms can survive over colder water for considerable distances, or else Canada wouldn’t have to worry about them.

  279. chrisl
    Posted Jan 3, 2007 at 12:09 AM | Permalink

    george h : some excellent points you make
    It seems that there are three time frames in climate science
    1. The geological record
    2. The data record
    3. The news cycle

  280. Willis Eschenbach
    Posted Jan 3, 2007 at 1:48 AM | Permalink

    Steve B, sorry, I missed this one. You ask:

    Re #269: Willis, how did you figure those SST numbers? I ask that because for these example we would be interested in SSTs running northeasterly from the Gulf Stream into NYC and Chesapeake Bay. Also, perhaps I’m not reading the graphs right but it looks as if you’ve lumped NYC and Chesapeake Bay together.

    The SST numbers are the zonal averages for July-Sept for 80°-50°W, from 0 to 55°N in 5° increments.

    Chesapeake Bay stretches from about 37 – 39N (Washington, DC is 38.5), NYC is about 40N. They were close enough for the purposes of my analysis that I just put in NYC, you can see where Chesapeake Bay is on the graphs. At that latitude, one degree N/S translates to about 1°C water temp.

    All the best,

    w.

  281. Steve Bloom
    Posted Jan 3, 2007 at 4:08 AM | Permalink

    Re #281: So it sounds as if inshore SSTs might shift things a little farther north, but not too much. Thanks for the time you put into this.

    Re #180: chrisl, don’t forget:

    4. The giant hissy fit to be thrown on this blog beginning February 2nd.

  282. Posted Jan 3, 2007 at 6:50 AM | Permalink

    #27, TAC:

    Having said that, I have a hard time understanding why one would begin by taking a lag-9 moving average “¢’‚¬? which would typically introduce correlation out to lag 9 “¢’‚¬? and then try to fix things by using a correction for serial correlation (AR(1), or lag 1). Without the correction, the standard trend tests will be hugely biased in the direction of finding significance when it isn’t there. So the question is: Does the “correction for serial correlation” take care of this problem? It might, I suppose. Without seeing any equations, it is hard to say.

    Maybe it is some kind of optimal smoothing in Mann’s way. In MannGRL04 he warns against false conclusions based on unobjective statistical smoothing approaches. Maybe 9-year moving average is objective non-statistical smoothing approach.

    #194, Willis

    However, claims of correlations on averaged values must be treated with some skepticism. As two series are averaged, the correlation increases. The only meaningful number is the correlation between the datasets themselves. Sea temperature only explains about a quarter of the variance in Atlantic hurricane numbers (r^2=0.27).

    And r^2 is 0.60 after 9-year averaging, I presume. The correlation is indeed higher for the lower frequencies.. Hmmm.. Was there a correction for serial correlation somewhere?

  283. Steve McIntyre
    Posted Jan 3, 2007 at 8:39 AM | Permalink

    UC- I notice that about a year ago I mentioned the Slutsky effect in a post about Klemeà…⟠but never returned to an exposition of this. The Slutsky effect, well known in economics, is (briefly) that repeated averaging yields cyclical time series and high correlations between unrelated series. Klemeà…⟠observed that hydrological series already incorporate natural “integrations” through precipitation and through formation of bodies of water.

  284. Steve Sadlov
    Posted Jan 3, 2007 at 9:10 AM | Permalink

    RE: “Canada wouldn’t have to worry about them.”

    Apples and oranges – (Maritime) Canada is affected by the Gulf Stream and eddies thereof. Compare summer SSTs between British Columbia and Nova Scotia.

  285. Posted Jan 3, 2007 at 11:34 AM | Permalink

    #284

    Google says: Slutsky-Yule Theorem: moving average of a random series may generate oscillatory movement when no oscillations exist in the original data. Interesting.

    There was something related to variance of R somewhere .. lets see .. here:

    http://www.climateaudit.org/?p=903 comment 53,

    var\{ R_{12}(s) \} \sim \frac{1}{n-s}\sum _{-\infty} ^{\infty} p_{11}(v)p_{22}(v)

    I think this is relevant in here as well.

  286. Posted Jan 3, 2007 at 12:09 PM | Permalink

    Steve:

    The Slutsky effect is also well understood by those working in signal processing, instrument development fluid dynamics, and heat and mass transfer. To avoid introducing the anamolies associated with linear averaging, analysits presented with noisy data usually transform data into the frequency domain, filter out the high frequencies and then perform the inverse transform.

    This must be done with careful attention not to filter out physics or signal of interest.

  287. J Edwards
    Posted Jan 3, 2007 at 12:59 PM | Permalink

    RE #287,

    Margo, excellent point. Has anyone tried running a DFT or z-transform against the data to see what it looks like in the frequency domain?

  288. bender
    Posted Jan 3, 2007 at 4:16 PM | Permalink

    Re DFT: like this?

    http://www.climateaudit.org/?p=825

  289. Posted Jan 3, 2007 at 8:31 PM | Permalink

    J Edwards:

    On page 7 of their Nth draft of their manuscript, Holland and Webster tell us they examined the spectrum; they do not reveal the results to the reader. They tell us they detected peaks at 2-3 year, 5 year, 9 year and “multi-decadal periods”. Reading the text, I gather they applied the 9 year filter (and sometimes a 5 year filter and sometimes no filter at all) to remove the El Nino effect which, they tell us, has a 2-3 year time period.

    Since I did not apply a nine year filter to eliminate a 2-3 year frequency effect, I will not attempt to explain this data processing decision.

  290. David Smith
    Posted Jan 3, 2007 at 8:51 PM | Permalink

    bender, I see the words “Bayesian” and “hurricane” in several titles on this list . Caution: it is from the reviled FSU Seminoles website.

  291. Posted Jan 3, 2007 at 10:15 PM | Permalink

    Re: 277.

    San Diego was struck by a hurricane in 1858.

  292. bender
    Posted Jan 3, 2007 at 11:23 PM | Permalink

    Re: #291
    FSU junk science. (Kidding!)

    Over dispersion in coastal hurricane counts does a decent job fitting truncated possion distributions to hurricane counts.

    P.S. Congrats LSU. Impressive win over stinky ND.

  293. David Smith
    Posted Jan 4, 2007 at 4:24 AM | Permalink

    RE #293 My guys survived the Irish. Whew! Gators turn.

  294. Brooks Hurd
    Posted Jan 4, 2007 at 12:30 PM | Permalink

    Re: 294 & 293
    Watching the game on Monday with another Buckeye. I hope that OSU has not forgotten how to play like Michican did against USC. Did some of the Wolverines get stuck in DIA’s snow?

    LSU’s win over the Irish was decisive!

  295. Steve Sadlov
    Posted Jun 13, 2007 at 8:23 PM | Permalink

    RE: #292 – Not true. San Diego was affected by a border line TS / TD not a hurricane, in 1858. And 1858 was not the most recent instance, there have been others since. Sometimes the feeder bands of a dying hurricane wet down the dust even up here in NoCal. But unless the California Current reverses and changes from a cold, 56 deg F current into something a lot warmer, there will be no hurricanes hitting California, at least not Alta California.

One Trackback

  1. [...] 3. Climate Audit: New Holland and Webster Paper S. McIntyre. [...]

Follow

Get every new post delivered to your Inbox.

Join 3,381 other followers

%d bloggers like this: