Holland and Webster Jan 2007 PPT

Roger Pielke pointed out that Holland and Webster have presented a PPT presentation posted up at UCAR (the home of IPCC WG1), the content of which is relevant to recent discussions at climateaudit and prometheus. The entire presentation is about data problems relating to storm trends in the eastern Atlantic and to landfall hurricanes, topics discussed at length at our respective sites.

The presentation is entitled “Heightened Tropical Cyclone Activity in the North Atlantic: Natural Variability or Climate Trend?” It asks: “Where have the Increases Occurred?” and reports:

Increases have occurred in all regions except the western Caribbean and southern Gulf of Mexico, but the largest proportional increases have been in the eastern Atlantic.

They ask: “Can Eastern Atlantic trends be due largely to data errors?” and answer:

There is little doubt that the eastern region was poorly observed in earlier years, but there are other factors; several studies have shown independent relationships, eg:

  • Kossin and Vitmer (2007) have shown that there is a close relationship between positive phases of the AMM and the eastward extension of the genesis region; and
  • Mann and Emanuel (2006) have shown that the AMM is potentially increasing because of Greenhouse Warming;
  • We have shown a strong and stable relationship between frequency and SSTs;

They go on to ask: Can Landfalling Storms be used to infer Missing Oceanic Data? noting that:

The high proportion of landfalling storms in the pre-satellite era has been used to imply poor observations of cyclones in the eastern Atlantic (e.g. Solow and Moore, 2002, Landsea 2006).

They then propose a boundary line between east and west Atlantic about which they state:

We assume that all cyclones that cross to the left of the black line were highly likely to have been observed in some form.

Note that this says nothing about how accurate the genesis points are or the observed intensity, just whether they were known to have existed. …

Using the proximity to land and ship route approach implies that almost all storms were within a good observing area for some part of their life. Note the steady increase of non-landfall storms, implying that this has a real component.

They conclude the following:

Maximum Potential Error Prior to 1960 is < ~1 storm per year (or around 6% prior to 1960), which agrees with the assessment by Neumann (1999).

Most of the potential error impact is on tropical storms: ~0.5 per year (10%) prior to 1945;
Minor hurricanes: ~0% error;
Major hurricanes: <10% per year.

Their conclusions:

  • The current Atlantic data base indicates a substantial trend in tropical cyclones and hurricanes and both a multidecadal oscillation and trend in major hurricanes;
  • The trend is strongly related to Atlantic SST and thus to Greenhouse Warming;
  • This trend is robust to all known potential problems with the data base,

At some point, I’ll try to wade through what they’ve done. As a quick comment, you’ll notice that they keep repeating that everything is strongly significant and that much of the rhetorical force derives from repetition of the comment, rather like an incantation. Likewise, the incantation that the trend is “strongly related to Atlantic SST and thus to Greenhouse Warming” – dare I say stronnnnngggg-ly related – whether or not the statement is true – is not actually demonstrated in the presentation, merely asserted.


  1. Steve McIntyre
    Posted Jan 18, 2007 at 10:17 PM | Permalink

    David Smith writes: (Transfer)

    RE #6 Steve M, you may want to consider #6 for a new thread.

    I’m glad HW is looking at the data they use.

    It will take me a while, and some red wine, to understand the details of their presentation. In advance of that, here is some data for your consideration.

    Below are the counts of tropical cyclones which did not come within 100km of land, grouped in 10-year buckets.

    Plot these and see if a familiar shape emerges.

    1905-1914: 12 tropical cyclones
    1915-1924: 11
    1925-1934: 14
    1935-1944: 12
    1945-1954: 24
    1955-1964: 21
    1965-1974: 31 (40)
    1975-1984: 41 (49)
    1985-1994: 37 (38)
    1995-2004: 50 (53)

    (the number in parentheses include the non-tropical cyclones known as “subtropical storms”, in case HW still use subtropical storms in their cyclone counts)

    Since these always-at-sea tropical cyclones did not make landfall or brush close to land, they would have been detected by ship, aircraft or satellite (with a few exceptions). Consider that

    * Ships have been around, but are historically sparse outside the major commercial shipping lanes.

    * Aircraft patrolled part of the western Atlantic (Bermuda-Antilles-Miami triangle) from roughly 1948-1959, as best as I can tell, but not the vast regions of the eastern Atlantic.

    * Satellites became geosync, with nightime vision, in the 1970s, the timeframe at which their reliability for finding distant systems became robust (my opinion)

    My contention is that the hockeystick shape is mainly due to improvement in detection. Ship detection is not robust at all, aircraft was limited, early satellites were a big improvement and modern satellites a great improvement

    HW’s contention (I believe) is that the hockeystick is mostly real and due to AGW. (I scanned the presentation for a physical mechanism that would explain the near-steady increase in entirely-at-sea storm counts that didn’t follow the AMO while near-land storms followed the AMO, but saw none.)

    A remarkable thing is that, if one backs these entirely-at-sea storms from the total tropical cyclone counts, such that only the robust near-land storms are used, one gets a pretty oscillation (AMO?) that rises little over the last 100 years.

  2. Steve McIntyre
    Posted Jan 18, 2007 at 10:20 PM | Permalink

    Their assertion that:

    Increases have occurred in all regions except the western Caribbean and southern Gulf of Mexico, but the largest proportional increases have been in the eastern Atlantic.

    is simply false. Increases occurred only in the eastern Atlantic. I don’t know why they say such stupid things.

  3. Ken Fritsch
    Posted Jan 19, 2007 at 10:08 AM | Permalink

    It should, but does not, amaze me what some authors can continue to conclude even given that they are not required (and do not offer) to answer specific questions concerning the data analysis and particularly as it regards sensitivity testing.

  4. David Smith
    Posted Jan 19, 2007 at 11:03 AM | Permalink

    I just had my first full read of the presentation. It’s like watching someone drive from Toronto to Montreal by way of Ethiopia.

    I need a second, and probably a third and fourth read, to tie it together.

  5. Posted Jan 19, 2007 at 1:45 PM | Permalink

    I’m not sure it’s quite possible to tie the whole argument together based on the ppt slides.

    To be fair, an reader can rarely tie together arguments from slides. Slides generally contain the major points, but one would expect the authors to be providing detail during the talk. We may all need to wait for the next paper!

    At least it looks like they are getting closer to looking at things they should have examined before banging out the last one.

  6. Ken Fritsch
    Posted Jan 19, 2007 at 2:23 PM | Permalink

    At least it looks like they are getting closer to looking at things they should have examined before banging out the last one.

    When I see them spell out definitely and clearly that sensitivity tests indicate some significant dangers to conclusions based on using the data and/or at least adding much uncertainity to their results/conclusions, I will assume that they will continue to work backwards from a conclusion.

  7. Posted Jan 19, 2007 at 2:40 PM | Permalink

    I have to sort of agree with you on that . . .

    Looking at the slides, I had to ask: “Got error bands?” for some of their things. And “Why’d you pick the number of 6 hour observation bands in the western region?” (instead of something else).

    Still, I’ll wait until I see more in the form of some sort of manuscript!

  8. David Smith
    Posted Jan 19, 2007 at 4:06 PM | Permalink

    Margo, I agree with regards to Powerpoint presentations – they leave gaps, often big gaps. Nevertheless, enough threads of reasoning should be visible to see how the slides support the conclusions.

    My approach is to try to figure out their argument and restate it, in my words, here. That lays things out in public, where someone can challenge my interpretation (I’m glad to learn and make corrections if warranted). Usually it’s not so hard to figure out the reasoning, but this one is tough for me.

  9. David Smith
    Posted Jan 19, 2007 at 9:10 PM | Permalink

    Here’s my outline of the HW Powerpoint:

    * this is apparently a continuation of the earlier presentation, with a focus on data reliability. As Margo notes, it’s a Powerpoint, which has limitations.

    * pre-1900 data is discarded due to “analysis errors”

    * they see a meaningful ratio of hurricanes to all tropical cyclones over the last 50 years. Apparently the message is that, if the total number of storms increases, then the number of hurricanes will increase

    * they report that storms (genesis point) increased in the eastern and northern Atlantic in the second half of the 20’th century

    * they acknowledge that the eastern Atlantic was poorly observed pre-satellite

    * however, they report support (their work plus two papers)for a physical basis for higher activity in the eastern Atlantic

    * the percent of storms that hit land fell in the 20’th century, which some take to mean that there were unreported storms early and reporting became better later in the century

    *HW then do some kind of landfall analysis that loses me. They report that HURDAT data can’t be used to imply missing storms.

    * they draw a line in the middle of the Atlantic and say that ships and land on the left (west) side would likely detect storms. They say that almost all storms passed into this area and were therefore detected

    * they report that the storms that passed over this line into the “detected” region include most eastern Atlantic storms, which are therefore real even though they never hit land. The reported number of these for-real storms that never reach land steadily increases, meaning that there was a real increase.

    * they compare their “line in the Atlantic” to the HURDAT data to show that their line does something, apparently an improvement, but I lose their reasoning on this

    * they then use 90 years of data, split in two, to estimate an upper limit on missed storms

    * they then split the 20’th century data into three categories (ts, weak hurricane, major hurricane) and show the trends in both total and landfalling storms, and the trends in storms that stay remote in the eastern Atlantic

    * their summary states that tropical storms increased by 100-150% in the 20’th century; weak hurricanes by 100% and major hurricanes by 75-100%. These values are sort of their error bars and show the minimal effect of storm undercount.

  10. David Smith
    Posted Jan 19, 2007 at 10:18 PM | Permalink

    Where to start? Well, first, what physical mechanisms do they suggest to explain the key results?

    The key claim is that the upward trend in eastern Atlantic storms is mostly real, not data errors. So what could physically cause that? The underlying theme in HW is that the Atlantic storm frequency is increasing due to higher SST (“we have shown a strong and stable relationship between (Atlantic storm) frequency and SST”) so it seems reasonable to assume that SST must have risen in the area of interest (eastern Atlantic, 1950 – 2000) and driven the storminess increase. Has it?

    Well, no. The temperature trend in the eastern Atlantic (10-30N, 60-20W) was downward 1950-1975 then upward 1975-2000, returning to 1950s levels by 2000. Yet, the trend in HW’s eastern Atlantic storm count was strongly upwards beginning in 1950 (see slide 15, add the “non land” lines”), regardless of the cooling eastern Atlantic SST.

    HW states that, “We emphasize that the SST-TC relationship is not entirely direct…” and suggests something about the AMO and eastward extension of the genesis area and something about “other physical processes to support this regional aspect of the trend (to be reportd later)”.

    So, I have no clue as to the physical explanation they propose and why the Eastern Atlantic would run counter to their overall SST/storm count theme.


  11. Gerald Browning
    Posted Jan 19, 2007 at 10:19 PM | Permalink

    IMO any objective person would be very suspicious of an analysis from
    authors who had not considered the quality of the data that was so crucial to their previous claims, especially given that these data quality issues had to be raised by others in a open forum.

    Steve M.,

    Have you heard anything from Judith Curry lately?

    John A.,

    I am working offline on the tutorial post I referred to in the Solar Proxies thread. Any hope for LaTeX so I can write some mathematical expressions if needed? If that is an insurmountable problem, I will use plain text and refer to plots to write the tutorial (might be better for a general audience).

  12. David Smith
    Posted Jan 19, 2007 at 11:01 PM | Permalink

    Next, a look at their Summary page. They state a “substantial trend in tropical cyclones and hurricanes…” and “this trend is robust to all known potential problems with the database, which could change the century trends to… “all cyclones (125% +-25%), minor hurricanes (not at all) and major hurricanes (85%+-10%) (My interpretation of their words.

    They also state, “The trend is strongly related to Atlantic SST and thus Greenhouse Warming”. Wow.

    On the trend claim, apparently HW use page 15 as their basis (the “total” lines). It seems reasonable that any such claim should be based on peak-to-peak rather than comparing valley-to-peak. If one uses peak-to-peak then the stated increases shrink dramatically. If one uses peak-to-valley than an explanation of why the 1920s were not a low point (valley) in a natural oscillation seems in order.

    In my opinion, the “substantial trends” are in the eye of the beholder

  13. Judith Curry
    Posted Jan 20, 2007 at 7:43 AM | Permalink

    Not much time to post lately, i did a quick read of this thread. … Peter Webster did check in briefly, and even made a few posts, but hasn’t been here since. We suggested that Greg Holland check climateaudit and he started to but quickly gave up since there was too much stuff here with little chance of finding anything useful after spending much time. Holland and Webster were motivated in their analysis for the AMS presentation primarily by Landsea’s presentations (which has been making since early last fall) about the NATL data being flawed and that landfalling data should be used as a proxy. Roger’s post on the tropical listserv and subsequent comments from a few others did apparently motivate a few very minor changes in wording in the one paragraph in the H/W paper summarizing the literature. So the bottom line is that what is going here is not influencing much the mainstream hurricane-AGW research. I am trying to pay attention here when I have time (very short on that commodity at present), and I am telling people to clean up their statistics when I review papers and receive advance copies of papers. I of course encourage you to continue your analyses, but i often see you barking up the wrong tree and it is hard to sort through everything to see if there is anything useful.

  14. welikerocks
    Posted Jan 20, 2007 at 8:18 AM | Permalink

    #13 Right. Barking up the wrong tree, hard to see anything useful …yadda yadda; but you’ve told everyone to clean up their statistics. LOL!

  15. David Smith
    Posted Jan 20, 2007 at 8:27 AM | Permalink

    A key part of the HW argument, as best as I can tell, is the line they draw in the Atlantic. If a storm passes to the west of this line, they assume that the odds are very good that it will be detected by land or ship.

    I think their argument is that the number of east Atlantic storms which pass west of this line have increased, and since such storms are real and likely-detectable by pre-satellite methods, then the east Atlantic has indeed become more active in recent decades.

    In this line of argument the location of the line is critically important. My view is that their line is far too generous, incorporating way too much ocean.

    First, tropical cyclones have asymmetric windfields, especially storms which are less-developed. People in hurricane regions talk of the “weak side” (the west side) and the “dirty side” (east side) of storms, reflecting significant differences in strength. The land of the western Atlantic, such as the Virgin islands, experience the west side of storms to their east, and detection is problematic at any appreciable distance, especially weaker storms.

    I figure I’ve been within 100 km of the center of about 10 landfalling tropical cyclones. Major hurricanes leave no doubt that they’ve arrived, but the smaller storms’ especially on the western side, would have left me wondering if a storm had arrived had I not known from news reports.

    Second is the assumption that ships are good ways to detect storms. Sailors have rules of thumbs and, if their rules indicate a possible storm, they generally go elsewhere. They are not hurricane hunters, they are wise chickens.

    Now, a rule of thumb (say, a 15 knot north wind and choppy waves from the east, slowly falling barometer reading) may or may not indicate that a storm exists. This is especially true outside the deep tropics. It could (and often is) some other form of weather system. The rule of thumb might tell the sailor to go south, in which case he or she would never know 9or care) if they were near a tropical cyclone or a seedline or an old cool front or an upper low or other non-cyclone system.

    There is also the issue of ship coverage density, which I won’t get into here but which HW should have explored for their methodology.

    To me, a more proper approach would be to go to the database and divide the storms into two categories:

    1. Those which passed within 100km of land when they were of tropical storm (or higher) strength’ “Land” includes islands.

    2. Those which did not, and therefore would likely have been detected by ship, aircraft or satellite.

    I think that one would see that the category 2 (hard-to-detect, non-land) storms has increased from about 20% of all storms in the first half of the century to about 40% in recent decades. That is a much bigger pool of storms than seen in HW and, if I understanding their reasoning, would have a big impact on their methodology.

    If one breaks the category 1 storms into “US, Canada detection” and “Other land detection” (mainly Caribbean) then one sees a noticeable oscialltion (AMO?) in the US/Canadian detection over the last 150 years, but little rise. The Caribbean detection also shows the oscillations but is low in count in the 1800s, but there is reason to suspect an undercount 1860-1885, as has been discussed here at CA.

    The category 2 shows the rise shown from the data in post #1 above. It’s a big pool of storms, not the small, almost inconsequential, pool shown by HW.

  16. Posted Jan 20, 2007 at 8:39 AM | Permalink

    Judy (#13)-


    It is also interesting to observe the many hits at Prometheus from the NCAR MMM server (maybe one of Holland’s colleagues is a close reader!) and also yours at Ga Tech (maybe a misguided student!). Since we are dealing in strange coincidences, I’d bet that the CA logs show similar traffic patterns…

  17. Ken Fritsch
    Posted Jan 20, 2007 at 9:19 AM | Permalink

    We suggested that Greg Holland check climateaudit and he started to but quickly gave up since there was too much stuff here with little chance of finding anything useful after spending much time.

    I think I understand Dr. Holland, as I sometimes lack the patience to dig sufficiently deep to see all the possible avenues that could lead to less confidence in my conclusions and especially so when I think my conclusion must be correct.

    So the bottom line is that what is going here is not influencing much the mainstream hurricane-AGW research.

    This line reminds me of a long ago boss with whom I had a good working relationship and who invariably called me into his office after I had distributed a report. He usually would comment on that report and let me know when he agreed with my conclusions. When he did not, which was seldom, he would invite me in and hold the report with two fingers above the waste basket and let it go as I entered the room. I called this the two finger dunk. Judy, I believe we just received your version of the two finger dunk.

    Regardless of the influence on a group of scientists, sorting through these papers and having a civil discussion and analysis of them continues to be fun and enlightening for me.

  18. Jeff Weffer
    Posted Jan 20, 2007 at 9:28 AM | Permalink

    Judith Curry in #13 – I am telling people to clean up their statistics when I review papers and receive advance copies of papers. I of course encourage you to continue your analyses, but i often see you barking up the wrong tree and it is hard to sort through everything to see if there is anything useful.

    If you want to see how the statistics need to be cleaned up, you can read Judith’s paper from August 2006 published in the Bulletin of the American Meteorological Society which really started Climate Audit’s recent examination of hurricanes.

    It is not surprising that the other authors of the paper were Holland and Webster and there were significant deficiencies in how the data was presented.

    Click to access i1520-0477-87-8-1025.pdf

  19. Steve McIntyre
    Posted Jan 20, 2007 at 9:31 AM | Permalink

    Judith, we know that you were reading the thread in which the E-W issue was raised, because you made 15 comments on the thread and asked specific questions about the analysis and specific answers were given to you, together with code..

  20. Ken Fritsch
    Posted Jan 20, 2007 at 9:37 AM | Permalink

    And more importantly (than acknowledging Dr. Curry’s remarks) I want once again to thank David Smith for his review and comments on the HW presentation and continuing efforts to summarize these accounts. David, I find that you, rather uniquely, have a knowledge base that allows cogent analyses and summaries of these papers. It is good to see others’ analyses because I must admit that I read through some of these papers and I find myself saying is it just me but …

  21. Posted Jan 20, 2007 at 9:38 AM | Permalink

    So the bottom line is that what is going here is not influencing much the mainstream hurricane-AGW research.

    Strawman alert! ..

    Judy counters climateaudit.org is “not influencing much the mainstream hurricane-AGW research.”

    Hhhhmmm… did anyone in this thread make such a broad claim? No. Is it relevant to the actual claim posted? No.

    So, one must wonder: Why did Judy feel compelled to resort to arguing against a strawman?

  22. David Smith
    Posted Jan 20, 2007 at 9:54 AM | Permalink

    Another problem with comparing the first decades of the 20’th century with the satellite era has to do with a certain type of landfalling storm.

    Cyclones that exist as tropical storms for 24 hours or less rarely have symmetrical wind fields with a braod distribution of tropical-storm force winds. What that means is that the storm winds typically cover only a small area and (in modern times) require an aircraft to do a lot of searching.

    Cyclones that weaken to tropical depressions prior to landfall are also a problem. A person on land would see a thunderstorm, not a tropical storm and conceivably would classify it as such.

    I checked the 1910s and 1920s and found no such weak storms in the database.

    I checked 1995-2005 for such weak systems. Here are the candidates:

    Dean (1995)
    Gabrielle (1995)
    Kyle (1996)
    Charley (1998)
    Hermine (1998)
    Katrina (1999, not the 2005 version)
    Allison (2001)
    Grace (2003)
    Henri (2003)
    Bret (2005)
    Gert (2005)
    Jose (2005)
    Alpha (2005)
    Gamma (2005)

    There’s no way to know what fraction of these would have been detected and classed as tropical storms in the first three decades of the 20’th century. My guess is maybe 20%, if that. If they existed in earlier decades but 80% were not detected then that implies an undercount of a storm per year, which is not insignificant.

    If the thought is that these types of storms did not exist in earlier decades, but exist today, then a physical explanation is needed. Whatever that explanation is needs to align with the HW chart (slide 4) that shows a decrease in storm activity in the western Caribbean/southern Gulf of Mexico, where most of these weak, short-lived systems existed.

    Perhaps such storms exist only in active phases of the AMO and so they really didn’t exist in the 1910s-1920s. OK, but I’m following HW’s apparent methodology of comparing an active phase (1995-2006) with an inacive phase (1910-1930) on their summary page. Also, the active decade of the 1950s had but two of these weak, brief cyclones.

    As a side note, why do the weather analysts today even mess with such trash storms? Well, because they have the tools and the time, and just in case one might suddenly develop into a real storm. And, I’m sure there’s a desire to find and report everything, witness their reanalysis of the 10’th storm of 2006, where they reexamined, months later, reports and satellite images and determined that a spin in the Atlantic indeed had tropical characteristics for 18 hours and should be part of the 2006 record.

  23. welikerocks
    Posted Jan 20, 2007 at 11:22 AM | Permalink

    #16 and #21 Dr. Judy has a “quote page” too. here:link to a site to find things that “famous” people have said Wonder what you have to do to be added there?

    In Michael Crichton’s new book “Next” he gives his opinion (again) on the state of science these days as it pertains to the media and politics. He says that 99.9% of the time scientists who claim to have been misquoted in the media about certain issues will answer “no” if you ask them if they’ve made any effort to have their words corrected. I asked Dr. Judy if she tried to get them corrected and silence was the answer to my question-few months ago when she started to participate here. I only asked after somebody posted an interview, and Judy claimed her words in the article were inaccurate. interview (I’ve saved these to my bookmarks since then)
    Now here are conflicting words that same “air of authority” -plus the “I am too busy” “no body reads this site” thing plus a strange logic-foggy bottom reasons for this or not that, and strawman arguments! Holy Cow.

  24. David Smith
    Posted Jan 20, 2007 at 3:44 PM | Permalink

    Ken, I’m still struggling with this paper. I listed in #9 what I think I read but the reasoning does not flow smoothly for me. As Margo points out, that may be the nature of a Powerpoint presentation.

    There are lots of side streets in this presentation. For instance, slide #9.

    I think the goal slide 9 is to disprove any tight link between landfalling hurricanes and total basin hurricanes. This is important to HW’s attempt to validate the database.

    Slide 9 looks at the number of observations in the (presumably) well-measured western Atlantic. It reports that, if one uses the landfall/storm count ratio, then the observation per storm data prior to 1960 becomes screwy.

    First, I’m not sure that any screwiness in the observation statistics necessarily means that the landfall/storm count approach is invalid with respect to non-landfalling hurricane count. It’s a leap of logic for me.

    But, in addition to that, the number of observations per storm is another way of saying storm movement (the rate at which the storm plods across the ocean, say from Cuba to Texas). The “over-observed” result they report is probably simply saying that typical storm movement sped up by 20% ( about 2 miles per hour or 4 km/hr) post 1970.

    Such a speed up is not surprising, if their result for storm genesis (slide 4) is correct. Their slide implies fewer low-latitude storms and more higher-latitude storms in the western basin. Low-latitude storms generally move slower than their northern cousins, due to the northern storms’ proximity to the faster mid-latitude westerlies.

    There may well be something else, something of greater significance, in this slide that I’m missing. I’ll be happy if someone can point that out to me.

    As a footnote, there is a 600-lb gorilla sitting quietly in the corner of the slide 9 chart (my opinion). The chart (black line) shows the number of hours that storms are present in the western part of the basin. The western part is, by far, the inhabited region. The black line shows that our current high-activity times have not returned the inhabited regions to the activity levels of the 1940s. So, from a consequence standpoint, what’s the fuss about, if any storm increase affects just the fish of the open Atlantic? (Now, I’m being a devil’s advocate, because I think it is indeed important to learn all we can, and any open-Atlantic increase needs to be understood, but the gorilla and the question are real from a societal perspective.)

    More more footnote is that the numbers in the slide are “adjusted” and my eyeballs wonder how they got such a tight “smoothed adjusted” plot with such scattered data, but I don’t know that this impacts anything. I just note it.

  25. Ken Fritsch
    Posted Jan 20, 2007 at 5:07 PM | Permalink

    Re: #24

    I think the evidence presented at CA (your analyses and observations including portions of near landfall TCs in this thread and Steve M’s 68W analysis) including the landfall data pre-published and brought to this blog by RPJ taken alone or even as bigger case, together, would put a great deal of pressure on HW to accept that their data used has some major reliability issues or to reasonably explain all these observations.

    If they are willing to make a case for landfall hurricanes remaining constant while the total observed increases, then I agree that their case for increased adversity or even coming closer to predicting human harm goes out the window with it. No harm, no foul and no predictions.

    PS: I think that even the fit of the hurricane data to a Poisson distribution and a finding the fit better with divided data than data taken as a whole could be explained by more missing counts occurring from the earlier period.

  26. Gerald Browning
    Posted Jan 20, 2007 at 7:30 PM | Permalink

    Roger and Steve,

    Notice how quickly Judith responded to my post on this thread. She evidently was not too busy to defend her collaborators on the BAMS article?

    When I issue the post I am working on, all the references will
    be included so that there will be no doubt as to the source of the
    ideas or the true impact of those ideas. The text will be easily understandable to a general audience and without any flowery language that obscures the results.

  27. Douglas Hoyt
    Posted Jan 20, 2007 at 7:43 PM | Permalink

    Of the two choices: 1) hurricanes have had a decreasing probability of hitting land (HW position) or 2) early hurricanes had a greater of probability of not being counted, both as we go from 1852 to the present, I would say that 2 is the much more likely. It is just common sense that many early hurricanes would not be detected, particularly if they were far from land in the eastern Atlantic. It is a symptom of shaky science that the strongest signal is arising in the region where the fewest observers are located.

    As a refresher, I suggest people look at posts 395, 400, 422, and 436 in this thread http://www.climateaudit.org/?p=785

  28. Bob K
    Posted Jan 20, 2007 at 8:04 PM | Permalink

    Dave S.

    Re: your 15

    You might want to look at what I posted just a while ago in another thread. It shows you what is easily possible to do, other than just drawing a degree mark on a map to divide storms. Here’s the link. http://www.climateaudit.org/?p=1090#comment-78701

  29. David Smith
    Posted Jan 20, 2007 at 9:08 PM | Permalink

    Bob K, thanks for the analysis and approach! That is a neat graphic. It’s much faster, and with fewer errors, than my eyeball-and-tick mark approach. You could be a member of Occam’s Brigade.

  30. Bob K
    Posted Jan 20, 2007 at 10:19 PM | Permalink


    I’m just a guy with some time on my hands. Figured I’d fill some of it trying to learn a little bit of graphics coding. Still a newbie to graphics work. That’s not to imply it might be wrong. Because it isn’t. I’m anal about such things. It certainly wouldn’t be hard for the climatological community to do something similar and probably improve on it.

    If your interested, I can send the graphics to John A to forward to you. Might help with that tick problem you mentioned.:) There are 18 in that set. If you load a few selected ones into your browser using separate tabs, and then tab between them sort of like an animated gif, sometimes things pop out at you that you wouldn’t have previously thought about. You get your eyes involved as well as your brain. Different people get different meanings from the same images. Of course quite often there is nothing there to get.:)

  31. Steve McIntyre
    Posted Jan 20, 2007 at 10:46 PM | Permalink

    Bob K, what language do you do your graphics in?

  32. David Smith
    Posted Jan 21, 2007 at 3:35 PM | Permalink

    RE #30 Yes, please forward them to John A, who I ask to forward them to me. I’d like to have them for my file. Thanks.

  33. David Smith
    Posted Jan 21, 2007 at 4:45 PM | Permalink

    Footnote on HW: It appears that Holland Webster continue to classify subtropical storms as “tropical cyclones” and use them in their analysis. This is based on examining several of their graphs (on slides 3,7,13).

    “Subtropical storms” are not hurricanes, they are not tropical storms and they are not tropical depressions. They are not tropical cyclones.

    I count 21 subtropical storms from 1969-2006. That’s about 0.6 storms per year they’ve added to their post-1969 storm count. I counted none prior to 1969.

    My understanding is that these systems are mainly identified by satellite. Pre-satellite, my guess is that they were either not identified or ignored.

    Using subtropical storms in a 20’th century tropical cyclone count introduces another apples-to-oranges twist to the picture.

  34. Steve McIntyre
    Posted Jan 21, 2007 at 9:53 PM | Permalink

    Folks. make sure that you keep up with Margo’s postings on hurricanes at her blog.

  35. Ken Fritsch
    Posted Jan 21, 2007 at 10:08 PM | Permalink

    Re: #33

    Footnote on HW: It appears that Holland Webster continue to classify subtropical storms as “tropical cyclones” and use them in their analysis. This is based on examining several of their graphs (on slides 3,7,13).

    I would be quite certain that your observation is getting back to HW, so I would assume we will be hearing an explanation of why these events are included. Without an explanation one might suspect some sloppiness in their work –or some graduate student’s work.

  36. Steve McIntyre
    Posted Jan 21, 2007 at 10:15 PM | Permalink

    The calculations are so trivial that I don;t understand why they are delegated. I haven’t seen any HW calculations that can’t be done with a few lines of R code.

  37. johnmccall
    Posted Feb 23, 2007 at 9:58 PM | Permalink

    re: 27 “As a refresher, I suggest people look at posts 395, 400, 422, and 436 in this thread http://www.climateaudit.org/?p=785

    Appreciate the direction to those 3+1 brief posts — sufficient to trigger a reread of that entire thread.

  38. David Smith
    Posted Mar 8, 2007 at 9:38 PM | Permalink

    The Holland presentation has one slide worth noting. Slide #15 has a pop-up (click on the slide) which shows sea surface temperature in three regions of the Atlantic.

    The regions are East Atlantic (EA), West Atlantic/Caribbean (WA) and the Gulf of Mexico (GM).

    What the Holland temperature plot shows is that the West Atlantic/Caribbean and the Gulf of Mexico are about the same today as they were in the 1940s and 1950s. It is only the East Atlantic which has shown a significant temperature rise.

    I took a closer look at the East Atlantic SST. Here is a plot of the annual SST in the Eastern Atlantic (June-Oct), which gives unsmoothed and more-recent numbers. There are anomalously high SST in 2004 and 2005 but the rest of the recent era in the Eastern Atlantic really doesn’t look that unusual.

    But that doesn’t match the plot in Holland. What I found is that I have to add in the adjacent region between 5N and 10N in the East Atlantic, shown here , to get close to the Holland SST plot. (This region is south of the great majority of storm tracks, which is why I had initially excluded it.)

    Since the rest of the globe shows no-trend in hurricane count or intensity in the last 25 years, the only game in town is the Atlantic. And it looks like the only part of the Atlantic with a SST pattern of interest is the lower part of the tropical Eastern Atlantic. I figure that is about 10% of the tropical Atlantic.

    It’s a narrow ledge on which to construct a global warming/SST/Atlantic-storms-only hypothesis. Any acrobats attempting the feat should be wary of razors.

  39. David Smith
    Posted Mar 9, 2007 at 11:03 PM | Permalink

    Here are some notes on subsurface temperature patterns in the tropical Atlantic.

    These March, 2007 maps reveal the global ocean temperatures 150 metres below the surface. The bottom map shows the temperature anomaly (versus climatology).

    These March maps show a region of above-normal water in the Atlantic near South America, a somewhat cool region near Africa and near-normal elsewhere in the tropical Atlantic. Nothing extraordinary. Since this water tends to mix and upwell with surface waters, it looks like surface temperatures in the coming months should be near-normal as we enter the hurricane season. That is also the forecast from NOAA, with near-normal Atlantic SST expected by the heart of the hurricane season.

    These charts are available back to 1995. What I notice in flipping through them is that, since 1995, the area off the northeast South American coast is often anomalously warm while the area near Africa is often anomalously cool. This is versus climatology, which presumably mainly reflects pre-1995 conditions.

    It makes me wonder if something shifted in the tropical Atlantic circulation patterns in the 1990s, with one consequence being a shift in subsurface temperatures which then affect surface temperature patterns. Surface temperature patterns affect the ITCZ location and maybe certain winds (easterly jet), too, which can affect hurricane formation. Plausible, yes, poorly fleshed-out, yes, but something to ponder.

    I also took a look at subsurface temperatures in March, 2005 ,before the famous 2005 hurricane season. Note the much more substantial warm water along the northeast South American coast and coolness near Africa. By July, 2005 ,this anomalous subsurface warmth had become extreme in the tropical Atlantic while the area near Africa remained anomalously cool, creating quite a contrast across the subsurface tropical Atlantic, as well as reinforcing surface warmth near the Antilles.

    The year 2005 took the decadal subsurface temperature pattern to an extreme, at the same time that hurricane formation went to an extreme. Connection? I don’t know – it’s just something to think about.

  40. David Smith
    Posted Mar 10, 2007 at 9:02 AM | Permalink

    Pacific sea temperatures show an interesting trend. (This is based on ocular examination, as that’s all the data allows.)

    This link shows 2006 monthly temperature anomalies at a depth of 400 metres, in the Pacific and adjacent areas. (A depth of 400 metres is still the upper ocean, about 10% of the average depth of the ocean.) This water is an indication of the heat content of the upper ocean. Also, the 400-metre water tends to work its way towards the surface, affecting surface temperatures.

    For 2006, note the loss of anomalously warm water as the year progresses. The weak El Nino pops up along the equator late in the year, but doesn’t amount to much. Particularly note February, 2006, so that an apples-to-apples comparison can be made with February, 2007.

    This link shows 2007 anomalies . (The month of March is under construction, so observe it with caution.) Note the lack of vibrant warm colors versus 2006.

    The western (left-hand) side of the maps are the Warm Pool, where water is quite warm. Small temperature changes in this region create big impacts on evaporation (latent heat released into the atmosphere), due to the 15% increase in evaporation per C rise.

    Also, in the tropics, there tends to be a “magic temperature” around 28-29C at which atmospheric stability noticeably decreases (= more thunderstorms). Big parts of the Warm Pool are near this temperature.

    If the Warm Pool cools then it releases less heat into the global atmosphere. This tends to affect (cool) global temperatures. Worth watching.

  41. David Smith
    Posted May 20, 2007 at 9:07 PM | Permalink

    This quote from James Hansen is encouraging:

    About An Inconvenient Truth’s connection of rising hurricane activity to global warming — something refuted by storm experts — Mr. Hansen said, “We need to be more careful in describing the hurricane story than he is.”

    My hope is that the original 2005 authors (Emanuel, Webster et al) also take more-cautious positions in their public statements. We shall see.

  42. David Smith
    Posted May 20, 2007 at 9:10 PM | Permalink

    Re #41 Sorry, the full article is here .

  43. bender
    Posted May 20, 2007 at 10:52 PM | Permalink

    I couldn’t agree more with #41. I thought the Curry, Webster et al. BAMS paper showing error-free hurricane trends (we don’t need no stinkin’ trend statistics!) was highly irresponsible – and told her so, both here and at RC. Thanks again to Sadlov for alerting me to that paper.

  44. Ken Fritsch
    Posted May 21, 2007 at 10:37 AM | Permalink

    Re: #40

    Would these new inputs/conditions have an influence on an adjusted hurricane prediction — if that were part of your predicting contest? Predicting the medium term future conditions that will eventually lead to hurricanes to my mind has more of element of chance than, say, looking at near future term conditions and using that information to predict hurricane frequencies. Any chance that you might modify your contest to include adjusted predictions? That is how Gray and other agencies do their predictions.

  45. David Smith
    Posted May 21, 2007 at 1:58 PM | Permalink

    RE #44 Ken, some meteorologists believe there is an April “curtain” in tropical forecasting, meaning that reliable pre-April forecasts of post-April tropical conditions is not possible. If that’s true, then the best bet is to go with a climatological average pre-April. Otherwise, there’s usually a divergence problem (reality diverges far from forecast).

    Wisely, the people making predictions in our contest generally went with a climatological average. We brave Climate Audit contestants are clustered around either the 50-year climatology (10 storms, more or less) or the 10-year active-AMO pseudo-climatology (15 storms, more or less).

    (The wisest forecasters are the Europeans, who are very quiet about even making a prediction, and historically don’t reveal their forecast until the season is over.)

    Now that April has passed we can get a better view of things. There will be no El Nino and sea surface temperatures will be slightly above normal. The wind shear outlook is for favorable conditions. Seedlings are already scattered about. I stick by my forecast of a slightly above-normal (using 10-year climatology) season, with 15 to 18 storms (which is cowardly-close to the 10-year climatological average).

    We’ll allow anyone to modify their forecast as the season progresses, including anyone who later wants to claim they knew the actual storm count last January but hit the wrong dang key on the keyboard.

  46. MarkW
    Posted May 21, 2007 at 2:04 PM | Permalink

    I always thought that it was too dangerous to make any of my predictions before, say, December.

  47. Ken Fritsch
    Posted May 21, 2007 at 2:33 PM | Permalink

    Re: #46

    I always thought that it was too dangerous to make any of my predictions before, say, December.

    Would that be the December of the NATL season being predicted? I, like the Europeans, have had a difficult time getting my prediction calculations from my model and thus may have to make a later prediction than I would like. My model is to give a plus or minus range and confidence limits. If the range is too wide and uncertain I will have to back off any predictions.

    David Smith, on a serious note I much enjoy your running commentary on the developing hurricane conditions and find it creates much additional interest in a hurricane/TS science that I might otherwise find incomprehensible. I must admit that I am very sketical of prediction sytems, in gemeral, and particularly longer term ones without significant out-of-sample results.

  48. David Smith
    Posted May 21, 2007 at 3:25 PM | Permalink

    Re #46 A wise man, indeed.

    Re #47 I share your skepticism but do think that, by early June, we can forecast whether a season will be near-normal, below-normal or above-normal and have a better-than-coin-toss chance of being right. That’s about it.

    For me, a lot of this is personal. Two years ago my nephew got his 15 minutes of fame, appearing in a National Geographic photo ( link ), carrying a bucket of cats from his flooded home in New Orleans.

    As an aside, my relatives had another bucket of cats (not shown) and a big dog. When the rescue boat came there wasn’t room for the two buckets of cats, three humans and the dog. The outcome was that the cat buckets and the big dog rode in the boat while my relatives swam. What people do for their pets…

%d bloggers like this: