PDI and Hurricane-Days

Here’s a figure showing Judith Curry’s PDI as compared to calculations from my collation of the ATL track data previously archived, also comparing PDI – which is the integral of wind speed cubed against other measures: a count of hurricane-days, an integral of wind speed and an integral of wind speed squared (which I think is what they call “ACE”). I’ve plotted the data back to 1851 ti see what it looks like prior to truncations by Curry, HW, Eamnuel or others.

All of this is plotted here for the Atlantic and thus is biased through progressively improved measurements of eastern storms through the century. I haven’t done any smoothing since the data is simple enough that people should be allowed to look at the raw data – rather than take the risk that impressions are introduced through smoothing artifacts such as Emanuel’s end-point pinning.

As an impression, it looks to me like the transformations by which wind speeds are summed, then squared, then cubed, results in a difference between late 20th century levels and (say) late 19th century levels that isn’t present in the simple count. This indicates that the count of hurricanes is similar but that the average speed and/or average length of hurricane has increased in the later period. Obviously the measurement of wind speeds in the late 19th century was not done in the same way as in the late 20th century. Without parsing through the data on every storm, I don’t see how one can draw any conclusions based on this data as to whether the differences result from technical changes or from climatological changes.

One other observation from this data set: let’s say that Holland and Webster have hypothesized a model between SST and hurricanes based on the 1905-2005 period. This hasn’t used up all the available data as there is obviously a significant amount of 19th century hurricane data. I haven’t gone through the SST data yet, but my impression is that HadCRU and other temperature data show 19th century SSTs as lower than 20th century SSTs. This makes the high hurricane counts of the 1880s rather an interesting out-of-sample test. It’s hard to see that any plausible relationship between SST and hurricane counts will apply to the 1880s. Doesn’t this sort of verification seem like the sort of thing that a Holland and Webster article should have undertaken? They would presumably end up having to argue that 19th century hurricane counts have been overstated in Hurdat.


Figure 1. Atlantic Basin – hurricane (65 knots) days; and wind speed integrals.

West Division Only

If one looks at the same data using only West Division data which is almost certainly more homogeneous than total Atlantic data, we see a quite striking reversal of relative levels between previous high years 1886-87, 1933 and 2005.


Figure 2. Same for storms west of 69W (except for unavailable Curry comparandum)

East Division Only


Fig 3. Same for east of 69W

Here’s the calculation:

curry< -read.csv("http://data.climateaudit.org/data/hurricane/curry.natl44to05.csv&quot;,sep="\t",skip=6,fill=TRUE)
pdi<-tapply( (Track$wind[temp]*.51444)^3,Track$year[temp],sum,na.rm=TRUE)/4
y< -tapply( !is.na(Track$wind)[temp1],Track$year[temp1],sum,na.rm=TRUE)
x<-as.numeric(names(y));days<-rep(NA,2006-1850); days[x-1850]<-y/4
pdi1<-tapply( (Track$wind[temp]*.51444),Track$year[temp],sum,na.rm=TRUE)/4
pdi2<-tapply( (Track$wind[temp]*.51444)^2,Track$year[temp],sum,na.rm=TRUE)/4

Here’s how I made this particular plot (there are some quite amazing R graphical techniques, but I’m only familiar with very workmanlike methods). This particular method gives control over axes and panels tho.

nf< -layout(array(1:4,dim=c(4,1)),heights=c(1.1,1,1,1.3))
text(1851,32,"Hurricane Days",font=2,pos=4)

x0<-800;text(1851,x0,"Wind Speed Integral",font=2,pos=4)

x0<-30000;text(1851,x0,"Wind Speed^2 Integral",font=2,pos=4)#9*10^4

x0<-1000000;text(1851,x0,"Wind Speed^3 Integral",font=2,pos=4)#5*10^6


  1. Posted Jan 5, 2007 at 11:57 PM | Permalink

    Thanks much Steve!

  2. TAC
    Posted Jan 6, 2007 at 6:39 AM | Permalink


    While I still find it surprising that the PDI time trend is so weak, there are simple statistical explanations, having to do with S/N ratios, for why we should expect to observe significant AGW in the mean statistics long before it is visible in the frequency of extreme events (see a related piece here, whose abstract concludes: “…human-induced greenhouse warming is more likely to produce noticeable and significant changes in the mean state of hydrological regimes than in hydrological extremes.”).

  3. Judith Curry
    Posted Jan 6, 2007 at 8:34 AM | Permalink

    Calculating ACE and PDI prior to 1944 is a very dodgy proposition, since the intensities are hugely prone to error (much worse than the 10-20% or so we are talking about in terms of undercounting).

  4. David Smith
    Posted Jan 6, 2007 at 8:58 AM | Permalink

    I believe there was an undercount of weaker storms in the western Caribbean / Gulf of Mexico in the 19’th century, in addition to the undercount in the eastern Atlantic.

    Some data on storms which made landfall in Mexico or Central America:

    Hurricanes 1870-1899 (active period): 13
    Hurricanes 1900-1929 (inactive period): 8
    Hurricanes 1930-1959 (active period): 21

    Tropical storms 1870-1899: 12
    Tropical storms 1900-1929: 17
    Tropical storms 1930-1959: 38

    Tropical storms were almost surely undercounted in the 19’th century.

  5. Michael J
    Posted Jan 6, 2007 at 11:21 AM | Permalink

    I really don’t think it is possible to have any meaningful statistics on TC numbers or intensities for any period of time prior to 1970 or so. There is no formula capable of estimating either one with the more recent documented, verifiable numbers with the invention and use of satellites and wind measurement devices.
    Even today some are re-adjusting the numbers for frequency and intensity of storms like Andrew which made landfall in an area with well sited instruments to measure them. It is much like the Mann crowd who attempt to use questionable proxies to actual observations and records know to be accurate and then determine that a .6 C rise in “temperature” is a recipe for disaster caused by man.

  6. Brooks Hurd
    Posted Jan 6, 2007 at 11:42 AM | Permalink

    Re: 3


    I agree with you about the uncertainties of TC intensity data pre 1945. I believe that the uncertainties increase significantly as we go back further.

  7. Ken Fritsch
    Posted Jan 6, 2007 at 11:48 AM | Permalink

    While I doubt that your tropical storm audits will bring any immediate response as regards necessary caution in using the data and drawing conclusions, I personally appreciate your efforts in presenting the data in forms (and with sensitivity tests) that allow readers possessing a range of technical abilities to judge its merits/message for themselves.

  8. David Smith
    Posted Jan 6, 2007 at 12:45 PM | Permalink

    Some storms are easy to detect, like hurricanes that strike a major city. Others (weaker storms, storms in sparsely settled areas, at-sea storms) are harder to detect, especially in the past.

    For fun, I took a look at those storms which remain at-sea for their entire existence. (Process: this was an eyeball exercise, using the Unisys storm plots. I defined at-sea storms as those which did not strike mainland or major islands. I defined the Greater and Lesser Antilles and The Bahamas as “major islands”. I did not count Bermuda or the Canaries as major islands. I excluded subtropical storms. if a storm got within 1 degree of shore, I counted it as a landfall.)

    Here are decadal totals. (My apology for not plotting this here, but I haven’t mastered that skill):

    At-sea tropical cyclones (hurricanes + tropical storms):
    1860s – 24
    1870s – 21
    1880s – 24
    1890s – 17
    1900s – 18
    1910s – 11
    1920s – 13
    1930s – 10
    1940s – 15
    1950s – 28
    1960s – 31
    1970s – 34 (41)
    1980s – 43 (45)
    1990s – 49 (51)
    2000s (prorated) – 51 (56)

    At-sea tropical storms:
    1860s – 7
    1870s – 4
    1880s – 4
    1890s – 4
    1900s – 10
    1910s – 6
    1920s – 5
    1930s – 5
    1940s – 10
    1950s – 9
    1960s – 17
    1970s – 9 (16)
    1980s – 21 (23)
    1990s – 20 (22)
    2000s (prorated) – 24 (29)

    At-sea hurricanes are the difference betwen the two.

    The counts in parenthesis include “subtropical storms”, which are neither fish nor fowl, not fully tropical. They are part of the Unisys raw data and can get incorporated into counts if one is not careful. I think that HW and others have inadvertently included subtropical storms in their counts.)

    Observations and opinions:

    * The ratio of hurricanes to tropical storms was out of whack in the 19’th century, indicating an undercount of the weaker storms

    * Something not shown in the data presented above is the percent of at-sea hurricanes which affected the major shipping lanes (New York to London) in the 19’th century. There were relatively few at-sea hurricanes in other parts of the Atlantic. This makes me suspect an undercount of hurricanes outside the major shipping lanes and it also makes me wonder if the late 1800s was actually an extremely active period, moreso than is generally thought.

    * The total declined in the early 20’th century, reflecting less-active times, which is expected. I am surprised it did not resume circa 1930 as activity resumed.

    * The total ramped up dramatically in the 1950s, as recon came into being and as the US Weather Bureau added staff and emphasis.

    * The total continued to ramp up in the satellite era, even though the 1970s and 1980s were rather dormant.

    * The at-sea tropical storm count seems to plod upwards, especially once satellite coverage began.

    If you plot these on a spreadsheet the patterns show up pretty well.

    I’ll send the yearly data to Steve M.

  9. TAC
    Posted Jan 6, 2007 at 1:50 PM | Permalink

    A couple of questions: Michaels’s Figure 4 [Michaels et al., 2006] shows no correlation between recent hurricane peak wind speed and SST when SST exceeds 28.25 degrees C:

    Figure 4. The relationship between maximum wind speed
    and the highest SST encountered prior to (or concurrent
    with) reaching the maximum wind speed at temperatures
    greater than or equal to 28.25 C. The relationship is
    statistically insignificant (N = 195, p = 0.513, R2 = 0.002,
    slope = 1.32).

    First of all, is this result believed to be correct?

    Second, this suggests that SST is no longer an important determinant for hurricanes once SST exceeds a threshold. That would seem to raise a lot questions of the form: What is preventing continual formation of hurricanes over warm tropical waters? Instead of wondering why we see so many hurricanes, perhaps it would be worth asking why we see so few.

    I can imagine a bunch of possible explanations — lots of physical processes are at work — but if SST is not important, then I am curious to know: What is?

    I apologize in advance if this is a bone-headed line of thought 😉

    PJ Michaels, PC Knappenberger, and RE Davis, “Sea-surface temperatures and tropical cyclones in the Atlantic basin”, GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L09708, doi:10.1029/2006GL025757, 2006

  10. Posted Jan 6, 2007 at 2:36 PM | Permalink

    I agree that eyeballing the data suggests there is no significant difference in the PDI’s over time.

    I know that some might like to decide whether or not the PDI has increased by simply comparing the PDI between 1944-1964 to that after 1994 using statistics developed based on these two periods only. (That is: means, standard deviations and all higher order moments based only those two periods). However, as I see it such a test is only valuable in the following sense:

    To show PDI has increased generally, the PDI in the later period must be shown to be earlier than in the first using the restricted data set. So, showing this is a necessary condition to showing PDI has increased. It is not, however, sufficient to demonstrate that.

    I’ll skip that explaining why it’s an insufficient condition, because it turns out that doesn’t matter for now because I applied the the hypothesis tests Judy suggested to test her two claims. (That PDI’s from 1994-2004 really are larger than those from 1944-1964 and that the Landsea correction makes do difference to the outcome. )

    Generally: I found the claim that PDI during the second period is higher than during the first period fails.

    First, I found the Landsea correction does make a difference to the results.

    If you use the correction, the recent PDI is higher to a confidence level of 5% (the break point was 4.8%). If you don’t use the correction, there is no significant difference between the current PDI and the past one.

    The sensitivity of this result to the Landsea correction suggests that we can’t base any conclusions about the increase in the PDI until after any uncertainties associated with the Landsea correction are resolved. Details here.

  11. David Smith
    Posted Jan 6, 2007 at 2:39 PM | Permalink

    Re #9 TAC, you’ve raised excellent questions.

    Perhaps the current Big Mystery is why are there consistently 85 (+ or – 10) tropical cyclones a year? Why not 40 this year, 140 next year and 85 the year after that? See this plot from Webster et al (2005). The black line in the left graph represents total global cyclone count.

    If anyone has an answer, they’re being rather quiet about it.

    Michaels’ finding on 28C is probably controversial. To me it says that, once 28C is reached, other factors, particularly wind shear and surrounding air features, come into play and determine just how severe the storm will become. It’s like a car: even though you put higher octane fuel into your car, it probably won’t go noticeably faster because of the car’s mechanical constraints.

    Hurricanes are like machines, with moving parts that need to be aligned just-so for maximum efficiency. Sea temperature is only one part of the machine.

    There may be something special about 28C, perhaps having to do with lapse rates or other thermodynamic factors.

    I think the 28C finding reported by Michaels is probably about right, though there may still be some benefit to a storm from even higher temperatures, especially if the warmer water runs deeper, too.

    Hurricane specialists probably see 28C SST as a necessary-but-not-sufficient condition for a severe hurricane and pay more attention to wind shear, upper air features and depth of warm water than they do to SST, once a storm is over 28C water.

  12. Bob K
    Posted Jan 6, 2007 at 2:49 PM | Permalink

    Here’s a count of all 65kt and greater readings separated into 10×10 grid boxes and
    grouped into three time periods. 1851-1900, 1901-1945, 1946-2005. There may be a
    couple readings not counted due to being out of range of what I did.

    I imagine the formating will be trashed by wordpress. Should be able to copy/paste them to a text editor then save as .csv and load them into a spreadsheet for easy viewing.

    I noticed that 1851-1900 had many more readings than 1901-1945. More than can be accounted
    for by the five year difference in time length. Very curious. Didn’t temps keep rising
    through the early 40’s??

    1851-00, 0-10W, 10-20W, 20-30W, 30-40W, 40-50W, 50-60W, 60-70W, 70-80W, 80-90W, 90-100W, Totals
    0-10N, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
    10-20N, 0, 0, 17, 67, 95, 150, 222, 148, 123, 3, 825
    20-30N, 0, 0, 1, 24, 59, 188, 367, 478, 507, 285, 1909
    30-40N, 0, 0, 5, 21, 63, 156, 396, 496, 96, 22, 1255
    40-50N, 0, 0, 5, 37, 87, 161, 162, 20, 0, 0, 472
    50-60N, 3, 3, 11, 17, 7, 4, 0, 0, 0, 0, 45
    60-70N, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1
    G’total 4507

    1901-45, 0-10W, 10-20W, 20-30W, 30-40W, 40-50W, 50-60W, 60-70W, 70-80W, 80-90W, 90-100W, Totals
    0-10N, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1
    10-20N, 0, 0, 0, 0, 42, 96, 167, 168, 201, 8, 682
    20-30N, 0, 0, 0, 1, 35, 141, 287, 363, 433, 307, 1567
    30-40N, 0, 0, 6, 15, 71, 111, 291, 258, 41, 12, 805
    40-50N, 0, 0, 6, 13, 44, 47, 60, 12, 0, 0, 182
    50-60N, 0, 5, 4, 2, 2, 3, 0, 0, 0, 0, 16
    60-70N, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1
    G’total 3254

    1946-05, 0-10W, 10-20W, 20-30W, 30-40W, 40-50W, 50-60W, 60-70W, 70-80W, 80-90W, 90-100W, Totals
    0-10N, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
    10-20N, 0, 0, 24, 110, 227, 255, 239, 168, 208, 7, 1238
    20-30N, 0, 0, 0, 26, 124, 305, 567, 544, 432, 350, 2348
    30-40N, 0, 1, 8, 95, 309, 476, 516, 517, 68, 23, 2013
    40-50N, 7, 12, 23, 72, 105, 131, 100, 14, 0, 0, 464
    50-60N, 1, 3, 8, 12, 18, 3, 0, 0, 0, 0, 45
    60-70N, 1, 1, 0, 2, 0, 0, 0, 0, 0, 0, 4
    G’total 6112

  13. Steve McIntyre
    Posted Jan 6, 2007 at 3:04 PM | Permalink

    #11. David, why 80? was the question discussed by Webster at AGU, with some interesting calculations on the total energy dissipiated by hurricanes – which he compared to energy amounts in Malkmus.

    This seems like such an obvious question that I’m surprised that it’s only now being discussed, but at least it’s being discussed now. HAving said that, it seems bizarre to be discussing trends in the ATL without visiting the question of whether this is just a decadal re-distribution of storms from (say) the Indian Ocean where there seems to be a decline over the past 30 years to the Atlantic.

  14. TAC
    Posted Jan 6, 2007 at 5:04 PM | Permalink

    #11 David, that was what I was imagining. However, it still seems that rising SST should create more TCs, if only because there will be more ocean above 28.25 C for a larger part of the year. Hmmm…the only thing I’m sure of at this point is that I still don’t see the correspondence between the data and what I understand about the physical process.

    In any case, thanks! 🙂

  15. David Smith
    Posted Jan 6, 2007 at 10:38 PM | Permalink

    Re #13 I did not know that Webster was discussing this – I’ll check to see if he has anything available online. It’s a fascinating topic.

    Re #14 My belief is that there indeed is a link between higher SST and intense storm count, but it’s small and gets lost in the noise.

  16. David Smith
    Posted Jan 7, 2007 at 9:23 AM | Permalink

    One suspicion I’ve had is that the US did a pretty good job of recording landfalling storms in the 19’th century (at least the stronger ones) while storm reporting from the Caribbean/Mexico was rather poor.

    So, I took a look at reported landfalling cyclones in the Caribbean/Mexico. This is defined as the area south of 25N and includes the islands of the Antilles as well as mainland.

    For the 25 years 1861 to 1885, I counted 36 landfalling storms. Sixty-five percent (65%) were hurricanes.

    For the 25 years 1886 to 1910, I counted 112 landfalling storms. Forty percent (40%) were hurricanes.

    That’s a remarkable jump. The jump occurred rather suddenly, in 1886.

    What this indicates to me is that reporting of landfalling storms, especially weaker storms, in the Caribbean was spotty in the earlier years. Those which proceeded to hit the US were likely counted anyway, but those which dissipated in the region or exited into the mid Atlantic may have been missed.

    The sharp transition at 1886 is odd. I wonder if that is due to the availability of records or marks the extent of someone’s reconstruction efforts. In the 19’th century Catholic priests in Cuba, and possibly other parts of the Caribbean, were quite active in observing and predicting hurricanes. Perhaps 1885 marks the extent of their record-keeping.

  17. Posted Jan 7, 2007 at 10:52 AM | Permalink


    Perhaps the current Big Mystery is why are there consistently 85 (+ or – 10) tropical cyclones a year? Why not 40 this year, 140 next year and 85 the year after that?

    Poisson(85) is almost normal N(85,85), and 40 would be more than four sigmas: rare event.

  18. Bob K
    Posted Jan 7, 2007 at 3:54 PM | Permalink

    From 1851-1900 there were 100 of the 377 storms had initial wind readings of 65kt or higher. From 1901-1945 there were only 5 of the 341 storms with initial wind reading of 65kt or higher. None after that. I calculated travel by using the inital location and final location readings. I suppose some may have doubled back a bit.

    Given the number of storms not detected prior to hurricane strength there was likely poor detection. A look at the mean storm travel for 35kt and higher indicates to me detection was a problem through 1950 at least.

    Here are the figures. Columns following the date are initial readings 65kt or greater. Total storms recorded. Percent of total. Mean travel of storms with initial reading of 65kt.

    65kt or higher.
    1851-1900, 100, 377, 26.5%, 1585nm
    1901-1945, 5, 341, 1.5%, 1696nm

    35kt or higher.
    1851-1900, 369, 377, 98%, 1690nm
    1901-1945, 297, 341, 87%, 1452nm
    1946-2005, 114, 635, 18%, 1304nm

    Here is mean storm travel for all storms.
    1851-1900, 1688nm
    1901-1945, 1446nm
    1946-2005, 1555nm

    Anyone know why we hear little about Mediterranean Sea hurricanes? It’s 150% the size of the Gulf of Mexico with waters just as warm. Pretty deep too. Makes me wonder about the effect of SST.

  19. Bob K
    Posted Jan 7, 2007 at 6:09 PM | Permalink

    Below has only storms with initial reading 20N,80W or farther N. and W. Gulf spawned.
    Seems to be about two hurricanes every three years are spawned there.

    Initial reading 65kt or higher.
    1851-1900, 18, 377, 5%, 1823nm

    all years, 18, 377, 5%, 1823nm
    max category achieved, 0=0, 1=8, 2=8, 3=1, 4=1, 5=0
    Initial reading 35kt or higher.
    1851-1900, 77, 377, 20%, 1686nm
    1901-1945, 39, 341, 11%, 861nm
    1946-2005, 25, 635, 4%, 777nm

    all years, 141, 1353, 10%, 1297nm
    max category achieved, 0=71, 1=45, 2=18, 3=3, 4=3, 5=1
    Initial reading any wind speed
    1851-1900, 81, 377, 21%, 1659nm
    1901-1945, 47, 341, 14%, 869nm
    1946-2005, 104, 635, 16%, 929nm

    all years, 232, 1353, 17%, 1172nm
    max category achieved, 0=131, 1=72, 2=19, 3=4, 4=4, 5=2

  20. David Smith
    Posted Jan 7, 2007 at 6:56 PM | Permalink

    Re #18 The Mediterranean is a few degrees too cool, is near a lot of dry air, is near a lot of wind shear and has few seedlings. Nevertheless, there are cyclones which somewhat resemble tropical storms from time to time. They may be some kind of hybrid.

    Someone on Wikipedia assembled this page on cloud systems that resembled tropical storms. I have no idea if this Wikipedia posting is accurate, but it’s interesting nevertheless.

    These kinds of cloud patterns would have been hard/impossible to detect prior to satellites but now they can be seen. So, 1950s Med storms = 0, 2000s Med storms = 5 or whatever: we have two data points and thus a trend!

    I won’t be shocked to see a Nature or Science article headlined, “Conclusive Evidence:
    Global Warming Is Spawning Mediterranean Hurricanes – Venice Vulnerable”

  21. David Smith
    Posted Jan 7, 2007 at 7:21 PM | Permalink

    Re #19 Bob K, I’m curious as to whether Caribbean storms have increased, as a percent of all reported storms. These would be storms that have at least one plot point inside the box 10N-23N; 60W-85W. If your technique allows that without much effort, I’d appreciate knowing is there is any long-term pattern.

    My guess is that the percentage increased to 1885 and then may have dropped somewhat after 1950, as father at-sea storms were detected.

  22. Steve McIntyre
    Posted Jan 7, 2007 at 8:25 PM | Permalink

    #21. Just run this and you’ll get the required series:

    Track< -Track.collation$ATL
    temp100)&(Track$lat<230)&(Track$long -850)
    #10N-23N; 60W-85W
    carib< -tapply(!is.na(Track$wind[temp]),Track$id[temp],sum)
    y< – rep(NA,length(unique(Track$id)))

  23. Bob K
    Posted Jan 7, 2007 at 9:33 PM | Permalink

    Here you go Dave,

    158 of 377, 42% crossed that box 1851-1900.
    Those storms at some point achieved the following categories. Not necessarily while in that box.
    TS=44, 1=27, 2=41, 3=35, 4=11

    220 of 391, 56% from 1901-1950.
    TS=83, 1=28, 2=38,3=41,4=25,5=5

    238 of 598, 40% from 1951-2005.
    TD=76, 1=38, 2=23, 3=42, 4=39, 5=20

    I don’t see much there.

  24. Bob K
    Posted Jan 7, 2007 at 9:49 PM | Permalink


    Re: 20

    lol. You’re probably right about the attention grabbing headline.

    Thanks for the info on the Med.

  25. David Smith
    Posted Jan 7, 2007 at 10:05 PM | Permalink

    RE #23, #16

    Thanks, Bob

    Here is my interpretation:

    The Caribbean is “landlocked”, such that storms have to pass over inhabited areas, where they can be detected. This detection system is pretty good and has been good for a long time, stretching back into the 19’th century. The key is to have them reported. Reporting has not always been done, despite detection.

    The period 1851-1885 had an undercount of Caribbean storms. Storms occurred and were detected by locals, but were not recorded into the database. This underreporting was corrected about 1885. For the entire 50 years (1851-1900), there was underreporting, but it was not as bad as it might have been thanks to the 1885 improvement.

    The period 1901 -1950 was one where Caribbean storms were both detected and recorded. Per Bob K’s work, there were 220 Caribbean storms.

    The period 1951-2005 was also one where Caribbean storms were detected and reported. Per Bob K’s work, there were 238 Caribbean storms. Adjust that for the extra 5 years, and we get 216 storms, almost identical to 1901-1950.
    No increase in frequency.

    However, Bob’s work shows that Caribbean storms, as a percent of all Atlantic storms, dropped. This was because more storms at-sea, in the eastern half of the basin, were detected in 1951-2005 than in earlier decades. (In other words, the denominator grew while the numerator held constant, giving a lower percentage.)

  26. Bob K
    Posted Jan 8, 2007 at 5:03 AM | Permalink


    I broke the first 50 years down into 1850-1880 and 1881-1900.

    You’re right about the difference pre 1881.

    1850-1880 had only 65 of 211 total storms recorded in that area. 31%
    1881-1900 had 93 of 166 storms recorded in that area. 56%

    Quite an uptick in detection.

    Looking over all the figures, and adjusting for detection quality, it appears the 1800’s were the worst 50 years out of the bunch.

  27. David Smith
    Posted Jan 8, 2007 at 5:56 AM | Permalink

    Thanks, Bob.

    The 56% you found for 1881-1900 is identical to the 56% you found for 1901-1950.

    Then along came aircraft, and especially satellites, and more at-sea and weaker storms were detected and recorded, lowering the (already-detected-and-recorded) Caribbean storms as a percent of all storms post-1950.

    The hypothesis hangs together.

  28. Bob K
    Posted Jan 8, 2007 at 6:45 AM | Permalink


    In case your interested. Here are two Atlantic basin tracking charts with all the readings in your defined grid plotted out for the two time periods.

    The color code is red for cat. 3 readings and higher. Blue for cat. 1 and 2. Green for tropical storm.

    This one is pre 1881 plots.

    This one is 1881-1900.

    The second one is denser, even though it’s for a shorter period.

  29. David Smith
    Posted Jan 8, 2007 at 9:30 AM | Permalink

    RE #28 Bob, I can’t get the links to work.

    I see you use something called Imageshack. I’ve never heard of that. Is that a place where one can park a plot and thus get an http address for posting? That would be great.

  30. beng
    Posted Jan 8, 2007 at 10:02 AM | Permalink

    RE 9: TAC says:

    What is preventing continual formation of hurricanes over warm tropical waters? Instead of wondering why we see so many hurricanes, perhaps it would be worth asking why we see so few.

    There’s prb’ly negative feedbacks operating. Possibilities are that TSs shade the ocean underneath them, dump cool rainwater in it & more importantly churn-up the upper layers, cooling them. The depth of atmosphere “behind” them is also stabilized temporarily. I’ve seen a study that monitored the track of a TS just east of the US east coast, & easily detected the cool ocean temps left behind it for some time (many days).

  31. Bob K
    Posted Jan 8, 2007 at 3:04 PM | Permalink

    RE #29 Dave, I just tried it now and it worked for me. It was a little slow compared to earlier though. Might want to try it again. The images are only about 350k each.

    Yeah. It’s free parking for images. Haven’t used it before myself. I didn’t even have to register to put them there. I guess if I wanted to keep a portfolio of images there, I would have to register, otherwise they’re just randomly stored. They say they don’t delete them. I’m kind of wondering how they make a buck since registration is free. Apparently they’ve been around for a few years and they do have a couple ads.

  32. Posted Jan 9, 2007 at 2:43 AM | Permalink

    Bob : The problem is that those are invalid JPEGs. Firefox can’t display them. Are you using Adobe Photoshop to manipulate them? I’ve run across the problem with Adobe products saving invalid JPEG formats in the past.

    In my experience, if Firefox won’t display them, many other programs won’t either. The problem is usually the that JFIF chunk is not the first chunk in the file. Perhaps you can load them into an image manipulation program (not Photoshop) and save them again, then they will become standard JPEG files.

  33. Posted Jan 9, 2007 at 2:47 AM | Permalink

    Bob: Yup. I checked. The images were created with Adobe Photoshop 7.0 and the first chunk in the file is an EXIF chunk. Most programs expect JFIF to come first, and EXIF to be at the end. You need to load them and re-save them with another program.

    I tried loading the files with a number of programs. feh – displayed blach image. qiv – would not load. xzgv – displayed normally. gimp – displayed normally. So with these type of invalid files it’s a bit hit and miss. It basically depends on whether the program knows to skip unrecognized chunks at the start of the file or not. The JPEG standard is a mess, so this type of thing is not very surprising, but you’d think Adobe would test that the files they produce can be opened by common applications.

    Photoshop 6.0 never made that mistake. What are Adobe thinking?

  34. Bob K
    Posted Jan 9, 2007 at 4:55 AM | Permalink


    Thanks for the info. I would have been befuddled for quite some time if you didn’t clue me in.

    You’re right. I did use photoshop 7.0 to save them. I use the Opera browser and it loaded them with no problem. I just tried loading one into Windows Picture and Fax Viewer. Had no problem. I also just tried IE6 and it wouldn’t load it. How screwy is that. MS recognizes the format for one of it’s other programs but doesn’t support it with it’s browser.

    I’ll have to try another program or change to a different format. I originally did the plots on a .bmp image and just wanted to sharpen the color by converting to cmyk color which .bmp doesn’t support. Figured .jpg would decrease the size also. Oh well. Back to the drawing board.

  35. Posted Jan 12, 2007 at 4:34 AM | Permalink

    Bob K, one thing you might want to try is see if Photoshop 7 still has the “save for web” feature that Photoshop 6 had.

    It was under the “file” menu and did two main things. One, it let you tweak the compression/colours in order to get a pleasing image that’s as small as possible for web use. Two, it stripped down the file so it didn’t include unnecessary bloat. I suspect that this mode may save a JPEG image that will load properly in IE6/Firefox. Yes, it is a bit odd that it works in some MS programs but not in IE6, but perhaps IE6 has a stripped down JPEG loader in order to speed up the browser. That’s just a guess, though.

    I’ll be curious to know if “Save for Web” is still there and whether it fixes the problem.

  36. Bob K
    Posted Jan 12, 2007 at 10:55 AM | Permalink

    Nicholas, it does have the “save for web” feature for jpg and other formats. Just tried it. File loaded into Firefox with no problem. I’m a novice when it come using Photoshop. So, I just experiment my way through doing something. Don’t have the inclination to really get into the program. Thanks again for the pointers.

  37. Liz
    Posted Jan 30, 2007 at 6:34 PM | Permalink

    Hi, I had a question for a project I am doing for school. What is the average length of a hurricane? Thank you.

  38. David Smith
    Posted Jan 30, 2007 at 7:24 PM | Permalink

    Hello, Liz. Here is some information:

    Over 1950-2000 there were an average of 6 Atlantic hurricanes per year, each having an average existence of 4 days. A typical hurricane moves at about 12 miles per hour. So, the path length of a typical hurricane is about

    12 MPH times 24 hrs/day times 4 days = 1,100 to 1,200 miles

    I hope this answers your question.

    The information comes from Wikipedia, Dr Bill Gray’s 2006 report and a book titled, “The Hurricane and Its Impact” (Simpson/Riehl, LSU Press, 1981)

%d bloggers like this: