Earlier we noted that the number of hurricane-days in 1933 actually exceeded the number of hurricane-days in 2005. However, the 2005 PDI was significantly higher than 1933 PDI, which indicates fairly trivially that hurricane speeds in 2005 were estimated to be higher than 1933. So here is a histogram of 1933 compared to 2005 wind-speed counts in 5 knot increments – first a 1933 histogram, then a 2005 histogram and then the same data side-by-side. Obviously the 2005 distribution is skewed towards higher wind speeds. When you cube the wind speeds in a PDI calculation, this difference is exacerbated.

The shape of the 2005 histogram is actually a little odd – it’s got a bimodal look that surprises me. Does anyone know what is the explanation for this?  There definitely seems to be a bias to measurement in even 10s that is more pronounced in the 2005 measurements and which seems surprising given that these are supposedly scientific measurements. The peak in the 110 knot tranche is also very remarkable in the 2005.

1933 was a notable year in other respects. As I recall, it is still the record year for U.S. temperature. It was the bottom of the Great Depression, with a dust bowl in the Prairies.



  1. Pat Frank
    Posted Jan 6, 2007 at 4:26 PM | Permalink

    Steve, if you plot the 2005 data according to month or season, do the two modes separate or is it a scatter? Also, in either case, if you fit the separate data sets with two Gaussians, I’m wondering if you’d get two similar Gaussians for each set, one centered near 70 and one near 120, with the 2005 data showing Gaussian areas that are more equal. I’d wonder, then, whether hurricanes are inherently bimodal, with every year showing something similar. I.e., areas shifting back and forth between the two Gaussian modes, where the modes really reflect energy dissipation.

  2. TAC
    Posted Jan 6, 2007 at 4:49 PM | Permalink

    SteveM, if I remember correctly, in 2005 the “loop current” rapidly transformed Cat 1 hurricanes into Cat 4-5 hurricanes, so perhaps there really were two distinct populations (pre- and post-loop). It would be interesting to look at the wind speeds wrt the loop current — I’ll try to put up a plot of wind speed versus latlong (possibly not until after I get the Christmas ornaments put away).

  3. David Smith
    Posted Jan 6, 2007 at 9:37 PM | Permalink

    In 1933 they guessed a storm’s intensity. In 2005 the intensities are measured.

    In 1933 there were a total of 5 pressure recordings of storms, all at landfall.

    In 2005 there were 649 composite pressure recordings of storms, almost all taken at sea, where the storms are generally more intense. The composites were assembled from thousands of pieces of data (satellite, aircraft, buoy, etc)

    In 1933 they guessed whereas in 2005 they measured. Apples-to-oranges.

    The 2005 bimodal appearance may have a physical basis. Even the 1933 data has a slight hint of it. There is evidence that intense hurricanes might be a “different breed of cat”. Klotzbach showed data that intense storms are often formed all-of-a-sudden from relatively weak hurricanes. Pressures fall rapidly and winds increase suddenly – there is not a smooth strengthening, it is all of a sudden over maybe 18 hours. Once they become intense, they stay that way, with 100+ knot winds, for days. I think this explains part of the bimodal appearance of the data.

  4. Nicholas
    Posted Jan 6, 2007 at 10:00 PM | Permalink

    Steve, I think you need to somehow process all the years from say 1900 to 2000, and work out if there is a systematic bias in the distribution which changes over time, or possibly even has hard transitions (e.g. when weather satellites were initially launched, etc.)

    I don’t know a way to do this without eyeballing all 100 distribution graphs, but perhaps you do.

    I think the result would be a lot more robust than just comparing two seemingly extraordinary years.

  5. George
    Posted Jan 6, 2007 at 11:46 PM | Permalink


    The warmest year was actually 1934, not 1933.

  6. Dane
    Posted Jan 6, 2007 at 11:49 PM | Permalink


    Heres one for you. There was a large earthquake in southern california that year, caused a REAL Tsunami in LA and Long Beach Beach Harbors. Also did much damage in the region as well. Any statistical relation to the climate issues? It may come up as a test question (still need to pass the CALIF test to get my Geologist licence) for me in March, you never know.

  7. David Smith
    Posted Jan 7, 2007 at 8:20 AM | Permalink

    RE #4 I’d bet that the earlier seasons, where people guessed, will have that Manx-cat appearance, with a stubby tail. Seasons that included actual measurements will have a longer, more natural-looking tail.

  8. Francois Ouellette
    Posted Jan 7, 2007 at 8:31 AM | Permalink

    I know it’s anecdotal evidence, but my grandfather used to pilot a ship down the rapids in the St-Lawrence river (between Cornwall and Montreal). This was a big tourist attraction at the time (see here for some details). However, for six years between 1934 and 1940, the the water level was too low (in the rapids) due to the high temperatures of the days, and there were simply no more trips until 1940. NOte that ships had been doing that trip for well over a hundred years so that was really an exceptional period weatherwise.

    We can’t compare with today, however, because since then they built the St-Lawrence Seaway and a few dams upstream that control the water level (which is now too low for any ship).

  9. McCall
    Posted Jan 8, 2007 at 12:22 AM | Permalink

    re: 5 “The warmest year was actually 1934, not 1933.
    And the warmest year at the NYC Central Park temp station was 1932, 33, 34 other?

  10. ET SidViscous
    Posted Jan 8, 2007 at 2:28 AM | Permalink


    India (and I assume the surrounding areas) are recieving a record cold spell. http://www.hindu.com/thehindu/holnus/002200701070313.htm. It doesn’t say in the article, but the by-line gives the previous record as 1935.

    So while the Eastern portion of NA is in a record warmth, Southern asia is in Record cold. The same was true 80 years ago. I realize it’s a small amount of data points, but is there a possibility of some kind of inverse correlation.

    The desnity of weather stations in North America compared to Asia is a well known issue.

    I wonder if it would be possible to compare the instrument record, and it’s trend, with the increase of data points (weather stations) comparing developed areas to under-developed areas.

  11. Posted Jan 8, 2007 at 5:22 AM | Permalink

    David Smith: I think that’s likely, but to be scientific someone should actually check that.

    IMO if it turns out that there are clear shifts in distribution of hurricanes at various times, it indicates either that the data is inconsistent due to changes in the way it was determined, or that there was a climate shift. Either way, I think it’s worth writing a paper on this discovery, if as we suspect it’s a robust one and not just a fluke of the data points picked above.

  12. David Smith
    Posted Jan 8, 2007 at 5:50 AM | Permalink

    RE #11 I agree that it would be a good thing for someone scientific to check it out. Perhaps that has already been done, and dismissed.

    When I get time later this week I’ll plot the wind histograms for cat 4 and 5 storms.

  13. Jeff Weffer
    Posted Jan 8, 2007 at 6:34 AM | Permalink

    The climate models need to be able to explain to MWP, the LIA, the 1930s warmth, the temperature decline of the 1940s to 1970s and even the ice ages before we should trust them.

    I mean it should be a basic requirement shouldn’t it? How can you build a model predicting the future when your model doesn’t even take into account trends of the recent past?

  14. Francois Ouellette
    Posted Jan 8, 2007 at 7:40 AM | Permalink

    The 1930’s are interesting in that there is plenty of evidence that they might have been warmer than today. Yet that’s not what the current surface temperature data are telling us. If one were to investigate the reliability of today’s data w/r to, for example, urban heat island effects etc., a revisitation of the 1930’s would be a good place to start. How good was the coverage in the 1930’s (e.g. in the Arctic and Antarctic), and what assumptions were made to estimate the global temperatures that might have under- or overestimated it? The fact that the coverage and the data were not then as good as today, but that we nevertheless have good enough records to know by how much, would make it a good exercise. For example, if you were to undersample today’s data to emulate the 1930’s, and give them the same kind of error bars (add noise to them), what would you get for today’s temperatures? Always remember that we’re talking fractions of a degree here. My own feeling is that it wouldn’t take much of a correction to make the 1930’s warmer than today.

  15. Michael J
    Posted Jan 8, 2007 at 10:16 AM | Permalink

    Francois you point out the real problem with comparing measurements between time periods. There are so many variables that would have to be factored in to get a “true” figure and even then there would be a rather large +/- necessary. The accuracy of the measuring device, heat island effect, location changes, more locations coverage, etc… make it virtually impossible to come up with answers with any confidence. Unless, and until, the measurements are using the identical standards, then the amnount of increase/decrease is subject to wide variations and/or interpretation.

  16. Nordic
    Posted Jan 8, 2007 at 10:32 AM | Permalink

    Having grown up in Vermont, when I think of hurricaines and the 1930’s I immediately think of the great hurricaine of 1938. A Wikipedia article is here: http://en.wikipedia.org/wiki/New_England_Hurricane_of_1938

  17. Steve Sadlov
    Posted Jan 8, 2007 at 10:42 AM | Permalink

    RE: #14 and 15 – During this period, my grandparents were on a farm in the midwest. We know its current owners and can make a direct comparison. Back in the 30s, there was no electrical grid there yet, and the house was heated using a lone wood stove. Neither was there indoor plumbing. During late 1940s, the grid finally arrived. During the 50s, central forced air heat and indoor plumbing were installed. The outbuildings never had any heat while my grand dad had it and since he sold out, the land has been rented to nearby farmers and there have been no farmers living on the actual property. But newer owners have added lighting to the barn, and possibly heat. The current owners are truckers and have upgraded the old farm house into a typical suburban dwelling, complete with whiz bang lighting, a big screen TV, etc. They also converted all the old gravel drives and pads into concrete ones. Some of the nearby major highways have become suburban strip development zones. The road the farm is on was unpaved until the 1980s. Imagine biases and local energy flux affecting temperature readings in the area, then and now.

  18. Chris H
    Posted Jan 8, 2007 at 11:15 AM | Permalink

    About 15 years ago I was doing some work for the gas company in a medium sized city. One of the projects I did for them was a software package to predict gas consumption. While I was looking at their historical data, I noticed that the mean value of one of their three temperature sensors had jumped by about half a degree a couple of years previously. When I asked about it, I was told that a three story building had been put up on the opposite side of the street from the sensor. It really doesn’t take much to change average readings by a half a degree.

  19. trevor
    Posted Jan 8, 2007 at 12:03 PM | Permalink

    Re #14,15: What’s up with you guys? Don’t you know that Phil Jones has addressed all of these issues for us, and come up with the answer at a high degree of confidence?

    Only problem is that Uncle Phil refuses to explain his analysis, adjustments, release his data etc. “I am a climate scientist. You can trust my work”. Yeah right, Phil.

  20. Steve Sadlov
    Posted Jan 8, 2007 at 12:56 PM | Permalink

    RE: #19 – According to Dr. Jones, “rural” locations like gramps’ farm constitute some sort of stable baseline, against which “urban” locations can be corrected. Oooops! (or should that be “DOH!”?)

  21. beng
    Posted Jan 8, 2007 at 1:10 PM | Permalink

    RE 14:

    The 1930’s are interesting in that there is plenty of evidence that they might have been warmer than today.

    Some winters in the 30s in the US mid-atlantic were remarkably warm (interspaced w/some snowy ones). Virginia’s all-time January high occurred ~1930 in Roanoke — an astonishing 87F (31C).

    Imagine the media weeping & knashing if that happened today.

  22. Steve Sadlov
    Posted Jan 8, 2007 at 2:28 PM | Permalink

    RE: #21 – Another place well known for winter warm / hot outbreaks is coastal and near coastal Southern California. It can literally get into the 80s any given day of the year there, when the conditions are just right. That’s what we can blame certain stereotypes about California weather on – LOL!

  23. David Smith
    Posted Jan 8, 2007 at 7:20 PM | Permalink

    Another thought on the odd wind distribution – many stronger storms crash into land and weaken quickly. There’s no smooth decline from intense levels, instead the decline is rather fast, perhaps 12 hours.

    So, the intense storms ramp up quickly (mentioned before) and many ramp down quickly, due to hitting land, distorting the appearance of the histogram.

%d bloggers like this: