The Atlantic Hurricane season

‘Tis nearly the Atlantic hurricane season, a traditional time for battening down the hatches, making sure the family are safe in the storm shelter, and making sure you are safely marooned in the local bar (this is what I did during one hurricane in the Bahamas in 1995).

‘Tis also the time for people to pop out of the wordwork to exclaim that this is exactly what we expect from our climate models, and that the climate models predict more storminess, more severity, more damage, more "Durm und Strang" and so on.

Since we like to check facts here on climate audit, I thought I’d check the latest posting on realclimate regarding hurricanes by looking back at the original data. Here’s what I found:

Referring back to the page containing the historical data regarding hurricanes in the 20th Century and the latter part of the 19th Century, here are the records (with my emphasis on the dates):

1. Busiest Hurricane Season Ever for the U.S.: The 1886 hurricane season has been analyzed to be the busiest on record for the continental United States. Seven hurricanes were recorded to have hit the U.S.: a Saffir- Simpson Hurricane Scale Category 2 hurricane into Texas and Louisiana in June, two Category 2 hurricanes into northwest Florida in June, a Category 1 hurricane into northwest Florida in July, the Category 4 "Indianola" hurricane into Texas in August, a Category 1 hurricane into Texas in September, and a Category 3 hurricane into Louisiana in October. The previous busiest hurricane season for the United States was 1985 with six landfalling hurricanes.

2. Extremely busy Decade for the U.S. Atlantic seaboard: The 1890s were one of the busiest decades on record for the Atlantic seaboard of the United States. Four major hurricanes impacted the coast from Georgia northward – the 1893 Category 3 "Sea Islands Hurricane" in Georgia and South Carolina, another 1893 Category 3 in South Carolina and North Carolina, an 1898 Category 4 in Georgia, and a 1899 Category 3 in North Carolina. Only the decade of the 1950s had more strong hurricanes making landfall along this part of the coast, going back to 1851 when reliable records began.

3. Cycles of hurricane activity: These records reflect the existence of cycles of hurricane activity, rather than trends toward more frequent or stronger hurricanes. In general, the period of the 1850s to the mid-1860s was quiet, the late 1860s through the 1890s were busy and the first decade of the 1900s were quiet. (There were five hurricane seasons with at least 10 hurricanes per year in the active period of the late 1860s to the 1890s and none in the quiet periods.) Earlier work had linked these cycles of busy and quiet hurricane period in the 20th Century to natural changes in Atlantic Ocean temperatures.

4. Georgia major hurricanes: During the 20th Century, Georgia did not have even a single major hurricane make a landfall along its coast. However, such absence did not continue back to the 19th Century. In contrast, Georgia experienced three major hurricanes in the later half of the 19th Century: a Category 3 in 1854 near Savannah, the Category 3 "Sea Islands Hurricane" in 1893 that killed 1000-2000 people near Savannah and a Category 4 in 1898 near Brunswick. Knowledge that such strong hurricanes have impacted this portion of the coast (and will undoubtedly hit again) is important for residents of Georgia to plan for the future.(See new NOAA Technical Memorandum by Sandrik and Landsea(2003).)

5. New England major hurricanes: Despite records showing six major hurricanes impacting New England in the 20th Century, the extension of hurricane analyses back to 1851 only show one major hurricane for the region in the second half of the 19th Century: 1869 hurricane which impacted Rhode Island and Connecticut. Thus it was a relatively quiet period for New England from the 1851 to 1910.

6. First time categorization of catastrophic 19th Century U.S. landfalling hurricanes: Several catastrophic hurricanes in U.S. history were categorized for the first time by the Saffir-Simpson Hurricane Scale. These included: the "Chenier Caminanda Hurricane" that struck Louisiana in 1893 and killed about 2000 people was assigned a Category 4 at landfall; the 1893 "Sea Islands Hurricane" killed 1000-2000 people in Georgia and South Carolina was ranked a Category 3 for its impact in both states; a hurricane in 1881 that also impacted Georgia and South Carolina and killed about 700 people was assigned Category 2 status. These hurricanes rank #2, 4 and 5, respectively, in the largest number of fatalities for U.S. landfalling tropical storms and hurricanes ever.

7. Strongest U.S. landfalling hurricane of the 1851 to 1910 era: The 1886 "Indianola" hurricane was analyzed as having 155 mph maximum sustained winds, a Saffir-Simpson Hurricane Scale Category 4 (approaching Category 5) and was the strongest to strike the United States between 1851 and 1910. This hurricane destroyed the town of Indianola, Texas due to its winds and 15′ storm surge and the town was never rebuilt. This was also the strongest hurricane of record anywhere in the Atlantic, Gulf of Mexico or Caribbean Sea during the same time period. (No Category 5 hurricanes were recorded to have hit the United States between 1851 and 1910. However, records are somewhat incomplete along in Gulf coast and Florida because there were some coastal regions with few to no inhabitants, thus there may have been some systems mis-diagnosed in intensity in that period.) 31 major (Category 3, 4 and 5) hurricanes are recorded to have hit the United States from 1851 to 1910.

8. Longest lasting hurricane on record: Storm #3 (also known as the "San Ciriaco" hurricane for its impact in Puerto Rico in 1899 has been re-analyzed to now tie the record for longest lasting tropical cyclone in the Atlantic basin. It began on August 3 in the tropical North Atlantic, hit Puerto Rico as a Category 4 hurricane on the 8th, hit North Carolina as a Category 3 hurricane on the 18th, transformed into an extratropical system north of Bermuda on the 21st, redeveloped into a tropical storm on the 26th, went through the Azores Islands as a Category 1 hurricane on the 3rd of September and finally dissipated as an extratropical storm on the 4th. It was a storm system for 33 days and a tropical storm or hurricane for 28 of those days. This ties the record with Hurricane Ginger of 1971, which also was a tropical cyclone for 28 days.

9. Most hurricanes ever in one day: On August 22, 1893, four hurricanes were occurring simultaneously: storm #3 approaching Nova Scotia, Canada, storm #4 between Bermuda and the Bahamas, storm #6 northeast of the Lesser Antilles, and storm #7 west of the Cape Verde Islands. Storm #4 would end up making a direct hit on New York City as a Category 1 hurricane two days later and storm #6 ending up hitting Georgia and South Carolina as a Category 3 hurricane (the "Sea Islands Hurricane") and killing 1000-2000 people. The only other known date with four hurricanes occurring at the same time was September 25, 1998, when hurricanes Georges, Ivan, Jeanne and Karl were in existence.

All fascinating, but are the number of storms getting larger and more severe (as we are being told to expect due to global warming)?

The reason I ask is that these events appear correlated with cold periods rather than warm, which is what you’d expect under the hypothesis that the greater the difference between the warmth of the tropics and the cold of the major polar airmasses, the greater the storminess.

25 Comments

  1. Michael Mayson
    Posted Jun 10, 2005 at 11:04 PM | Permalink

    Roger Pielke, Chris Landsea and others have a recent paper accepted by the Bulletin of the American Meteorological Society on this very subject

    http://sciencepolicy.colorado.edu/admin/publication_files/resourse-1762-hurricanes%20and_global_warming.pdf

    “”… claims of linkages between global warming and hurricanes are misguided for three reasons.”

  2. Joseph
    Posted Jun 30, 2008 at 12:01 PM | Permalink

    I take a look at the available data here.

  3. Bob Koss
    Posted Jul 1, 2008 at 8:05 AM | Permalink

    Joseph,

    You might want to consider how observations have changed since 1851. The older the data the more suspect the true storm counts become.

    Very sparse observations east of 60W and more than 300nm from land in the early record. Thirty storms have only one observation between 1851-1870 with ten of those category 1 and another four of category 2. It is unknown how many other storms were missed entirely.

    Prior to 1871 they recorded no storms having wind speed of less than 40kt. With satellites now able to monitor the entire Atlantic basin, they start tracking them at 10-15kt wind speed and record them even if they stay subtropical. Prior to 1968 no subtropical storms were entered into the database. There are 24 of that type since 1968. Those also appear in the yearly storm count.

    Lots of things to adjust for before making a statistical analysis and no straightforward way to make the adjustments. Browse some of the other hurricane threads for further information.

  4. Posted Jul 2, 2008 at 1:33 PM | Permalink

    The older the data the more suspect the true storm counts become.

    That’s one of many reasons why I suggest controlling for trends. You’ll note that in my analysis
    I completely standardize both temperature and number of storms for their time trends (given by
    third-order polynomial models).

    What I compare are residuals (“more or less than expected relative to time trend”). To illustrate
    this graphically, I put the residuals in a scatter and then do a least squares fit of the scatter,
    which, as it turns out, has a statistically significant slope. (I suspect that if I were to
    use only a July-August average temperature for each year, the correlation would be even
    clearer).

    To further clarify that, if I try to correlate year vs. named storm residuals (or “standardized
    named storms” if you will), there is no trend at all. It is flat.

    I’ve done the same analysis for earthquakes, BTW. In that case, the residual scatter trend
    is completely flat, which is exactly what you’d expect if the null hypothesis is true.

  5. Kenneth Fritsch
    Posted Jul 2, 2008 at 5:34 PM | Permalink

    Joseph,

    I think Bob Koss was referencing the changing detection capabilities over time that can readily confound the correlation of SST with TC activity since both are probaly increasing at the same time.

    Now, if you have a statistical method that overrides some of the observations of east and west of 60W longitude to which Bob Koss is referring, that would indicate little or no trend in TC activity over time, perhaps you could make a clearer case for your methods here and/or contact the climate scientists Webster and Holland. They might be interested in using your methods.

  6. Posted Jul 4, 2008 at 12:52 PM | Permalink

    I think Bob Koss was referencing the changing detection capabilities over time that can readily confound the correlation of SST with TC activity since both are probaly increasing at the same time.

    I understand. What I’m saying is that by detrending, these sorts of time-dependent measurement
    errors should not bias the correlation. They are noisy data, but they should not be biased
    data, since the time trend is controlled for. In other words, instead of helping the
    correlation, the noisy data would tend to prevent a statistically significant finding.

  7. Kenneth Fritsch
    Posted Jul 4, 2008 at 1:26 PM | Permalink

    Joseph,

    What is the R^2 of your regression of SST residual versus storm residual? Also could you plot your SST residual versus the years of the time series you studied and do the same for the storm residuals?

    How would one determine that a third order polynomial captures the trends involved in either the SST or number of storms? Why not a second order or higher order polynomial or several linear plots over the time period of interest? Why could not the residuals from a third order polynomial be an artifact? A sensitivity test would require you to test other fitting procedures unless you can a prior justify a third order polynomial.

  8. Posted Jul 5, 2008 at 12:42 PM | Permalink

    What is the R^2 of your regression of SST residual versus storm residual? Also could you plot your SST residual versus the years of the time series you studied and do the same for the storm residuals?

    R^2 would be 0.5 and 0.2 respectively.

    I’ve found a third order polynomial fit to be somewhat better than a second-order fit.
    In detrended fluctuation analysis, the detrending is usually done with a simple linear fit.
    I think closer fits are better, in terms of controlling for potential coincidence.

    I do have plots of year vs. temperature residuals and storm residuals. This is
    a basic confirmation I do. The trend is flat in both cases. If it isn’t flat,
    this invariably indicates there’s an error or imprecision in the equation. If
    I post a new analysis that addresses some of the points raised, I’ll be sure to
    include the detrended graphs.

    Evidently, there should be no statistical association between temperature residual
    vs. some independent property of the year. If there is one, you have to ask
    yourself why there would be one.

    How would one determine that a third order polynomial captures the trends involved in either the SST or number of storms? Why not a second order or higher order polynomial or several linear plots over the time period of interest?

    Like I said, I’ve found the third-order polynomial fit to be slightly better.
    You could do the analysis with a linear fit or a second-order polynomial, or
    a sixth-order polynomial fit if you will, and I don’t think the result will
    be in a different direction. I suppose a basic check that might be done to
    rule out coincidence is whether the correlation starts to disappear as the trend
    fits get closer. I don’t think they will.

  9. Kenneth Fritsch
    Posted Jul 5, 2008 at 3:16 PM | Permalink

    Joseph,

    I could generate your basic data from scratch I suppose, but just to be sure we are talking about the same data could you please link me to the data that you used in your analysis, i.e. the storm frequecies and SST time series.

    I was confused by your answer to R^2 for storm frequency to SST residuals. That would be one R^2 not two as you indicated. It would appear that your residuals are merely indicating a correlation on an annual basis between SST and storm frequency, but without an R^2 we would not know how good that correlation is.

    When I have looked at these relationships over time using what I call Easy to Detect Storms (in an attempt to remove the detection bias) I have found that while storm activity can correlate with SST over some time periods when I look over longer time periods the correlation breaks down. This was not the case when using wind shear and prevailing wind where the correlations help up. What I think I see with the SST versus storm activity over short periods is a confounding of the long term cyclical nature of storm activity (it being on the rise) with a coincidental rise in SST.

    Most of the correlations I have seen with storm activity versus SST have been better when a moving average was used and thus indicating that the annual storm activity to SST relationship (if it exists to any extent) was suffiently noisy to obscure a signal.

  10. Kenneth Fritsch
    Posted Jul 5, 2008 at 3:27 PM | Permalink

    Joseph,

    Another I point I failed to mention is how well is your method removing any cyclical component of the times series and would not that component as well as trend have to be removed for your finding the residuals. Willis E did that here at CA for storm activity over long time periods and found a cyclical component of approximately 60 years, as I recall. He used a trend and cyclical component removal to obtain a much improved fit of the storm frequencies to a Poisson distribution. I have done similar analyses using the reoccurring AMM effects on Easy to Detect TCs to obtain improved fits to Poisson distributions for TC frequencies.

  11. Posted Jul 6, 2008 at 7:58 AM | Permalink

    Yes, matching cycles could be an issue, but then you have to ask yourself why there
    are matching cycles. I think it’s because there’s an effect, not because of
    coincidence. That’s also why closer fits should be better. If you do something like
    a sixth-order polynomial fit, the cycles should become part of the trend.

    I plan to do a follow-up post addressing some of the points raised. I’ll upload the
    spreadsheet then.

  12. Kenneth Fritsch
    Posted Jul 6, 2008 at 6:12 PM | Permalink

    Re: #11

    Joseph, I took the Main Development Region SSTs for August, September and October and the NATL TC counts from the Mann authored paper linked below.

    http://www.meteo.psu.edu/~mann/Mann/articles/articles.html

    The data are for the time period 1870-2006. I plotted the SST and TC counts against years for the time period and than attempted several different trending schemes (which are reported below) to determine which gave the largest R^2. It can be seen that a sixth order polynomial gave the best fit for those attempted.

    TC Counts:

    Trend = linear and R^2 = 0.14; Trend = Poly and R^2 = 0.19; Trend = Poly3 and R^2 = 0.19; Trend = Poly 4 and R^2 = 0.21; Trend = poly5 and R^2 = 0.28; Trend = Poly6 and R62 = 0.28.

    SST:

    Trend = Linear and R^2 = 0.31; Trend = Poly2 and R^2 = 0.38; Trend = Poly3 and R^3 = 0.38; Trend = Poly4 and R^2 = 0.44; Trend = Poly5 and R^2 =0.50; Trend = Poly6 and R^2 = 0.54.

    Using the 6th order polynomial for each the SST and storm counts, I than calculated the residuals for each and plotted with TC count residuals versus SST residuals for the period 1870-2006. I obtained an R^2 = 0.11 and trend rate of 2.4 storms per century.

    I also plotted the residuals for SST and TC counts versus year over the period, 1870-2006, and found absolutely no trend in the scatter plots.

    The plot of TC counts over time shows no trend from 1870-1940 and then a trend of 6.4 TC counts per century with an R^2 = 0.10. A plot of TC counts versus SST over the time period 1870-2006 shows a trend of 6.3 storms per degree C temperature anomaly with an R^2 = 0.31. Over the period 1870-1940 the trend is 4.8 with an R^2 = 0.16, while for the period 1941-2006 the trend is 6.9 with an R^2 of 0.29.

    Based on these trend differences I looked at the period of 1941-2006 for the TC count residuals versus SST residuals and obtained a trend of 4.4 TC counts per degree C temperature anomaly with an R^2 = 0.09. There were three data points which were obviously highly leveraged with all three having a TC count residuals less than -5 counts. When these data points were removed from this plot the resulting trend went to 1.9 TC counts per degree C temperature anomaly and the R^2 to 0.02.

    Based on these results and particularly when drilling down to what one might consider the more critical data, I do not see much in the way of a good correlation between residual TC counts and SST. That should not discourage anyone from looking at these data in different ways with this approach. Neither am I qualified to pass judgment on the statistical validity of the methods used here by Joseph.

  13. Joe Solters
    Posted Jul 7, 2008 at 6:13 AM | Permalink

    Re: Recent hurricane classification (or Mis-classification). Today’s HRC/NOAA Discussion no. 17, (5:00am 7-7-08) on hurricane Bertha states that while the “objective Dvorak estimates remain below the hurricane threshold”, the “subjective Dvorak intensity is 65kt”. HRC thus calls tropical storm Bertha a hurricane and the media go whacko again. It only takes an estimate of one degree to make a hurricane out of a tropical storm. Like last season, this storm is probably tracking straight up the mid-Atlantic, far offshore, into relatively colder SST, and will die there. Again, like last season, Bertha may never actually attain hurricane intensity, or may do so for only a few days. But it’s fovever in the books and stats as a hurricane. HRC needs a better classification system to describe these weak mid ocean tropical storms.

  14. Tony Edwards
    Posted Jul 7, 2008 at 7:12 AM | Permalink

    Joe Solters #13, do you suppose that Bertha is the first of this season’s “Tiny Tim” storms? See

    http://www.climateaudit.org/?p=2529

  15. Joe Solters
    Posted Jul 7, 2008 at 10:58 AM | Permalink

    Re: 14 Bertha may be more than a “tiny tim”; it’s intensity has been upgraded to 80kt on the basis of satellite estimtes and is forecast to go to 90kt shortly. My point was that HRC is very losey-goosey about categorizing low-intensity (tropical storms) hurricanes. Maybe Steve’s ACE index is more meaningful since it appears to use duration as an additional criterion. Thanks for the site tip.

  16. Posted Jul 9, 2008 at 10:14 AM | Permalink

    Kenneth: I went ahead and tried that 6th-order analysis with the data from 1851 to 2006.
    You are correct that it loses statistical significance, even though there’s still
    a trend in the expected direction. I get a slope of 1.2. My initial impression
    was that either we’ve in fact controlled for a subtle coincidence with the closer
    fit (and there’s really no association) or that my conjecture that closer fits better
    control for coincidences is mistaken, or rather, in trying too hard to control for
    coincidence we’re adding noise. In fact, ‘This approach has been criticized by some
    authors because of loss of valuable information in course of such “detrending”.’
    [ http://www.statistics.com/resources/glossary/d/dtrndca.php ]

    But then I tried the following, and you can try it too with your data. Correlate the
    temperature residuals with the named storms residuals one year later. The slope I get
    in this case is 4.892, enough to push it over statistical significance. It’s 3.126
    two years later, and it deteriorates after that, although there’s somewhat of a cycle
    which I think is expected due to autocorrelation.

    I can’t really explain why there would be an effect 1 year later, but I’ve found this
    to be the case in a few different analyses.

    When you do the sixth-order fits, do you see that the shapes of the fits match between
    the trends? Don’t you find that interesting?

    I also did the analysis with a standard linear detrending starting in 1900 (in order
    to righ censor data that is bad as suggested by Bob Koss). In this case the effect
    found is about 6 more storms for every 1 degree.

  17. Kenneth Fritsch
    Posted Jul 13, 2008 at 9:49 AM | Permalink

    Joseph, I did some more analyses which I will summarize here. Before I start I have to note that I could not find, on searching the internet, an analysis for detrending (or removing cyclical content) from a regression using the residuals for two dependent variables from their own time series. Are there any statisticians here willing to comment on this?

    I looked at the residuals from a sixth order polynomial for TC counts to determine how well they might fit a Poisson distribution over the period 1870-2006. The distribution using the raw TC counts over this period yields a chi square goodness of fit p = 0.00, while using the residuals gives a fit with a p = 0.59. (I had to determine a zero point for the residual counts since a Poisson distribution cannot have negative values. For that point I used the mean count for the raw count and added it to the residual counts). This fits with my previous analyses using Easy to Detect storm counts to remove trends due to detection changes and AMM positive or negative to remove some of the cyclical component in storm counts.

    I also noted that the TC count residuals from the sixth order polynomial have seven high leverage points, all of which are the highest of all the count residuals. Two of the seven TC residuals had corresponding SST residuals which were negative (low ranking) while the other five had corresponding SST residuals that were in the 75th percentile and higher rankings, although these apparent outliers did not have corresponding outlier SSTs.

    As a further check on the residual method I used the 1870-2005 hurricane count times series with a sixth order polynomial. The results, which are listed below, show that the residual methods give a positive trend for hurricane counts with SST and very low R^2 values. The fit to a Poisson distribution using hurricane count residuals is very good which is again in agreement with previous analyses I made using a positive and negative AMM index. I need to analyze the residual hurricane counts in more detail as I did for the residual TC counts.

    Raw hurricane count versus time:

    1870-2006:

    R^2 = 0.04 and Trend = 1.3 counts per century.

    1870-1940:

    R^2 = 0.06 and Trend = -3.1 counts per century.

    1941-2006:

    R^2 = 0.03 and Trend = 2.2 counts per century.

    Hurricane residual counts versus SST residual counts:

    For a sixth degree polynomial for hurricane counts, R^2 = 0.18.

    1870-2006:

    R^2 = 0.08 and Trend = 2.7 counts per degree C.

    1941-2006:

    R^2 = 0.09 and Trend = 3.3 counts per degree C.

    Poisson chi square goodness of fit for period of 1870-2006 for hurricane counts:

    For raw counts, p = 0.15

    For residual counts, p = 0.95

    With a p = 0.95 for the fit to a Poisson distribution for the residual hurricane counts versus residual SST, I see little room for an actual hurricane count to SST relationship.

    What bothered me about using the residual TC counts (I need to analyze the residual hurricane counts) is that influential points (in the regression of residual counts versus residual SST) were all on the high side of the residual counts and it appears that a smooth curve, like a sixth degree polynomial, cannot approach connecting with these blips. One might expect that the corresponding SST values would be at the extreme end of those values, but that was not the case, even though they were on the average on the higher side.

  18. Kenneth Fritsch
    Posted Jul 14, 2008 at 8:59 AM | Permalink

    Re: #17

    I neglected to note previously that if one uses residual TC or hurricane counts and those counts fit a Poisson distribution than one cannot properly regress residual counts versus residual SST using linear regression. I have listed excerpts from a Sabbatelli paper below that describes the use of generalized linear models and estimates of a maximum likelihood values. The authors also talk about SST autocorrelation and the need to adjust the statistics for that relationship.

    I am sufficiently curious that I plan to simulate a perfect Poisson distribution randomly applied to the SST residuals and then linearly regressed to determine the probability of the occurrence of a substantial positive trend of the residual count to the residual SST with an R^ between 0.05 and 0.10.

    In the paper titled and linked below “The influence of climate state variables on Atlantic Tropical Cyclone occurrence” rates by Thomas A. Sabbatelli and Michael E. Mann and published Sept 15, 2007, the authors note that:

    Any statistical approach to analyzing TC counts must respect the Poisson distributional nature of the underlying process (that is, that TC counts are characterized by a point process with a low occurrence rate). Our first approach employs Poisson regression [see e.g., Elsner et al., 2000, 2001; Elsner, 2003; Elsner and Jagger, 2006], a variant on linear regression which is appropriate for modeling a conditional Poisson process in which the expected occurrence rate co-varies with some set of state variables (e.g., indices of ENSO, the NAO, and MDR SST)…

    ..Poisson regression is a variant on linear regression appropriate for data such as TC counts for which the null hypothesis of a Poisson distribution is appropriate [see Elsner et al., 2000, 2001; Elsner, 2003; Elsner and Jagger, 2006 for further discussion]. Given a count series Y with unconditional mean rate m believed to follow a state dependent Poisson distribution, Poisson regression estimates a generalized linear model for the conditional expected rate of occurrence..

    .. Unlike ordinary linear regression, a closed-form analytical solution to equation (5) is not possible. However, it is straightforward to numerically estimate maximum likelihood values for the regression parameters..

    http://www.meteo.psu.edu/~mann/Mann/articles/articles.html

  19. Posted Jul 14, 2008 at 9:27 AM | Permalink

    I found a much easier and convincing way to illustrate the association, with a new graph,
    which you will find here:

    http://residualanalysis.blogspot.com/2008/07/hurricanes-and-global-warming-revisited.html

    Kenneth: I’m not sure how you can use a Poisson distribution to determine
    the association between two time series. You lost me there. I probably require
    further explanation.

    You also didn’t comment on the apparent 1 year lag.

  20. Kenneth Fritsch
    Posted Jul 14, 2008 at 2:36 PM | Permalink

    First of all I want to commend you for warning your blog readers before proceeding into your discussion mentioning my analysis here at CA with the following excerpt:

    I understand Climate Audit is one of the major AGW denial blogs.

    I think my statistical observations (and confirmed by Mann and Sabbatelli, who would never, ever, be accused of being denialists) on not using a linear regression when the dependent variable is shown to follow a Poisson distribution with time while the independent variable no doubt follows a normal distribution. It would not matter to the statistics how the distribution was derived. I think you need to read the paper I linked. The authors attempt to make the case that the state variables SST and NATL associated climate cycles when removed from the NATL TC counts significantly improve the fit of the counts to a Poisson distribution. It is my judgment that changes in SST are being confounded with changes in detection capabilities and have obtained some substantial improvement in the Poisson fits for TC and hurricane counts by looking at Easy to Detect storms (to remove detection changes) and AMM cycles in the MDR for NATL TCs. Also it is interesting to note that hurricanes (which should have been easier to detect going back in time) have smaller trend with time and with much reduced R^2 than TC counts. Using ACE for TCs from the west of longitude 60w (easier to detect going back in time) over time shows no trend while that to the east of 60w (more difficult to detect going back in time) shows a significant trend.

    Moving averages in my view are simple techniques acting as a visual aid for the graph readers when a trend is involved. Statistically they will not tell you any more than a regression will about a trend – they just sometimes make it easier to see. If the time series have much noise in them such that a MA is needed to see a trend that is also telling us that there is much noise, i.e. other effects are acting to significant extents. For TC and hurricane counts I am assuming that slate is wiped clean at the end of the old season and a new one appears for the current season. If this is the case than one would have a difficult time explaining lag effects and even using MAs. Your SST and TC count MA dips and valleys do not correspond well and as one might expect if SST was being confounded with detection capability changes.

    Joseph, I would feel better about your residual methods if you could provide me with a literature reference to a method used as you applied your method.

  21. bender
    Posted Jul 14, 2008 at 4:45 PM | Permalink

    #19

    I’m not sure how you can use a Poisson distribution to determine the association between two time series. You lost me there. I probably require further explanation.

    Poisson regression simply assumes the distribution of errors is poisson, as opposed to standard linear regression, which assumes a normal distribution. Easy to specify using R glm. Once you specify a poisson error structure, the dependent variable is regressed onto the independent variable as per usual.

  22. DeWitt Payne
    Posted Jul 14, 2008 at 5:18 PM | Permalink

    I still maintain that when looking for small changes (less than 2 sigma) in trends of time series with noisy data the well established technique (in industry) of Exponentially Weighted Moving Average control charts should be used. When the variable of interest is shown to follow a Poisson distribution, then Poisson CUSUM charts should be used. They would also be useful to show the effects of attempts to detrend data.

  23. Kenneth Fritsch
    Posted Jul 15, 2008 at 8:45 AM | Permalink

    Re: #22

    I still maintain that when looking for small changes (less than 2 sigma) in trends of time series with noisy data the well established technique (in industry) of Exponentially Weighted Moving Average control charts should be used. When the variable of interest is shown to follow a Poisson distribution, then Poisson CUSUM charts should be used. They would also be useful to show the effects of attempts to detrend data.

    DeWitt, I do not disagree with your methods, but my main concerns with looking at TC/hurricane counts and TC activity in general over time has been the need, in my mind, to remove cyclical effects like AMM, Nino and AMO and other effects like changing detection capabilities and SST and any confounding of detection capabilities with changes in SST. I would assume that if these effects are removed with reasonable completeness the count distributions should fit significantly more closely to an expected Poisson distribution and this is basically what I have found. The ACE index with similar corrections fits better to a normal distribution as one would expect for that variable.

    By the way, when I look at unadjusted TC raw counts versus SST there are long periods with little or no correlation, e.g., over the time period 1940-1990 the NATL TC count to SST shows a trend of 2.3 counts per degree C, but with an R^2 = 0.04. Compare that to 1870-2006 where the trend is 6.3 counts per degree C with an R^2 = 0.31. I say that observation is consistent with SST being confounded with some other influential variable.

  24. Jan Willett
    Posted Mar 4, 2009 at 8:38 PM | Permalink

    I’m trying to locate information about a storm in the spring of 1968 east of Bermuda which damaged my husband’s navy ship, the USS Swerve MSO 495. Do you have any suggestions where I might look for this information? Thanks.

  25. Bob Koss
    Posted Mar 5, 2009 at 11:40 AM | Permalink

    Memories can get hazy about details after such a long period of time.
    Here is the closest match in the NHC database.
    Hurricane Brenda passed 180 n. miles north of Bermuda on June 23, 1968 heading east. Nothing earlier in the season.
    Perhaps there is a ship history or website online.

Follow

Get every new post delivered to your Inbox.

Join 3,191 other followers

%d bloggers like this: