RealClimate logo

Tropical cyclone history – part I: How reliable are past hurricane records?

Filed under: — group @ 18 February 2008

Guest Commentary from Urs Neu

When discussing the influence of anthropogenic global warming on hurricane or tropical cyclone (TC) frequency and intensity (see e.g. here, here, and here), it is important to examine observed past trends. As with all climate variables, the hurricane record becomes increasingly uncertain when we go back in time. However, the hurricane record has some peculiarities: hurricanes are highly confined structures, so you have to be at the right place at the right time to observe them. Secondly, hurricanes spend most of their life in the open oceans, i.e. in regions where there are very few people and no fixed observations. This means that the reliability of the long-term hurricane record is dependent on who was measuring them, and how, at any given time. The implementation of new observation methods, for example, might have altered the quality of the record considerably. But how much? This crucial question has been widely discussed in the recent scientific literature (e.g. Chang and Guo 2007, Holland and Webster 2007, Kossin et al. 2007, Landsea 2007, Mann et al. 2007). Where do we stand at the moment? This post will concentrate on the North Atlantic, which has the longest record.

The official Atlantic hurricane record provided by the U.S. National Hurricane Center (HURDAT) represents the reference data base for most of the studies and contains all observed TCs, their individual tracks and intensity. The record has been extended back to 1850, and the earlier periods (until 1914) have been re-analysed in recent years. Reanalysis work continues, and updates and corrections are regularly reported.

This record contains two important abrupt inhomogeneities. The introduction of air reconnaissance flights in 1944 and the launch of the first geostationary satellite ATS-I in December 1966 mark two important improvements of measurement facilities and thus the observational coverage of the area under examination. Landsea (2007) claims a third one in 2002, since the new advanced microwave sounding unit (Quikscat) has lead to the retrospective detection of additional tropical cyclones in the last few years. Some also argue for placing the start of the satellite area later, in the mid-1970s relying on the launch of the GEOS-satellites. Furthermore, there are some changes in ship track patterns after 1914 with the opening of the Panama Canal and during the two world wars.

In addition, there are also a number of gradual observational improvements over time, e.g. the increasing quality of satellite images, or in earlier times the increase of the number of ship tracks or the growing population density on the coastlines, both of which enhance the probability that a TC would have been observed. And last but not least, there might be inhomogeneities due to the subjective component in the classification of tropical storms (the so-called Dvorak method, Dvorak 1984) which might lead to systematic differences between different forecasters. However, the homogeneous reanalysis of the last 23 years has shown that this subjectivity and improved observations has not lead to a noticeable alteration of the long-term trend for Atlantic storms (Kossin et al. 2007).

Climate change impacts on hurricanes generally focus on two key quantities, the frequency and intensity of storms. There is no reason to believe that both quantities will change similarly. In the discussion of past activity, the frequency is described by the number of tropical storms that occur annually in each basin. The maximum intensity (or even more complex metrics such as the Power Dissipation Index which integrates intensity over space and time; Emanuel 2005) is harder to measure, because it requires detailed information about the storm along the storm track over its entire lifetime. Therefore most discussions of historical trends have focused on tropical storm frequency/number or in some cases the number of intense storms (e.g. the number of major hurricanes).

Thus the key question is: how many storms did we miss in the past? Recently there have been a number of attempts to estimate this ‘undercount bias’ in the tropical storm record. These attempts have included:

  • reconstruction of the observational bias by relating past observation density (e.g. ship tracks) to modern storm tracks
  • using the relation between total TC number and better known subsets of the TC record (e.g. landfalling storms)
  • using relationships of known underlying variables (e.g. relevant climate indices) to annual TC numbers to create a ‘predicted’ TC record, and compare it to the observed record.

All these approaches have a common caveat, namely the assumption that the relationships they rely upon are constant over time. The validity of this assumption therefore has to be examined in any studies using such approaches. Let us consider some recent such studies:

(1) Landsea [2007] performed a simple analysis to estimate the observational bias for the time from 1900 until the begin of the satellite period in 1966. He examined the percentage of tropical cyclones that struck land (PTL) and notices a considerable difference between the time periods 1900-1965 (pre-satellite period, PTL=75%) and 1966-2006 (PTL=59%). He suggests that this difference indicates an underestimation of about 2 tropical cyclones per year before 1965.

Unfortunately, Landsea does not discuss the evolution of PTL before 1900 (left side of the red dashed line in Figure 1). If PTL really is a proxy for underreporting due to decreasing observation density, PTL should further increase before 1900. However, there is a decrease. The period 1851-1899 has an average PTL of 67%, the period 1851-1885 even has an average PTL of 61%, which is not significantly different from the satellite period after 1966 (59%).

Figure 1. Percentage of all reported tropical storms, subtropical storms, and hurricanes that struck land 1851-2006. Extension of Fig. 2b in Landsea [2007].

Therefore it is questionable if PTL really is a reliable proxy of underreporting. One might argue, as Landsea implicitly does, that after 1900 the population density on the coasts and islands was high enough to catch all tropical cyclones, and the underreporting is only due to decreasing density of shipping tracks, while before 1900 also some tropical cyclones that struck land were missed. However, before 1900 not only population density but also shipping track density was lower and therefore PTL likely should be about at the same level as after 1900 but not significantly lower. In addition, Landsea contradicts that argument himself by stating that even in 2005 a retrospective analysis reveals that there was a tropical cyclone that made landfall in a sparsely populated area and was therefore not initially included as a landfalling storm.

Moreover, in another study (Holland, 2007) it has been shown, that the natural variability of TC numbers is different for different regions of the tropical Atlantic. Therefore, the proportion of TCs over the open sea also varies naturally, altering the landfall proportion for reasons unrelated to observation bias. Thus PTL seems not to be a good proxy for observational biases. The study showed that the decrease in PTL in the 1960s is mainly due to a decrease of TC number in the Caribbean and the Gulf of Mexico. Since these regions are well observed by dense ship tracks and a number of islands, this decrease is very unlikely to be mainly an observational bias.

(2) Sticking to ship tracks, Chang and Guo (2007) performed a different type of analysis: They compared the ship tracks of the years before the satellite era with TC tracks of recent years. For example, they took the ship tracks of the year 1917 and overlaid the TC tracks of 1999 and determined how many of the 1999 tropical cyclones would have been observed if the ships had navigated as in 1917. By comparing the years before the satellite era to all the ‘satellite’ years (after 1965) they obtained statistics for how many TCs would likely have been missed in earlier years if the distribution of TC tracks had been similar to that during the satellite period. In this way they estimated a TC undercount of about 2 per year in the period 1903-1914, 1-2 per year 1915-1925 and of less than 1 per year from 1925-1965.

The adequacy of this estimation depends on several assumptions: a) that the distribution of the hurricane tracks is about the same in the satellite period than in the periods before. As we have seen before (Holland 2007), there are shifts in the regional distribution over the 20th century. This could influence the estimation of underreporting both in a positive or negative way; b) that all landfalling storms have been detected correctly. It is likely that some of the landfalling storms (or their true strength) might have been missed, which would bias the estimate artificially low; c) that ships did not circumnavigate the storms (e.g. on the basis of predictions). If this was the case, the undercount estimate might be too conservative; d) that ships measured wind correctly. Because this error is random, a significant systematic bias of the undercount estimate is unlikely; and e) that all ship tracks are recorded in the database used. There may have been other, unrecorded ship tracks, which might have detected a storm missed by known ship tracks. This would lead to an exaggerated undercount estimate. Altogether these assumptions likely tend to somewhat underestimate the undercount.

(3) In a third study (Mann et al. 2007; in full disclosure, I was a co-author of this paper), an alternative approach was used employing the statistical relationship between Atlantic TC numbers and three climate variables influencing Atlantic TC activity (1. August-October sea surface temperatures over the main development region (“MDR”); 2. the El Niño/Southern Oscillation, and 3. the North Atlantic Oscillation) during the modern period of reconnaissance flights and satellite observations (1944-2006). This relationship was then used to predict TC numbers for the period 1870-1943 as would be expected from the behavior of the three climate variables used over that period. These estimates were then compared to the observed TC record, the difference providing an estimate for the underreporting. The results yielded an undercount before 1944 of 1.2 TCs per year (best estimate), with a range of 0.5-2 TC per year.

This analysis also relies on several assumptions. Namely, that a) the underlying climate variables do not contain artificial trends or other inhomogeneities that might bias the results. That the results were insensitive to using different alternative SST datasets, or switching the role of training period and prediction period (i.e. training on 1870-1943 and predicting for 1944-2006) was taken, however, as evidence against this being a significant issue. b) that there are no long-term trends in other climate variables not included in the statistical model, but that do influence TC numbers (e.g. wind shear or vertical stability) or might influence the relationships between TC numbers and the variables that are used. Although such an influence cannot be excluded, cross-checks that were performed such as statistical validation and switching the order of training and prediction intervals, seem to argue against this being a problem.

In summary, according to current knowledge, the best estimate for the underreporting bias in the hurricane record seems to be about one tropical cyclone per year on average over the period 1920-1965 and between one and three tropical cyclones per year before 1920. With only a few years of data available, the influence of Quikscat analyses after 2002 as discussed by Landsea, is difficult to as yet meaningfully estimate.


Chang, E. K. M., and Y. Guo (2007): Is the number of North Atlantic tropical cyclones significantly underestimated prior to the availability of satellite observations? Geophys. Res. Lett., 34, L14801, doi:10.1029/2007GL030169.

Holland, G. (2007): Misuse of landfall as a proxy for Atlantic tropical cyclone activity. Eos Trans. AGU, 88, 349.

Holland, G.J., and P.J. Webster (2007): Heightened tropical cyclone activity in the North Atlantic: natural variability or climate trend? Philos. Trans. R. Soc. Ser. A, 365, 2695– 2716, doi:10.1098/rsta.2007.2083.

Kossin, J. P., K. R. Knapp, D. J. Vimont, R. J. Murnane, B. A. Harper (2007): A globally consistent reanalysis of hurricane variability and trends. Geophys. Res. Lett., 34, L04815, doi:10.1029/2006GL028836.

Landsea, C. W. (2007), Counting Atlantic Tropical Cyclones Back to 1900. EOS, 18, 197-208.

Mann, M.E., T.A. Sabbatelli, U. Neu (2007): Evidence for a modest undercount bias in early historical Atlantic tropical cyclone counts. Geophys. Res. Lett., 34, L22707, doi:10.1029/2007GL031781.

43 Responses to “Tropical cyclone history – part I: How reliable are past hurricane records?”

  1. 1
    Michael Chenoweth says:

    This was a nice summary of recent work. Another approach that would be of interest would be to remove all satellite, aircraft, radar, fixed and drifting buoy data and use only the surface record from land stations and ships to count the number of storms in the 1944-2007. This would be a better apples with apples comparison of historical data with the modern data.

  2. 2
    Tim McDermott says:

    The approach in Michael Chenoweth’s comment (#1) is only valid if reporting requirements remain the same. It is possible that ships no longer report hurricanes, because of satellite surveillance. I don’t know if this is true of not, just a caveat.


  3. 3
    Louis-Philippe Caron says:

    Another good piece of work on the topic is
    Vecchi G.A. and T.R. Knutson, 2008: On Estimates of Historical North Atlantic Tropical Cyclone Activity. J. Climate, in press.

    They used the past observation/modern storm tracks approach similar to Chang and Guo but pushed the analysis back to 1878 and attempted to reconstruct the activity during the 2 WWs. They estimated the number of missing cyclones to be around 2-3 during the period 1878-1900. From their results, whether or not an upward trend in TC number is statistically significant depends on the starting point: from 1878 it is not, but from 1900 it is (Atlantic only).

    However, as it was pointed out, TC number is only one measure of TC activity, and maybe a lousy one at that: it is quite conceivable that we get less TCs in future climate, but stronger ones. In that case, TC number itself would be quite misleading when evaluating the TC threat. A reconstruction of PDI, which addresses number/duration/intensity, would probably be more useful, but I don’t think anyone as tried to tackle that problem yet.

  4. 4
    Steve Horstmeyer says:

    Very nice summary of the inherent pitfalls of studying the historical occurences of a natural phenomenon that occurs in discrete units and is restricted spatially.

    Another source of error is storms that were reported as being tropical in nature but were not. Probably the most notable storm of which I am aware is the great storm of 1704 in southern England, which was extensively documented by Daniel DeFoe. It is still undetermined if it originated as a tropical cyclone or a powerful mid-latitude cyclone.

    Though it is a bit farther back in time than the present discussion, it illustrates the caveats of meteorological re-construction, (i.e. using individual or discrete atmospheric events) as an index of climatological variations.

    The good news is that all the studies cited in this post were within the same general range of error.

    To me the big problem is not frequency of occurrence of tropical systems but he intensity. Maybe another way of putting this is that if fewer tropical storm systems occur, but those storms are of greater average intensity the potential human impact may be greater if we assume no major change from the long term tropical cyclone track climatology.

    The big three then in tropical cyclone historical analysis are 1. frequency of occurence, 2. average intensity and 3. average cyclone track. Inherent in each are major difficulties for historical reconstruction but is not each also an important benchmark for climate change?

  5. 5
    Roger Pielke. Jr. says:

    Steve McIntyre and I looked at the spatial distribution of trends in tropical cyclones in the Best Track dataset in a paper presented at the 2007 AGU. The entire increase in the basin occurs east of 69W. Landfall in the western part of the basin is very highly correlated with activity in that region (our slide 10), as would be expected.

    So explaining trends requires explaining what happens in the far eastern part of the basin. if you are going to rely on Holland (2007) then there are several other interesting patterns to be explained (our slides 12-14).

    Our powerpoint presentation can be viewed here:

  6. 6
    bigcitylib says:

    A very wonderful summary, but it could use a paragraph in the end as a conclusion saying what it all means (and maybe one at the beginning too).

    My own shallow reading of the Lit would suggest that Landsea is arguing that much of the alleged increase in hurricane numbers would disappear if we were to adjust for previous undercountings. It would also suggest that you are arguing that Landsea is wrong and the recent increases are real. But there is nothing really in the argument that SAYS that. Again, I think I know what you’re talking about, but the post would be better as a piece of education Pop Science if you made explicit what the stakes are.

  7. 7

    Re: #3
    If anyone is interested in the Vecchi and Knutson manuscript mentioned above, preprints are available here:
    From the AMS early release website
    from my website

    Regarding the issue of other metrics of storm activity (e.g., power dissipation index – PDI – and accumulated cyclone energy – ACE), extending statistics like these into the late-19th Century can become more problematic since they depend on more than correctly identifying the existence of a storm. ACE and PDI are defined using the wind speed over the life of a cyclone, so one must have an adequate description of a cyclone’s life-cycle. Since the indices depend on a power of the wind speed (square for ACE and cube for PDI), errors in wind speed measurements can have a large effect.

    Vecchi and Knutson (2008) explored (in addition to storm counts) the total number of cyclone-days in a year, which is intermediate between PDI/ACE and storm counts, and found no century-scale trend in the HURDAT database (raw or adjusted). This is because HURDAT exhibits at least a nominal decrease in the average duration of its tropical storms – whether this decrease in average duration represents a real change in the character of the storms or a change in recording/observing practices remains to be adequately understood.

    I think it’s also worth noting that a reanalysis of HURDAT is currently underway (maybe somebody more familiar with it could comment on the ongoing progress). I believe that a few new storms have been identified (though not necessarily approved to be in the official HURDAT dataset) over the period 1910-1920.

    It would also be interesting and helpful to see efforts like HURDAT extended to the other tropical cyclone basins – to the extent that it’s feasible. A recent manuscript describes an attempt in the Eastern Pacific:
    Englehart, P. J., M. D. Lewis, and A. V. Douglas (2008),
    Defining the frequency of near-shore tropical cyclone activity in the eastern North Pacific from historical surface observations (19212005), Geophys. Res. Lett., 35, L03706, 15 February 2008.

  8. 8
    Tom Woods says:

    This is my first post here.

    I was wondering if there is a record of SST measurements taken during shipping voyages (which I’m sure there were) and if so how often?

    Since upwelling events in the tropical Atlantic are rare, almost non-existant, events other than in the wake of a tropical cyclone, perhaps an abnormal drop in SST’s recorded in ship measurements could also be used in reconstruction.

    Perhaps this method has been employed for I had not read the above mentioned papers. If not, it is worth looking into for those with the resources to do so.

  9. 9
    Urs Neu says:

    Thanks Gabriel for the link and comment. I have read the paper, but RC prefers to post only on published papers. Since the principle of your analysis is described above (similar to Chang and Guo) and your results do not diverge significantly from the mainstream, it does not alter the general picture. But it is an important step further in the analysis, of course.
    A feature in your paper that has not been looked at by other authors is the underestimate during the two world wars due to altered ship tracks. This might have a small influence on long-term trends.
    As you mention, unfortunately it is hard to imagine any possibilities to reconstruct parameters like ACE ore PDI due to the necessary temporal and spatial resolution. And if ever we find one, the uncertainties will probably be much greater than the trends we are looking for. Thus we probably will have to live with the fact that we don’t really know the past evolution of the factors we think to be most interesting, except for a few decades.

    Maybe the most promising thing to get more information on the past will be to look more closely at the number of the most intense tropical cyclones, e.g. major hurricanes. There is not much analysis yet. More on that in part II.

  10. 10

    It is quite difficult to make decisive conclusions about the trend in long-term hurricane records, not only because of changing observational bias but also because these are records of extreme events (with non-Gaussian distributions) and application of conventional techniques to determine linear trends do not provide robust results. In addition when analyzing time series of extreme events it is important to consider that extremes could be associated with chaotic behavior of the climate system and therefore advanced statistical methods can provide valuable information. Sorry for self-publicizing but you can find two pdfs of articles in press ( ) that show statistically significant causal linkages between tropical cyclones and sea surface temperatures (and ENSO and Gulf Steam dynamics) and demonstrate that the conclusions are robust against changing observational bias.

  11. 11
    S2 says:

    the period 1851-1855 even has an average PTL of 61%,

    Should that read 1851-1885?

    [Response: Oops–yes, thanks. Fixed. -mike]

  12. 12
    Urs Neu says:

    Re 5
    Roger, we certainly agree, that the regional distribution and regional trends are very interesting and worth to look at. However, I have some problems with your interpretations.

    First, some remarks about your slides:
    You pretend that the entire increase in cyclone counts has taken place in the mid Atlantic. However, your graph shows that there is also a positive trend west of 69W, although less pronounced than east of 69W. What are both trends in numbers? Both regions, east and west of 69W, show a clear positive trend from the 1970s to the 2000s. And there is almost a doubling in both regions after 1995. How does that match with all increase happening in the mid-Atlantic?

    It seems very unlikely that the increase in the mid-Atlantic (and in the west-Atlantic) is due to an observational artifact, since almost the whole increase after 1960 (and a big part of the long-term increase) occurs after 1995. How would you explain that?

    As far as I know, air reconnaissance started in 1944, not 1960 (slide 6).

    What is the source of the number of vessels over time (slide 16)? I would be interested in the whole series, not only a few years.

    Second, some general remarks:
    We all know very well, that there is much regional variability, on the annual as well as on the decadal or even multi-decadal scale, which is much stronger than the variability of the global or continantal scale. This regional variability is due to atmospheric or ocean circulation variations and due to other stochastic internal variability. As Holland (2007) has shown, not only trends but also time series are very different for different regions. We also know that feature from all the local and regional temperature series over the globe. If we see a region, where there is only a small trend, this does not allow any conclusion about large-scale forcing and its influence in the future. It seems rather likely that this trend has been influenced by circulation processes which often oscillate and change. Thus it might be very dangerous to assume, that the relatively small positive trend in the western Atlantic will hold in the future. In contrary, a continuation of the trend is not more likely than a change of the observed trend pattern (small in the west) or even a reverse. At least during the last few decades, the trend differences between east and west of 69W have become very small. Therefore no danger of more landfalling hurricanes in the future is not more likely than that there might be a strong increase in store.

  13. 13
    Roger Pielke. Jr. says:

    Hi Urs, thanks for these comments. A few replies.

    1. These figures in particular are discussed in some detail on Steve McIntyre’s blog here: Our comment referred to the entire dataset, and not any starting point within that dataset. I’ll email Steve for the summary statistics.

    2. I urge caution in looking at a graph and then picking starting and ending points for analysis. You are right that there are changes in the basin starting around 1970 and also around 1995. One could eyeball the chart and suggest similar changes that happened around 1900 and 1930 as well. Why did these occur? I don’t think there is a single satisfactory explanation that accounts for trends in spatial distribution, overall frequency, and observational uncertainties. There are I think several competing arguments that by themselves may help to explain the observed data. Given multiple valid explanations, further appeals to the data probably cannot resolve the existence of multiple competing models for what has been observed. One of the things Steve and I have done by highlighting trends in spatial statistics is highlight another factor that might help to judge between competing explanations. A lot of attention has been paid to over trends in the basin, but very little by comparison has been devoted to explaining trends in location of observed changes. Whatever the ultimate explanation, the vast majority (I’d say all) of the observed increase in activity has occurred far out to sea. Is this observations? Climate change? both? I don’t know, but any satisfactory explanation will account for these patterns, and I’m sure you’d agree.

    3. The Figure showing change in median longitude refers to both aircraft and satellite. The change is even larger if one places the breakpoint at 1944.

    4. I’ll ask Steve on the shipping figure source.

    5. While I don’t disagree with you that future landfall rates may very well be higher than over the past 100 years, and certainly over the period 1970 to present, I do find your argument for such a possibility to be pretty unconvincing. Over the period of record US landfalls have exhibited remarkable stationarity. This provides no guarantee of the future, of course, but surely we must have a reason to expect more landfalls if we are to argue – as you have – that a “strong increase” is of equal probability (“not more likely” you said) to a continuation of stationary landfall statistics. I think that our ignorance is quite a bit larger than this, and I’d submit the hurricane seasons of 2006 and 2007 as evidence of that.

    The good news is that most policy makers don’t need to know the answers to these issues to be better prepared for future hurricane impacts. However, if I was in a business that tries to finely manage catastrophe risk, I’d have the distinct feeling of being in a c-a-s-i-n-o (disallowed word;-) with no understanding of the odds of winning the various games.

    [Response: Roger, thanks for the reply.
    1. and 2. I do not say that your calculations are wrong. I am speaking about interpretations. Your observation of higher trends in the mid-Atlantic (still waiting for the numbers) is for about 150 years. No problem there in principal. However, if you want to discuss processes and forcings it makes not much sense to look at a time interval where the weight of different forcings has changed considerably. This only complicates things. Different forcings might result in different patterns. If you want to look at what anthropogenic forcing might bring, you should look at the time frame where CO2 forcing is the most important factor. And that is for the last few decades.
    We know from temperature patterns, that it is much more difficult to model and explain local and regional patterns, because there is much more chaos and variability in local evolutions. Therefore I doubt that you will be better off looking at regional patterns to explain the influence of the large scale forcings.
    3. The average 1944-1960 is not higher than in the 1880s and 1890s.
    5. Once again, we do not see the same things in the data. I can see a lot of variability in the landfall data.
    I agree, that there is much to do until we know what has happened and why, and what to expect in the future. However, I once more draw other conclusions. We know that the climate system will change and that there might be more intense storms. We are only at the beginning of a change. The risk of strong changes is undoubtedly higher if we alter a system considerably, as is the risk of high impacts. If it was my business, I would take precautionary measures if the risk rises and not lay back and say, oh well, we do not know, so lets go on (being not a manager who gets a golden handshake when he ruins the business with high risk deals) -Urs]

  14. 14
    Ray Ladbury says:

    One is reminded of the classic story of observational bias–the myth that dolphins will rescue shipwrecked sailors and carry them toward shore. Of course, this cannot be verified versus the null hypothesis that it’s 50:50 whether the dolphin carrys the sailor toward or away from shore, as you’ll only hear stories told by the survivors.

  15. 15
    Eric (skeptic) says:

    Two questions please, does the current La Nina jive with last year’s hurricanes statistics (shouldn’t there have been more Atlantic hurricanes)? Second, what is your opinion of the recent alleged upgrading of tropical storms, is there a way to apply your statistical analysis to storm intensity rather than just numbers?

    [Response: Actually, we precisely predicted the 2005 named storm total in advance of the season (15 named storms) based on three causal climate factors, one of which was the predicted La Nina. Perhaps we were a bit lucky, but certainly it is not the case that the observations were in any was inconsistent with expectations, at least for this quantity. Some additional discussion here by Chris Mooney. -mike]

  16. 16
    Jim S. says:

    What’s the impact on under reporting due to communications? Prior to the early 1900’s there was no shipboard radio communications. Is this accounted for in the analysis of 19th century data?

  17. 17
    Eric (skeptic) says:

    Thanks Mike. A few more questions, storms cool the ocean surface temporarily, correct? How much of this effect if any would be captured in the SST measurements (e.g. your figure 1b)? Second, why are Atlantic SST and El Nino completely uncorrelated (e.g. 1b and 1c)? Wouldn’t there be a connection via weather patterns?

  18. 18
    David B. Benson says:

    Jim S. — Inspection of ship’s logs, I suppose.

  19. 19
    Bob North says:

    Mike – Regarding your method for predicting the # of named storms (i.e. TS’s and Hurricanes), how different is your approach from that used by NOAA and by Gray? I note there were 14 named Atlantic storms this year and you predicted 15 plus/minus 4 while NOAA predicted 13-17, which are essentially identical predictions and were both equally accurate. Gray was a bit high at 17. Also, any thoughts on the fact that, other than the two Cat 5 Hurricanes, the remaining named storms were all Cat 1 or TS?

    Thanks in advance for your feedback,
    Bob North

    [Response: To my knowledge, the others do not use the statistical approach that we employed (Poisson regression). James Elsner of FSU has employed this approach in past work, but not for the quantity we were looking at (total named storms). I’m fairly certain that our study is also the only one to use the three particular statistical predictors we used (Aug-Oct SSTs in main development region for North Atlantic TCs, winter Nino3.4 and winter NAO indices). As for the historical distribution of the strength of storms, well we haven’t attempted that. Other researchers such as Kerry Emanuel and Peter Webster have been investigating this, but it is fraught with additional uncertainties (its somewhat easier to say that a TC of some sort existed, then to classify it in strength on the Saffir-Simpson scale). -mike]

  20. 20
    J. Solters says:

    Counting tropical cyclones appears somewhat arbitrary by definition. The last two seasons exhibited at least two or three storms each that stayed far at sea causing limited opportunity for actual measurement. Review of the published inter and daily analyses of NHC forecasters reveals great difficulty in estimating wind intensity of these distant storms. Estimating a few knots of windspeed over or under the category threshold means hurricane classification or not. Both the 2006 and 2007 seasons contained problematical hurricane classification where distant storms were estimated to exceed the threshold windspeed for only a few days. The point is that even current NHC numbers are very iffy when applied to mid-Atlanic events. I can’t imagine any value going back 100 years and guessing about hurricane numbers.

    [Response: I agree that the classifications include estimates which contain uncertainties (Besides: for the last decades the basis is not only measurements, but also classification of appearance on satellite pictures, which have full geographical coverage). This is important for the classification of individual storms. However, the influence of estimation uncertainties diminishes when averaging over a lot of storms, because under- and overestimates will compensate to a considerable extent. Thus, for long-term trends, the importance is not that big. On the other hand, there will be some bias towards underestimation if observations of cyclone tracks are full of gaps which enhances the probability to miss the time frame of maximum speed. -Urs]

  21. 21
    Leon Palmer says:

    Interesting analysis. The assertion that prior to 1900 there were not enough populated areas to catch all landfall hurricanes might be testable for the 1860-1864 time period. During the Civil War the union navy blockaded the confederacy from virginia to mexico… in other words the land may have been under populated, but the coastal waters were well populated. Has anyone compared the two to validate that the decrease in landfall hurricanes during the 1860’s was due to under reporting due to under population of the coasts? Or has the union navy records already been factored into the landfall reporting?

  22. 22
    Michael Chenoweth says:

    With reference to the question on worldwide shipping totals, a spreadsheet with the yearly numbers for 1870-1916 was posted (at my request) by Jeremy Lowe, of New Zealand, late in January 2008. The webpage is available at
    and can be found under “Registers and other general sources” line item R7. I don’t know if this is the source used by Climate Audit.

    What is most interesting is the rapid decline in sailing ships (wind-propulsion) relative to the rising number of steamers (which also carried sails [to supplement steam power or as an emergency back-up when engines failed] for many decades after conversion to steam). Sailing ships had to follow the trade wind routes while steamers could take great circle routes. The trade wind routes in the North Atlantic run southwest and west of the Azores for ships in-bound to Europe whereas steamers can sail directly up along the northwest coast of Africa where they run a lower risk of encountering tropical cyclones than ships in the 30-40W area. Sailing ships also sail much slower than steamers and have more opportunity to encounter tropical cyclones in a given voyage. The decline in sailing ships appears to have been fairly steady since 1870 and by the end of the 1930s there numbers were probably negligible. Therefore, not only are the number of opportunities falling from the decline in the number of sailing ships, but the number of days of potential observations being lost are even larger than the number of ships indicates.

    This goes against the common perception (repeated by Urs) that ship observations fall in number as one goes back in time. Certainly the number of observations in datasets like COADS and ICOADS drop off but the total number of POTENTIAL ships of opportunity actually does increase before the early 20th century minimum in ship totals.

    For this reason newspaper shipping news accounts are invaluable because they provide the reports from the ships that were in the area and whose logbooks are no longer extant. This was the impetus for Fernandez-Partagas and Diaz to do the first systematic probe of The New York Times, The London Times and Gaceta de la Habana. Unfortunately, these three newspapers are not the best sources of shipping news as my own on-going research has revealed. It takes many more newspapers from throughout the Caribbean region as well as from the US and Europe to find these reports. This work is being done and results will be forthcoming. After 1910, there is no systematic search of newspapers throughout the Atlantic basin that has yet been performed. This, in part, may account for some of the low counts relative to years prior to 1911.

    One final point on low counts. Tropical storms that formed north of 30N were often detected and depicted in daily weather map series but were not counted in tropical cyclone statistics. For an example, see the September 1929 issue of the Monthly Weather Review (MWR), which is available on-line, and shows once-daily weather maps for the North Atlantic for late September 1929. A hurricane affecting south Florida is depicted while another tropical cyclone, which formed east of Bermuda and then re-curved off of the US coast is plainly apparent from the plotted ship observations. This storm will, no doubt, be included in HURDAT once the HURDAT committee have worked their way up to 1929. Another storm, not presently in HURDAT, that passed through the Azores is documented in the September 1940 MWR.

    Late 19th and early 20th century meteorologists either did not yet know that tropical cyclogenesis could occur at relatively high latitudes and took decades of observations before this notion finally set in or they chose not to count them. Nine of the twelve “new” tropical cyclones that Fernandez-Partagas and Diaz discovered in the years 1878-1886 were first observed north of 30N but were never counted as tropical cyclones until they performed their data search and re-analysis. This is probably another contributing reason to the under-count in tropical cyclones and this from a well-sampled area of the basin.

  23. 23
    Eli Rabett says:

    It seems an obvious idea, but there were a lot of small fishing vessels in the Atlantic (Azores, Canaries, etc,) and if nothing else looking at church records for drownings/lost ships, should be a useful exercise. There might even be diaries.

  24. 24
    Urs Neu says:

    Michael, thanks for your comments.
    Maybe I was not clear enough in my argument. We are discussing the possible undercount bias of the existing hurricane database, and thus have to consider the ship data that was used for HURDAT, and not potential data that could be analysed. If I’m not mistaken (HURDAT people please correct me), HURDAT is not based on a great deal more data than you find in ICOADS. What is important for reanalyses is the available observational data, not the number of vessels or ship tracks. It would be interesting to have information about the amount of data analysed for HURDAT over time. Anybody knows?

    I fully agree that there still might be a lot of data out there which could improve the existing data set. However, it will be increasingly difficult to dig up the data, if there has been any recording at all and the data has not been lost, when going back in time. After a lot more of reanalyis, we also have to reconsider the undercount bias estimates, of course.

  25. 25

    Here’s something re the future of oceans & wind patterns. The anoxic zones off Oregon and elsewhere are thought in some cases to be caused by GW. See article based on a study from SCIENCE:

    So the theory goes that if the oceans become super-anoxic in a globally warmed world, if GW were to go up to 5 or 6C warmer, then the oceans would start outgassing hydrogen sulfide and kill off life in the oceans and on land, as thought happened during the end-Permian 251 mya.

    It happened before, it could happen again.

    Anyone for turning off that light not in use?

  26. 26
    Petewsh61 says:

    Sorry if this is a bit off topic, but I’ve been having a discussion with a skeptic and I need some info.

    There are several indicators of climate change in the NE region of the US that point to a warming period since about the mid-1960’s. My skeptic friend claims that the most likely explanation of this is the PDO.

    What is our current understanding of the PDO and it’s affect on regional and global climate. Can it be the main causal mechanism of increasing NE US climate?

  27. 27
    Michael Chenoweth says:

    Re #24:
    Urs, yes, you are correct to mention that the focus of your piece has to be on the HURDAT database that we have and not the one we wish we had. I just want readers and active researchers to better understand the underlying issues on data availability that can influence our interpretation of presently existing data sets.

    There is a lot of work being done on retrieving the 19th century data and there are 82 new candidate storms that I have found for 1851-1888 that are not in HURDAT (and I’m not done yet and yes, I’m also suggesting eliminating and merging a handful of storms now in HURDAT). It takes a great deal of time to gather, de-conflict sources, plot and analyze, and then write up accounts that can be used by the HURDAT committee or be readied for the peer-reviewed literature.

    The HURDAT team can speak for themself but as I understand it they are using the historical daily weather map series for the Northern Hemisphere along with COADS data in the 20th century re-analysis, official published records and any other data that anyone is willing to provide (I have provided some of my 20th century data to them). Daily mapping is essential to maintain continuity on individual systems and understanding the likely mechanism for tropical cyclogenesis and is just as important as the raw observations. The HURDAT Re-Analysis web page provides spreadsheets of the raw data that they used and your question on the amount of data analyzed by HURDAT can in part be answered by carefully scanning their data section on the Re-Analysis web site.

  28. 28

    RE #5 and the Q, if hurricane intensity/activity is only increasing in areas away from land, should policy-makers care?

    Depends on what type of policy-makers, I suppose. If they are dealing with ALL the impacts of AGW (not only hurricanes, but the myriad of other GW effects as well) now and into the long-term future, then of course they should care about reducing our GHGs as fast as possible & as much as possible in a orderly fashion that takes advantage of all the economic benefits from energy/resource conservation/efficiency and alternative energy.

    If the policy-makers are solely concerned with hurricane-proofing the hurricane zones, or being better prepared to respond to land-fallng hurricanes, then it seems the answer is they should care a lot more than they do now, even if GW is not increasing land-falling hurricanes at present. Just think about Katrina. And some people say that wasn’t even caused by GW. So there is plenty of room for policy makers to improve a great great deal in their policies and actions. And I live in a hurricane zone, so I’d hope they hop to it.

    So either way, policy-makers have their work cut out for them.

  29. 29
    Eric (skeptic) says:

    African Dust Storms May Cool Atlantic, Lessen Hurricanes (

  30. 30
    Steve Bloom says:

    Re #15 response: 2005 should be 2007.

  31. 31
    Steve Bloom says:

    Re #29: This effect is well-known. The difficulty is that it’s very intermittent from year to year, as the text of the article discusses.

  32. 32
    AdeV says:

    I’ve only seen it mentioned once in #23 – but the incidence of ships being sunk by hurricanes and other large storms would surely rise the further back in time one goes? And, of course, in the days before radio, ships would be unable to report the storm, which – if it doesn’t landfall – would (potentially) lead to it going unreported.

    I don’t know enough about HURDAT to know if that’s likely to be of significance, though. Any ideas?

  33. 33
    Hank Roberts says:

    I recall one big change happened when wooden ships were replaced by steel. Steel ships sink; wooden ships abandoned, burned, or broken up floated all but submerged. I read somewhere recently that the old shipping records track these, identified by name and location sighted, over long stretches. They were real hazards to navigation, so they were reported and watched for. I don’t have a cite at the moment, nor know if it’s been looked at. But before steel ships, there might actually be more information rather than less in this kind of record.

    Somewhere (Lloyds?) there must be insurance records tallying how many ships sailed on what routes, when, and how many were lost and on what routes and approximate times.

  34. 34
    Chuck Booth says:

    Re # 33 Hank Roberts

    Probably more informative than the tracking of derelict ships were the detailed position, wind, and current records compiled by Matthew Fontaine Maury, Superintendant of the U.S. Naval Observatory, from the logs submitted by Naval and merchant ship captains for a couple of decades in the mid-19th century. Maury was one of the pioneers of physical oceanograph and meteorology, convening the first international meteorology conference in Brussels in 1853, and conducting some of the first systematic analyses of hurricane formation and tracking. Some interesting accounts of Maury’s contributions to oceanography and meteorology are:

    The Physical Geography of the Sea, by M.F. Maury, 1855 (,M1)

    Tracks in the Sea: Matthew Fontaine Maury and the Mapping of the Oceans, by Chester G. Hearn, International Marine-McGraw Hill, 2002.

    Matthew Fontaine Maury: Benefactor of Mankind: – History of Oceanography:

  35. 35
    Michael Chenoweth says:

    The percentages of tropical cyclones that struck land (PTL) is lower in 1851-1885 partly because this period benefits from the work of Fernandez-Partagas and Diaz for the years 1851-1910. They used the shipping news to find new storms at sea in addition to new landfalling storms. However, their main source (The New York Times) stopped providing weather reports from ships in their shipping news section after the 1877 hurricane season. From 1878 onward, most of the ship data comes from the Monthly Weather Review and most of their new storms involve tropical cyclones that never affected land. Very few new storms come only from newspaper accounts after 1877. Therefore, until a full compilation of ship reports is added into HURDAT (work currently underway) the PTL figure of 61% should be considered a sampling artifact due to a better (certainly not comprehensive) range of reports included into HURDAT up to 1877 relative to that of 1878-1910.

    [Response: What is the source of your hypothesis? In the description of their work (BAMS, 77, pp. 2899-2906) Fernandez-Partagas and Diaz neither mention the New York Times as their main source (but as one of many), nor do they mention a change in data sources as you discuss here. -Urs]

  36. 36
    Michael Chenoweth says:

    The New York Times analysis is part of my present work doing re-analysis of the Atlantic hurricane record which so far does not extend past 1888. Log into the New York Times web page and you can access their past issues from 1851 for free. The only ship news after 1877 that includes weather reports is usually in the occasional general news section (a feature common throughout the period of record). The bulk of the NEW storms found in 1851-1877 is buried in the very small font of the shipping news section. It is this section which stopped carrying the weather comments of arriving ships although it continued in the New York Herald after 1877 and is the New York newspaper that I am using to find new storms along with other newspapers throughout the Caribbean region.

    The only newspapers that Fernandez-Partagas and Diaz had access to at the University of Miami was the New York Times, and the London Times for 1851-1910 and Gaceta de la Habana for at least part of 1851-1870. This is all described in their publications which are available on-line at the HURDAT Re-Analysis web site

    In addition to these two to three newspapers they used secondary references well-known in the hurricane community and which are also provided in their accounts at the above referenced web site.

    They did not mention the change in dominant source types, but it can be discovered by closely looking at the source of the references in the write-up to each piece. Just as your re-examination of Nyberg et al. (2007) revealed interesting features not previously noticed so does a careful look at the sources used by Fernandez-Partagas and Diaz.

    If you have any other questions please let me know.


  37. 37
    Cary Mock says:

    Urs, regarding your response in #35, Fernandez-Partagas and Diaz published the raw data in several “Parts” (they are available on the HURDAT website in PDF form) with the title “A Reconstruction of Historical Tropical Cyclone Frequency in the Atlantic from Documentary and Other Historical Sources…” It is pretty obvious that the NY Times is their main source.

  38. 38
    Urs Neu says:

    Mike, Cary
    Thanks for the link. Ok, now I see what you mean. I have an additional question: After looking at the documentation, I got the impression that, while the New York Times as a source got much rarer after 1977, Monthly Weather Review was cited much more after 1977 than before. Do you know if some of the information previously published in the New York Times was published in Monthly Weather Review afterwards?

    And one more question: Thus, do you agree that the PTL for 1851-1877 (59% on average) is the most reliable for the 19th century, because we had most information for that period? If I understand you correctly, you think that 61% for the period 1851-1886 and the 67% for 1851-1900 are probably to high, because less open sea storms might have been detected during the period 1877-1900 (which enhances PTL artificially). This does not alter any of my conclusions, does it (59% would be exactly the same as after 1966)?

  39. 39

    Re: 38
    Urs, as I had understood it, your argument was that Landsea’s methodology could be dismissed because PTL could not be thought of as ~constant on decadal timescales – since PTL returned to it’s 1966-2006 values in the mid-19th Century. However, if one can attribute the low values of PTL (59%, same as over 1966-2006) in the period 1851-1877 to having sufficient data to reconstruct open ocean storms, that particular argument against Landsea’s methodology is no longer necessarily valid. In fact, that PTL over the “well reconstructed” mid-19th Century (assuming for argument that the mid-19th century is indeed adequately reconstructed) is the same as in the satellite era is suggestive that perhaps PTL is ~constant at around 60%. Is there a physical basis to distinguish between these two (or other) interpretations – that the real PTL has had large swings or that it has been ~constant on multi-decadal timescales?

    Unfortunately, our incomplete physical understanding of the controls to tropical cyclone characteristics (frequency, seasonal landfall, etc), coupled with the fact that the observational records are imperfect, makes multiple and conflicting interpretations of the data possible. It may be the case that the multi-decadal variations/changes in PTL (and other cyclone characteristics) are due to variations in the climate system (forced or internal), but it can also be the case that changes in observing and recording practices (and incomplete reconstruction) have led to multi-decadal changes in PTL (and other cyclone characteristics). In addition to working on improvements to the observed record (including estimates of uncertainty), the community needs continue working at understanding and quantifying the physical controls of the large-scale climate system on cyclone properties (and possible feedbacks from cyclones to climate).

    I suspect that until – in addition to an observational record of sufficient quality to resolve the changes in cyclone statistics – we have a solid physical understanding of climate-cyclone controls/feedbacks, and are able to represent them well in comprehensive climate models, the story will remain somewhat confused.

  40. 40
    Cary Mock says:

    Urs, I assume you mean 1877 in the first paragraph. Some of the shipping reports in the NY Times are also in Monthly Weather Review, but I wouldn’t think that it would be a tremendous obvious significant amount. Monthly Weather Review got its data from a variety of sources. I will let Chenoweth elaborate more, however, on the 1870s and later periods since he is more familiar with them. I will admit, I stick my business more in the pre-1870 period and in particular try to go as detailed as possible before 1851. I will add that I think it is tough to generalize percentages for the whole 1851-1886 and 1851-1900 periods, as some shorter subperiods can vary quite a bit in more or less historical information. For example, HURDAT currently have relatively little data in HURDAT during the US Civil WAr years due to historical reasons.

  41. 41
    Michael Chenoweth says:

    I don’t know the exact source the Monthly Weather Review (MWR) used in obtaining its ship reports. They could have read one of many coastal newspapers that published marine intelligence or even other sources. However, even though I do find accounts in the NY Herald that match the MWR, I also find still other new ship reports.

    As for PTL, my point is that because F-P and Diaz had shipping news as their main source of data they found more storms that remained at sea, or could only be documented at sea, and so I am not surprised that the number is lower than after 1885. This is the feature of the currently-existing HURDAT that led you to your current conclusion.

    Having said that, my own provisional work on 1851-1888 is finding more storms making landfall than those at sea. This is in part because I read local Caribbean newspapers which cover the weaker storms much better than a remote New York or London newspaper looking for newsworthy items to reprint, and also because some storms in HURDAT that are presently documented only at sea can be traced earlier or later on in their life cycles to land regions. This includes new US landfalling tropical cyclones as well. So the PTL in HURDAT changes in my re-analysis to a higher percentage.

    The years 1851-1870 are further along in my re-analysis process than later years. The PTL figure in my re-analysis for 1851-70 is 70% compared to about 55% in HURDAT at present, per your figure 1. The increase, if any, in PTL will likely be less dramatic after 1870.

  42. 42
    Urs Neu says:

    Re 39

    Gabriel, my argument is the following: if (as Landsea assumes) 1) PTL is constant over time, and 2) PTL is a measure for the missing of tropical cyclones in the record because TCs are mainly missed over the open sea, then it would follow that, because the database used for HURDAT is smaller in the 19th century than in the first half of the 20th century, PTL should be higher in the 19th century than it is afterwards. This is obviously not the case, no matter if PTL is 59, 61, or 67% in the 19th century.
    One could argue, that in the 19th century the density of land observations was also bad and lead to a similar missing ratio as on the open sea, but I doubt that. Maybe Mike and Cary can appraise that better.
    My only point is, that I think the link between PTL seems not to be a very good predictor for estimations of missed TCs, because there are other factors involved (be it trends or variability in PTL or others).

  43. 43
    James Elsner says:

    It is useful to note that PTL is a random variable constructed as the ratio of two other random variables. As such even under static conditions (climate and no changes in observational technology) there will be fluctuations that can be quite large. Below is a simple model that can be used as a null hypothesis when making claims about changes in PTL. It is written in R, which is a free, open source tool for statistical analysis and modeling. We have a tutorial on using R for climate research on our blog ( Simply copy the code and paste it into your R session.

    TrueBasin=rpois(156,12) #True basin count Poisson w/rate = 12 stms/yr
    for(i in 1:156) ObsBasin[i]=sum(rbinom(TrueBasin[i],1,0.90)) #Detect Pr=.9
    for(i in 1:156) ObsLand[i]=sum(rbinom(ObsBasin[i],1,0.6)) #Land/Basin=.6

    To get a plot that looks like Figure 1, use


    But change the quotes around the h and Year to their ascii equivalent.

    Of course detection probability is certainly a function (likely discontinuous) of year. And we could add a detection probability to landfall counts.