Slower decay of landfalling Hurricanes in a warmer world — really?

by Frank Bosse

A recent paper published in “Nature” made some excitement in the media, see here or here.

In the paper by Li & Chakraborty (L&C 2020 thereafter), the authors find a statistically significant increase of the decay time when a North Atlantic hurricane makes a landfall due to warmer SST in a warming environment. They also undertake some model-related research about the impact of this observations.

The key point of thepaper is the finding that warmer SSTs lengthen the decay time of hurricanes after landfalls.

In L&C 2020, this is shown by figure 1f:

Fig.1: The reproduction of Fig.1 f in L&C 2020. The ordinate reflects the decay time τ in hours.

In the legend the authors state: “We note that the τ time series echoes the SST time series with Pearson correlation r = 0.73.”

The authors describe the way they found the relation, which declares an increase of the decay time of more than 40 hours per 1K SST increase:

“We average τ for all the landfall events in a given year and apply a 3-year smoothing, twice in a row, to this time series.”

They made a regression with strongly smoothed time series, a procedure that is normally frowned on.

In the supplementary data (freely available) one can download an Excel sheet where the raw data used can be found.

For the deduction of the increasing τ with increasing SST (in the area 10°N…35°N ; 100°W…75°W ,  the authors take advantage of the data for 71 landfalling hurricanes during 1967 to 2018.

The SST for each event is determined as follows:

“We average the SST in time over the hurricane season, June–November, …”

The result “R=0.73”, see Fig.1, of the linear regression implies that 53% of the variation of τ comes from the variation of the SST.

I had a look at the raw data and a few questions arose:

  1. The use of the average SST of the whole hurricane season for a single event?

The actual named hurricane develops over a few days in an actual environment, not the average SST of the actual whole season. It makes a difference if the landfall happens during July or November, the average SST difference is 1.8 K in this case, which is much more than the range of the abscissa in Fig.1 .

The use of the seasonal average SST  for all hurricanes during that season, rather than the actual SST applicable to each hurricane, has the potential to produce highly misleading results. The average SST applicable to each hurricane might have little relationship with the average SST during the whole season

  1. The use of the average of all τ in a year if more than one hurricane is involved?

Every hurricane event is a discrete event. In the raw data many years have only one hurricane per year, these events are not averaged of course.

  1. Applying a double 3 years smoothing before making the regression shown in Fig.1. ?

The authors state:

“this approach lessens the effects of non-climatic factors and random noise”.

However, the whole research is about the point:”To what degree impact the actual SST the decay time of landfalling hurricanes?” There will be some other influences and it’s not appropriate to smooth over several years partly out to elicit a strong climate related signal. Applying a 3-year smoothing to both decay time and SST data twice in a row is unjustifiable.

I decided to recalculate the regression shown in Fig.1 but I used the actual SST for every hurricane from the monthly ERSSTv5 data for the described area. I included every hurricane because this is the physical approach: It’s not justified at all to use an average in some years and in some years not, as that gives radically different weightings to hurricanes depending on how many are included in the raw data in each year.

I also use the unsmoothed data to avoid spurious correlations due to the applied smoothing.

This is the result:

Fig.2: The regression of the decay time on the SST without the data-preconditioning in L&C 2020.

There is only a tiny non significant trend in the raw data- p=0.1, so the slope does not reach the standard 95% confidence level.

On twitter  Ryan Maue questioned the raw data selection; that issue is beyond the scope of this post.

The outcome of L&C 2020 is very overconfident when it comes to the dependency of the decay time on the SST. The R²=0.53  found in LC 2020 vanishes to an insignificant 0.04 if one uses the physical data, without the applying of unjustified averages and smoothing actions prior to the regression.

This means:

The SST impact on the decay time is negligible, other influences accounting for almost all variability in the decay time.

The peer review process of “Nature” for L&C 2020 lasted more than 8 months, it makes wonder if there was no reviewer with some fundamental skills in statistics involved.

However, this must be the case unfortunately: In the “methods-statistical significance” section the authors mention a test for autocorrelation and there is written: “(which we test using the Dublin–Watson test)”. This must be a typo, the name of the test is “Durbin– Watson”.

One should hope that the peer review process of “Nature” would be improved soon to avoid overconfident, obviously flawed papers like L&C 2020.

67 responses to “Slower decay of landfalling Hurricanes in a warmer world — really?

  1. Thank you for your research and critical analysis of mainstream climate science claims Frank.

    • Yes, indeed. This paper is so full of malpractice and novice errors it is clear the NEITHER the authors nor the reviewers are even remotely competent in data processing.

      Lamentable as this state of affairs is for the supposedly “much respected” Nature title, this is now par for the course. Any pseudo-scientific babble gets published as long as there is a positive correlation with CO2.

      • When anyone refers to “smoothing” the data they are invariably referring to a running average, which is one of the crappiest forms of low-pass filter around but is great for those with no more knowledge of data processing than trivial fiddling with spreadsheets.

        Why did they choose this filter ? How did they chose three point length? Why did they apply it twice ? This heavily reduces the number of independent points in the dataset and thus reduces the significance of the resulting correlation.

        How did they account for that in calculating the significance of the slope…. oh wait, they didn’t.

        How the hell can junk like that get through so-called peer-review ?

        End of the age of enlightenment is upon us.

  2. Here’s an article without the paywall:

    For Chakraborty, the Nature study’s senior author, the  “aha!” moment came even before the research started.

    “We were studying the evolution of landfalling hurricanes using simulations and kept finding features that could not be explained using the prevailing models,” he said. “Long story short, we realized that the prevailing models are missing a key component: the moisture stored in a hurricane.”

    Hurricanes hold more moisture in a warmer climate because the atmosphere can hold about 7 percent more water for every 1.8 degrees Fahrenheit warming. Part of a hurricane’s total energy is stored in the water it carries, and that extra fuel helps storms overpower the weakening effect of friction over land, he said.

  3. The key issue here is, are those data points really suitable for feeding into a (no doubt computerised) program for rigorous statistical analysis? I am no statistician but as a mature research scientist do have 60 years experience gazing at various kinds of experimental data. What I see here in the chart is a fairly random set of data that would look even more random (not an acknowledged statistical term, I know) were it not for just two outlying points. I would never attach anything significant to them, let alone a conclusion with significant consequences for highly publicised climate science.

  4. There is something that keeps hurricanes alive over land … but it’s water, not temperature.


    • “There is something that keeps hurricanes alive over land … but it’s water, not temperature.” – Willis Eschenbach

      If evaporation is the fuel for tropical storms then it is both. Interesting article though and very relevant to the recent hurricanes of Eta and Iota unfortunately.

      • YouTube channel host Deciphering Weather replied:

        “I remember back in the late 90’s, when I first started tracking the hurricane season, Hurricane Danny made landfall in the gulf and then moved north and the northeast towards NC & VA. It dissipated into a remnant low and then reformed into a tropical storm while over land due to the excessive rainfall.”

  5. Nature’s “Peer Reviewed” has been exposed time and time again. The whole “Peer Reviewed Process” is flawed.

    • “Peer review is at the heart of the processes of not just medical journals but of all of science. It is the method by which grants are allocated, papers published, academics promoted, and Nobel prizes won. Yet it is hard to define. It has until recently been unstudied. And its defects are easier to identify than its attributes. Yet it shows no sign of going away. Famously, it is compared with democracy: a system full of problems but the least worst we have.”

      Imagine the problem of someone like myself who’s starting point is a conceptual error made over 350 years ago by the founders of modern gravity theory.

  6. “As noted in the main text, we smooth all the time series to lessen the
    effects of non-climatic factors and random noise. Further, smoothing yields more reliable s.d. (and, therefore, s.e.m.) by increasing the
    number of samples per data point (Extended Data Table 3a). However,
    smoothing also induces serial correlation. The unsmoothed time series
    has either no serial correlation (for example, SST time series) or a small
    decorrelation timescale (<2 years; for example, τ time series). With an
    increase in the time window of smoothing, the decorrelation timescale
    monotonically increases. Accounting for the decorrelation timescale,
    we find that the statistical significance of unsmoothed to variously
    smoothed time series remains robust to the the specifics of the smoothing (Extended Data Table 3c)."

    It may be frowned upon by Frank and his unspecified congregation of statisticians. The extra water holding capacity of warm air is mathematically demonstrable – and there has been considerable discussion over the years of an accelerated hydrological cycle. On the other hand – the infiltration capacity of soils is 25mm or so initially – and some 2.5mm/hour thereafter. This is soon overwhelmed in any cyclone I have ever seen.

  7. Frank Bosse, thank you for the essay.

  8. Dr. Chakraborty in the media is more “determined” in his conclusions.
    “The decay has slowed down TREMENDOUSLY over the last 50 years …”, “Once we understand that moisture plays a key role, the connection with climate becomes evident…” “Overall, our study challenges widely-held ideas about hurricane decay…” (…)

    The media… The media should remember, e.g.:

    “…two alternative statistical models of hurricane activity vs SST, both of which perform comparably during the historical period, give DRAMATICALLY different projections of late 21st century activity…”
    “The example for Atlantic power dissipation illustrates how dynamical and statistical downscaling techniques, or different statistical approaches, can differ substantially in their projections of the tropical cyclone response…”
    “First, it is possible that 21st century changes in tropical cyclones will be less potentially damaging… …TC activity in some basins, such as… …North Atlantic, could shift eastward away from current landfalling regions and thus perhaps reduce the percentage of storms that make landfall in major population regions.”
    (Tropical Cyclones and Climate Change, Knutson et al., 2010., DOI: 10.1038/ngeo779)

  9. Frank Bosse – it would be interesting to include the data from the applied statistician who correlated hurricane activity with the solar cycle:

    “James Elsner, a climatologist at Florida State University in Tallahassee, has analyzed hurricane data going back more than a century. He says he has identified a 10- to 12-year cycle in hurricane records that corresponds to the solar cycle, in which the Sun’s magnetic activity rises and falls.”

  10. Alan Lowey: You can make your own picture with the help of the “KNMI climate explorer”, your search engine will find it, there you can select “monthy climate indecies”, in the navigation you’ll also find many time series for Hurricanes ect…. select one of those and under “investigate this series”, you’ll find “wavelets”. Klick… thereafter you can look if there some energy at 11-12 years ( sunspots…) … I didn’t find somewhat, however I didn’t try it with every index for tropical cyclones.

    • Thanks for the tip, I had a quick look. Tbh I’m a bit sceptical in using these kind of tools. Too many assumptions and ‘normalization of the data’ imo.

      I tried to find data from global Argo of ocean temperature at 2000m, changing over time relative to latitude ie. distance from the equator. I made an enquiry about how I wanted to test the hypothesis of increasing solid body tides moving equatorial warm water, so as broadening the tropics. I was circular game of pretending to be helpful but ultimately unachievable.

      If anyone can provide that data in a graph, it would be most appreciated.

  11. Climate science says that it is not possible to test for impact of global warming on tropical cyclones in a single cyclone basin. Pls see

  12. Tropical cyclones ( unresolvable by climate models ) were not part of the global warming narrative until Dr. Trenberth inserted them. This resulted in the resignation from the IPCC by Dr. Landsea, a lead tropical cyclone investigator.

    Since neither TC frequency:

    nor TC intensity:

    appear to be unusual with the global warming to date, advocates appear more and more desperate to find something TC related that increases with global warming.

    Arguments about intensity are suspect when one considers the NE Pacific. Storms there intensify while traversing cooler waters, indicating at least and exaggeration of temperature:


      “Graph showing the number of severe and non-severe tropical cyclones from 1970-2017 which have occurred in the Australian region. Severe tropical cyclones are shown here as those with a minimum central pressure less than 970 hPa.“

      Australian Bureau of Meteorology

      • The data looks useful but again how does central pressure relate to overall energy of a tropical cyclone season?

        Why isn’t a map included which shows where and how the tropical depressions form?

        Is there a distinction between Pacific coast formation and that of the Indian ocean for example?

    • Are just considering data from tropical cyclones bordering the US?

    • “Arguments about intensity are suspect when one considers the NE Pacific. Storms there intensify while traversing cooler waters, indicating at least and exaggeration of temperature:” – Turbulent Eddie

      Have you got a reference for that claim please?

      • The data are from ERSST and NOAA HURDAT.

        In figure 4c above:
        ‘F’ represents the mean location of TC formation,
        ‘P’ represents the mean location of TC peak intensity, and
        ‘T’ represent the mean location of TC termination.

        The mean intensification phase ( from formation to peak ) is indicated over the mean TC season SSTs in figure 4d. There is much path variance, of course, and it’s not every storm, but most NE Pacific TCs intensify as they traverse cooler waters.

        This is spatial, of course, owing to the orientation of the temperature gradient on the Eastern ocean basins but it is a reminder that other factors are important and at least for the NE Pacific, other factors are more important than SST.

      • I followed what you say in relation to the graphs and see that your initial statement isn’t very compelling. The tropical storms soon dissipate over the cooler water.

      • The line segments in figure 4d are of the mean intensification phase ( from formation to peak wind speed ).

        NE Pacific tropical cyclones intensify while traversing cooler waters.

      • Yes, but I think you take the schematic difference in SST too literally. The line isn’t so distinct in reality. The water is still evaporating but at a reduced rate compared to warmer water nearer the equator.

        It’s the difference between saying tropical storms intensify over ‘cool water’ as opposed to ‘cooler than equatorial water’.

      • Btw I just remembered from watching a documentary many years ago that the magic number for SST was 14°C. I don’t see this in the current literature anywhere.

      • It’s the difference between saying tropical storms intensify over ‘cool water’ as opposed to ‘cooler than equatorial water’.

        Yes, in the NE Pacific, most tropical cyclones intensify over waters cooler than those over which they form.

      • I’m guessing that’s because of their size and that they’ve attained a more efficient rotation, thereby updrafting more evaporated seawater??

  13. So, the cloud that hangs over our heads– that won’t blow away– and guarantees the, “Very Dark Winter,” the Left warns us about, is the the superstition and ignorance that results from the Left’s politicization of science.

  14. It’s difficult not to relate the devastating floods in Central America with those occurring simultaneously in the west Pacific during a Grand Solar Minimum:

  15. “One should hope that the peer review process of “Nature” would be improved soon to avoid overconfident, obviously flawed papers like L&C 2020.”

    Are you kidding? It’s not in Nature‘s nature to do peer-review when there’s a chance to promote climate change alarmism.

  16. My 1st observation (re: article in LAT’s about this ‘study.’

    “examined 71 Atlantic hurricanes made landfall since 1967. In the ’60s, ’wind strength declined by 2/3 within 17hrs of landfall.”

    What? “In the 60’s?” Study starts in 1967. Comparing 50 yrs of storms to a base period of 3 yrs?

    • That’s interesting but the Marshall Islands are in the west Pacific, close to the equator. The series of cold snaps known as The Little Ice Age were famous for freezing over the Thames river in the UK.

      The findings still fit my hypothesis of the previous warm period of approx 950-1200 being due to tidal forcing pushing warm equatorial water to higher latitudes. This then warms the atmosphere which gets pushed upwards and trapped in the Polar Vortex for hundreds of years, releasing it’s energy in periodic cold spells. These huge events could keep equatorial winds from moving to higher latitudes and therefore form more tropical storms.

      • Findings based on frozen seal skins from the hut of the famous explorer Shackleton inform us that, the LIA apparently was… global.

      • Wagathon – not exactly, it’s evidence that the series of cold snaps termed The Little Ice Age occurred in both polar regions.

        This is commensurate with tidal forcing around the equatorial regions *before* the LIA. Warm waters are pushed towards both poles and the extra warm atmospheric energy then pushed upwards into both polar vortexes. It’s only after hundreds of years that this extra energy is released in a series of cold spells, not necessarily at exactly the same time in each polar region.

    • There conclusions seem somewhat of a joke compared to recent super events in the Atlantic equatorial regions and the west Pacific.

      “So we can expect to see the opposite effect in the deep tropics—a decrease in tropical cyclones close to the equator.”

      They must have written the report a few weeks or more before Hurricane Eta and Iota have stolen the headlines and pushed the hurricane season into the record books.

  17. Alan Lowey November 18, 2020 at 8:59 pm: The Wikipedia article on Hurricane Danny of 1997 gives a different explanation from high rainfall being the reason Danny regained tropical storm status (from a tropical depression, not a remnant low) while over land: It went into a favorable baroclinic zone associated with a developing trough. The area this happened in is infamous for storm development due to lee side cyclogenesis. Another example is Hurricane Hazel of 1954 which underwent strong extratropical cyclone development and became able to maintain hurricane force sustained winds inland in that area (and well afterwards) when Hazel ran into a favorable baroclinic zone associated with a cold front.

    • Okay, you’re replying to the comment made by YouTube host Deiphering Weather:

      “I remember back in the late 90’s, when I first started tracking the hurricane season, Hurricane Danny made landfall in the gulf and then moved north and the northeast towards NC & VA. It dissipated into a remnant low and then reformed into a tropical storm while over land due to the excessive rainfall.”

      I’m just starting to take an interest in such activity and found this:

      “After stalling near the mouth of Mobile Bay on July 19, Hurricane Danny turned to the east, and made its final landfall near Mullet Point, Alabama later that day. The storm rapidly weakened as it continued northward, and was only a tropical depression by the 20th. The weak depression moved through Alabama, Georgia, South Carolina, and North Carolina, maintaining a well-defined cloud signature. Due to a front behind the system, Danny strengthened baroclinically over North Carolina to a tropical storm on the 24th. It quickly reached a secondary peak of 60 mph (96 km/h), and continued rapidly northeastward.”

      Maybe it was a combination of the two, who knows?

    • This was the original article of debate:

      “If it has already rained, it’s going to continue to pour, according to a study of how ocean-origin storms behave when they come ashore. More than 30 years of monsoon data from India showed that ground moisture where the storms make landfall is a major indicator of what the storm will do from there. If the ground is wet, the storm is likely to sustain, while dry conditions should calm the storm.”

    • I’m just starting to learn the basics of meteorology and found this useful:


      BAROTROPIC – Region of uniform temperature distribution; A lack of fronts. A perfect example of a barotropic environment is the southeast U.S. in the summer or the tropics. Everyday being about the same (hot and humid with no cold fronts to cool things off) would be a barotropic type atmosphere. Part of the word barotropic is tropic. The tropical latitudes are barotropic. There are no fronts in the tropics.

      BAROCLINIC – Distinct air mass regions exist. Fronts separate warmer from colder air. In a synoptic scale baroclinic environment you will find the polar jet in the vicinity, troughs of low pressure (mid-latitude cyclones) and frontal boundaries. There are clear density gradients in a baroclinic environment caused by the fronts. Any time you are near a mid-latitude cyclone you are in a baroclinic environment. Part of the word baroclinic is clinic. If the atmosphere is out of balance, it is baroclinic, just as if a person felt out of balance they would need to go to a clinic. 

      You may run across the term quasi-barotropic- This means the fronts in the region are existent but weak. Often weak cool fronts will move into the southeast U.S. in Summer. Since the circulations created by weak fronts are weak, the atmosphere is not as dynamically unstable as it is in the case when the atmosphere is baroclinic. In most cases the atmosphere has some sort of minor temperature gradient in the troposphere outside of baroclinic regions.”

  18. frankclimate, thanks for the analysis. I agree that using moving averages in an OLS regression has to impose some autocorrelation into the regression residuals and should be taken into account for adjusting the standard error and probabilities. I have not read the paper and was thus wondering whether the authors addressed the autocorrelation issue to any degree at all or actually measured it.

    I did use the Excel data that you provided and obtained an annual mean storm duration for the reported years. I then OLS regressed the yearly means versus the reported years. After accounting for autocorrelation in the regression residuals I obtained a probability of 0.097 and with an adjusted R squared less than 0.05. I did a second regression after eliminating a single data point for the year 2011 which had a single storm with a duration of 50.1 hours. That regression resulted in a probability of 0.183 and an adjusted r square less than 0.03.

    • Thanks Ken for stepping by and replicating some of my operations. As I wrote, I didn’t made annual averages ( IMO not justified) and I used the actual monthly SST for every Hurricane. If you are interested in my calculations you can download it:

      • Frank, I had some time to look at what occurs when a synthetic series is simulated 1000 times in 2 different manners. The first 1000 time simulation was on a synthetic series with a known ar1, standard deviation and trend. From the simulations the slope was recorded with each simulation and at the end the quantiles of the slopes were determined in 0.01 increments. The second 1000 time simulation took the series result from the first simulation and did a 2 step centered 3 year moving average on it and then processed the results as in the first simulation. The quantile results from the first and second simulations were compared. Those results were very close to being the same for all quantile increments.

        What occurs in cases like this one is that the moving averaging increases the autocorrelation to the extent that it exactly negates the gain in uncertainty probability from taking a moving average. I suspected that this should be the case, but I wanted to show myself that it was true.

        If the authors of the paper being discussed here are proposing that their use of moving averages somehow decreases the slope uncertainty they are wrong. I have not yet tried another simulation that might give a better result from a moving average. It would be the case of having one or two outlier data points towards the end of the series and on the high side where the moving average could smear those points into a higher trend. That is hardly an appropriate approach in my mind but it might be what occurred.

      • Ken, thanks for your pursued investigations.

      • Frank, I did the same Monte Carlo simulations accounting for increased autocorrelation from taking moving averages with an outlier towards the end of the original series and found that the original and moving averaged series had the same quantile probabilities for the slope.

        I need to go back to the paper and attempt to see what the authors’ rational was for doing the moving averages. As I recall they attempted to adjust the probabilities of the moving average results by reducing the degrees of freedom. A Monte Carlo simulation would have been a more direct and precise method and particularly with modern day computer power available.

  19. A desperate flail for something that’s actually getting “worse” in regard to climate. Storms in the US and globally are the same or falling. Global harvests carry on growing, even fisheries bouyant.
    Why bother with this storytelling? No one cares.
    The progressive masses are fully on board with a green revolution and don’t care about data. Why should they? Just to salve some vestige of conscience? Such activity could get them in trouble in future.

    • Phil – “Global harvests carry on growing, even fisheries bouyant”. Yes, this is consistent with the increasing earth tides hypothesis.

      The tropics have got wider from the equator. The data has been found in oceans and biodiversity. Are you sceptical of this mainstream scientific claim?

  20. I think it is indisputable that Nature has been fervently on the side of global warming advocates for many years, so it is not altogether surprising that papers with questionable rigour but desired messages get past the editorial review process.

  21. Pingback: Slower decay of landfalling Hurricanes in a warmer world — really? |

  22. Pingback: A Relation Between Hurricanes and Climate Change? Really? – Econophysics2020

  23. Only one graphic is needed to show how climate history supports the alarmist and green agenda:

    • “Slower decay of landfalling Hurricanes in a warmer world…”

      I’m not interested enough to read the paper, but hypothetically, seems reasonable the decay should at least be a little slower (non-zero). Consider a land temperature of 40 F compared to 90 F.

      Same rate of decay – really?

  24. I had a look at landfall hurricane acceleration over 6, 12 and 18 hours
    for data period given 1967-2018
    (note: hurricane’s slowing down calculates to a negative acceleration)

  25. Pingback: Bundesnetzagentur: Deutsche Stromimporte in 2020 gestiegen – Climate-

  26. Kaplan, J. DeMaria, M. A simple empirical model for predicting the decay of tropical cyclone winds after landfall. J. Appl. Meteorol. Climatol. 34, 2499–2512 (1995).

  27. contribute to the increasingly slow decay. Our findings suggest that as the world continues to warm, the destructive power of hurricanes will extend progressively farther inland.