Statistics

Disproving Kossin’s Increasing Hurricane Intensity Claims: Update — Guest Post by Greg Kent

Editor’s note I have strong views on statistical “significance”, confidence intervals, p-values, and all that, finding them harmful and worse. However, Kossin used these concepts, and a criticism of his methods on his terms has value. There’s much more to Kent’s critique than just that. It is a brilliant and devastating refutation of Kossin’s claims. Science at its best.

In mid-May 2020, the Proceedings of the National Academy of Sciences (PNAS) published an article on hurricane trends over the past 39 years. That study, Kossin et al 2020, found “a clear shift toward greater intensity” storms that is manifested in the observational record as “increased probabilities of exceeding major hurricane intensity.” Both theory and climate models have long projected a shift to a higher proportion of high intensity storms but support for this theory has been difficult to detect in the observational record. Kossin et al 2020 provided a new homogenized dataset of hurricane wind intensity estimates for global tropical cyclones from 1979-2017. Based on this new dataset, the article claimed that a statistically significant shift to more intense storms has finally been detected in the observational record. By identifying “significant global trends in [tropical cyclone] intensity over the past four decades,” as projected by the models, the paper’s findings were said to “increase confidence in projections of increased [tropical cyclone] intensity under continued warming.”

The mainstream media trumpeted the results of this new study to a broader audience. The New York Times stated that “Climate Change Is Making Hurricanes Stronger.” Similarly, the Miami Herald urged south Floridians to “Expect stronger, deadlier, more frequent hurricanes in the years to come.” The Washington Post headline claimed that “The strongest, most dangerous hurricanes are now far more likely because of climate change.” As if the headline wasn’t scary enough, the Post’s article warned the public: “With powerful hurricanes on the increase, one can expect damage costs, in dollar terms and in potential loss of life, to skyrocket.” The lead author of the study, Dr. James Kossin, told the Post that “we have high confidence that there is a human fingerprint on these changes.” All of this was quite worrisome news, coming as it did just before hurricane season started.

The problem is that most of the claims of these newspapers simply weren’t true. Unfortunately, the Kossin et al 2020 paper contained a rather significant error: the paper reported erroneous values for the “metric of interest” on which the paper’s conclusions rest. Once corrected, the purported increase is no longer statistically significant. If that wasn’t bad enough, the media hyped-up the results of the paper to reach conclusions that are far more worrisome than what the paper actually found.

SUMMARY OF THE PAPER

Kossin et al 2020 updated a homogenized hurricane intensity database based on satellite imagery. The standard “best track” dataset is based on observational records with known variations across timeframes and regions, so it is difficult to make long-term comparisons. The new homogenized dataset provides estimates for storm intensity based on satellite imagery. Although these are just estimates (placed in 5-knot bins), they are at least consistent and allow apples-to-apples comparisons across basins and timeframe for the 39 years for which available data exists. The dataset contains hurricane wind speed estimates every 6 hours during the lifetime of each storm. These every-six-hour measurements of wind speed in knots are known as “fixes.” Essentially, the Kossin study split the 39-year record into two periods (1979 + 1981-1997 vs. 1998-2017) and counted the number of “fixes” in each period that are greater than or equal to 65 knots (indicating a hurricane) and the number of “fixes” that are greater than or equal to 100 knots (indicating a major hurricane of Category 3 or higher). The study then used these two counts to create a proportion: the percentage of all hurricane-strength “fixes” that are at major hurricane intensity. This proportion of 6-hour wind speed measurements is the study’s “metric of interest.”

The bottom-line finding of the study is that that proportion of 100 kt or greater intensity “fixes” increased from 27% (CI=[25%, 28%]) of all hurricane-force fixes in the early period to 31% (CI=[29%, 32%]) in the later period. Because the confidence intervals don’t overlap, the increase in the proportion of major hurricane intensity wind speeds is statistically significant. Hence, the paper declared that the long-expected shift to more intense storms was finally observed in the record. As will be seen, it is critical to keep in mind what this “metric of interest” really says. It is strictly about the proportionate share of all 6-hour intensity measurements. It is a relative measurement, not a measure of absolute numbers.

IRREPRODUCIBLE RESULTS

Unfortunately, the results reported in the paper are inconsistent with the underlying dataset that was supplied with the paper’s supplemental information. To derive the “metric of interest,” the paper reported a total of 2362 major-hurricane-force wind speed measurements out of a total of 8848 total measurements of hurricane wind speed for the period 1979 + 1980-1997. These numbers are incorrect. The correct counts, which are easily derived from the supporting data on the PNAS website, are 3202 major-hurricane-force winds speeds and 9420 total hurricane-level wind speeds. The counts are similarly erroneous for the later period 1998-2017. Since the counts were wrong, the proportions used in the study were also wrong, as was the change in the proportion between the early and later period. The original paper reported that major storm wind speeds accounted for 27% of all hurricane wind speeds in the early period but rose to 31% in the later period, or a 15% increase. In fact, the proportion of major hurricane wind speeds only increased 10%, rising from 34% in the early period to 37% in the later period. I notified Dr. Kossin of these error before Memorial Day, and he replied shortly thereafter confirming the errors and indicating that an erratum would be issued. In November, a Correction was published in PNAS acknowledging the original paper contained calculation errors due to a shift in the research team’s code. After correcting for that error, the Correction reports that the proportion of major hurricane wind speeds had in fact only increased by 10%, rather than the 15% originally reported.

Table A. Shows key metrics (including confidence intervals) from the original paper and the correction published in PNAS, as well as metrics recalculated from the dataset.

The Kossin et al correction revised several sentences in the paper and completely revised Table 1, which contains the detailed numeric results of the study globally and for each hurricane basin individually. Unfortunately, the revised Table 1 still contains errors. Kossin et al failed to change the global results for the late period so that the erroneous values from the original paper (Ntot = 9275 and Nmaj = 2842) still exist in the Correction’s version of the table. It is clear that this error was a minor oversight because the proportion, which is the “metric of interest”, for the late period was updated to the correct value (e.g., Pmaj = 0.3725), and each of the individual basins was properly updated (as will be seen below in Table C).

STATISTICAL SIGNIFICANCE, OR NOT?

More important is the claim in Kossin et al’s Correction that “none of the errors alter any of the key results or messages of the manuscript.” In other words, the Kossin et al Correction continues to maintain that the difference in the proportion of major wind speed remains statistically significant. Statistical significance is of primary importance to the value of the paper, since Kossin et al is the first paper to claim to have found a significant increase in hurricane intensity in the observational record. Unfortunately, there is reason to suspect that the results are not actually significant. As noted previously, the change in the proportion of global major hurricane winds was erroneously reported in the original article as increasing by 15%, which just barely crossed the threshold for statistical significance (e.g., the upper bound of the early period was 28% and the lower bound of the late period was 29%). With the correct calculations, the real difference between periods is significantly smaller (only 10%), which implies that the statistical significance of the change would likely decline. Since the original confidence intervals were contiguous, it would seem doubtful that sufficient wiggle room would exist to continue to claim statistical significance.

Kossin et al attempted to supply some wiggle room by adding two extra decimal places to their confidence intervals shown in Table 1 of the Correction. The Correction shows the upper bounds of the early period as 35.55% and the lower bounds of the late period as 35.59%. If the conventions of the original paper had been followed, these both would round to 36%, showing overlapping confidence intervals and rejecting statistical significance. By adding two additional decimal places to their calculations in the Correction, Kossin et al can show a miniscule gap of 4/10,000ths between the confidence intervals in order to claim statistical significance.

However, the confidence intervals shown in the Correction appear to be erroneous. The original paper describes the method for calculating the 95% confidence intervals. The standard process for calculating confidence intervals of proportions is followed with one exception. The full sample size of hurricane wind speed measurements cannot be used in the calculation of the standard error because the data points are not independent. According to the “Methods,” the degrees of freedom “are reduced by a factor of 3, which assumes a decorrelation of 18 hours.” In other words, testing for statistical significance uses 1/3 of the population size when deriving the standard error.

When I calculate the confidence intervals following this methodology, the results are different from what is shown in the Correction. The actual confidence intervals of the two periods overlap. The upper bound of the early period is 35.65% and the lower bounds of the late period is 35.49%, which means that the confidence intervals overlap by 16/10,000ths. Overlapping confidence intervals, of course, means that the difference in the proportions is not statistically significant.

Table B. Shows statistics for determining significance via confidence intervals as reported in the Correction, as well as statistics recalculated from the dataset following the “Methods” of the original paper.

It seems likely that another Correction may be needed to retract the main claim of the paper since, in fact, a statistically significant shift toward major hurricanes is still not observable in the record. This isn’t to dispute that the mix of hurricane winds most likely has shifted. Eventually, the shift in the proportion of intense winds will very likely be significant in the observational record. The point is that it was premature to make that claim for the observational record through 2017.

What are we to make of the errors in the paper and the Correction? After all, Dr. Kossin is one of the leading hurricane scientists in the world and the Proceedings of the National Academy of Sciences is no slouch outfit. If fundamental errors can occur in a paper like this, then apparently no scientific paper is infallible. To anyone who has been paying attention, it’s clear that science in general, and climate science in particular, needs better quality control. Fifteen years ago, Stephen McIntyre sounded the alarm that climate science desperately needed an audit, but sadly not much has changed. The traditional peer review process, in which reviewers generally don’t look at code or actually reperform analysis, isn’t up to the task of ensuring that the results are correct before they are published. We live in a time when “science” is the new religious dogma and skepticism, once a sine qua non of scientific inquiry, is denounced and vilified as “denialism.” In such a context, it is more important than ever that “science” get it right. Too often, it doesn’t.

GLOBAL RESULTS ARE DRIVEN BY ONLY ONE HURRICANE BASIN

The pervasive erroneous calculations in the original paper and the invalid claim of statistical significance are not the only issues with Kossin et al. There is also reason to question whether the 10% increase in the proportion of major hurricane force winds was a global or largely regional phenomenon. Kossin et al presented results for each of the hurricane basins around the world. The data shows that the global results are driven largely by a single basin, the North Atlantic. The proportion of major wind speeds increased by 72% in the North Atlantic, far more than in any other hurricane basin. Western Pacific, which accounts for over 40% of the major hurricane force winds over the last 4 decades, showed a smaller proportion of intense storms in the later period (indicating a negative change). The other basins either showed no change at all between periods or the change was so small as to fail tests of statistical significance at traditional levels of confidence. Across all three Pacific basins, which account for 3/4ths of global major hurricane force winds, the proportion of intense winds increased less than 2%. The increased proportion of major storm “fixes” is statistically significant only in the North Atlantic basin. Unfortunately, the North Atlantic basin contains only a small percentage of global hurricanes (i.e., about 12% of global major-hurricane force winds) and is heavily influenced by natural variation. The cut-off year between Kossin et al’s two periods (1997/1998) occurs within a year or two of a flip of the Atlantic Multidecadal Oscillation (AMO) from a cool period to a warm period, and it is widely known that the AMO has a substantial impact on the intensity of North Atlantic basin tropical cyclones.

Table C. The change in the proportion of major hurricane wind speeds to all hurricane wind speeds from the early to late period by hurricane basin, and the percentage of major hurricane wind speeds for each basin. The values are copied from the Kossin et al Correction, except for the global wind speed counts for the late period (which was erroneous in the Correction) and values for the South Atlantic basin (which were not reported in the Correction). Those values are recalculated from the dataset.

Because Kossin et al’s increase in the proportion of major storms is driven largely by the North Atlantic basin, and the intensity of North Atlantic basin storms are largely driven by a natural variability, two inevitable conclusions result: (1) the observed global shift to higher intensity winds may be merely temporary and will likely reverse when the AMO flips to a cool phase in the next few years and (2) the affect observed by Kossin et al has not been shown (and likely cannot be shown) to be caused by human-influenced global warming. Despite the claims in news articles, the paper itself specifically discusses the impact of natural oscillations and other natural factors. Kossin et al recognizes that the multidecadal variability, particularly in the North Atlantic where the observed affect is strongest, “complicates detection” of human influences. The paper acknowledges that “many factors contribute” to the phenomenon discussed in the paper, and the paper “makes no attempt to formally disentangle all of these factors.” In other words, the paper does not try to determine how much of the observed change in the proportion of major hurricane fixes is due to natural vs. man-made causes. The paper even explicitly disclaims that it constitutes “a traditional formal detection” of human influence.

That Dr. Kossin spoke of “high confidence that there is a human fingerprint” to the Washington Post is rather disingenuous because his study explicitly falls short of such claims, almost certainly because those claims cannot be established with sufficient rigor to pass scrutiny in peer review. It is unfortunate that the Post reporter failed to check whether Kossin’s interview matched the findings in the paper. In fact, there is a glaring disconnect with the disclaimer language appearing in the paper. The actual claims made in the paper are much more subdued. Whereas confidence in a theorized “link increasing TC intensity to a warming world” had been “compromised” by the absence of a supporting evidence, the new dataset “increases confidence that TCs have become substantially stronger, and that there is a likely human fingerprint on this increase.” The paper only claims “increasing confidence” in a theoretical “link,” but falls short of claiming that the threshold of “high confidence” has been established. In climate science, “high confidence” has a specific meaning, and the paper fails to establish anything like it. In essence, all the paper is saying that the theorized “link” between warming and increasing intensity as a proportion of total winds is no longer completely untenable because it is at least clear that proportionately more storms have intense winds. That is a minimum but essential element to establish a causal link. Showing correlation is the first step toward demonstrating causation, but it is still a far cry from proving a causal link. The Kossin et al study merely indicates that his results go in the same direction as the models and theory, but do not address causation directly. Of course, all of those statements were based on the results from the original paper. Take away statistical significance and even the more subdued claims of the paper collapse. At least for now.

THE MOST INTENSE HURRICANE WINDS ARE UNCHANGED OR SLIGHTLY DECREASING

Another interesting factor from the Kossin et al dataset is that the increasing proportion of intense storms has occurred for only the least intense levels of major hurricanes, namely category 3 storm winds, and not for the highest intensity category 4 or category 5 winds. The paper itself acknowledges, there is an “absence” of an increased proportion of intense storms “at the most intense part of the intensity spectrum.” In other words, the most intense storms have not increased in the observed record. This is easily seen by a careful examination of the paper’s figure 1, shown below left. Because of the calculation errors in the original paper, I have recreated the figure with corrected values derived from the dataset, shown below right. The overall pattern is the same. For each 5-knot bin of wind speed, the graphs show the different proportion of “hurricane fixes” in the early period (blue) and late period (red) that exceed that wind speed. For 100 knots (the lowest wind speeds for Category 3 storms), the red line (late period) shows a higher share of “fixes” exceeding 100 kts than the blue line (early period). Then, moving to the right for each successive wind speed bin, the increase in the proportion from the early to late periods (that is, the spread between the red line and the blue line) becomes smaller and then relatively quickly disappears. See the green circled area of the graphs. By the time wind speeds reach 115 kts, indicating category 4 force winds, Kossin et al’s data shows no increase whatsoever between the early period and the late period. In fact, the graph shows that the proportion of the most intense storms, specifically Category 5 storms exceeding 135 kts, is actually slightly less in the late period compared to the early period. The aqua circle in the graphs show that the red line (later period) is slightly below the blue line (early period), as the proportion of Category 5 storms have slightly decreased over the 40-year study period.

Figure A. Graphs showing the proportion of major hurricane force winds to all hurricane force winds by 5-knot bins for the early (blue) and late (red) periods. Left is from the original PNAS article Figure 1. Right is recalculated from the dataset. The green circled regions of the graphs show that the increase from early to late periods for the 100-knot bin rapidly diminishes to zero by 115 to 120 knots. The aqua circled regions of the graphs show that for the intense winds over 130 knots, the late period falls slightly below the early period.

This decrease in the proportion of Category 5 winds is not statistically significant, but it is rhetorically significant. If Kossin et al had chosen to highlight the “absence of a probability shift at the most intense part of the intensity spectrum” in language that lay folks (like media and politicians) could understand, the rhetorical impact of his findings would have been significantly damped. The truth is that Kossin et al’s results about the overall mix of hurricane winds shows an increased proportion of Category 3 winds and only Category 3 winds. The proportion of storms of Category 4 winds is unchanged in both periods, while the proportion of Category 5 winds slightly decreased. Category 3 winds are quite destructive, but Category 4 and 5 winds, though less common than Category 3 winds, are significantly more destructive. Pielke et al 2008 found that Category 3 storms cause 35% of the total damage, so it is troubling that their proportion is increasing. But Category 4 and 5 storms cause about 50% of the total damage, so it is reassuring that the proportion of those storms has held steady or even decreased slightly. With full disclosure of all results, the true picture of Kossin’s findings is more nuanced than the hyped media story and is at least partly good news.

RELATIVE PROPORTIONS VS. ABSOLUTE NUMBERS OF MAJOR HURRICANE WINDS

One final but crucial discussion point remains. The Kossin et al paper is strictly about major hurricane wind speeds as a proportion of all hurricane wind speeds. That is, it is about a proportion that measures the mix of hurricanes in a relative sense. That paper makes no claims about the absolute number of storms or wind speeds – it is only about the proportionate share that is reported. Although proportions are interesting, they ought not to be reified into a thing that exists. In the real world, it is the absolute frequency of winds that matters. Unfortunately, Kossin et al makes it hard for readers to keep their eye on the pea because the article refers to the “metric of interest” as both a proportion (11 times) and a probability (15 times), and the language used to describe that probability is rather opaque. “The probability of major hurricane exceedance” is the probability that the every-6th-hour wind speed measurements of hurricane force winds will exceed the threshold for a major hurricane. It is still about the mix of major hurricane forces winds in a relative sense (i.e., relative to all hurricane force winds). It is not about the absolute probability of major hurricanes or major hurricane winds.

Although this distinction is critical, the mainstream media failed to grasp it. The media (and likely most of the public) assume that an increasing proportion of major hurricane strength winds/storms necessarily implies that there the frequency of major hurricane strength winds/storms will increase. That is false. Mainstream media like the Washington Post interpreted the paper’s statements regarding increasing “major hurricane exceedance probability” as meaning that the odds of society experiencing the most destructive hurricanes has increased. However, that is not necessarily true. What may be true about the relative mix of hurricanes does not necessarily mean the absolute number of major hurricanes has increased. Why not? If the number of hurricanes is decreasing, then the mix of hurricanes could become more intense in a relative sense while the absolute number of intense hurricanes could decrease.

Think of a pizza parlor that used to serve only one size: an 8-slice large pizza for 3-4 people. Because COVID restrictions have kept larger groups away, the parlor changes its pizza size to a 4-slice small pizza for one. As a proportion of the total pie, the slices of the small pizza (25% of the whole) are much larger than the slices of the large pizza (12.5% of the whole). Despite the relative increase in the size of the slices, because the small pizza is so much smaller than the large pizza, one of its slices is still much smaller in an absolute sense than a slice from the large pizza. In the real world of pizza, it is the absolute size of the slice that matters — that’s what determines whether you’ll feel hungry or stuffed when you leave the restaurant. Knowing only how the proportion per slice has changed is not very helpful, unless you also know how the overall size of the pizza has changed. A relatively larger slice of a much smaller pizza can be smaller in an absolute sense than a relatively smaller slice of a much larger pizza. And what is true for pizza is also true for hurricane winds.

Figure B. One slice of an 8-slice large pizza (accounting for 12.5% of the whole) is larger in an absolute sense than one slice of a 4-slice small pizza (accounting for 25% of the whole), even though it is smaller in a relative sense.

In a very real sense, the relative share or proportionate metric, which is what Kossin’s paper is all about, is not the most important metric. The real question is whether there will be more major-hurricane-intensity winds in an absolute sense. Unfortunately, the models provide little help there. The same climate models that project an increase in the proportionate share of more intense hurricanes also project an overall decrease in the number of hurricanes. For example, Murakami et al (2020) published in PNAS just a few weeks earlier than Kossin et al, is a climate model study that projected a significant decrease in the number of tropical cycles. Essentially, the climate models project that major hurricanes will be a proportionately bigger slice of a much smaller pie.

Figure C. Figure S3 from Murakami et al 2020 showing projected declines in tropical cyclones through 2100 both globally and in the North Atlantic basin.

However, the latest modelling efforts haven’t reached a consensus on which effect (the increasing size of the relative slice or the decreasing size of the overall pie) will have the greatest impact. For example, the World Meteorological Organization expert team assessment of tropical cyclone projections was published in the Bulletin of the American Meteorological Society as Knutson et al 2020. The expert team, which included Dr. Kossin, was “divided on whether the global frequency of very intense TCs will increase or not” because of “competing influences.” On the one hand, the expert consensus is for “a general decrease in overall TC frequency.” On the other hand, the experts also agree that there will be “a generally increasing average TC intensity.” It is the “combination” of these two offsetting factors that befuddles the experts and makes estimates of the net effect on the frequency of intense storms “less robust.” In layman’s terms, this means the experts have no idea whether the frequency or absolute number of very intense storms/winds will increase or decrease. As a member of the WMO panel, Dr. Kossin believed that the net effect will “likely” increase the absolute frequency of very intense storms, but the majority of Dr. Kossin’s colleagues on the panel did not share his confidence in that projection.

KOSIN ET AL ON WHETHER MAJOR HURRICANE FORCE WINDS ARE INCREASING IN ABSOLUTE TERMS

The new Kossin et al study focused strictly on a relative metric concerning the proportionate share of intense wind speeds in order to determine if the expectation of the models were correct. It did not address the more pressing concern about expectations for the absolute number or frequency of intense storms or winds. Fortunately, these measurements can be derived from the supplemental dataset that Kossin provided with the paper at PNAS. Rather than examining the every-6th-hour wind speed measurements as proportions, we can annualize those wind speed counts at different intensities to see exactly how global wind speeds have changed in an absolute sense over the last 40 years. This analysis, which Kossin et al failed to do, will answer the critical societal concern of whether the most intense wind speeds are increasing in absolute terms.

The only complication to annualizing the data is the difference in years between the two periods. The early period consists of 1979 and 1981-1997, or 18 years. (The year 1980 was excluded because satellite coverage was incomplete.) The later period from 1998-2017 contains 20 years. The results showing annualized frequency of hurricane fixes (in absolute numbers) is shown below. The data show that the number of intense hurricane “fixes” is declining, and the most intense wind speeds have declined the most. Category 3+ winds speed “fixes” declined 9%, Category 4+ declined 13%, and Category 5 have declined a whopping 29%. One might wonder how the absolute number of major storm force winds could decrease 9% while the relative proportion of major storm force winds (shown in the shaded column below) has increased 10%. The answer to this apparent paradox is that the change in the total number of all hurricane force winds (i.e., a 17% drop) was much larger than the increase in the proportionate share of major storm winds. The net affect has been a decrease in intense winds and a significant decrease in the most intense winds over the last 40 years. Dr. Kossin’s latest data show that his stated concern about projected future increases in the frequency of very intense storms are not supported by observations to date and may be unfounded, as the majority of his WMO colleagues suspected.

Table D. The number of wind speed “fixes” above category 1, 3, 4, and 5 wind speed thresholds in knots for the early and late periods as a whole and annualized. The proportion of major-hurricane winds to all hurricane force winds is also shown highlighted in gray. These numbers are derived by the Kossin et al dataset.

Another interesting question is how these changes in the intensity of wind speed measurements equate to destructiveness. Pielke et al (2008) looked at normalized losses from US hurricanes over the long term after normalizing for changes in economic development. That study provided estimates of the destructiveness of storms by Saffir-Simpson category level. Category 3 storms caused 37% of the destruction, category 4 storms caused 41%, and category 5 storm caused 7%. Although a bit dated and focused solely on the US, Pielke et al 2008 remains a useful source of information about the relative damage caused by storms, as the recent WMO study by Knutson et al referenced it for precisely this reason. Applying Pielke’s destructiveness percentages to the reductions in wind speed measurements between Kossin et al’s early and later periods produces a method to approximate the change in damage potential or destructiveness over the last 40 years of the observation record. The result shows that the destructive potential of global intense hurricane force winds has fallen by 10%. In terms of social welfare, that reduction in destructive potential is very good news.

Table E. Percentage change in annualized wind speeds and destructive potential between early and late periods by Saffir-Simpson hurricane category. Destructive potential for each category of storm is based on Pielke et al 2008.

Recall the worrisome words of the Washington Post’s article on the Kossin et al paper: “The strongest, most dangerous hurricanes are now far more likely because of climate change…With powerful hurricanes on the increase, one can expect damage costs, in dollar terms and in potential loss of life, to skyrocket.” Not only was that conclusion wrong, it was completely wrong. The truth is that major hurricane winds have decreased since 1979 about 9%, resulting in lower destructive potential: about 10% lower than if hurricanes winds had behaved as they did in the early period. Besides being good news for the present, this finding also sheds some light on what the future may hold. Recall that the models and experts are unsure of how the frequency of the most intense hurricanes could change in the next few decades in an absolute sense. Thanks to Kossin et al’s dataset, we have the data to know how hurricane winds were affected during the first 40 years of climate change. The decline in major hurricane forces winds that has occurred provides us some preliminary reason to be hopeful that intense storms/winds will continue to become less frequent (even if the proportion of intense winds to all hurricane winds continues to rise), and that is truly good news worth reporting widely.

Interestingly, two weeks after Kossin et al was published in PNAS, a hurricane modelling paper was published in Geophysical Research Letters. Stansfield et al (2020) provided hurricane projections specifically for the North Atlantic basin for the years 2070-2100. Stansfield et al’s model forecasts a 22%-32% decrease in the total number of North Atlantic tropical cyclones and a 27%-32% decrease in the number of landfalling cyclones, but also a “slight shift toward more intense storms.” The median lifetime maximum windspeed is projected to increase about from 0-2% (about 1.3 mph). That increase in intensity is so small as to be virtually undetectable. A 2% increase in windspeed is a bargain to exchange for a 32% reduction in the frequency of storms, especially when the most intense storms (namely, Category 4 and 5 storms) are expected to decrease, as Stansfield also found. Surprisingly, the study even showed that the most intense storms (the 95th percentile) in the future are projected to have windspeeds that are 10 mph (7%) less than present windspeeds. In fact, the probability of having lifetime maximum intense winds exceeding any threshold over 45 m/s will be less in the future than it is today. Since the threshold for a Category 3 storm is 50 m/s, that means that major hurricanes in the North Atlantic will be less likely in 2070-2100.

Figure D. From figure 1 of Stansfield et al 2020 showing the probability of North Atlantic hurricanes in 2070-2100 exceeding lifetime maximum intensity in 2.5 m/s bins. The blue line is the reference scenario based on current (1984-2014) conditions. The two “global warming” scenarios are based on representative concentration pathway (RCP) scenarios of carbon emissions used by the UN’s International Panel on Climate Change (IPCC). The green line is for moderate warming based on RCP45 (a realistic carbon emissions scenario based on current energy policies), while the red line is for significant warming based on RCP 85 (a highly unlikely “worst case” emissions scenario). The graph shows that the likelihood of level 1 storms increases under global warming (e.g., the green and red lines are higher than the blue line). However, for wind speeds above 45 m/s, the green and red lines are below the blue line, indicating that category 2-4 storms are less likely to occur in a warming future, regardless of whether that future entails moderate or significant warming.

The Stanfield et al projections appear to indicate that the beneficial pattern that is detectable in Kossin et al’s dataset may continue in the future: namely, that a significant decrease in the frequency of storms may more than compensate for a slight increase in the proportion of intense storms. All in all, hurricanes may very well be substantially less dangerous in a future world than today.

As far as I can tell, Stansfield et al received almost no press coverage. The New York Times and Miami Herald did not report on these favorable projections for the future. The Washington Post, whose official slogan is that “Democracy dies in darkness,” chose to leave to the public in the dark rather than allow the good news of Stansfield to see the light of day on its pages.

Update This is from Kent in answer to criticisms.

The data I used was definitely not smoothed – the dataset involved thousands of every-6th-hour wind speed estimates. Of course, creating a proportion over an 18 year and then 20 year period is a kind of smoothing I suppose. Kossin did present a “triad time series” as an alternative way of looking at the data that did involve smoothing (centered 3 year average) and then creating a trend. That would seem to be a no no. But the time series was a relatively minor part of Kossin et al. The main focus of the article was on the proportions of the early and late period, so that’s what I choose to focus on. Unless I’m missing something, I don’t see this as an additional minor criticism that could be made of Kossin et al, but I don’t believe it is relevant to my analysis.

Subscribe or donate to support this site and its wholly independent host using credit card or PayPal click here

Categories: Statistics

12 replies »

  1. Thanks as always for such a tour-de-force, Matt. However, the clear exposition of the statistical error wasn’t the best part.

    I fell out of my chair laughing upon reading your contender for the best math line of the decade, viz:

    “This decrease in the proportion of Category 5 winds is not statistically significant, but it is rhetorically significant.”

    Too good, my friend.

    w.

  2. Kossin used time series data (measured with error) smoothed and average into two data points. Kent failed to note that but instead made the same mistake.

    Figure C from Murakami is a proper display of time series data.

    Junk science critiqued with junk science is still junk science.

    Bottom line: minuscule change in “storm intensity”, even if it was occurring, is an insufficient excuse to destroy civilization via authoritarian oppression.

    Kossin and the NYT are tools. They need to be censored and cancelled. They have no place here at Briggs where standards have been set and the bar is much, much higher. If this is the promised exploration of the Climate Change Hoax, it disappoints and falls pathetically short of expectations.

  3. Disclaimer: my sincerest sympathies for folks born with 47 chromosomes and the hardships that come with it.

    Picking 165+ vs 100+ wind speeds measurements as a “metric” for whether hurricane intensity is increasing is ******* retarded. Why not 163 and 92? Does nature somehow care how we classify hurricanes? Why are you picking numbers at all? The wind speed data is the data. When you pick numbers you’re decreasing knowledge, not increasing it.

  4. Interesting .. not so much because the media misrepresented the original study’s conclusions and then ignored the corrections, but because something this shoddy passed through many reviews before somebody bothered to check the arithmetic and sources. Kudoes etc on doing it!

    On the other hand.. the underlying issue with the Kossin et al paper is a common one: developing extremely precise conclusions from very imprecise data. Cyclone wind speed estimates are, for example, obtained by averaging two somewhat correlated sets of measures and adjusting (extensively) for measurement geometry.

    Developing stats to three decimal places to describe data measured to, at best, +- 20% is reasonable, drawing conclusions about the processes these data describe is ok too, but silently assuming that the precision of the stats describing the data pertain to the conclusions drawn from that data is absurdly illogical, what both Kossin and Kent do, and one of the things wrong with many papers in this field.

  5. I had the same thought Ryan. I guess these are some sort of standard classification but I would want to run a decision tree or clustering process to group the storms or see this on a continuous scale. It’s just easier to run it this way and if it gets a ‘significant’ result, all the better. Just lazy.

  6. Let’s use the method to see if snow fall has changed. Take all your snow records from 1981-1997 and average them. Then take all the snow fall records from 1998-2017 and average them. Do a means comparison, say a Duncan’s Multiple Range Test.

    Then throw it all in the trash bin because you cannot test Time Series Data that way! Time series go up and down like a roller coaster. Sometimes they trend down then up. Sometimes they trend up then down. Sometimes they jiggle. Sometimes they’re flat. In any case, their averages over two pre-selected time periods are meaningless, without inferential or predictive power.

    Duh! Hey all you whiz bang stat experts, tell me why I’m wrong. Go ahead, make my day. Brilliant my left nut. Come on, man.

  7. Mr. Kent, that is an exacting demolition, thank you.

    Confirmation that “climate change is making hurricanes stronger” is a Big Lie:

    ”The New York Times stated that “Climate Change Is Making Hurricanes Stronger.”

  8. Thanks Greg

    As you pertinently point out all of the increase comes from the N Atlantic. But it is well known that this has been driven by the Atlantic Multidecadal Oscillation – here’s what NOAA say:

    “During warm phases of the AMO, the numbers of tropical storms that mature into severe hurricanes is much greater than during cool phases, at least twice as many. Since the AMO switched to its warm phase around 1995, severe hurricanes have become much more frequent and this has led to a crisis in the insurance industry.”
    https://www.aoml.noaa.gov/phod/faq/amo_faq.php

    The AMO was in cold phase till the mid 1990s, since when it has been warm.

    This fact alone is enough to sink Kossin’s claims

  9. What an excellent article. Thank you. I found it very clear and easy to understand.
    It appears to be very logical.

    Taken on face value, the criticism extremely scathing, Kossin has committed a biased review and cherry picked the data to support a low value statistic. Various media have then amplified that to give a false alarmist message of climate change.

    However I have learned from experience of lawsuits that you have to wait until both sides have spoken.
    I would be interested in what Kossin’s response to this is – has he posted any sort of reply?

Leave a Reply

Your email address will not be published. Required fields are marked *