Thongchai Thailand

Climate Change and Hurricanes

Posted on: September 18, 2018
















  1. Sea surface temperature (SST) is the link that connects climate change research with tropical cyclone research. Rising SST is observed (Hadley Centre, 2017) and thought to be an effect of Anthropogenic global warming or AGW (Hansen, 2005) . At the same time, the theory of tropical cyclones holds that cyclone formation, and particularly cyclone intensification are related to SST (Vecchi, 2007) (Knutson, 2010). Testable implications of the theory for empirical research are derived from climate model simulations (Knutson, 2010) and also from sedimentary evidence of land-falling hurricanes over a 1500-year period (Mann, 2009). These studies suggest some guidelines and testable implications for empirical tests of the theory that AGW affects tropical cyclone activity (Knutson, 2010). (Details in footnote below)
  2. A high level of interest in tropical cyclones derives from an unusually active hurricane season in 2004 when more than 14 tropical cyclones formed in the North Atlantic basin . Four of these storms intensified to Category 4 or greater and made landfall in the USA causing considerable damage. The even more dramatic 2005 season followed in its heels with more than thirty depressions. Four of them intensified to Category 5 and three made landfall. The most intense was Hurricane Wilma but the most spectacular was Hurricane Katrina which made landfall in Florida and again in Louisiana. Its devastation was facilitated by a breach in the levee system that was unrelated to AGW but its dramatic consequences made it an icon of the possible extreme weather impacts of AGW.
  3. The Emanuel paper (Emanuel, 2005) came in the heels of these events and is possibly best understood in this context. The assumed attribution by the media of the epic devastation to AGW set the stage for climate science to claim the destructiveness of hurricanes as extreme weather effects of AGW. The Emanuel 2005 paper was one of several published in the heels of these hurricane seasons. The paper presents a new measure of tropical cyclone intensity which the author calls “Power Dissipation Index” and to which he assigns the acronym PDI. The paper finds a statistically significant rising trend in the aggregate annual PDI of North Atlantic Hurricanes in the study period 1949-2004 in tandem with rising sea surface temperature (SST) for the appropriate zone where hurricanes form. The graphical depiction of this result is reproduced in Figure 1.
  4. The usual measure of tropical cyclone activity is the ACE or Accumulated Cyclone Energy. It is computed as the sum of squares of the maximum sustained wind speed in each 6-hour window during the life of the cyclone. It represents the total amount of kinetic energy generated by a tropical cyclone and this energy has been related to the energy in the ocean surface as measured by surface temperatures and temperature differentials such that the cyclone can be described as a heat engine (Rotunno, 1987) (Emanuel, 1987) (Goni, 2003) (Latif, 2007) (Klotzbach, 2006) (Emanuel, 2003).
  5. When the ACE measure did not show the trend that the author was looking for he decided to cube the velocities instead of squaring them as a way of increasing the differences among annual values. This innovation produced the desired result and a trend became evident over the last 30 years of the 55-year study period as seen in Figure 1.
  6. Since the sum of cubes could not be called ACE, the author gave it a new name and called it the Power Dissipation Index or PDI. The object variable in the hypothesis was thus changed from ACE to PDI. The PDI hypothesis was then modified to exclude the first 22 years of the study period where no trend and very little correspondence between PDI and SST are seen (Figure 1). Thus, the tailor made hypothesis to be tested was whether there is a rising trend in the PDI in the most recent 30 years (1975-2004) of the study period. This hypothesis was then tested with the same data over the same time span that was used to construct it. The procedure of testing a hypothesis with the data used to construct the hypothesis constitutes circular reasoning because the methodology subsumes and ensures the desired result.
  7. At this point, a rising trend is seen in the PDI time series 1975-2004 at an annual time scale but the trend is not statistically significant because of extreme year to year variability in cyclone formation and intensification (Knutson, 2010). To smooth out the variance, the author took a 5-year moving average of the PDI data; and when that also failed to show a statistically significant trend, he took 5-year moving averages of the 5-year moving averages (in effect a 10-year moving average) and was finally able to find statistical significance for rising PDI over the last 30 years of the study period (Watkins, 2007). The findings presented by the paper are based on this rising trend and the visual correspondence between PDI and SST seen in Figure 1.
  8. However, in his hypothesis test computations the author failed to correct for degrees of freedom lost in the computation of moving averages. When moving averages are computed some data values are used more than once. It can be shown that the average multiplicity of use is given by the relationship M = (λ/N)*(N-λ+1) where M is the average multiplicity, N is the sample size, and λ is the width of the moving window (Munshi, 2016) (VonStorch, 1995). In the case of a window with λ=10 years moving through a time series of N=30 years, the average multiplicity is M=7. The effective sample size is computed as EFFN=N/M or EFFN=30/7 = 4.285 and the degrees of freedom for the t-test for trend is DF=EFFN-2 or DF=2.285. The statistical significance reported by the author at N=30 and DF=28 is not found when the sample size is corrected for multiplicity. A false sense of statistical power was created by the methodology used when decadal moving averages were taken (Watkins, 2007) (Munshi, 2016). Full text download links for the paper on moving averages [SSRN.COM]  [ACADEMIA.EDU] .
  9. The North Atlantic basin is just one of six major cyclone basins around the world. The other five are The West Pacific, the East Pacific, the South Pacific, the North Indian, and the South Indian. The most active basin is the West Pacific. The theory of anthropogenic global warming as expressed in terms of climate models indicates that only long term changes in global averages of all six cyclone basins may be interpreted in terms of the impacts of climate change (Knutson, 2010). The study of a single basin over a brief 30-year period is unlikely to contain useful information relevant to AGW. Data for all six basins over a 70-year study period 1945-2014 does not show trends in total aggregate annual ACE that can be interpreted as an impact of warming as shown in three related posts on this site [LINK] [LINK] [LINK] . 
  10. The left frame of Figure 2 shows the relative activity of the six basins over the 70-year study period. From left to right the basins are East Pacific, North Atlantic, North Indian, South Indian, South Pacific, and West Pacific. The North Atlantic, although close to home for many climate researchers, is not a significant source of cyclone energy in a global context. The right frame of Figure 2 does not show a sustained trend in tropical cyclone activity on a decadal time scale particularly when the decade 2005-2014 is included.
  11. It is likely that (Emanuel, 2005) was a product of climate activism that had reached a high level of intensity in the years leading up to 2005 by way of the push for the ratification of the Kyoto Protocol for CO2 emission reduction as well as the European heat wave of 2003 that was claimed and widely accepted to be caused by AGW. It was a time when the extreme weather effect of AGW was given credence by the IPCC and generally taken for granted. Given the theoretical basis that connected SST to tropical cyclones, the truth of AGW driven hurricane intensity was thus taken to be a given (Emanuel, 1987) and then apparently proven by the 2004/2005 hurricane seasons. It remained for climate science only to tend to the details of presenting the data in the appropriate format.
  12. Thus the ultimate form of circular reasoning is found in (Emanuel, 2005) in which a high level of confidence ex-ante in the truth of the proposition that AGW causes extreme tropical cyclone activity left the presentation of empirical evidence of that relationship as mere detail. The role of confirmation bias in research of this nature is discussed in a related post [CONFIRMATION BIAS] .
  14. FOOTNOTE: Guidelines for attribution of tropical cyclone properties to climate change: 1. Globally averaged intensity of tropical cyclones will rise as AGW increases SST. Models predict globally averaged intensity increase of 2% to 11% by 2100. 2. Models predict falling globally averaged frequency of tropical cyclones with frequency decreasing 6%-34% by 2100. 3. The globally averaged frequency of “most intense tropical cyclones” should increase as a result of AGW. The intensity of tropical cyclones is measured as the ACE (Accumulated Cyclone Energy). 4. Models predict increase in precipitation within a 100 km radius of the storm center. A precipitation rise of 20% is projected for the year 2100.
  15. Complications of empirical tests in this line of research (Knutson, 2010) that that introduce excessive uncertainty in research results: 1. Extremely high variance in tropical cyclone data at an annual time scale suggests longer, perhaps a decadal time scale which in turn greatly reduces statistical power. 2. Limited data availability and poor data quality present barriers to research. 3. Limited theoretical understanding of natural variability makes it difficult to ascertain whether the variability observed in the data is in excess of natural variability. 4. Model projections for individual cyclone basins show large differences and conflicting results. Thus, no testable implication can be derived for studies of individual basins. It is necessary that empirical studies have a global geographical span. 5. Advances in data collection activity, methods, and technology create trends in the data that must be separated from climate change effects (Landsea, 2007) (Landsea, 2010).



6 Responses to "Climate Change and Hurricanes"

[…] The Regression Summary table shows much stronger regression results at the longer time scales of 11 years and 22 years as suggested by John Casey. We use α=0.001 as the critical p-value as suggested in the Valen Johnson study as a way of reducing the rate of irreproducible results in research {Johnson, Valen E. “Revised standards for statistical evidence.” Proceedings of the National Academy of Sciences 110.48 (2013): 19313-19317.}. In computing p-values for 11-year and 22-year moving averages, the degrees of freedom are adjusted for a loss of degrees of freedom as explained in paragraph#10 of this post [LINK] . […]

[…] expect to see only at the undergraduate level. The Emanuel paper is discussed in a related post [LINK] .  The failure of the case by climate scientists that climate change is making tropical cyclones […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: