In September 2005, a paper published in Science magazine reported that worldwide, the number of Category 4 and 5 hurricanes had increased 80% in the past 30 years. The paper, (Webster et al., 2005), titled "Changes in Tropical Cyclone Number, Duration, and Intensity in a Warming Environment", linked the rise in storms to increasing sea surface temperatures and concluded that "global data indicate a 30-year trend toward more frequent and intense hurricanes." The authors, led by Dr. Peter Webster of Georgia Tech and Dr. Greg Holland of the National Center for Atmospheric Research, argued that this was consistent with climate models that have predicted a future increase in frequency of the most intense hurricanes due to human-emitted greenhouse gases. This paper, along with another paper published in August, "Increasing Destructiveness of Tropical Cyclones over the past 30 years", by Dr. Kerry Emanuel of MIT, showing an increase in hurricane power and longevity in recent years, created a huge stir in the media. However, there is a large amount of uncertainty in the hurricane intensity data used by both papers, and their findings should be considered as preliminary evidence that the global incidence of Category 4 and 5 hurricanes may be increasing. There are good reasons to believe that the actual increase in Category 4 and 5 hurricanes is far lower than the 80% increase found by Webster et al.
The papers by Webster et al. and Emanuel have created considerable controversy in the hurricane science community. Many hurricane scientists disagree with the new results, and have disputed them in new papers submitted for publication. I will examine the arguments of three of these scientists here. Keep in mind that the Webster et al. paper went through peer review--it was revised based on the recommendations of at least two anonymous reviewers who read the paper before publication. The arguments of the other scientists disputing the paper have not been subject to peer review, and may have more errors or omissions than peer-reviewed work would have. The three scientists are:
Dr. Bill Gray of Colorado State University, who is famous for his successful long-range hurricane predictions and nearly 50 years of hurricane research and forecasting, submitted his critique to Science for publication, but the journal rejected it, since the document had already been published (on Dr. Gray's web site). Journals typically do not publish material that has been published elsewhere. In his abstract, Dr. Gray says: "I do not agree that global Category 4-5 tropical cyclone activity has been rising, except in the Atlantic over the last 11 years. The recent Atlantic upsurge has explanations other than global temperature rise".
Dr. Chris Landsea, Science and Operations Officer at the National Hurricane Center, chaired a standing-room only session exploring the hurricanes/global warming connection at the 2006 meeting of the American Meteorological Society (AMS). He presented a talk with additional evidence supporting Dr. Gray's position.
Dr. John Knaff, a hurricane researcher at the NOAA/Cooperative Institute for Research in the Atmosphere Colorado State University, has performed an extensive re-analysis of Northwest Pacific typhoons, and questions the intensity estimates used by Webster et al. for typhoons during the period 1966-1987. His paper, "Reexamination of Tropical Cyclone Wind-Pressure Relationships" has been accepted for publication to Weather and Forecasting, and will probably be published in late 2006.
Lets examine the credentials of the Science paper's authors. The primary author, Dr. Peter Webster of Georgia Tech, holds a Ph.D. from MIT and has received the most prestigious award issued by the American Meteorology Society--the Carl Gustav Rossby Research award. Webster's primary expertise is not hurricanes--he has mostly studied monsoons. However, the second author, Dr. Greg Holland of the National Center for Atmospheric Research, is a hurricane expert. He earned his Ph.D. in 1983 at Colorado State as a student of Dr. Bill Gray, and has authored over 100 hurricane-related journals articles or book chapters. One of the other co-authors, Dr. Judith Curry, is the Chair of the Georgia Tech School of Earth and Atmospheric Sciences. So, the paper's authors have a track record of producing high-quality research that should be taken seriously.
Hurricanes act as giant heat engines, so it is logical to assume that an increase in sea surface temperatures (SSTs) will make more intense hurricanes. Indeed, there is a general consensus among hurricane scientists that an increase in SSTs due to global warming, should, in theory, lead to more intense hurricanes. Theory predicts that hurricane wind speeds should increase about 5% for every 1 degree Centigrade increase in tropical ocean temperature (Emanuel, 1987). Computer models confirm this tendency, but assign a slightly smaller magnitude to the increase (Knutson and Tuleya, 2004). Given the expected 1.5° to 4.5° C warming of Earth's climate expected by 2100, theory predicts a gradually increasing frequency of Category 4 and 5 storms.
Global warming in the past century has increased ocean temperatures about 1°F (0.5°C) which should correspond at most to about a 2.5% increase in hurricane wind speeds. If this theory is correct, an upper-end Category 3 hurricane with wind speeds of 130 mph--like Hurricane Katrina at landfall--owes 2-3 mph of its sustained winds to global warming. Hurricane wind speeds are estimated to the nearest 5 knots (5.8 mph), and one can get a general idea of what percent increase we've seen in Category 4 and 5 hurricanes due to global warming by looking at the number of high end Category 3 hurricanes (winds of 130 mph) and low end Category 4 hurricanes (135 mph winds). If we assume a 2-3 mph increase in winds of these storms is due to global warming over the past 35 years, one would expect to see a 5% increase at most in Category 4 and 5 hurricanes. An increase this small is not detectable given the current accuracy of estimating hurricane winds, and the relatively few number of of these storms that occur each year. This expected maximum 5% increase is quite a disagreement with the 80% increase found by Webster et al.! So, either the measurements are wrong, or the theory is wrong--or a combination of the two. I believe it may well be a combination of the two. The fact that the originator of the intensity theory (Kerry Emanuel) is one of the scientists who is advocating that the theory may be in error, is reason enough to doubt the theory. The formation and intensification of hurricanes are not well understood, and it would be no surprise if major revisions to intensity theory are made in the future. However, such a wide difference between the theory and the reported trends should make us suspicious of the observed data, as well.
Webster et al. show a plot (Figure 2) of the sea surface temperature (SST) in the six major ocean basins that support tropical cyclones. Since 1970, SSTs in all the oceans have risen by up to .5° C. The paper chooses to look only at the period from 1970 to the present, since 1970 is the approximate time when global satellite measurements of tropical cyclone intensity became available. Before 1970, there are reliable intensity measurements only in the Atlantic and Northwest Pacific, thanks to the Hurricane Hunters. These measurements began in 1944 in the Atlantic and 1945 in the Northwest Pacific (but stopped in 1987 in the Pacific).
Webster et al. also present plots of the global frequency of tropical storms, hurricanes, and the number of days those storms are present (Figure 3). No trend is apparent in these plots, and the paper states that "against a backdrop of increasing SST, no global trend has yet emerged in the number of tropical storms and hurricanes." So far, all hurricane scientists are in agreement.
Webster et al. present a plot (Figure 4) where of the number of Category 1, Category 2 and 3, and Category 4 and 5 storms, averaged into 5-year "pentads". The results show little change in the statistics of Category 1, 2, and 3 hurricanes, but a startling increase in Category 4 and 5 hurricanes. These most intense and dangerous storms on Earth have increased from 50 per five-year period in the 1970s, to 90 in the past decade--a near doubling!
Here's where the critics of Webster et al. differ. Let's look at the criticisms one ocean basin at a time. First: the Southern Hemisphere oceans, and the part of the Indian Ocean in the Northern Hemisphere. These regions are responsible for 25% of the world's Category 4 and 5 hurricanes. Webster et al. found a huge increase in these storms, from 34 during the 1975-1989 period, to 79 during succeeding 15 year period. Dr. Gray criticizes the quality of the data in the Indian Ocean and Southern Hemisphere ocean basins during the period 1975-1989, remarking: "In the late 1970s I visited all the global tropical cyclone centers and observed their satellite capabilities and the training of their forecasters as part of a World Meteorological Organization (WMO) tropical cyclone trip that I was commissioned to make. The satellite tools and forecaster training in the tropical cyclone regions of the Indian Ocean and Southern Hemisphere during the 1975-1989 period was not adequate for the task of objectively distinguishing Category 4-5 hurricanes from Category 3 hurricanes or to always be able to confidently distinguish Category 4-5 hurricanes from Category 1-2 hurricanes." Dr. Gray does not provide any details about how how this lack of training could have led to a systematic error in classifying too few storms at Category 4 and 5 intensity. However, Dr. Landsea did cite an example of this in his talk at the 2006 American Meteorological Society meeting--for a number of years after 1974's Tropical Cyclone Tracy devastated Darwin, the Australians never assigned an intensity higher than Tracy to any storm because they believed Tracy was the strongest a storm could be in the Australian region. Tracy was a weak Category 4 cyclone with 135 mph winds and a 954 mb pressure. This was a false assumption, as many cyclones stronger than Tracy have formed in Australian waters.
This brings up the most serious weakness in the Webster et al. paper--they do very little discussion of the uncertainty in hurricane intensity measurements. Hurricane intensity is characterized by a 1-minute measurement of maximum sustained winds at ten meters above the surface. In practice, this quantity is virtually never measured, but must be inferred indirectly from other measurements. All of these inferred measurements of wind speeds have errors. Satellite estimates of hurricane intensity are often wrong by a full category on the Saffir-Simpson scale--or even two categories. This is particularly true when there are only two geostationary satellites covering the Earth, as was the case for much of the 1970s, and part of the 1980s. I learned this the hard way during my stint with the Hurricane Hunters when I flew into Hurricane Hugo in 1989 expecting a Category 3 hurricane based on satellite estimates. The problem was that one of the GOES satellites had failed earlier that year, leaving just one satellite to cover all of the U.S. and Atlantic Ocean. This one satellite was positioned much farther west in order to see all the way to California, and thus had a poor, oblique view of hurricanes out over the Atlantic. If a satellite can't see all the way to the bottom of the eye of a hurricane because of an oblique viewing angle, it will come up with an eye temperature that is too cool, and thus an intensity estimate that is too low. Hugo turned out to be a Category 5, and made us pay dearly for our mistake. Similarly, a systematic underestimation of hurricane intensities in the 1970s is quite possible, due to the oblique viewing angle that the relatively few geostationary satellites afforded during that period.
Furthermore, Dr. Landsea argued, the technique used to perform satellite estimation in all ocean basins of hurricane intensity (the Dvorak technique) did not even get invented until 1972--two years after the start of the data used by Webster et al. It wasn't until 1984 that the Dvorak technique was extended to infrared satellite imagery. So, between 1972 and 1984, all satellite hurricane intensity estimates were done on visible satellite imagery, and were unavailable at night. In addition, measurement errors from the first generation of infrared satellite instruments was much higher, since their resolution was a relatively poor 9 km--compared to the 4 km resolution of today's instruments. Dr. Landsea called for a major re-analysis effort of the old satellite photos from the 1970s and 1980s to try to quantify some of these errors. He argued that his preliminary look at old satellite images from the Northern Indian Ocean from 1970 to 1989 had already revealed five additional storms that may have been Category 4 or 5. Webster et al. found only one Category 4 or 5 hurricane for this time period there. Dr. Landsea suggested that until a full re-analysis took place in all ocean basins, the quality of the historical global hurricane intensity data was not high enough to be able to see a possible increase in the number of Category 4 and 5 hurricanes.
Dr. Holland spoke after Dr. Landsea, and agreed that a major re-analysis effort was needed in order to help quantify some of the errors in hurricane intensities. He rebutted Dr. Landsea's arguments by pointing out that given the very large rise in Category 4 and 5 hurricanes in the Southern Hemisphere and Indian Oceans, about 50% of all Category 1 and 2 hurricanes would have to be re-analyzed as Category 4 and 5 storms to invalidate their results in that region. While acknowledging the the data was bad in the 1970s, he maintained that nobody thought it was that bad.
If we restrict ourselves to just looking at the Southern Hemisphere and Indian Ocean data since 1989--the date when everyone agrees that the data quality is reasonable--there is not enough data to be sure of any kind of trend (Figure 5). So until better intensity estimates of hurricanes in the Indian Ocean and Southern Hemisphere ocean basins from 1970-1989 are available, one should view the results of Webster et al. with caution for these regions. Still, there is such a large increase in Category 4 and 5 hurricanes that it is unlikely that re-analysis will find that all of this increase in unreal.
The question of the integrity of the typhoon intensity data in the Northwest Pacific is critical, since this ocean basin accounts for fully 46% of the global Category 4 and 5 hurricanes since 1970. Dr. Gray and Dr. Knaff both question typhoon intensities measured by reconnaissance aircraft in the Northwest Pacific during the 1973-1986 period. The technique used to determine typhoon intensities during this period (the "Atkinson-Holliday" or AH technique), is thought to have significantly underestimated the maximum winds. Looking at a plot of all Category 4 and 5 activity since 1945 in the Northwest Pacific (Figure 6), one can that intense typhoons were about as common in the 1950s and 1960s as they were during 1990-2004, but took a major dip in the 1970s and 1980s during the period the AH technique was used. I asked Dr. Webster and Dr. Holland about the intense typhoon activity back in the 1950s and 1960s, and they argued that this activity was the result of high SSTs in the Northwest Pacific during that period. On his website, Kerry Emanuel argues that typhoon intensities were overestimated in the 1950s and 1960s. However, Knaff and Zehr (2006) make some convincing arguments that typhoon intensities during the 1973-1986 period were too low due to measurement error, and the number of Category 4 and 5 storms in the region have been roughly constant for the past 50 years. This paper has been accepted for publication in Weather and Forecasting, and will likely be published late this year. Dr. Knaff and Charles Sampson of the Naval Research Laboratory have performed a preliminary re-analysis of maximum typhoon intensities for the period 1966-1987 based on the Knaff and Zehr (2006) results, and this re-analysis will be presented at the upcoming 27th Conference on Hurricanes and Tropical Meteorology (April 24-28, 2006). In the extended abstract, they show that after correcting for the AH technique errors, the number of Category 4 and 5 typhoons during the 1966-1987 period increased by 1.5 per year, leaving only a slight upward trend in Category 4 and 5 typhoons during the period 1970 - 2004. The 16% increase in Category 4 and 5 typhoons found by Webster et al. during the past 15-year period is reduced to just 3%. Based on this new research, the results of Webster et al. may have to be modified. In particular, their global increase in storms from 1990-2004 compared to 1975-1989, as presented in that paper will be reduced from 57% to 42% if Dr. Knaff's typhoon re-analysis is accepted.
Dr. Gray formulates the reasonable hypothesis that if one compares global major hurricane activity for the most recent ten years (1995-2004) with the previous ten years (1985-1994), one should see a significant difference, since global surface temperatures increased about 0.4° C between the two periods. He shows that the number of Category 3-4-5 hurricanes stayed exactly the same between these two periods--218 for each time period--if one excludes the Atlantic. I tabulated the results for just Category 4 and 5 hurricanes, and the results were very similar--135 storms storms globally (excluding the Atlantic) from 1985-1994, and 142 for 1995-2004. As most of you are aware, the Atlantic has seen a big increase in the number of intense hurricane the past ten years. Dr. Gray attributes this to the Atlantic Multidecadal Oscillation (AMO), a natural cycle I discussed in an earlier blog. Dr. Gray offers another comparison, but just for Category 4 and 5 storms. The most reliable comparison one can make is using data from the Northeast and Northwest Pacific from the past 20 years. This excludes the issues of dealing with the natural AMO cycles in the Atlantic, and the poor data quality in the other ocean basins. Again, the data show essentially no difference between time periods. Indeed, when looking at the plot of Category 4 and 5 hurricane for the Northeast Pacific--the ocean area off the west coast of Mexico (Figure 7), and responsible for 19% of the world's Category 4 and 5 hurricanes--one sees no increasing trend in recent years.
The Atlantic contributes only 9% of the world's Category 4 and 5 hurricanes, so is not much of factor when considering global numbers of these storms. Dr. Gray shows that the number of Category 4 and 5 hurricanes in the Atlantic has remained constant when one compares numbers from the past 15 years with an earlier active period from 1950-1964. However, this is a poor comparison. The period 1950-1964 fell entirely within a time when the warm phase of the AMO dominated the Atlantic, and had significantly enhanced intense hurricane activity (see Figure 8). The period 1990-2004 includes five years from the cold phase of the AMO, when intense hurricane activity was significantly down. Thus, comparison of 1950-1964 with 1990-2004 in the Atlantic is poor. One should make the comparison between data from the 11 years from the most recent warm phase of the AMO (1995-2005), and the previous warm AMO period we have good data for (1944-1969). This comparison shows that Category 4 and 5 hurricanes in the Atlantic have increased by 60% in the past 11 years compared to the previous active period 1944-1969. One can make a similar comparison for the cold phase of the AMO, contrasting the years 1970-1982 with 1983-1994. This comparison show no increase in Category 4 and 5 hurricanes for the later period with warmer SSts. I asked Dr. Landsea about the 60% increase in Category 4 and 5 hurricanes during the most recent warm phase of the AMO, and he thought that at least part of the increase could be explained by inadequate information from the Hurricane Hunters during that period. He explained that during that time, it was common in intense hurricanes for the Hurricane Hunters to get close enough to the eye to fix the storm on radar, but not actually penetrate through the eyewall into the eye. Who can blame them! The older aircraft like the DC-6 used during that time period were not safe to fly into Category 4 and 5 hurricanes. Dr. Landsea is working on a re-analysis project for the entire Atlantic hurricane database, but has only made it to the 1930s, and hopes to have a more definitive answer on the intensities of hurricanes during the 1950-1969 period in a few years.
So who's right? Given the uncertainties in estimating tropical cyclone intensity presented by Drs. Gray, Landsea, and Knaff, plus the very large disagreement with the theory of hurricane intensification, it is unlikely that the large 80% increase in Category 4 and 5 hurricanes found by Webster et al. is real. There does appear to be some increase, but it is likely much smaller. Many troubling questions need to be answered, such as why comparison of the most recent ten years (1995-2004) with the previous ten years (1985-1994) shows almost no increase in Category 4 and 5 storms globally, during a period when a substantial increase in SST occurred.
All the scientists involved in this debate have stated the need for a rigorous re-analysis of all historical tropical cyclone data. However, there is currently little funding for such work. Dr. Knaff told me that his typhoon re-analysis work was unfunded, and that he did it because he felt strongly that the results of Emanuel (2005) and Webster et al. were inaccurate and needed to be challenged. Dr. Landsea's reanalysis of Atlantic storms is funded, but something he can only devote time to when his duties at NHC allow him. Dr. Knaff wrote me, "While I realize there are plans to reanalyze the Atlantic, the West Pacific, Southwest Pacific, and Indian Ocean are all being done piece by piece as part of several unfunded projects with little general support. If people are going to use the data for global studies, then NOAA, NSF or some other entity should fund a global reanalysis." I agree completely! Before I am willing to conclude that Category 4 and 5 hurricanes are indeed showing a significant increase, I want to see the science done with a better dataset, and covering a longer period of time. The NOAA Office of Global Programs or National Science Foundation needs to step in and fund this research.
While Category 4 and 5 hurricanes may indeed be increasing in frequency globally, one cannot yet say that global warming is the cause. Webster et al. close with the sentence, "attribution of the 30-year trends to global warming would require a longer global data record and, especially, a deeper understanding of the role of hurricanes in the general circulation of the atmosphere and ocean, even in the present climate state." Furthermore, global warming cannot be cited as the cause of recent intense storms, such a Hurricane Katrina, Hurricane Wilma, or Australia's Cyclone Larry and Cyclone Glenda.
Author: Dr. Jeff Masters
Emanuel, K.A., "The dependence of hurricane intensity on climate" (Abode PDF), Nature, 326, 483-485, 1987.
Emanuel, K.A., "Increasing Destructiveness of Tropical Cyclones over the past 30 years (Adobe PDF), Nature, 436, 686-688, 4 August 2005.
Hoyos, C.D., P.A. Agudelo, P.J. Webster, and J.A. Curry, "Deconvolution of the Factors Contributing to the Increase in Global Hurricane Intensity", www.scienceexpress.org, 16 March 2006, 10.1126/science.1123560.
Knaff, J.A., and R.M. Zehr, "Reexamination of Tropical Cyclone Wind-Pressure Relationships", accepted to Weather and Forecasting, 2006.
Knutson, T.R., and R.E. Tuleya, "Impact of CO2-Induced Warming on Hurricane Intensity and Precipitation: Sensitivity to the Choice of Climate Model and Convective Parameterization," (Adobe PDF) Journal of Climate 17, 18: 3477-3495, 2004.
Virmani, J.I., and R. H. Weisberg, "The 2005 hurricane season: An echo of the past or a harbinger of the future?", Geophysical Research Letters 33, L05707, 2006 doi:10.1029/2005GL025517.
Webster, P.J., G.J. Holland, J.A. Curry, and H.-R. Chang, "Changes in Tropical Cyclone Number, Duration, and Intensity in a Warming Environment", Science, 309, 1844,1846, 16 September 2005.