Increased tropical cyclone activity: natural cycles or global warming

It is widely accepted within the hurricane science community that hurricane activity in the North Atlantic follows a multi-decade cyclicity that saw, for example, a higher level of activity between the 1930s and 1960s than in the 1970s to early 1990s. This cyclicity is traditionally attributed to the Atlantic Multidecadal Oscillation (AMO), a series of long-term changes in the sea-surface temperature (SST) of the North Atlantic, characterised by an alternation of relatively cool and warm phases that each last for between 20 and 40 years. The period of enhanced hurricane activity between the 1930s and 1960s coincided with a warm phase AMO, while the quieter interlude that followed in the 1970s and 1980s corresponded to cool phase conditions. The dramatically increased Atlantic hurricane activity since around 1995, culminating in a very active 2004 and a record-breaking season last year, coincides with another AMO warm phase. Recently, however, the hurricane science community has been split by suggestions that the AMO may not be the primary driver of the exceptional activity of recent years.

In June 2005, just prior to the record breaking Atlantic hurricane season, Kevin Trenberth⁶⁹, of the US National Center for Atmospheric Research, fired the opening shot in a war of words between hurricane scientists that is becoming increasingly intense and, at times, acrimonious. The war is being fought over two issues; firstly, whether or not there has been an increase in the number of powerful tropical cyclones over the last few decades, and secondly, if this is the case, what the cause is most likely to be. In his paper, published in the journal, Science, Trenberth proposes that ‘human influences’ are already evident in the environments in which hurricanes form. These, he suggests, are reflected, for example, in an upward trend in sea-surface temperatures in the tropical Atlantic that has become more marked over the last 35 years (figure 2). Trenberth points out that higher SSTs will favour enhanced convection and therefore more thunderstorms, but not necessarily their organisation into the tropical storms that are hurricane precursors. Once a tropical storm forms, however, the warmer seas and other environmental conditions associated with global warming will provide more energy to fuel the storm, thereby increasing its potential intensity and amount of associated precipitation, both hurricane properties that already appear to be occurring. Looking ahead, Trenberth admits that there is no guarantee that we will see more hurricanes in the future, but we are likely to see a shift towards more extreme, and therefore more damaging, storms.




Figure 2. Hurricane Katrina's power and destructive potential was at least partly the result of very high sea-surface temperatures (SST) in the Caribbean. For the period from August 23 - 30 2005, the yellow, orange and red colours show those areas of the Caribbean and adjacent Atlantic Ocean that were above the 82° F temperature required for a hurricane to strengthen. In broader terms, the current debate centres on whether or not high SSTs constitute the main driver of recent high levels of tropical cyclone activity, both in the Atlantic and worldwide, and whether the warmer oceans are a reflection of global warming.

Courtesy: NASA

Trenberth’s argument for a human-induced, global warming signal in recent hurricane activity appears to be supported by two further papers published in August and September last year. In the first, Kerry Emanuel ²⁰ of MIT, defines - in the journal Nature - an index of potential hurricane destructiveness (the Power Dissipation Index or PDI, which measures the power released by tropical cyclones), the application of which suggests that tropical cyclones in the North Atlantic and western North Pacific have more than twice the destructive potential nowthan they had during the 1970s (figure 3). This he relates to rising sea-surface temperatures since 1975 that are generally explained by global warming, suggesting that the increase in potential destructiveness is ‘at least partially anthropogenic’. The rise in destructiveness, suggest Emanuel, may be a reflection of more intense storms, or of storms that maintain themselves at a higher intensity for longer. Emanuel feels that both are important, and backs this up by pointing to a roughly 60 percent rise in accumulated annual duration of Atlantic and North Pacific storms since 1949, and - during the same period - to a 50 percent increase in annual average storm peak wind speed over the same regions. Emanuel concludes by suggesting that future global warming may lead to an upward trend in tropical destructive potential and, given increasing population and wealth concentration in coastal zones, to a ‘substantial’ rise in hurricane-related losses in the 21st century.



Figure 3. (Left) A measure of the total power dissipated annually by tropical cyclones in the North Atlantic (the Power Dissipation Index, PDI), compared to September sea-surface temperature. (Right) Annually accumulated PDI for the North Atlantic and western North Pacific, compared to annually-averaged sea-surface temperature.

Courtesy: Nature

In the second paper, published in Science a month after Emanuel’s work, Peter Webster ⁷⁴ of the Georgia Institute of Technology, and co-workers, report that against a background of rising sea-surface temperatures, no global trend has yet been detected in the number of tropical storms or tropical cyclones. They do, however, recognise a large increase in both the number and proportion of category 4 and 5 tropical cyclones over the last 35 years (figure 4 and table 1).


Figure 4. (Left) The total number of tropical cyclones (black), category 1 storms (blue), the sum of categories 2 and 3 (green), and the sum of categories 4 and 5 (red) are shown for 5-year periods. The horizontal dashed lines show the 1970-2004 average numbers in each category. Note that globally, the total number of storms is lower than since the late 1970s, but the proportion of more powerful storms is significantly higher. (Right) Numbers of tropical cyclones in the same categories, shown as a percentage of the total number. Dashed lines show average percentages in each category over the 1970-2004 period.

Courtesy: Peter Webster.

This, state the authors, is most pronounced in the North Pacific, Indian and South West Pacific Oceans, and weaker in the North Atlantic. Globally, however,
Webster and colleagues propose that the number of powerful hurricanes have almost doubled since the 1970s, and now make up around 35 percent of all hurricanes, rather than about 20 percent 35 years ago. While not categorically implicating climate change as the driver for this change, Webster and his co-authors note that the trend is ‘not inconsistent with recent climate model simulations that a doubling of CO2 levels in the atmosphere may increase the
frequency of the most intense cyclones’.


Table 1. Change in the number and percentage of hurricanes in categories 4 and 5 for the 15-year periods 1975-1989 and 1990-2004 for the different ocean basins

(Webster and others, 2005).

While the above statement is true, a trend towards stronger tropical cyclones was not expected to be seen after just a few decades of warming, which is one reason why the findings of Trenberth, Emanuel and Webster and his colleagues, aroused scepticism in some circles. In a riposte to Emanuel’s paper, published in Nature (see also a paper published by Pielke ⁵³ and others in November 2005 in the Bulletin of the American Meteorological Society), Roger Pielke ⁵² of the University of Colorado points out that if tropical cyclones are becoming more destructive over time, then this should be manifested in more destruction. For the US, Pielke, indicates that his findings show no upward trend in destruction once the data are normalised to remove the effects of societal changes - e.g increases in population and wealth. Pielke concludes that either Emanuel’s Power Dissipation Index (PDI) is actually a poor indicator of hurricane destructiveness, or the trend identified by Emanuel is an artefact of the available data. This line of argument is also followed up in the same issue of Nature by Chris Landsea ³⁸ of NOAA’s Hurricane Research Division in Miami, who questions Emanuel’s analysis of the data, even to the extent of suggesting that one third of Emanuel’s rise in PDI for the Atlantic Basin is incorrect. Citing further problems he has with Emanuel’s data analysis, Landsea questions the bias-removal method used to standardise the data, which involves reducing tropical cyclone wind speeds by 2.5 to 5.0 metres per second for the 1940s - 1960s, which he feels may not be warranted as applied. Without the bias removal, Landsea notes that the Atlantic hurricane activity of the last 10 years is not unprecedented, and was in fact equalled or even surpassed in the mid-20th century. Landsea also concludes that the most recent decade has a PDI for the United States that is close to average, rather than displaying an increase in the overall intensity and intensity of hurricane strikes.

Responding to Pielke and Landsea, again in the pages of Nature, Kerry Emanuel ²¹ stands by his conclusions about his trend in tropical cyclone power dissipation. He reiterates that the trend is large and universal, has about the same value in all ocean basins, and is well correlated with sea-surface temperature. Specifically in response to Pielke’s and Landsea’s criticisms that no increase in hurricane-related destruction is observed in the US in recent decades, Emanuel points out that only a fraction of hurricanes ever affect the US coastline. He stresses that because his PDI is accumulated over all storms throughout their entire lives, it contains about 100 times more data than an index related to hurricanes at landfall, and presents this as an explanation of why the real trend is detectable in the power dissipation but not in landfalling statistics. Emanuel takes on board Landsea’s criticisms on bias removal and other aspects of data analysis and applies Landsea’s suggestions to his data. He stresses, however, that these do not affect the strong correlation between hurricane activity and tropical SST, and reiterates the fact that the SST record is long enough to show the influence of global warming. Emanuel questions Landsea’s suggestion that the recent multi-decadal variability in tropical SST and Atlantic hurricane activity is due to a natural cycle (the AMO). Rather, he
maintains that current levels of storminess are unprecedented in the historical record and that a global warming signal is now emerging in tropical cyclone activity. In particular, he points out that this is most evident when global tropical cyclone activity is addressed, rather than just the 12 percent or so of storms that occur in the Atlantic Basin.

The trends in storm strength published by Webster and colleagues for the western North West Pacific (NWP), are also challenged; this time in the journal Science by Johnny Chan ¹³ of the University of Hong Kong, who presents data suggesting that there is no trend in Western NWP typhoon intensity and that apparent trend observed by Webster and co-workers is actually part of a large, natural variability known as the Pacific Decadal Oscillation (PDO). Chan’s data also suggests that there is a negative correlation between typhoon intensity in the region and local tropical SST temperature, but a positive correlation between typhoon intensity and other parameters including wind shear and so-called atmospheric vorticity. Replying in the same journal, Peter Webster ⁷⁵ and his fellow researchers agree that there are strong signals of natural variability in both hurricane statistics and SST in the region. They point out, however, that this does not refute the central tenet of their study, namely, that there is a change in hurricane intensity with increasing SST.

A pause in the flurry of papers and written replies on the causes of recent tropical cyclone activity was ended in February by the issuing of a statement by the steering committee on the effects of climate change on tropical cyclones ¹⁴ of the World Meteorological Organisation & Commission for Atmospheric Science. The statement issued by the steering group, which includes the aforementioned Chan, Emanuel and Landsea, is intended to provide an updated assessment of the current state of knowledge, with respect to the influence of anthropogenically-induced climate change on tropical cyclones. Inevitably, given the widely disparate views of members of the group, the statement adds nothing new to the debate, and does little more than highlight the reasons that underpin those disagreements that have been discussed here in preceding paragraphs.

The run-up to the 2006 Atlantic hurricane season has seen publication of another clutch of relevant papers. In April, Carlos Hoyos ³⁰ and colleagues at the Georgia Institute of Technology, reported in Science the results of another study supporting both the idea of an increasing worldwide trend in the numbers of category 4 and 5 tropical cyclones (for the period 1970 to 2004), and its direct linkage to a synchronous increase in SST. They further point out that while other aspects of the tropical environment, such as wind shear, do influence shorter-term variations in tropical cyclone activity, they do not contribute substantially to the observed global trend.

In May, Patrick Michaels ⁴¹ of the University of Virginia, and colleagues, presented in Geophysical Research Letters, their examination of the relationships between SST and tropical cyclones in the Atlantic Basin. Michaels and co-workers apply higher resolution data than in previous studies to show that peak wind speeds do not occur over locations of highest SST, and note that a better indicator of the final intensity reached by a tropical cyclone is the highest SST encountered prior to the time of peak wind. They also recognise an apparent threshold SST (28.25ºC) that must be exceeded for a tropical system to achieve wind speeds in excess of 50 metres a second (i.e. as characterises a category 3 storm). Once this threshold is exceeded, the authors find no relationship between storm intensity and SST, suggesting that increasing SST will act to increase the percentage of major hurricanes, but not change their final intensities. Michaels and colleagues conclude that the effect of higher SST on tropical cyclone intensity is far less than could possibly explain the enhanced Atlantic Basin activity since the mid-1990s.

Michaels and his team also take a speculative look at what could happen in a warmer world, within which all Atlantic tropical cyclones might encounter temperatures that exceed the 28.25ºC threshold, which would require a general warming of 2-3ºC in the tropical cyclone source region of the Atlantic. Between 1982 and 2005, 195 out of 270 storms encountered threshold temperatures, with 58 (29.7 percent) achieving category 3 status or above. Using the
period as an analogue for a warmer world scenario, the authors assume that all 270 storms encounter the threshold temperature, and that the same percentages apply. This would result in 80 (29.7 percent of 270) storms becoming major tropical cyclones, and would also see a 6 percent rise in peak wind speed averaged across all 270 storms.

In another Geophysical Research Letters paper, published the same month, Philip Klotzbach ³⁷ of Colorado State University, examines trends in global cyclone activity over the past 20 years, and comes to the conclusion that both Emanuel and Webster are wrong. Klotzbach’s findings lead him to suggest that the new data provide no support to the idea that global tropical cyclone frequency, intensity and longevity have increased over the past two decades. The author does admit to a small apparent rise in the number of category 4 and 5 storms, but dismisses this as being the likely result of improvements in observational technology. Klotzbach concludes that the role of SST is a relatively minor one, and reiterates that vertical wind shear is a much more fundamental component for the development and maintenance of major hurricanes.

Just a week after the beginning of the 2006 Atlantic hurricane season, Ryan Sriver and Matthew Huber ⁶⁵ of Purdue University, revisited - in Geophysical Research Letters - Emanuel’s (2005) work on global tropical cyclone power dissipation; this time using surface wind and temperature records from the European Centre for Medium-Range Weather Forecasts 40-year Reanalysis (ERA-40) Project. The study is complementary to that of Emanuel in that it utilises a different data source and adopts a more objective and rigorous analytical approach that makes fewer assumptions. The study also concentrates only on low frequency power dissipation variability, so that effects of shorter-term climatic signals, such as ENSO (El Niño - Southern Oscillation) and QBO (Quasi-Biennial Zonal Wind Oscillation) are eliminated. Despite the different approach, the results show good agreement with Emanuel’s after 1978, which corresponds approximately to the use of satellite data in ERA-40. Sriver and Huber conclude that the ERA-40 estimates of tropical cyclone trends are in agreement with those of Emanuel (2005), Webster et al. (2005) and Trenberth (2005). They find that the global mean power dissipation exhibits a 25 percent rise between 1958-78 and 1979-2001, suggesting that the recent observed increase in tropical cyclone activity may be related to observed trends in tropical SST.

A week after Sriver & Huber’s contribution to the debate, a paper in EOS by Michael Mann of Pennsylvania State University and Kerry Emanuel ⁴⁰ is set to liven up the debate even further. Mann and Emanuel statistically revisit the concept of the Atlantic Multidecadal Oscillation in controlling sea-surface temperature, and the role it plays in hurricane activity. They conclude that while there is a strong historical relationship between tropical Atlantic SST and tropical cyclone activity that extends back at least to the late 19th century, the AMO apparently plays no role. Instead, they suggest that the underlying factors appear to be the influence of ‘primarily anthropogenic forced, large-scale warming and an off-setting regional cooling overprint due to late 20th century anthropogenic aerosol forcing’. This controversial conclusion is certain to spark renewed argument and discussion that is certain to be translated into print. Such papers will be addressed in HRSR2007. Meanwhile, it is perhaps appropriate that the final paper covered in this review, should be co-written by Kevin Trenberth (along with Dennis Shea ⁷⁰ of the National Center for Atmospheric Research in Boulder, Colorado), who started the whole debate 12 months ago. In another Geophysical Research Letters paper, published at the end of June, Trenberth and Shea also address the issue of the AMO, and its role in Atlantic hurricane activity. They note that during the record-breaking 2005 season, SST in the tropical North Atlantic were also at record levels (0.9ºC above the 1901 - 70 norm), and suggest that this was a major reason for the record hurricane season. In agreement with Mann and Emanuel, the authors propose that the AMO made a minimal contribution to the record tropical North Atlantic SST anomaly on the order of 0 – 0.1 ° C. In contrast, the global SST anomaly of the last five years was about 0.45 ° C. i.e. much larger than the AMO contribution The story continues…….in HRSR2007

Is it the end for the Saffir-Simpson Hurricane Scale?

Given the recent high levels of tropical cyclone activity, particularly in the Atlantic Basin, and the continuation of the lively debate on the impact climate change may be having on hurricane frequency and intensity, Lakshmi Kantha ³⁵, of the University of Colorado, suggests in EOS, that now might be an appropriate time to revisit and revise the system used for categorising hurricanes. The author notes that the Saffir-Simpson Hurricane Scale (SSHS) (table 2) has now been used for three decades in the US for hurricane emergency response decisions. He points out, however, that it has many drawbacks and can be confusing, and proposes its replacement with a scale that provides more consistent estimates of hurricane intensities and hazards.


Table 2. Saffir-Simpson Hurricane Scale

Kantha notes that, unlike the Richter and other hazard scales, the SSHS is quantized; in other words, each category is characterised by particular values of pressure, sustained wind speeds etc. This means that a change of just 1 km h or one mb in central pressure can lead to a hurricane jumping an entire category. For example, a drop in sustained wind speed from 210 to 209 km h, would lead to a storm being downgraded from a category 4 to 3, although its intensity and destructive power will be immeasurably less. Kantha also points out that the SSHS ‘saturates’ at the top end. In other words, no matter how much wind speeds exceed 250 km h, the storm will remain at category 5. To minimise these, and other, problems, the author proposes alternative scales; a Hurricane Intensity Index (HII), related to the dynamic pressure exerted by hurricane winds, or a Hurricane Hazard Index (HHI), based upon the maximum, sustained near-surface rotational wind speed, the radius to which intensity winds extend, and the travel velocity of the storm. Unlike the SSHS, both scales would be continuous and open-ended. While the HII would be better suited to predicting wind damage to structures, the HHI would be far more useful than the current SSHS to estimating the scale of a potential disaster and thus more useful for emergency managers. Applying the scales to previous hurricanes, Category 5 Andrew in 1992 would have scored 5.0 on the HII and 10.4 on the HHI. In contrast, category 3 Katrina last year, would have registered at 2.9 on the HII, but a massive 19.3 on the HHI (table 3).



Statistical aspects of hurricane forecasting and risk assessment
In a very comprehensive paper published in The American Statistician, Ronald Iman ³², of Southwest Technology Consultants, and colleagues, give their impressions of the state-ofthe- art in hurricane forecasting and planning for hurricanes within the insurance market. Most pertinently and useful for the insurance sector, the authors outline potential contributions from statisticians that could improve both hurricane forecasting and mitigation. Proposed contributions include:

• Mixing and meshing data in order to come up with a simple 5-minute forecast model that runs with ‘current’ information, rather than the current, more sophisticated 6-hour advisory, that takes 5 hours to recycle, and which renders it close to obsolete at its announcement.
• Replacing standard ‘one-size-fits-all’ wind field models with ‘morphing’ wind field models that better capture hurricane behaviour by letting the present conditions of a storm dictate the model selection of wind-field components.
• Application of a multi-model approach, as described above, to forecasting storm tracks, such that the ‘winning’ model could be provided as the next predictor.