by Judith Curry
The 2005 Atlantic hurricane season saw hurricane activity and devastation that was unprecedented in the historical record.
In August 2005, Kerry Emanuel published a paper in Nature associating the increase in sea surface temperature (SST) since 1950 with an increase in maximum hurricane potential intensity and the destructive capacity of hurricanes, focusing on hurricanes in the North Atlantic and North Pacific. Webster et al. in an article published in Science (published 5 yrs ago this week) showed that while the total number of hurricanes has not increased globally since 1970, the proportion of category 4 and 5 hurricanes had doubled, implying that the distribution of hurricane intensity has shifted towards more intense hurricanes. The coincidental (edit) timing of the publication of these papers within weeks of Hurricane Katrina’s devastation focused intense media attention on the topic of greenhouse warming and increasing hurricane intensity, although neither Emanuel or Webster et al. directly attributed the tropical cyclone changes to greenhouse warming.
Five years after Hurricane Katrina, hundreds of papers have been published that relate in some way to the subject of hurricanes and climate variability/change. We also have 5 years of additional data. Major assessments have been undertaken by three different international and national groups. What have we learned? What don’t we know? How should we assess the risk from future hurricanes?
Assessing the assessments
Since 2005, assessments of climate change detection and attribution research with regard to tropical cyclones have been undertaken by two international groups (the IPCC and the WMO) plus the U.S. Climate Change Science Program. Below, I highlight the major conclusions from each of the three assessment reports; for full context of the statements cited here, please see the original sources. All likelihood statements follow the conventions of the IPCC.
The IPCC AR4 WG1 Summary for Policy Makers (published in 2007) concluded the following:
• Detection and attribution: “There is observational evidence for an increase in intense tropical cyclone activity in the North Atlantic since about 1970, correlated with increases of tropical sea surface temperatures. There are also suggestions of increased intense tropical cyclone activity in some other regions where concerns over data quality are greater. Multi-decadal variability and the quality of the tropical cyclone records prior to routine satellite observations in about 1970 complicate the detection of long-term trends in tropical cyclone activity. There is no clear trend in the annual numbers of tropical cyclones.”
• Projections: “Based on a range of models, it is likely that future tropical cyclones (typhoons and hurricanes) will become more intense, with larger peak wind speeds and more heavy precipitation associated with ongoing increases of tropical sea surface temperatures. There is less confidence in projections of a global decrease in numbers of tropical cyclones. The apparent increase in the proportion of very intense storms since 1970 in some regions is much larger than simulated by current models for that period.”
The CCSP Synthesis and Assessment Report (published in 2008) focused on tropical cyclones in the North Atlantic and North Pacific:
- Detection and attribution: “[I]t is likely that the annual numbers of tropical storms, hurricanes and major hurricanes in the North Atlantic have increased over the past 100 years, a time in which Atlantic sea surface temperatures also increased. . . This evidence suggests a substantial human contribution to recent hurricane activity.”
- Projections: “For North Atlantic and North Pacific hurricanes, it is likely that hurricane rainfall and wind speeds will increase in response to human-caused warming. Analyses of model simulations suggest that for each 1oC (1.8oF) increase in tropical sea surface temperatures, core rainfall rates will increase by 6-18% and the surface wind speeds of the strongest hurricanes will increase by about 1-8%. It is presently unknown how late 21st century tropical cyclone frequency in the Atlantic and North Pacific basins will change compared to the historical period.”
A review article published in 2010 by the World Meteorological Organization Expert Team on Climate Change Impacts on Tropical Cyclones (Knutson et al.) concluded the following:
- Detection and attribution: “It remains uncertain whether past changes in any tropical cyclone activity (frequency, intensity, rainfall, and so on) exceed the variability expected through natural causes, after accounting for changes over time in observing capabilities.”
- Projections: “It is likely that the global frequency of tropical cyclones will either decrease or remain essentially unchanged owing to greenhouse warming. . . Current models project changes ranging from −6 to −34% globally, and up to ±50% or more in individual basins by the late twenty-first century. Some increase in the mean maximum wind speed of tropical cyclones is likely (+2 to +11% globally) with projected 21st century warming.”
Differences among these assessments are described by Knutson et al. in the supplementary material. Relative to the CCSP report:
“[Knutson et al.] do not assign a likely confidence level to the reported increases in annual numbers of tropical storms, hurricanes and major hurricanes counts over the past 100 years in the North Atlantic basin. . . Specifically [Knutson et al.] do not conclude that there has been a detectable change in tropical cyclone metrics relative to expected variability from natural causes, particularly owing to concerns about limitations of available observations and limited understanding of the possible role of natural climate variability in producing low frequency changes in the tropical cyclone metrics examined.”
Relative to the IPCC AR4:
“[Knutson et al.’s] conclusions–that it is more likely than not that global tropical storm frequency will decrease and more likely than not that the frequency of the more intense storms will increase in some basins–are more specific than IPCC AR4, which concluded that there was “…less confidence [than likely] in these projections [of a decrease in the overall number of tropical storms] and in the projected decrease of relatively weak storms in most basins, with an increase in the numbers of the most intense tropical cyclones.”
How should we interpret these differences in conclusions and confidence levels of the three different assessments? One issue is the time derivative factor of comparing assessments published in 2007, 2008, and 2010. The differences in confidence levels of the three different assessments reflect different groups of experts (with some overlap). The basis for assessing confidence levels used in all three assessments (following Moss and Schneider 2000) expresses quantitative levels of confidence based on the amount of evidence (number of sources of information) and the degree of agreement (consensus) among experts. Given the explicit reliance on the consensus among experts in the confidence assessment, it is not difficult to see how different groups of experts can assess the same information in different ways.
The challenge to making and interpreting assessments could be improved by using the Italian flag representation of three-valued logic (Blockley and Godfrey 2000, see also this document; h/t Michael Welland )
in which evidence for a proposition is represented as green, evidence against is represented as red, and residual uncertainty is represented as white. The white area reflects uncommitted belief, which can be associated with uncertainty in evidence or unknowns. This representation gives uncertainty an explicit place in the assessment and requires consideration of both the known unknowns and the unknown unknowns. The differences in the three assessments can then be interpreted as different views on the size of the white area and what it actually contains.
So, given that there have been three major assessments in the past three years, what is the point of an individual (i.e. me, with my individual interests, biases, misconceptions and ignorance) conducting yet another assessment, and posting it on a blog? I am providing a different type of assessment, one that focuses on the white region of the Italian flag diagram, i.e. the uncertainties and the knowledge frontier. Sitting squarely in the white region are a host of issues that have barely been addressed in the other assessments, issues that relate to physical processes and the climate dynamics of hurricanes. I view consideration of these issues to be essential for developing confidence in our statements about attribution and projections of tropical cyclone activity. And most importantly, I am posting my assessment on a blog to stimulate dialogue on this topic.
But before pondering the knowledge frontier, first an update on the Webster et al. (2005) paper.
Detection of an increase in hurricane intensity?
Since the Webster et al. paper was published in 2005, two major developments have occurred that have improved the global hurricane database. The first is the development of the International Best Tracks (IBTrACS) dataset (Knapp et al.), which reflects a cleaning up and homogenization of the data relative to what was used by Webster et al. The second advance is the development of a new homogeneous satellite-based intensity analysis (developed by Kossin et al.)
Data from 1980-2009 from the IBTrACS (DATA) are used to recreate the Webster et al. analysis. Note, data from the 1970’s are not used in this new analysis owing to the substantial uncertainty of data in the southern hemisphere. No statistical analysis of the data is presented here; I leave that to others to decide the best way to assess the magnitude and significance of the trend.
The total number of hurricanes shows no overall trend, with 2007, 2008, and 2009 ranking among the 4 lowest totals in the record. The numbers of category 4 and 5 hurricanes during 2007 2008, and 2009 are also correspondingly low. However, consideration of the % of category 4 and 5 hurricanes shows a continued overall increase, with 2002, 2004 and 2006 ranking as the highest years in the record, and all years since 2004 are above average except for 2008. Further support for an increase in the proportion of the most intense hurricanes is provided by Elsner et al. (using the Kossin et al. satellite dataset), see esp Fig 1b. This increase in the % of most intense hurricanes is dominated by increases in the North Atlantic and also the North and South Indian Oceans.
Hence, the available information that we have on global hurricane intensity supports the original findings of Webster et al. of a global increase in the % of category 4 and 5 hurricanes, although the trend is somewhat smaller than that found by Webster et al. The increase in the % of category 4 and 5 hurricanes reflects a shift in the distribution of hurricane intensities, i.e. a fattening of the tail of the intensity distribution.
Back to the white area in the Italian flag diagram: uncertainty and unknowns that comprise the frontier of knowledge. My selection of topics at the knowledge frontier is by no means complete, but reflects questions that have accumulated in my own mind regarding the causal chain that links the formation and intensification of hurricanes to climate dynamics. And I note in advance my bias for citing recent papers from the Georgia Tech group. I anticipate that others will expand on this list of papers.
The relationship between sea surface temperature and hurricane intensity. The causal chain for global warming to increase hurricane intensity has been argued to occur via the increase in sea surface temperature (SST) (e.g. Curry et al. 2006 ). A nominal SST threshold of 26.5-27oC is used as a criterion for tropical cyclogenesis, and a nominal threshold of 28.5oC for intensification to a major hurricane. Some new insights into the relationship between a warming climate and hurricane intensity is suggested by Hoyos and Webster (2010, forthcoming). During the latter half of the 20th century, the tropical warm pool defined by the isotherm of 28oC has expanded in area. However, the region of tropical cyclogenesis has not expanded, owing to the area of convective activity remaining nearly constant. Hoyos and Webster argue that the temperature threshold for tropical cyclogenesis increases as the average tropical ocean temperature increases. It is the increasing intensity of the convection with warmer temperatures that seems to be the link between SST increase and hurricane intensity, rather than the absolute value of the SST itself. Further, the location of the intense convection is related to the difference between the local SST and global tropical average SST, rather than to the absolute value of the SST itself (Vecchi et al. 2008)
Maximum potential intensity theory. In Emanuel (2005) and Webster et al. (2005), the theory of maximum potential intensity (MPI; Emanuel 1988) was invoked to explain the link between increasing sea surface temperature and increasing hurricane intensity. MPI theory is widely accepted by the hurricane community, and is used in operational forecasts of hurricane intensity. However, several critiques of MPI have raised questions about its validity. Smith et al. (2008) argue that MPI is derived from a flawed model that makes the tacit assumption of gradient wind balance in the atmospheric boundary layer, a layer that owes its existence to gradient wind imbalance in the radial momentum equation. Russian biophysicists Makarieva and Gorshov (2009 , 2010) critique the application of the dissipative heat engine model used by Emanuel used in the development of MPI (note I understand that Emanuel has submitted a response to this paper). Cione et al. (2000) questions the conventional wisdom of the TC inflow layer used to derive MPI, finding that low level inflow is not isothermal, not constant with respect to surface specific humidity, and not in near thermodynamic equilibrium with the sea surface. Lackman and Yablonsky (2004) point out the importance of the precipitation mass sink in the pressure drop in a hurricane, an effect that is not included in MPI theory. Rethinking of the maximum potential intensity of hurricanes seems to be in order.
Hurricanes and modes of natural climate variability. Interannual and multidecadal modes of climate variability have long been known to influence hurricane activity. Recent research by Kim et al. has highlighted the impact of the increasingly frequently Modoki El Nino on hurricane activity in the Atlantic and the Pacific. The Pacific Decadal Oscillation (PDO) has a dominant impact on hurricane variability in the Pacific, and modulates the frequencies of El Nino/La Nina. DiLorenzo has identified a new multidecadal oscillation in the Pacifc, the North Pacific Gyre Oscillation (NPGO) that is independent of PDO. In a follow up paper, DiLorenzo et al. identify the central Pacific warming (Modoki) as a different model from ENSO, that is the high frequency expression of the NPGO. I suspect that the combination of the PDO and NPGO can explain much of the variability in Ryan Maue’s analysis of Accumulated Cyclone Energy diagram (reference to Maue’s paper added – edit), given that the majority of global hurricanes occur in the Pacific. In the Atlantic, the AMO, NAO, and AMM have all been invoked to explain variability in the Atlantic in numerous papers, and I have argued (in an unpublished presentation) that the PDO also influences Atlantic hurricane activity on decadal time scales. And finally, Elsner presents intriguing evidence for a relation between Caribbean hurricane activity and the sunspot cycle.
Inferences of hurricane activity from climate model simulations. Because of the coarse resolution of climate models, such inferences require either statistical relations between the model predicted fields and hurricane activity, or some sort of downscaling with a seeding mechanism. The seeding is required because even at much higher resolutions, models do not do a good job of predicting genesis (formation) of tropical cyclones. Prediction of hurricane intensity is very challenging, even in an operational forecasting environment using high-resolution models for a hurricane that has already formed. Given these issues with understanding and modeling tropical cyclogenesis and intensification, what kind of confidence should we have in inferences of hurricane statistics from climate model simulations? While such inferences are becoming increasingly sophisticated (e.g. Bender et al. 2010, Emanuel et al. 2008), I find it difficult to take them with more than a grain of salt: the seeding is artificial and intensification mechanisms are uncertain. Further, issues of climate model structural deficiencies of relevance to hurricane development have been raised by Emanuel (paper 4A.4) with regards to the substantial disagreement between modeled and observed values of upper topospheric temperature and humidity.
Other uncertainties. Other uncertainties include:
- the role of tropical cyclones in the general circulation of the atmosphere and ocean
- lack of basic understanding regarding the intrinsic nature of tropical cyclone genesis and evolution (I will post on this topic in the future)
- basic storm energetics and water cycle (particularly the evaporative flux from the ocean surface)
- the role of hurricane horizontal size on energetics and damage
Here is what I think we can state with some sort of confidence:
- Hurricane frequency and intensity in the North Atlantic has likely [>66%] increased since 1970. The transition from the cold to warm phase of the AMO is a plausible explanation for this increase. Attribution of a portion of the increase in hurricane intensity to AGW would require (at minimum) resolution of the problems with the intensities during the period of 1945-1970 (encompassing the previous warm period of the AMO).
- It is more likely than not [>50%] that the % of category 4 and 5 hurricanes has increased globally since 1980. Increased confidence requires continued efforts to reprocess the data. Attribution of any portion of this increase to anthropogenic global warming would require careful examination of the data and modes of natural variability in each of the regions where hurricanes occur.
- It is more likely than not [>50%] that the maximum intensity of the strongest hurricanes would increase in a warmer climate. While there is an absence of evidence against this hypothesis, there are also substantial uncertainties in the observations and theory (i.e. a preponderance of green with no red, but a substantial white region).
- I do not currently place any confidence in climate model projections of future hurricane activity. At the same time, I recognize the substantial advances made on this in recent years particularly by the GFDL group and ongoing efforts by several other groups as well.
How should decision makers respond to this issue? We got it right in our 2006 statement (kudos to Kerry Emanuel for taking the lead on this), the statement summary is given below:
“We are optimistic that continued research will eventually resolve much of the current controversy over the effect of climate change on hurricanes. But the more urgent problem of our lemming-like march to the sea requires immediate and sustained attention. We call upon leaders of government and industry to undertake a comprehensive evaluation of building practices, and insurance, land use, and disaster relief policies that currently serve to promote an ever-increasing vulnerability to hurricanes.”
Note regarding comments: This thread will be heavily moderated for relevance to science of hurricanes and the assessment of the science. More general discussion on the hurricane topic is welcome on the Open thread. I will participate on both this thread and on the Open thread. A good litmus test for whether you should submit your post to this thread or the Open thread is whether or not you have read any of the scientific papers linked to in this thread. If you would like a relatively easy to read “primer” on this subject, i recommend this.