What is Required to Improve Detection and Attribution

At present, the possibility that anthropogenic warming has affected tropical cyclone behavior in the Atlantic is a plausible hypothesis. Hegerl et al. (2007) indicate that it is more likely than not that anthropogenic warming has affected tropical cyclone behavior. This is a fairly weak conclusion but it is the best that we can obtain at present. It should be noted that the detection and attribution of a human influence on global climate has been an evolutionary process, with relatively weaker conclusions based on less formal approaches reached in the IPCC Second Assessment (Santer et al 1995). The Summary for Policymakers of that assessment concluded "The balance of evidence suggests a discernible human influence on global climate'' (Houghton et al. 1995), which is much weaker than the conclusions of the IPCC Fourth Assessment: ''Most of the observed increase in globally averaged temperatures since the mid-20th century is very likely due to the observed increase in anthropogenic greenhouse gas concentrations" (Hegerl et al. 2007).

The formal attribution process as clearly defined by the IPCC is the best way to make strong conclusions about climate change effects. Due to the current inadequacies of tropical cyclone models and theories, formal attribution following the IPCC approach is not possible at this time. Therefore we must also employ other methods, as has been done in the past for assessing the possible effects of global warming on the future behavior of tropical cyclones (e.g. Henderson-Sellers et al. 1998; Walsh 2004). It is inevitable that this will involve making hypotheses about the physical reality of statistical relationships between variations in variables that have already been formally attributed and variations in tropical cyclone characteristics. The level of confidence for attribution of these statistical relationships is directly related to the level of confidence that we have in the hypothesized physical relationship that explains them. If these physical relationships are well-established, either by theory, simulation or observation at shorter time scales, then this confidence can be reasonably high. Moreover, there must also be some reason to believe that this physical relationship will remain the same in a warmer world. These ideas could provide the basis for a more structured approach to attribution until such time as simulations and theory improve to the point when the much stronger formal attribution process becomes possible.

The first step in any attribution process, formal or otherwise, is detection of a trend. Improved tropical cyclone data records would increase our confidence that trends had actually been detected and were simply not due to data inhomogeneities. There are a number of approaches that could be undertaken: a consistent reanalysis of the polar-orbiting satellite record, for instance, could be performed similar to the method used Kossin et al. (2007) for the geostationary satellite data. Given the finer resolution or the polar-orbiting satellites, this may lead to a more accurate determination of intensity trends. There are a number of limitations of any reanalysis procedure, however. As mentioned previously, storms that were never observed by anyone are gone forever, and only estimates can be made of their effect on any detected trends. For a reanalysed tropical cyclone data set to most useful for climate analysis, there have to be no artificial trends in the data caused by changes in observing practices. The reanalysis of Kossin et al. (2007) attempts this but at the cost of a degraded resolution of recent satellite data. One possibility would be instead to create a best track dataset with all available data but include error estimates that are larger for earlier storms. In this way, climate trends could still be analysed with statistical techniques that take into account the change in the error distribution with time when statistical significance of trends is calculated. Additionally, change points in the observing systems should have created change points in the data, and these can be corrected using well-established methods (e.g. Lanzante et al. 2003).

Once a robust trend is detected, the attribution step would ideally utilize an excellent climate model that produces tropical cyclones of about the right intensity and numbers, run with and without anthropogenic forcing, that would reproduce with reasonable fidelity the observed intensity trends, particularly in the Atlantic. The work of Knutson et al. (2007) is an important step in this direction, as their results imply that the increase in tropical cyclone numbers in the Atlantic is related to the pattern of the observed SSTs that were used to force their model. Since the SST anomalies are likely related to global warming, at first glance this suggests a causal link between global warming and tropical cyclone numbers in the Atlantic. Similar models will be used to run coupled climate runs that could then help identify the anthropogenically-forced transient climate response of tropical cyclones in the Atlantic and elsewhere.

In the absence of excellent climate model simulations, studies such as those of Emanuel (2007) could be further analysed to strengthen their conclusions. Specifically, it is presently unclear whether all of the individual components of his PDI parameter (MPI, vorticity and vertical wind shear) are varying in a manner consistent with an anthropogenic cause. An anthropogenic influence on MPI is likely, based on its formulation and our theoretical understanding of influences on tropical cyclone intensity, but this is not clear for vorticity or vertical wind shear. For instance, Vecchi and Soden (2007) show that multi-model projections of vertical wind shear trends in the Atlantic over the 21st century are strongly positive in parts of the tropical Atlantic (i.e. more hostile to cyclone formation), although trends are neutral in the main development region. This issue can be addressed by examining changes in the large-scale atmospheric fields between two sets of GCM simulations, with and without anthropogenic forcing, to determine whether the observed trends in vorticity and vertical wind shear are similar to those expected from anthropogenic forcing. Similar studies could be performed with other hypothesized combinations of variables. If a quantitative theory of tropical cyclone formation were to be developed, studies along these lines could also address the issue of the relative responses of formation and intensification to anthropogenic forcing. Important in all of these type of studies is whether there are good reasons to believe that relationships between parameters will remain the same in a warmer world. Such reasons would include a theory successfully tested at shorter time scales, such as the Emanuel MPI theory or a well-established observed relationship that is not expected to change in a warmer world, such as that between vertical wind shear and tropical cyclone intensification (e.g. Vecchi and Soden 2007).

Excellent climate model simulations have the potential to suggest where and when the detection of an anthropogenically-forced tropical cyclone signal might be achieved. Leslie and Karoly (2007) examine this issue using multi-member ensembles of simulations with a variable-resolution climate model, including both control and climate change simulations. They show that there is large natural decadal variability in the simulated number of strong tropical cyclones per decade in the northeast Australian region but that the simulated increase in strong tropical cyclones due to anthropogenic climate change should appear above the noise some time in the 2020s or later. The confidence of this prediction would be substantially increased if other independent models were to make similar predictions.

The formal detection and attribution methods described above and in Hegerl et al. (2007) use a null hypothesis of no expected change in the climate variable being considered, apart from that due to natural internal climate variations. Now that there is a substantial body of scientific research supporting the conclusion that most of the observed global average temperature increase since the mid-20th century is very likely (more than 90% certain) due to the increase in anthropogenic greenhouse gases in the atmosphere (Hegerl et al., 2007), it may be more appropriate to use a different null hypothesis. It is now appropriate to use a null hypothesis that global scale temperature increases, including sea surface temperature increases, over the past fifty years have a significant anthropogenic influence and then apply the same attribution methods to detect and attribute an anthropogenic climate change influence on tropical cyclones. The problem is substantially changed, now making use of the prior information that anthropogenic climate change is causing large scale warming and then seeking to quantify the specific changes expected to occur in the frequency and intensity of tropical cyclones. This is essentially a Bayesian statistical approach (e.g. Lee et al. 2005).

The use of Bayesian statistics has the potential to increase the sensitivity of detection and attribution studies and make it easier to increase the confidence that observed changes are due to anthropogenic influences. Bayesian techniques are being increasingly employed in atmospheric and oceanic statistical models (Wikle 2000; Berliner et al. 2002; Katz 2002). Elsner et al. (2004) apply Bayesian statistics to detect discontinuities (''change points'') in hurricane data, while Elsner and Jagger (2004) show that the inclusion of 19th century data as priors improved the significance of relationships between indices of ENSO and the NAO and 20th century North American coastal hurricane incidence. Jagger and Elsner (2006) used Bayesian extreme value statistics to show that warmer global temperatures were associated with larger numbers of intense hurricanes, although this result was not highly significant. Their results could also be interpreted to show that observed increases in global temperature and increases in maximum hurricane intensity are consistent with MPI theory.

Was this article helpful?

0 0
Survival Treasure

Survival Treasure

This is a collection of 3 guides all about survival. Within this collection you find the following titles: Outdoor Survival Skills, Survival Basics and The Wilderness Survival Guide.

Get My Free Ebook


Post a comment