If a real trend is detected, attribution of such trends could be accomplished in a number of ways. A well-tested theory of tropical cyclone numbers or intensities could be compared with observed trends. Alternatively, a numerical simulation of tropical cyclones could be performed analogous to previously performed simulations of 20th century climate and the results with and without anthropogenic forcing compared with observations. Less confidently, statistical links could be made between well-attributed variables and tropical cyclone characteristics.
Studies that are largely statistical can give indications of associations that need to be investigated, but as tools for attribution they naturally provide less confident results. For example, Holland and Webster (2007) demonstrate a very plausible causal connection between the observed global warming, the warming of sea surface temperatures (SSTs) in the Atlantic and the subsequent changes in tropical cyclone behaviour in that region. A mechanism for this is proposed by Vimont and Kossin (2007), who show that there are apparent strong relationships between variations in tropical cyclone characteristics and the Atlantic Meridional Mode (AMM; see the review by Xie and Carton 2004). This is due to the circulation changes induced by the AMM, including changes to SST anomalies, whereby the main genesis regions of tropical cyclones tend to move equatorward to regions where the MPI is larger and where they are more likely to reach their MPI due to lower wind shear during positive phases of the AMM. One way to improve the confidence of the attribution to global warming in the analysis of Holland and Webster (2007) would be to employ physically-based modelling studies to show that a consequence of warming in the late 20th century is that changes in atmospheric circulation in models forced by changes in anthropogenic factors are consistent with a southward move in the main Atlantic tropical cyclone genesis regions, and that such circulation changes do not occur in unforced simulations. In this way, confidence would be improved in the Holland and Webster (2007)
conclusion that tropical cyclone trends in the Atlantic are due to global warming. Thus such attribution studies need not take the form of direct simulation of tropical cyclones in climate models - and, given the current state of the art, this would be difficult (Walsh 2008). But they should include, where possible, an assessment of how anthropogenic climate change is likely to affect the crucial variables used in detected statistical relationships, employing either simulations or theoretical techniques to do so.
Was this article helpful?