Before detailing the numerous obstacles facing detection and attribution studies of tropical cyclone behavior, we illustrate the process of detection and attribution through the use of a few examples. One of the easiest variables that can be used to demonstrate successful detection and attribution is global mean near-surface air temperature. Such a study is relatively straightforward for a number of reasons. The detected signal for global mean temperature increase in the past 100 years is highly statistically significant (e.g. Trenberth et al. 2007). The data used to estimate this trend have been extensively analyzed over many years and have small error bars. The main tool used for attributing this trend to man-made climate change, the global climate model, simulates global average temperature variability well (Randall et al. 2007). When global climate models are driven by the best available estimates of the radiative forcing of the 20th century, they reproduce well the observed temperature global average increase in the latter part of that century. Finally, when the key man-made elements of the forcing are removed, leaving only the naturally-varying components such as solar forcing, the models fail to reproduce the observed temperature increase. Thus the observed increase in global average temperature in the 20th century can be confidently ascribed to man-made global warming (Hegerl et al. 2007). Numerous earlier studies showed this (e.g Tett et al. 1999; Stott et al. 2001); more recently (Hegerl et al. 2007), this work has been extended to continental-average temperatures over most areas of the globe, demonstrating that these temperature increases can also be attributed to anthropogenic climate change.
Other observed climate trends have been formally ascribed to anthropogenic climate change. Barnett et al. (2005) and Pierce et al. (2006) analyzed trends in upper ocean temperatures in various ocean basins over the period 1960-2000, examining the observed change of temperature with depth and comparing it with the results of climate model simulations. They found that the oceanic warming over this period had been most pronounced in the upper part of the ocean and that this profile of temperature change was well simulated by numerical models using anthropogenic forcing, and could not be simulated when this forcing was removed. An increasing number of attribution studies have considered variables other than temperature. An anthropogenic influence has been formally identified in the increasing height of the tropopause over the last 3 decades (Santer et al. 2003), associated with stratospheric cooling due to ozone depletion and tropospheric warming due to increasing greenhouse gases. Observed multi-decadal changes in global patterns of mean sea level pressure have been attributed to anthropogenic forcing (Gillett 2005), as the observed changes cannot be explained by natural variability and are consistent with the response to anthropogenic forcing. However, the simulated pressure response to anthropogenic forcing is much weaker than the observed pressure trends, even though there is general agreement in the large scale spatial pattern of pressure changes.
These standard techniques have not been successfully applied to all climate variables, however. Problems have been encountered in detection and attribution studies of variables such as precipitation, due to the generally poor ability of models to simulate precipitation trends, the expected strong regional variations in trends due to climate change and large interannual variability in current and future climates (e.g. Lambert et al. 2004). Similar problems would be encountered in any similar formal attribution studies for tropical cyclones, beginning with the effect of data quality on signal detection.
Was this article helpful?