As noted in Section 12.3.2, increases in greenhouse gases produce a distinctive change in the vertical profile of temperature. Santer et al. (1996c) assessed the significance of the observed changes in recent decades using equilibrium GCM simulations with changes in greenhouse gases, sulphate aerosols and strato-spheric ozone. This study has been extended to include results from the transient AOGCM simulations, additional sensitivity studies and estimates of internal variability from three different models (Santer et al., 1996a). Results from this study are consistent with the earlier results - the 25-year trend from 1963 to 1988 in the centred correlation statistic between the observed and simulated patterns for the full atmosphere was significantly different from the population of 25-year trends in the control simulations. The results were robust even if the estimates of noise levels were almost doubled, or the aerosol response (assumed linear and additive) was halved. The aerosol forcing leads to a smaller warming in the Northern Hemisphere than in the Southern Hemisphere.
Tett et al. (1996) refined Santer et al.'s (1996a) study by using ensembles of transient simulations which included increases in CO2, and sulphate aerosols, and reductions in stratospheric ozone, as well as using an extended record of observations (see Figure 12.8). They found that the best and most significant agreement with observations was found when all three factors were included1. Allen and Tett (1999) find that the effect of greenhouse gases can be detected with these signal patterns using optimal detection (see Appendix 12.1).
Folland et al. (1998) and Sexton et al. (2001) take a complementary approach using an atmospheric model forced with sea surface temperatures (SST) and ice extents prescribed from observations. The correlation between the observed and simulated temperature changes in the vertical relative to the base period from 1961 to 1975 was computed. The experiments with anthropogenic forcing (including some with tropospheric ozone changes), give significantly higher correlations than when only SST changes are included.
Interpretation of results
Weber (1996) and Michaels and Knappenburger (1996) both criticised the Santer et al. (1996a) results, quoting upper air measurements analysed by Angell (1994). Weber argued that the increasing pattern similarity over the full atmosphere (850 to 50 hPa) resulted mainly from a Southern Hemisphere cooling associated with stratospheric ozone depletion. Santer et al. (1996b) pointed out that when known biases in the radiosonde data are removed (e.g., Parker et al., 1997), or satellite or operationally analysed data are used, the greater stratospheric cooling in the Southern Hemisphere all but disappears. Weber (1996) is correct that stratospheric cooling due to ozone will contribute to the pattern similarity over the full atmosphere, but decreases in stratospheric ozone alone would be expected to produce a tropospheric cooling, not a warming as observed. This point should be born in mind when considering a later criticism of the pattern correlation approach. Both Weber (1996) and Michaels and Knappenburger (1996) note that the greater warming of the Southern Hemisphere relative to the Northern Hemisphere from 1963 to 1988 has since reversed. They attribute the Southern Hemisphere warming from 1963 to the recovery from the cooling following the eruption of Mount Agung. Santer et al. (1996b) claim that this change in asymmetry is to be expected, because the heating due to increases in greenhouse gases over the most recent years has probably been growing faster than the estimated cooling due to increases in aerosols (see Section 126.96.36.199). Calculations of the difference in the rate of warming between the Northern and Southern Hemispheres vary between different climate models and as a function of time, depending on the relative forcing due to greenhouse gases and sulphate aerosols, and on the simulated rate of oceanic heat uptake in the Southern Hemisphere (Santer et al., 1996b; Karoly and Braganza, 2001).
Assessing statistical significance of changes in the vertical patterns of
There are some difficulties in assessing the statistical significance in detection studies based on changes in the vertical temperature profile. First, the observational record is short, and subject to error, particularly at upper levels (Chapter 2). Second, the model estimates of variability may not be realistic (Section 12.2.2), particularly in the stratosphere. Third, because of data and model limitations, the number of levels used to represent the stratosphere in detection studies to date is small, and hence may not be adequate to allow an accurate representation of the stratospheric response. Fourth, all models produce a maximum warming in the upper tropical troposphere that is not apparent in the observations and whose impact on detection results is difficult to quantify. Nevertheless, all the studies indicate that anthropogenic factors account for a significant part of recent observed changes, whereas internal and naturally forced variations alone, at least as simulated by current models, cannot explain the observed changes. In addition, there are physical arguments for attributing the changes in the vertical profile of temperature to anthropogenic influence (Section 12.3.2).
Other reports in this collection