9.2.3 Implications for Understanding 20th-Century Climate Change
Any assessment of observed climate change that compares simulated and observed responses will be affected by errors and uncertainties in the forcings prescribed in a climate model and its corresponding responses. As noted above, detection studies scale the response patterns to different forcings to obtain the best match to observations. Thus, errors in the magnitude of the forcing or in the magnitude of the model response to a forcing (which is approximately, although not exactly, a function of climate sensitivity), should not affect detection results provided that the large-scale space-time pattern of the response is correct. Attribution studies evaluate the consistency between the model-simulated amplitude of response and that inferred from observations. In the case of uncertain forcings, scaling factors provide information about the strength of the forcing (and response) needed to reproduce the observations, or about the possibility that the simulated pattern or strength of response is incorrect. However, for a model simulation to be considered consistent with the observations given forcing uncertainty, the forcing used in the model should remain consistent with the uncertainty bounds from forward model estimates of forcing.
Detection and attribution approaches that try to distinguish the response to several external forcings simultaneously may be affected by similarities in the pattern of response to different forcings and by uncertainties in forcing and response. Similarities between the responses to different forcings, particularly in the spatial patterns of response, make it more difficult to distinguish between responses to different external forcings, but also imply that the response patterns will be relatively insensitive to modest errors in the magnitude and distribution of the forcing. Differences between the temporal histories of different kinds of forcing (e.g., greenhouse gas versus sulphate aerosol) ameliorate the problem of the similarity between the spatial patterns of response considerably. For example, the spatial response of surface temperature to solar forcing resembles that due to anthropogenic greenhouse gas forcing (Weatherall and Manabe, 1975; Nesme-Ribes et al., 1993; Cubasch et al., 1997; Rind et al., 2004; Zorita et al., 2005). Distinct features of the vertical structure of the responses in the atmosphere to different types of forcing further help to distinguish between the different sources of forcing. Studies that interpret observed climate in subsequent sections use such strategies, and the overall assessment in this chapter uses results from a range of climate variables and observations.
Many detection studies attempt to identify in observations both temporal and spatial aspects of the temperature response to a given set of forcings because the combined space-time responses tend to be more distinct than either the space-only or the time-only patterns of response. Because the emissions and burdens of different forcing agents change with time, the net forcing and its rate of change vary with time. Although explicit accounting for uncertainties in the net forcing is not available (see discussion in Sections 184.108.40.206 and 220.127.116.11), models often employ different implementations of external forcing. Detection and attribution studies using such simulations suggest that results are not very sensitive to moderate forcing uncertainties. A further problem arises due to spurious temporal correlations between the responses to different forcings that arise from sampling variability. For example, spurious correlation between the climate responses to solar and volcanic forcing over parts of the 20th century (North and Stevens, 1998) can lead to misidentification of one as the other, as in Douglass and Clader (2002).
The spatial pattern of the temperature response to aerosol forcing is quite distinct from the spatial response pattern to CO2 in some models and diagnostics (Hegerl et al., 1997), but less so in others (Reader and Boer, 1998; Tett et al., 1999; Hegerl et al., 2000; Harvey, 2004). If it is not possible to distinguish the spatial pattern of greenhouse warming from that of fossil-fuel related aerosol cooling, the observed warming over the last century could be explained by large greenhouse warming balanced by large aerosol cooling or alternatively by small greenhouse warming with very little or no aerosol cooling. Nevertheless, estimates of the amplitude of the response to greenhouse forcing in the 20th century from detection studies are quite similar, even though the simulated responses to aerosol forcing are model dependent (Gillett et al., 2002a; Hegerl and Allen, 2002). Considering three different climate models, Stott et al. (2006c) conclude that an important constraint on the possible range of responses to aerosol forcing is the temporal evolution of the global mean and hemispheric temperature contrast as was suggested by Santer et al. (1996a; see also Section 18.104.22.168).