IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis Comparing Simulations of Northern Hemisphere Mean Temperatures with Palaeoclimatic Observations

Various simulations of NH (mean land and marine) surface temperatures produced by a range of climate models, and the forcings that were used to drive them, are shown in Figure 6.13. Despite differences in the detail and implementation of the different forcing histories, there is generally good qualitative agreement between the simulations regarding the major features: warmth during much of the 12th through 14th centuries, with lower temperatures being sustained during the 17th, mid 15th and early 19th centuries, and the subsequent sharp rise to unprecedented levels of warmth at the end of the 20th century. The spread of this multi-model ensemble is constrained to be small during the 1500 to 1899 reference period (selected following Osborn et al., 2006), but the model spread also remains small back to 1000, with the exception of the ECHO-G simulation (Von Storch et al., 2004). The implications of the greater model spread in the rates of warming after 1840 will be clear only after determining the extent to which it can be attributed to differences in prescribed forcings and individual model sensitivities (Goosse et al., 2005b). The ECHO-G simulation (dashed red line in Figure 6.13d) is atypical compared to the ensemble as a whole, being notably warmer in the pre-1300 and post-1900 periods. Osborn et al. (2006) showed that these anomalies are likely the result of a large initial disequilibrium and the lack of anthropogenic tropospheric aerosols in that simulation (see Figure 6.13c). One other simulation (González-Rouco et al., 2006) also exhibits greater early 20th-century warming in comparison to the other simulations but, similarly, does not include tropospheric aerosols among the forcings. All of these simulations, therefore, appear to be consistent with the reconstructions of past NH temperatures, for which the evidence (taken from Figure 6.10c) is shown by the grey shading underlying the simulations in Figure 6.13d.

It is important to note that many of the simulated temperature variations during the pre-industrial period shown in Figure 6.13 have been driven by assumed solar forcing, the magnitude of which is currently in doubt. Therefore, although the data and simulations appear consistent at this hemispheric scale, they are not a powerful test of the models because of the large uncertainty in both the reconstructed NH changes and the total radiative forcing. The influence of solar irradiance variability and anthropogenic forcings on simulated NH surface temperature is further illustrated in Figure 6.14. A range of EMICs (Petoukhov et al., 2000; Plattner et al., 2001; Montoya et al., 2005) were forced with two different reconstructions of solar irradiance (Bard et al., 2000; Y.M. Wang et al., 2005) to compare the influence of large versus small changes in the long-term strength of solar irradiance over the last 1 kyr (Figure 6.14b). Radiative forcing related to explosive volcanism (Crowley, 2000), atmospheric CO2 and other anthropogenic agents (Joos et al., 2001) were identically prescribed within each model simulation. Additional simulations, in which anthropogenic forcings were not included, enable a comparison to be made between ‘natural’ versus ‘all’ (i.e., natural plus anthropogenic) forcings on the evolution of hemispheric temperatures before and during the 20th century.

The alternative solar irradiance histories used in the simulations differ in their low-frequency amplitudes by a factor of about three. The ‘high-amplitude’ case (strong solar irradiance forcing) corresponds roughly with the level of irradiance change assumed in many of the simulations shown in Figure 6.13b, whereas the ‘low-amplitude’ case (weaker solar irradiance forcing) is representative of the more recent reconstructions of solar irradiance changes (as discussed in Section 6.6.3). The high-amplitude forcing history (‘Bard25’, Table 6.3) is based on an ice core record of 10Be scaled to give an average reduction in solar irradiance of 0.25% during the Maunder Minimum, as compared to today (Bard et al., 2000). The low-amplitude history (‘Bard08-WLS’) is estimated using sunspot data and a model of the Sun’s closed magnetic flux for the period from 1610 to the present (Y.M. Wang et al., 2005), with an earlier extension based on the Bard et al. (2000) record scaled to a Maunder Minimum reduction of 0.08% compared to today. The low-frequency evolution of these two reconstructions is very similar (Figure 6.14) even though they are based on completely independent sources of observational data (sunspots versus cosmogenic isotopes) and are produced differently (simple linear scaling versus modelled Sun’s magnetic flux) after 1610.

Table 6.3. Simulations with intermediate complexity climate models shown in Figure 6.14.

Bern2.5CC  Plattner et al., 2001  
Climber2  Petoukhov et al., 2000  
Climber3a  Montoya et al., 2005  
Volcanic  Forcing from Crowley (2000) used in all runs  
Solar  ‘Bard25’ runs used strong solar irradiance changes, based on 10Be record scaled to give a Maunder Minimum irradiance 0.25% lower than today, from Bard et al. (2000) 
‘Bard08-WLS’ runs used weak solar irradiance changes, using sunspot records and a model of the Sun’s magnetic flux for the period since 1610, from Y.M. Wang et al. (2005), and extended before this by the 10Be record scaled to give a Maunder Minimum irradiance 0.08% lower than today 
Anthropogenic  ‘All’ runs included anthropogenic forcings after 1765, from Joos et al. (2001) 
‘Nat’ runs did not include any anthropogenic forcings 


a Models: Bern2.5CC = Bern 2.5D Carbon Cycle-Climate Model, CLIMBER = Climate Biosphere Model.

The EMIC simulations shown in Figure 6.14, like those in Figure 6.13d, fall within the range of proxy-based NH temperature reconstructions shown in Figure 6.10c and are compatible with reconstructed and observed 20th-century warming only when anthropogenic forcings are incorporated. The standard deviation of multi-decadal variability in NH SAT is greater by 0.04°C to 0.07°C for the stronger solar forcing (Bard25, Table 6.3) compared to the weaker solar forcing (Bard08-WLS). The uncertainty associated with the proxy-based temperature reconstructions and climate sensitivity of the models is too large to establish, based on these simulations, which of the two solar irradiance histories is the most likely. However, in the simulations that do not include anthropogenic forcing, NH temperatures reach a peak in the middle of the 20th century, and decrease afterwards, for both the strong and weak solar irradiance cases. This suggests that the contribution of natural forcing to observed 20th-century warming is small, and that solar and volcanic forcings are not responsible for the degree of warmth that occurred in the second half of the 20th century, consistent with the evidence of earlier work based on simple and more complex climate models (Crowley and Lowery, 2000; Bertrand et al., 2002b; Gerber et al., 2003; Hegerl et al., 2006; Tett et al., 2007; see also Chapter 9).

An overall conclusion can be drawn from the available instrumental and proxy evidence for the history of hemispheric average temperature change over the last 500 to 2,000 years, as well as the modelling studies exploring the possible roles of various causal factors: that is, greenhouse gases must be included among the forcings in order to simulate hemispheric mean temperatures that are compatible with the evidence of unusual warmth observed in the second half of the 20th century. It is very unlikely that this warming was merely a recovery from a pre-20th century cold period.