The recent global temperature hiatus has been explained by the IPCC AR5 as partly due to natural radiative forcings (solar & volcanic effects) and internal variability. Recently, other effects such as CFCs and biases in the observational coverage have also been suggested, as well as continuing uncertainty about the regional effects of aerosol forcings. When comparing simulations and observations, the CMIP5 simulations tend to use projected forcings rather than observed forcings after 2005. But what effect does this have?
The figure below compares global temperatures in four versions of the NASA GISS GCM. There are 5 simulations with historical forcings up to 2005 (grey), 5 simulations using RCP4.5 from 2006-2012, and 5 simulations using historicalExt forcings from 2006-2012. The historicalExt forcings use observed greenhouse gas (GHG) concentrations, solar cycle and volcanic aerosols, but use the aerosol emissions from RCP4.5. The observed GHG concentrations are higher than in the reference RCP4.5 scenario.
Note that the differences between the mean of the simulations (thick lines) are small with no clear differences. Perhaps this is unsurprising given the relatively small ensemble size and small possible effects. However, this suggests that the higher GHG concentrations may help cancel out the negative forcing from the natural factors somewhat, in this model. It would be great to see other groups perform similar experiments (with multiple ensemble members), but to my knowledge only NASA GISS have done this so far.