We show that atmosphere-only general circulation models (AGCMs) are a useful tool for understanding recent climate change and that they can be used to detect an anthropogenic effect on climate. Ensembles of AGCM runs are forced with the observed change in sea surface temperatures (SST) and sea-ice extents and various combinations of radiative forcings. The anthropogenic forcings of increasing greenhouse gases, increasing sulphate aerosols, decreasing stratospheric ozone and increasing tropospheric ozone are successively added to the experiments one at a time. The AGCM method aims to detect a signal against climate noise measured as the internal atmospheric variability. Note that the signal does not include that part of the anthropogenic forcing that has arisen through feedbacks with the ocean.
We use general linear regression theory to separate out the various anthropogenic signals from the effects due to changing SSTs. We apply this to the simulated land surface temperature and show that increasing greenhouse gas concentrations has caused a warming of 0.15³C since 1949 on top of a warming of 0.2³C due to increasing SSTs. Note that the latter warming already includes any indirect effect from increasing greenhouse gases through feedbacks with the ocean. Furthermore, we show that it is unlikely that the warming of the land in the increasing greenhouse gas experiment over this period could be accounted for by internal atmsopheric variability.
In the framework of the general linear model we also investigate
various aspects of the experimental design. We show that the
experimental design of successively adding forcings may not be
optimal. This is because all anthropogenic forcings are slowly changing
in time. Therefore, in the experimental design described above it is
difficult to distinguish between the different forcings.