Results are reported from simulations with an atmospheric general circulation model (GCM) covering the 20th century. These results focus primarily on the continental surface air temperature (SAT) record, with some discussion of large spatial scale precipitation averages as well. These experiments were conducted using the Max Planck Institute for Meteorology ECHAM3 GCM configured at triangular-21 truncation, giving a spatial resolution of approximately 5.5 degrees. The surface boundary conditions for the model were constructed by forming decadal climatologies of sea surface temperature (SST) data from the Global SST Atlas (GOSSTA) data set. No changes in greenhouse gas concentrations were made during the experiments. The GCM was then integrated for 5 years with each climatology. Because the high southern latitudes are poorly sampled and show considerable high-amplitude variability in the GOSSTA data, two complete experiments were conducted. In one experiment the time-varying GOSSTA data were prescribed globally (the GLBL simulation), in the other, SSTs in the high southern latitudes were held at the 1950-1969 climatology (the NO-SH simulation). The results show both GCM simu(l)ations reproduce important features of the observed 20th century global temperature record, suggesting that the low-frequency variability in global average SAT is modulated to an important degree by variability in SSTs. In the model results, the changes in SST are transmitted to the atmosphere via latent heat flux, i.e., by changes in the flux of water through the hydrologic cycle. The most notable difference between the GLBL and NO-SH SAT records is that during the 1920-1940 period the NO-SH results agree more closely with observations, while the GLBL results are superior from the 1950s forward. At least in part, these differences appear to be linked to differences in simulated sea-ice extent in the two simulations. Analyses of large-scale, averages in continental precipitation show good model-observation agreement in the tropics (+/-30 degrees latitude) and in the extratropical southern hemisphere; however, the upward trend in extratropical NH continental precipitation seen in the observed record is not reproduced in either simulation.
