Analyses are conducted to assess whether simulated trends in SST and land s
urface air temperature from two versions of a coupled ocean-atmosphere mode
l are consistent with the geographical distribution of observed trends over
the period 1949-1997. The simulated trends are derived from model experime
nts with both constant and time-varying radiative forcing. The models analy
zed are low-resolution (R15, similar to 4 degrees) and medium-resolution (R
30, similar to 2 degrees) versions of the Geophysical Fluid Dynamics Labora
tory (GFDL) coupled climate model. Internal climate variability is estimate
d from long control integrations of the models with no change of external f
orcing. The radiatively forced trends are based on ensembles of integration
s using estimated past concentrations of greenhouse gases and direct effect
s of anthropogenic sulfate aerosols (G+S). For the regional assessment, the
observed trends at each grid point with adequate temporal coverage during
1949-1997 are first compared with the R15 and R30 model unforced internal v
ariability Nearly 50% of the analyzed areas have observed warming trends ex
ceeding the 95th percentile of trends from the control simulations. These r
esults suggest that regional warming trends over much of the globe during 1
949-1997 are very unlikely to have occurred due to internal climate variabi
lity alone and suggest a role for a sustained positive thermal forcing such
as increasing greenhouse gases. The observed trends are then compared with
the trend distributions obtained by combining the ensemble mean G+S forced
trends with the internal variability "trend" distributions from the contro
l runs. Better agreement is found between the ensemble mean G+S trends and
the observed trends than between the model internal variability alone and t
he observed trends. However, the G+S trends are still significantly differe
nt from the observed trends over about 30% of the areas analyzed. Reasons f
or these regional inconsistencies between the simulated and the observed tr
ends include possible deficiencies in (1) specified radiative forcings, (2)
simulated responses to specified radiative forcings, (3) simulation of int
ernal climate variability, or (4) observed temperature records.