Resumen
The observed global, hemispheric, and meridional (zonal means) temperature fluctuations near surface covering the period 1851-1991 or 1861-1991, respectively, are reproduced by 'multiforced' regression models which take into account volcanic, solar, ENSO, and GHG (greenhouse gases) forcing. Moreover, the observed trends and trend-to-noise ratios are compared using different data sources. Different statistical tests including random forcing of detrended data (Monte Carlo testing) and extrapolations on the basis of 'calibration' subperiods demonstrate that the statistical approach works quite well Therefore not only assessments of GHG and natural signals in observational temperature time series are presented but also extrapolations based on the IPCC scenarios A and D. A comparison with the results obtained from the Hamburg Max-Planck-Institute for Meteorology (MPI) coupled atmosphere-ocean transient simulations show a somewhat larger magnitude of the statistically derived GHG signals where the difference, however, is not significant when the uncertainties of both methods are taken into account.