Unforced Surface Temperature Variability in Observations and Global Climate Models

The magnitude of unforced global mean surface temperature variability in global climate models (GCMs) is compared to that apparent in instrumental and reconstructed datasets over the period 1500-2010. Time series of unforced surface temperature variability are acquired by subtracting an estimate of the forced component of the variability (estimated with a box-diffusion model of global ocean mixed layer temperature) from several observational datasets. The Fourier spectra of observed unforced variability obtained in this manner is then compared to that of GCM preindustrial control runs from the CMIP3 and CMIP5 archives. Results indicate that most GCMs tend to underestimate unforced surface temperature variability across most frequencies compared to reconstructed datasets. Agreement is higher between GCM simulated unforced variability and the unforced variability extracted over the instrumental era.