The latest global climate models (GCMs) have performed pre-industrial control simulations as part of the CMIP5 coordinated experiments. In these simulations there are no changes to radiative forcings, which are kept fixed at year 1850 values – all the variability is therefore generated internally to the climate system. How different can this variability be?
The first post on this blog compared the pre-industrial control simulations from the previous CMIP3 GCMs. The figure below updates that comparison to CMIP5 – the same scale is used for each simulation.
It is instantly clear that there are large differences in simulated internal variability, just like in CMIP3. Several models have very little variability (e.g. GISS, inmcm4), others have significant decadal variability (e.g. GFDL CM3, CMCC-CM), and some have ‘drifts’ because they have yet to reach equilibrium (e.g. GFDL CM3). By eye, the amplitude of inter-annual variability in the observations sits between the extremes from the simulations, perhaps towards the lower end, but separating out the variability from the trend is non-trivial.
What should we make of this? Much of the difference in simulated inter-annual variability is likely due to different ENSO variability characteristics, but the decadal variability is likely to reside elsewhere, perhaps in the Atlantic or extra-tropical Pacific Oceans? Maybe that will be next to look at.