We have a new paper out in Nature Climate Change on potential changes in natural unforced variability of global mean surface air temperature (GMST) under global warming.
Paper
News and Views piece
Duke press release
Unforced GMST variability is of the same order of magnitude as current externally forced changes in GMST on decadal timescales. Thus, understanding the precise magnitude and physical mechanisms responsible for unforced GMST variability is relevant for both the attribution of past climate changes to human causes as well to the prediction of climate change on policy relevant timescales.
Much research on unforced GMST variability has used modeling experiments run under “preindustrial control” conditions or has used observed/reconstructed GMST variability associated with cooler past climates to draw conclusions for contemporary or future GMST variability. These studies can implicitly assume that the characteristics of GMST variability will remain the same as the climate warms. In our research, we demonstrate in a climate model that this assumption is likely to be flawed. Not only do we show that the magnitude of GMST variability dramatically declines with warming in our experiment, we also show that the physical mechanisms responsible for such variability become fundamentally altered. These results indicate that the ubiquitous “preindustrial control” climate modeling studies may be limited in their relevance for the study of current or future climate variability.
Another principal finding of our study is that global warming may cause local temperature variability to increase over low-to-mid latitude land regions at the same time that global temperature variability dramatically decreases. This represents a cause for concern, as it is precisely these low-to-mid latitude land regions that are characterized by the highest human population density and biodiversity.
Patrick T. Brown, PhD 
I’ll try to have a look at the paper, but is there a simple way to explain the physical processes responsible for the variability and why it alters, and then changes the magnitude of the variability, in a warming world?
The main mechanism that we focus on in the paper is the sea-ice-albedo feedback on unforced global mean temperature variability. In the model that we focus on (GFDL-CM3) the sea-ice-albedo-feedback appears to greatly enhance unforced global mean temperature variability in the preindustrial control run. In an equilibrated 2XCO2 run, however, there is less sea-ice and thus less sea-ice variability. The end result is that the reduced sea-ice-albedo-feedback in the warmer climate reduces global mean temperature variability.
Thanks, that’s interesting. I hadn’t appreciated that the sea-ice-albedo feedback could play such a role even in an unforced scenario.
Does the GFDL-CM3 model do well at reproducing various 60-80 year climate variability?
Good question. It is actually much more difficult to know than one might expect. One reason is that it is notoriously difficult to separate regional decadal unforced variability from forced variability (e.g., from time-varying anthropogenic aerosol emissions) in the observational record. Another problem is the spatiotemporal coverage of observations. In the GFDL model’s preindustrial control run, its low-frequency global variability originates mostly from the Southern Ocean. However, we do not have even a century of observations from the Southern Ocean.
Another point I would make is that the results from the GFDL-CM3 model should not be taken as a forecast for precisely what we expect to happen to global temperature variability. Rather, the results demonstrate that, in principle, global temperature variability can be very sensitive to the background temperature. This seems to manifest differently in different models.