There’s a very interesting new paper by Gregory and Andrews called Variation in climate sensitivity and feedback parameters during the historical period. The motivation seems to be to try and reconcile why estimates of the effective climate sensitivity using recent observations seems to suggest climate sensitivities that are somewhat lower than climate model estimates of the Equilibrium Climate Sensitivity (ECS). For simplicity, I’ll use ECS for both effective climate sensitivity and Equilibrium Climate Sensitivity, but – technically – they aren’t quite the same thing.
A standard way to assess ECS using observations is to use a simple energy balance model in which
where is the change in forcing over the time interval considered, is the change in system heat uptake rate, is the change in temperature, and is the feedback parameter. The assumption here is that is constant and – if one wants to equate effective and equilibrium climate sensitivity – that it will remain constant. The ECS is then simply
What this paper did was to use Atmospheric General Circulation Models (AGCM) and prescribe[d] time-varying observationally derived fields of [sea surface temperature] SST and sea-ice concentration, but fixed atmospheric concentrations at pre-industrial levels. In such a model, the change in forcing is 0, but the change in temperature, , will not only be a reasonable representation of the actual change (not quite, because the land is not warming as fast, but close – 85% to 95%), but the spatial distribution of the warming will also be similar to what actually happened. To determine the climate feedback parameter, you then use
To estimate , you simply regress against . As shown in the top left panel of the above figure, the result for the whole period is 1.72 Wm-2K-1 (ECS = 2.2K), where and are relative to 1979-2008. The bottom left panel, however, shows what happens if you consider shorter time intervals, and indicates that it can vary quite substantially, depending on the time interval considered. The right-hand panels show determined by regression in 30-year sliding windows. They also show during the first 20 and first 100 years of abrupt 4xCO2 simulations and show that it is smaller (and the ECS larger) than is the case when the spatial distribution of the surface warming is chosen to represent what we probably experienced over the last 100 years or so.
As the paper itself says
Our results suggest that the differential climate feedback parameter varied on multi-decade timescales during the historical period and that it was generally larger than abrupt4xCO2, in particular during the last three decades.
The reason being that essentially
[d]ifferent geographical pattern of SST that produce the same global-mean can give different .
There seem to be a number of conclusions that one can draw from this study.
- Climate models that have a reasonable representation of the pattern of surface warming do a reasonable job of estimating the resulting feedback response.
- The reason why observationally-based estimates for ECS tend to suggest lower values than other estimates (such as climate models) may well be simply because of the spatial distribution of surface warming that we have actually experienced, rather than because our climate is actually less sensitive than these other estimates suggest.
- Projections for future warming will likely be reasonable as long as the pattern of surface warming in climate models is a reasonable representation of what we will likely experience under increasing anthropogenic forcings.