FAIL (the browser should render some flash content, not this).

How Sensitive are Atmosphere-Ocean GCMs: Their Response to a Radiative "Push"

Olivé, D.J.L., Peters, G.P. and Saint-Martin, D. 2012. Atmosphere Response Time Scales Estimated from AOGCM Experiments. Journal of Climate 25: 7956-7972.
The Earth's climate is a complicated dynamic system in which each part has characteristic time-scales over which they evolve internally, or as they react to forcing external to that part of the system. The same would apply for the climate system as a whole. The atmosphere, being less massive than the ocean or the cryosphere, tends to respond more rapidly to forcing than the other parts of the system. Thus, when the climate of the atmosphere itself is modeled, we consider the atmosphere a boundary value problem. This basically means that the atmosphere will be a servant to the underlying Earth's surface.

Also, the Earth's climate can be considered to be differentially 'sensitive' to external forcing, that is some process will induce a certain temperature change per unit change in the amount of forcing. One of the continuing disagreements in the climate change debate is the differences between calculated and modeled sensitivities of the climate. Some scientists believe the climate is quite sensitive to forcing, while others believe the climate is more resilient.

Olivé et al. (2012) examine the changes in modeled global temperature to changes in CO2 concentration using two different models. This study also endeavors to quantify the uncertainties in the climate change scenarios. The authors use output from the Hadley Centre's Atmosphere - Ocean (UKMO-HadCM3) and the CNRM-CM3 Centre National de Recherches Meteorologiques Coupled Model, version 3 (CNRM-CM3) coupled Atmosphere Ocean General Circulation Models (GOGCMs). The simulations were relatively short, on the time-scale of 100-300 years, since the resolution was higher.

The authors also used CO2 scenarios where the increase was simulated to be sudden, corresponding to a factor of 6.5 times increase (10 W/m2), a sudden doubling, and then a gradual increase. The gradual increases (1% per year) resulted in a doubling and quadrupling of CO2 in 70 and 140 years, respectively. Additionally, the authors initially use a shorter (10 years) and long-scale (100 years) CO2 forcing as well as sensitivity values within the range of those given in the published literature for their a priori estimates. Then they solve the equations backward to get their estimates of the short and long time scale modes and sensitivity from the models runs.

The results from Olivé et al. showed that in their models, the short-time scale CO2 radiative forcing was on the order of 3- 4 years, and on the long time-scale of 100-300 years. The shorter mode was faster for the model simulations where the CO2 forcing was sudden (Fig. 1) rather than for the gradual CO2 increase simulations. The uncertainty was also lower for the sudden increase simulations. For the longer time-scale forcing, longer model integrations would be needed to estimate these and reduce the uncertainty. The model sensitivities were "when assuming two modes it varies between 0.49 and 0.83 K W-1 m2 or between 0.56 and 1.01 K W-1 m2" for the two different models, which were both within the published range shown early in the paper.

Figure 1. Adapted from Fig. 1 of Olivé et al. (2012). A time series of global- and annual-mean surface air temperature obtained with a one (red line) and two (blue line) mode approach for (top) UKMO-HadCM3 and (bottom) CNRM-CM3. The model data are represented by the symbols, and the modal results obtained using a priori (a posteriori) parameter values are represented by the dashed (solid) lines.

This kind of diagnostic work is a legitimate use of long term computer model simulations. Even though there is no predictive work being done in this study, it is remarkable that there is such variation in the estimates of the short and long-term radiative forcing as well as the climate sensitivity. This is especially true between the two models used. This demonstrates that the models used are less than perfect, as any of these model-to-model differences would have to be the result of the model physics used, especially that relating to the radiative forcing. Also, such long integrations would lead to the build-up of numerical error. Lastly, the fact that the authors could not appreciably reduce the uncertainty in estimating these quantities indicates that further improvement in model performance is not likely using the current technology.

Archived 28 November 2012