Skip to content

Some weeks ago, I invited readers to improve upon parts of a summary of global warming science, written by Julia Slingo for the benefit of readers in central government. The ground covered was mainly about surface temperatures. At some point I may well write this up into something more formal.

I think it would be interesting to also say something about climate models and their uncertainties and I have been giving this some thought. My knowledge of climate models is somewhat sketchy, so some of my understanding may be incorrect, but here’s the ground I think central government really ought to understand:

1. Climate models are based on well-understood physical laws. There is wide agreement that on its own a doubling of CO2 levels would produce an initial warming of around 1degC.

2. However, the knock-on effects of this initial warming (“the feedbacks”) are not well understood, particularly the role of clouds.

3. Most climate models suggest a warming of 2-6degC/century. It is not clear that this range actually covers the full envelope of possilities because of uncertainties over the feedbacks.

4. The temperature predictions of climate models cannot be tested in the short-to-medium term; 30 years is required to properly assess their performance.

5. However, climate modellers derive comfort that their models are reasonable approximations of the climate systems from a number of observations:

  • their models generally replicate the Earth’s temperature history (“hindcasts”), although it should be noted that even models ancapsulating very different sensitivities to CO2 can do this, demonstrating that the models are fudged.
  • some models spontaneously reproduce features of the real climate, such as the PDO and El Nino, although not well enough to make such models useful predictive tools.

6. However

  • when the detail of the “hindcasts” is examined, it is found that the models do not in general replicate regional temperature histories.
  • to the extent that models have had their predictions tested against outcome, their performance has been poor.
  • no model has been shown to be skillful at regional forecasts.

Have I got anything wrong? Have I missed anything? I also wonder if politicians actually need to know about feedbacks and physics and sciencey stuff like that. Don’t they just need to know how good the models are?