One of the ways of distinguishing between good science, and poor quality science and pseudo-science, is that good science leads to predictions that can be tested by experiments. Poor quality and pseudo-science does not. For example, we know there will be an eclipse of the Sun visible from parts of Europe on March 20th 2015. We can calculate the orbit of the Moon very accurately and so predict that the Moon will be between the Earth and the Sun then. This prediction is testable: Just go to the right part of Europe, and see if it goes dark.

Of course we have been calculating the orbit of the moon very accurately for a long time, so we know it will go dark exactly when and where astronomers predict it will. Our climate is more complex than the orbit of the moon, and climate science is controversial due to various climate-change deniers deciding they don’t want to believe the evidence that our planet is warming.

As our climate, and any changes in it, are so important, we need to test climate science, to distinguish good from bad, reliable from unreliable. A nice example of this is discussed in a post over at RealClimate.org. The post discusses a 1981 paper by Hansen et al. The paper is now 31 years old and made predictions for the global temperature for the 40 years from then to 2020.

So, as we now know the global temperatures from 1981 to 2011, we can test the predictions made in this paper. Over the 30 years their estimated range of temperatures is a little on the low side, i.e., they predicted global warming but underestimated its magnitude a little. To me this looks like a pretty good job by Hansen et al.

The fact that the less sophisticated (and using simple models on much slower computers) climate models of 1981 do so well is reassuring. Its give us confidence that modern predictions are very likely to be accurate.