Tuesday, January 20, 2009

Some time ago I was asking the question: ‘How good are these climate models? What sort of predictive value have they shown in modelling future climate? After all, we’ve been doing them for a few decades now.’
A nice person on realclimate.org (there are some, not all of them treat people who disagree with them as the demonised other) directed me to a classic paper by Hansen et al.
"If you want an indication of how well these models do you can go get (J. Geo Res. 93 (1988) 9341) the Hansen GCM paper that people talk about, and compare their results with observed patterns of warming and other things."
Here is the plot from that paper showing the response of overall global temperature (which the authors argue convincingly is a much better parameter than any subset of the data, e.g., whether it snowed at my house or not in a given year) for three different scenarios- A being continued exponential growth, B being a more subdued form of business as usual, and C if drastic cuts are implemented starting a few years ago.

I went and got the Hadcrut3 data set and plotted it on top of this one, as near as I was able, and got this.

There are other data sets out there. I shall plot some of the others and put them up for you.

The Hansen et al. model predicts the greatest degree of warming at high latitudes, fitting observations, but the model also reproduces another feature of observed weather, that those latitudes have the highest natural variability from one year to another.

Update 2012:
Here is another three years of data. I do realise I haven't plotted any of the other data sets. Bad me. The red points are the average of 13 monthly data points averaged on each month, while the blue points are the actual Hadcrut3 monthly global averages you can download yourself.