Tuesday, December 6, 2011
There's a very interesting new paper out from Grant Foster (who blogs at Open Mind) and Stefan Rahmstorf (who occasionally blogs at Real Climate). What the paper does is take five global temperature series that look like this:
Basically there is a clearly linear trend (the global warming signal) with a much smaller amount of noise left over that is a) fairly correlated between the different sources, and b) not very autocorrelated - ie it looks like the annual fluctuations in the weather.
I won't say too much about their analysis since it's well covered at Open Mind. What I was curious about is the extrapolation, since the data seem to cluster so tightly about a linear trend. I read off the average of the five adjusted series above and then did a simple linear least squares which gave the overall slope as 1.63 ± 0.09 C/century. This statistical error of about 5% is much much smaller than the uncertainties from climate models. Of course, the linear nature of the data is purely an empirical observation that could break down in the future (whereas the climate models attempt to capture the underlying physics of the global climate system).
In fahrenheit terms this is 2.94 ± 0.15 F/century. It's important to note that this is a global average - the rate will be lowest over the oceans and at low latitudes and higher over land and in the Arctic.
To give a feel for a different stability analysis - here's a comparison of the linear and quadratic extrapolations over the rest of the century:
Over that timeframe the uncertainty in extrapolation is dominated by the small non-linear component present in the data. Still - the uncertainty is not that large and will be rapidly reduced by another decade or two of data. So this gives us an idea of how much global warming we - and the planet we love - will have to adapt to.
It's worth adding the caveats that 90 years is a long time - if indeed solar power becomes cheaper than coal in a couple of decades time, or we manage to invent Freeman Dyson's carbon-eating trees, then the situation could be much better. On the other hand, since we are making fairly large changes in a system that we don't understand very well, there could well be thresholds in the climate system above which it starts to change much faster (eg a lot of the forests burning up and releasing their carbon into the atmosphere). Any of those kinds of things could render a straightforward extrapolation of the last thirty years pretty pointless.