David Rose Again

An Eccentric Anomaly: Ed Davies's Blog

David Rose is at it again with this article in the Mail on Sunday. As with last time lots of people have expressed their displeasure at various points. To find this commentary a good place to start is the update to and comments on the post from which the graph at the top of the article was originally taken (here) but replaced with Comparing global temperature observations and simulations, again by Ed Hawkins on the Climate Lab Book blog.

I just want to focus on an answer to one particular point. Rose writes:

But when the latest official global temperature figures from the Met Office are placed over the predictions, they show how wrong the estimates have been, to the point of falling out of the ‘95 per cent’ band completely.

He doesn't make explicit what the implications of this supposedly dramatic falling out would be. Guessing, I'd suggest that he hopes people will interpret any excursion as meaning there's a 95% probability that the models are wrong or something like that. If that is his intention then it would be ridiculously misleading. At the very least we should expect him to take more care not to be misinterpreted in this way.

Rather trivially, it's not a 95% band; it covers 5% to 95% so it's a 90% band.

More substantively, the bands reflect the estimated probability distributions for individual years. As such, we'd expect that over a long-enough run of years one in 10 would be outside the 90% band: 1 in 20 below and 1 in 20 above.

It's difficult to see on either version of this graph but it looks to me like the observations dropped out of the bottom of the 90% band for a few years in the mid 1970s and haven't done so since. Similarly, the observations touch the top of the band in a couple of years but don't seem to have gone out much or at all. Therefore, it would not be at all surprising to see a few years out of the band in the near future. In a way it's almost overdue - though thinking like that can result in the sort of awful statistical muddles gamblers get into.

Note also that the dark-red band in Rose's version of the graph is a 25% to 75% band so we should expect about 50% of the actual years to lie outside it. Looking at fat lines on small graphs is difficult but, again, it seems reasonable to me that we should have a couple of years now outside that band one way or the other. It's just unfortunate from a long-term policy perspective that it happens to be on the low side.

[[ Update: 2013-04-04: This National Review article, The New Climate Deniers, makes the same “mistake“ except that it spells out the intended conclusion. Referencing Ed Hawkins' work it says:

He has found that if global temperatures stay the same for a few more years, they will fall below the range of 20 climate models. In other words, the scientific “consensus” will have been proven wrong.

Not impressed. ]]