More “skillful predictions” in climate science

In a previous post I drew attention to a 2013 paper by a team of climate scientists claiming “forecast skill” in decadal predictions, following a paper predicting 0.3C of warming from 2004-14, which of course turned out to be completely wrong.

Three new papers

Several recent climate science papers have referred to “predictions” and their “skill”.  In fact, these are hindcasts, not predictions, in what appear to be more examples of climate scientists fooling themselves.

Decadal climate predictions for the period 1901-2010 with a coupled climate model is in Geophysical Research Letters. The climate models in this paper fit quite well for the Atlantic, but not well in the Pacific, which seems a bit like saying that predicting a coin will come down as heads works well some of the time.

Skilful Long Range Prediction of European and North American Winters has 22 authors and is in the same journal.  In a double you-couldn’t-make-it-up, (a) the authors are from the Met Office – the people who predicted the “barbecue summer” and who thought that the 2013/14 winter would be drier than usual, and (b) some of the authors are the same as those who responsible for the completely wrong global forecast mentioned above. The claim that the Met Office will now be able to do accurate long-range winter forecasts was enthusiastically reported by the gullible media, though commenters were less impressed, noting that the Telegraph article was written on April 1.

Another example is this post by Michael Tsamados at the Climate Lab Book blog. This refers to a paper in Nature claiming that the September Arctic sea ice minimum can be “predicted”  with “skill” by looking at melt ponds in Spring (May). Confusingly, some results are called forecasts and some are called hindcasts – in reality they are all, of course, hindcasts.

It’s interesting that none of these papers make any real predictions. Perhaps they have learnt lessons from Smith et al 2007. Tsamados tells me that his group will be making an Arctic ice prediction in early June.

Arctic ice prediction review

Here is a graph showing the Arctic ice extent predictions made last June, by climate scientists and others, with the actual outcome marked as a dashed red line. (Source)

The lowest and least accurate prediction, 3.4 million square kilometres, was from the Met Office (remember, these are the people who keep claiming that their predictions are skillful). Most of the other climate scientist groups made predictions of around 4.1. The sceptical blog Watts Up With That predicted 4.8, and the actual result was about 5.3 (of course, in the previous year, 2012, the minimum ice extent was very low and all the estimates were wrong in the other direction). This chart does not include the predictions of more extreme climate scientists such as Paul Beckwith, who predicted that the ice would disappear completely in 2013, or Marshall Shepherd, president of the American Meteorological Society, who gave a talk repeating this prediction (“Some scientists project that by the end of this summer the Arctic in the warm season will be totally ice-free”), misleading his TED audience.

Meanwhile, retraction watch tells us that another paper claiming high forecast skill in the prediction of European winters based on Arctic ice extent has been retracted (although the fact that it has been retracted is not immediately obvious from the paper at Nature).


2 thoughts on “More “skillful predictions” in climate science

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s