How Accurate Are Climate Models

By Paul Homewood

 

Back In March, I ran a post on claims in the Guardian that global warming predictions made by climate models had proved accurate.

There were a number of holes in their analysis, which can be found here. Their story, however, was based on some work done by Myles Allen.

The reason I am returning to this topic is that it has been quoted by the Department of Energy & Climate Change, or DECC, in reply to my letter to my MP.

 

[ To recap, I wrote a letter to my local MP in May, posing several questions and asking for her justification of the UK Climate Change Act. The letter is here. She in turn passed it onto DECC, for their comments]

 

Suffice to say that I have just received DECC’s reply, and will be posting on it in full in the next day or so. However, in response to my specific complaint that public policy is being based on climate model predictions, that have so far proved woefully inaccurate, DECC replied:-

 

However, a recent study in Nature notes that climate models appear to have useful skill at predicting global temperature change on decadal timescales. This study compares recent observations of global temperature with a forecast of the decadal mean temperature from 2001-10 made using the HADCM2 climate model.

  

This study, of course, turns out to be very same Myles Allen study that the Guardian had reported on, so I would take a further look at it.

 

Allen had originally run his model predictions in 1999, and forecast that 2001-10 temperatures would be about a quarter of a degree higher than the baseline of Sep 1986 – Aug 1996. When he came back and compared his numbers with the actual temperatures, he can up with a pretty close fit. Indeed, HADCRUT4 shows an increase of 0.27C.

As I pointed out at the time, though, much of this increase had actually occurred between 1986 and 1996,  and by 1999, when the original paper was published, temperatures had already exceeded his forecast of an increase of 0.25C. There was also the issue of Pinatubo, which had depressed temperatures during the baseline period. My full review is here

 

However, there are two specific points that I want to expand on today.

 

1) In his recent review of the original forecast, Allen states:-

“This forecast provided support for the 2001 IPCC report 10 , which stated that “anthropogenic warming is likely to lie in the range 0.1–0.2 °C per decade over the next few decades”.

 

This forecast, of course, lies well below most other predictions. In their letter to me, DECC state categorically:-

“However, the weight of evidence continues to indicate that climate sensitivity is very likely between 2C and 4.5C, with a best estimate of 3C”.

So why do they quote a much lower prediction from Myles Allen?

 

2) Allen also states:-

“Even if temperatures for the decade 2007–2016 remain no higher than those for the decade 2002–2011, the 1999 forecast would still not be falsified at the 10% confidence level. However, it would no longer be substantially better than the random walk. “

  

In other words, although the predictions would just about hang on at the bottom of the range of uncertainty, they would not be significantly more meaningful than the random set. (The range of uncertainty by 2010, is about 0.40C!)

So how do the latest numbers compare with 2002-11?

  

Temperature Anomalies – Centigrade

	HADCRUT4	RSS
2002-11	0.481	0.261
Current 10-Year Average	0.474	0.240
Last 12 Month Average	0.471	0.254

 

 So, both current temperatures and 10-Year averages are actually slightly lower than 2002-11. With just three years to go, it seems unlikely that the 2007-16 will alter this position drastically.

 

So it appears increasingly likely that even this model, which is very much at the lower end of the range of predictions, will soon be invalidated.

I don’t know what planet our politicians are living on, but it certainly is not this one.

 

 

References

Test of a decadal climate forecast – Myles Allen

http://www.inference.phy.cam.ac.uk/sustainable/refs/climate/ngeo1788.pdf