I thought I might add a new chapter to my series, which I’ve called helpful tips for the Global Warming Policy Foundation. The first was a quick science lesson for Lord Lawson. The second was the cheerfully titled Come on Andrew, you can get this. My new installment is an attempt to explain, to Matt Ridley, the significance of an uncertainty interval, something that someone with a science PhD should understand, but appears not to. I actually get paid to teach this kind stuff, so you’d think they might appreciate the free and friendly advice. I get the impression, though, that they don’t 🙂
It relates to a recent article that Matt has written for the Wall Street Journal, in which he says
As a “lukewarmer,” I’ve long thought that man-made carbon-dioxide emissions will raise global temperatures, but that this effect will not be amplified much by feedbacks from extra water vapor and clouds, so the world will probably be only a bit more than one degree Celsius warmer in 2100 than today.
When I asked Matt on what basis he was making this claim, he directed me to the GWPF report OverSensitive, written by Nic Lewis and Marcel Crok, and – in particular – to Table 3 (below).
So, here’s the basic problem, the values that Matt appears to think justify his claim are based on a single value for the Transient Climate Response (TCR). In fact, this appears to be based on the work of Otto et al. which uses observations (plus forcings from models) to estimate the TCR, and concludes that it has a 5 – 95% range of between 1 and 2 degrees (with a best estimate of 1.35 degrees). He also argues that the RCP8.5 emission pathway is completely unrealistic and so should be ignored. Therefore he’s concluding that the worst case scenario is RCP6.0 with a probable TCR of 1.35 degrees. However, the point is Matt, that you can’t just pick a single number; you should really consider the range.
I thought I would illustrate this using a basic one-box model
where is the change in anthropogenic forcing, is the heat capacity of the system, and is the climate sensitivity. I determined values that resulted in TCR values of 1, 1.35, and 2 degrees (in which the forcing was assumed to change because of a 1% per year rise in CO2). The main reason for this TCR range is uncertainties in the aerosol forcing, so I then adjusted the anthropogenic forcing from the RCP11 dataset so that my model results over the period 1880 – 2010 roughly matched the instrumental temperature record. I then extended the forcing dataset to 2100 along an RCP6.0 pathway (i.e., reaching a change in anthropogenic forcing of 6 Wm-2 by 2100). The basic result is in the figure below.
So, indeed, if the TCR is 1.35 degrees it would be a bit over 1 degree warmer than today in 2100 (the y-axis in the figure is relative to 1880, so take away about 0.8 degrees to get relative to today). However, the work that Matt is basing his views on suggests that the TCR is as likely to be above 1.35 degrees as below. Hence the warming is as likely to be higher than suggested by Matt as it is to be lower. Also, there is a non-negligible chance that it could be as much as 2 degrees higher than today in 2100. I should add that this is more like a 95% range, rather than the 66% range presented by the IPCC. Anyway, the point is, Matt, that this is roughly what everyone was getting at on Twitter; basing your estimate for future warming on a single TCR value from a single study is not very scientific. Some might call it a cherry-pick.
There are also some additional points, that scientists should really be willing to acknowledge. The range for the TCR from the study used by Matt (1 – 2 degrees) is lower than the IPCC estimate (1 – 2.5 degrees). However, the study used by Matt is an observationally-based approach that suffers from a number of possible issues. In addition to the uncertainty in the aerosol forcing, it’s also sensitive to variability in the surface temperature, cannot capture possible non-linearities in the feedback response, and cannot easily compensate for inhomogeneities in the forcings. This doesn’t mean that it’s wrong, but this is evidence that it might be underestimating the TCR. So, not only could we warm more than Matt suggests (using exactly the same evidence as he’s using), there’s additional evidence suggesting that even this could be an underestimate.
So, I hope this helps Matt to understand what people where getting at when they were questioning him on Twitter. I should add that I did this all rather quickly, so I’m not claiming that the range I’ve got for the warming by 2100 is exactly right, but I think it’s reasonable. This was just meant to be an illustration, rather than being exact. Of course, if anyone has any thoughts or corrections, feel free to make them through the comments.