Friday, November 25, 2011

A Lower Climate Sensitivity

The paper published this week in Science that finds a smaller value for climate sensitivity (2.3°C vs 3.0°C) relies on a climate model, which we know don't predict past temperatures and must be fine-tuned to keep from crashing. It neglects several crucial factors, such as dust and cloud albedo. The paleo dataset covers barely a quarter (26%) of the Earth's surface. And what about solar variability? Modelers have a saying: "All models are wrong, but some are useful." They've said it themselves: all models are wrong. Garbage in, garbage out. And why did Science wait until the week before the Durban climate conference to publish this paper? Who got to the editors? I've been told the CEI will be filing a FOIA on Monday for their emails....

Oh, wait. For a minute there I thought I was guest blogging the inverse-paper at Watts Up With Climate Depot?

Seriously though, this paper is good news, but not as good as you'd want. It shows that we will probably avoid the higher end range of temperature change, and have more time to address the CO2 problem and get it right. But not a lot more time.

I threw together a little spreadsheet (we all know the best science is done on a Friday night with a spreadsheet) that simply assumes atmospheric CO2 levels keep increasing the way they have been: at a yearly rate that was about 0.30%/yr in 1970 and is 0.55%/yr now, and I just linearly interpolated that into the future. Remembering to carry the 4 and avoid using the Comic Sans font, I got the following for the number of years warming delayed with a climate sensitivity (S) of 2.3 K vs 3.0 K:

 In other words, the warming that would have occurred in 2100 under a S=3.0 K scenario would occur in 2115 under a 2.3 K scenario. Etc.

The lead author, Andreas Schmittner (who is at Oregon State University) told the Oregonian:
"I think we should be worried, but we should not be desperate," said Andreas Schmittner, an Oregon State University researcher and lead author of the study, published online today by the journal Science. "It's not already too late to do something. We still have time to figure out a solution."
And, of course, this paper has no consequences for the problem of ocean debaseification (or what regular people call ocean acidification).

However, in an accompanying Perspective in Science, Gabriele Hegerl and Tom Russon don't sound especially convinced:
The work of Schmittner et al. demonstrates that climates of the past can provide potentially powerful information to reduce uncertainty in future climate predictions and evaluate the likelihood of climate change that is larger than captured in present models. However, given the remaining uncertainties, which the authors discuss in their paper, a firm upper boundary is still elusive. The study also shows that to take advantage of the opportunity offered by past climates for understanding future climate change requires not just high quality data but also appropriate physical climate models and statistical modeling. This is not an easy challenge, but it is an important one.


Anonymous said...

I am astounded at the level of coverage this single paper is receiving and the credulity being afforded its results, by people and media outlets that should know better than to take a single study as uncontroversial.

Speaking of the "inverse" to this paper, a few months ago, this paper was published in a very reputable climate journal, studied the same period in question (the LGM) to estimate climate sensitivity, and found ECS to be higher than the mainstream estimate by roughly the same emount that Schmittner et al. found it to be lower.

Did the BBC cover the Holden et al. paper? Did you?

I certainly didn't.

Why? Because the best estimate for ECS as ~3°C is constrained by a mountain of evidence, and a paper just a little to either side doesn't really change the peak distribution.

Yet the Schmittner et al. paper is getting international coverage, and oddly uncritical responses from some of the ostensibly best science and climate writers. Odd...

Anonymous said...

Your analysis of the implications of changing climate sensitivity from 3.0 to 2.3 is grossly wrong. Ignoring minor differences between transient and equilibrium climate sensitivity, a 25% reduction in climate sensitivity represents roughly a 25% reduction in future warming at any given time.

Obscuring (intentionally or accidentally?) this basic truth about climate sensitivity, you have tried to translate this temperature reduction into a delay in warming. Such a strategy was effectively and accurately used against the Kyoto Protocol. Unfortunately, you mistakenly assume that the annual increase in CO2 will continue to grow linearly from 0.30%/year in 1970 and 0.55%/year now (2010), eventually reaching 1.11%/yr in 2100 and 1.86%/yr in 2200. A constant % increase per year affords simple exponential growth; a linearly increasing % increase per year affords "exponential" growth with time squared in the exponent. All scientists should be aware the dangers inherent in extrapolating a purely empirical trend observed over 40 years through two centuries.

The IPCC has created emissions scenarios so that we can make realist projections about CO2 in the atmosphere. None of these scenarios show an exponential increase in CO2 emissions through even the 21st century - all scenarios show a lower growth rate in the second half of the century and most show constant or DECREASING emissions. The increase will NEVER reach 1.1% per year in 2100. This is because global population is expect to peak in 2050 and probably decline thereafter.

Furthermore, the IPCC's emissions scenarios may not include realistic expectations about how much fossil fuel can be economically extracted from the ground through even 2100. If we are not at peak oil production right now, we could be soon. Your graph - which runs to 2200 - almost certainly assumes that we burn far more fossil fuel than realistically be extracted from the planet.

Emissions are only part of the problem; accumulation of CO2 in the atmosphere depends on how much CO2 is absorbed by various "sinks" (a phenomena that currently negates about 50% of current emissions, but which is poorly understood). The IPCC projections of CO2 accumulation in the atmosphere all show that the increasing rate of accumulation (which can be approximated with some sort of exponential curve) ends around 2050 and becomes LINEAR OR DECREASES (to zero in the most optimistic cases). There is absolutely no justification for proposing any type of exponential increase in CO2 through 2100. If you would simply plot your projected CO2 levels vs the IPCC's, you'd see how unrealistic your calculation is.

Mathematically, your model proposes:

dCO2/dt = 0.00625%/yr2 * CO2 * t Eq 1

where t is the number of years after 1922, the year when your linearly extrapolated growth rate would be zero. In 1970 (t=48) the growth rate according to Eq 1 is 0.3%/yr * CO2 and in 2010 is 0.55% * CO2. Separating variables gives:

dCO2/CO2 = 0.00625% * t

and integrating gives:

ln[CO2(t)/CO2(1922)] = 0.003125% * t^2
CO2(t)/CO2(1922) = exp(0.003125% * t^2)