Sunday, September 19, 2010

Between the science and a hard place, Part Two: Confidence and Risk

As you sit in what is hopefully a comfy chair, blissfully consuming this fine, robustly flavored, 100% complimentary content, have a moment's sympathy for the lot of the climate blogger. Many are his (or her) woes.

Fer instance, I want to talk to you today about the Scylla and Charybdis of lukewarmism, confidence intervals and risk. I would normally pursue this much as I did in the last installment, by plugging in some lukewarmist assumptions and discussing the results. But the trouble with this, as we saw in Part One, is that even the most conservative lukewarmist predictions imply equilibrium warming greater than 2 degrees C – a condition not seen for millions of years, which climate scientists predict will lead to disaster. In order to talk about the problems of confidence intervals and risk, though, we need a projection that doesn't definitely lead to disaster.

My solution is to pull from the AR4 absolutely the lowest-warming model, based on the lower bound of the 95% confidence interval for the lowest-emission scenario family, B1. This family is described as follows:

The B1 scenarios are of a world more integrated, and more ecologically friendly. The B1 scenarios are characterized by:
• Rapid economic growth as in A1, but with rapid changes towards a service and information economy.
• Population rising to 9 billion in 2050 and then declining as in A1.
• Reductions in material intensity and the introduction of clean and resource efficient technologies.
• An emphasis on global solutions to economic, social and environmental stability.

This is a totally unjustifiable procedure on a couple of levels. Obviously at the moment we are not making anything that could be described as rapid progress towards an "integrated . . . ecologically friendly world." Put that to one side. You also cannot simply take the lowest bound of the lowest temperature rise scenario and call that your estimate; that's cherry-picking of the worst kind. But this is the only way to get an estimate this low, so . . . I'm going to have to ask you to play along.

So say I'm a lukewarmer, and I think we're in for about 1.1C in warming over the next century. Add that to the 0.7C seen since the start of the industrial revolution, and you end up with 1.8C of warming – a princely 0.2C below our maximum safe-ish warming of 2C. Provide we do not care anything about what happens after 2100, as for example, in 2120 or 2130, we're golden.

Having ascended to these lofty height of assumption, we can finally deliver the coup de grace (the stress is causing me to mix my metaphors – who would be dealing out mercy killings on a mountaintop?) We have our lukewarmer, and they have their 1.1C by 2100. First question: What is your level of confidence in that projection? Let's look at some real-life estimates of climate sensitivity, based on actual data:

Royer, et al. (2007)[24] determined climate sensitivity within a major part of the Phanerozoic. The range of values—1.5 °C minimum, 2.8 °C best estimate, and 6.2 °C maximum—is, given various uncertainties, consistent with sensitivities of current climate models and with other determinations. - cite_note-24

Annan and Hargreaves (2006)[22] presented an estimate that resulted from combining prior estimates based on analyses of paleoclimate, responses to volcanic eruptions, and the temperature change in response to forcings over the twentieth century. They also introduced a triad notation (L, C, H) to convey the probability distribution function (pdf) of the sensitivity, where the central value C indicates the maximum likelihood estimate in degrees Celsius and the outer values L and H represent the limits of the 95% confidence interval for a pdf, or 95% of the area under the curve for a likelihood function. In this notation their estimate of sensitivity was (1.7, 2.9, 4.9)°C.

Based on analysis of uncertainties in total forcing, in Antarctic cooling, and in the ratio of global to Antarctic cooling of the last glacial maximum relative to the present, Ganopolski and Schneider von Deimling (2008) infer a range of 1.3 to 6.8 °C for climate sensitivity determined by this approach.[12]

Andronova and Schlesinger (2001) found that the climate sensitivity could lie between 1 and 10°C, with a 54 percent likelihood that it lies outside the IPCC range. - cite_note-15

The first thing you likely noticed about these calculations is that to continue to test lukewarmism we have really, really underestimated the magnitude of the warming expected in the 21st century. But the second thing you should notice is that the ranges -- the confidence intervals -- are really large. Climate sensitivity is a fiendishly difficult thing to pin down with precision, and most honest scientists come up with a range of 4-9 degrees C for the sensitivity that may be implied by any given data set or model. Those assimilating the results of many experiments and model runs can do substantially better, with recent estimates of 2.6-4.1C, a range of only 1.5C, or 2.6C if we reach down and include the unlikely possible of a sensitivity of 1.5C.

So the critical question then becomes: what is the lukewarmers' range? Consensus scientists estimate climate sensitivity at about 3C, but concede that it might be 1.5C, 4.5C or even higher (and very unlikely to be much lower). What range do lukewarmers think is plausible?

So far, to my knowledge, no self-identified lukewarmer has been persuaded to answer this question. They will find it difficult. Because they have positioned themselves as participants in the scientific debate, they can hardly claim 100% confidence in X climate sensitivity, no error bars. If they are reasonable, they have to accept they they are as fallible as the rest of the scientific community, and although they think the climate sensitivity is 1.5C (say) it might be 1.0C, or 2.0C, or even (gasp!) 3.0C (where the consensus puts it).

So even if they see 1.5C as the most likely number, they have to concede the possibility the number may be higher or lower. Emissions, too, may be higher or lower -- and so could the 2C estimate for disastrous climate change, which is just an estimate, be low or, what is more likely, high.

And this leads us to the problem of risk. What amount of risk will we accept in the next century that our civilization will be severely affected by global warming? Five percent? One percent? A tenth of one percent?

Put it another way: rapid cuts in emissions are estimated to cost between 1-3% of the GDP. Suppose there is a revolver to your head with a thousand cylinders (it's a big revolver). For a 3% raise, how many chambers are you willing to have loaded before your boss pulls the trigger? One? Maybe if you're hard up. One in a thousand's not a huge chance to take. Ten? Probably not, if you have anything to live for. A hundred? Never in a million years.

People differ in their appetites for risk, but I think most people would agree that a 1% risk of major climate disruptions like multi-meter sea level rise, massive droughts, mass extinctions, and millions of climate refugees, is around the upper limit of acceptable. And that's the end of lukewarmism (again). Any remotely reasonable estimate of climate sensitivity, even if centered on 1.5C or even lower, even if it gives you a central estimate of less than 2C, will carry with it the very significant chance that the real value is 2C or 2.5C or 3C. Traveling with it will be uncertainties about the rate of growth of emissions, and the possibility that less than 2C will get us to planetary disruptions like the rapid melting of most of the ice sheets or large-scale methane release.

Once you've acknowledged the greenhouse warming caused by CO2 and other greenhouse gases, even a ludicrously low estimate of climate sensitivity will not save you from the iron logic of risk assessment: "maybe not" and even "probably not" are unacceptable for the kind of impacts we're talking about. Even 1% is too high. But, absent a new data set allowing a much, much more exactly calculation of climate sensitivity than we have been able to provide to date, there is no way even the most Pollyanna estimates of climate sensitivity and future emissions can provide any acceptable level of assurance that "business as usual" is anything but a road to ruin.


  1. Thanks for this. Nicely argued. Many of us who now believe that action is required were lukewarmers once.

  2. Thanks! I was never a lukewarmer per se, but in about 1990 I was a "Where's the proof?" person.

    I think an important point is that there's nothing especially heretical about saying maybe temperatures will rise more slowly than most scientists think. The correct response to that, in my thinking is "Sure, maybe. I guess we'll see. Even if you're right, though, the risks are not substantially different."

    It's like two muggers with a gun. One has a full clip, the other one has seven bullets. Which one are you safe with?

  3. I would like to see the assessment of the climate system sometimes more disaggregated from the issue of what to do about the changes. I fear that the discussion gets muddled. Also that there is a tendancy for deniers to argue the science, when they don't like the change. And alarmists, the reverse.

    I think if we are debating the science, it is better to just debate the science. I mean there is some physics going on. And at the end of the day, the molecules and the system are not Republicans or Democrats. Someone is right or wrong. Hard to tell and argue, since the system is so complex. But there IS an objective reality.

    Thus I would like to discuss things in a completely clinical and disinterested sense. Like someone talking about how an atom bomb works who doesn't give a rat;s ASS if it is used for good or evil. But just wants to calculate megatonnage as exactly as possible and to solve a debate on how the bomb works.

    Or at least who can sort of REALLY segregate his thought processes.

    I just think maybe well under 50% can do this. HAve that amount of scientific dispassion. At least of the people we see in coimment threadds.

    In contracst, most of the VC commentators are capable of doing this. But it is damned rare.

  4. "I think an important point is that there's nothing especially heretical about saying maybe temperatures will rise more slowly than most scientists think. The correct response to that, in my thinking is "Sure, maybe. I guess we'll see. Even if you're right, though, the risks are not substantially different."

    It's like two muggers with a gun. One has a full clip, the other one has seven bullets. Which one are you safe with?"

    Exactly. Precisely.

    "I think if we are debating the science, it is better to just debate the science."

    But what we are really debating is the policy. The science is being debated on an ongoing basis in the peer-reviewed literature. There is lots of uncertainty about what exactly the coming decades will look like, but very little uncertainty about the fact that business as usual will both cause expensive definite harms as well as significant risks of devastating harms. And that conclusion is relatively insensitive to what climate sensitivity or emissions scenario turns out to be correct.

    So let's argue the policy.

  5. I keep coming back to this stuff. I must have read it at least 5 times. It's that good. Thanks.

  6. talk to a personSeptember 14, 2021 at 1:49 AM talk to a person will provide you with all the information you need to contact Microsoft Support.