Another item in the latest Industrial Physicist is a piece on understanding the uncertainties in global warming models, and the public policy implications of those uncertainties. It's well worth a read if you care about global warming in particular or science and public policy in general.

One of the hardest things about ensuring that public policy is based on sound science is that sound science inherently involves uncertainties. Politicians like yes or no answers, but science only gives really reliable answers in the very long term, far longer than the relevant political timescales. In order to make policy based on sound science, politicians have to take uncertainty into account, and allow for the possibility that the policies may need to be adjusted as new information becomes available.

The government group I've seen handle uncertainty well is at Treasury. Not all of the government folks are afraid to deal with it. It does take a certain level of training that most politicians don't have, though.

It's almost entirely untrue that politicians like Yes or No answers.

A few moments consideration shows that they only like 'yes' answers.

While there are certain to be "fuzziness" in the predictions based upon various factors, this tends to obscure that fact the the models may be fundimentally wrong. By declaring a wide range of predicted outcomes in the output of the model, and defining this as 'normal' we are in a very unscientific region where there is little or no falsifiabilty to the model's predictions.

I don't mean to disparrage the models as they are a necessary work in progress in the climatology field, yet, we must recongnize that due to great uncertainties involved it would be very UNWISE to use the blunt instrument of intrusive public policy based upon such uncertain predictions.

This is anologous to "innocent until proven guilty". We must demand extremely good evidence before we massizely impinge on the economy to (possibly) mitigate some climate warming.

I have questions falling into two groups.

Group A:

A1) Does the climate model (or models) in question include water and water vapor?

A2) Does the climate model (or models) in question include variations in insolation?

A2.1) If the answer to (A2) is "yes," how are they predicting the value?

Group B:

B1) If given known values for 1980, how well does the model predict 2000?

B2) If given known values for 1900, how well does the model predict 2000?

After reading the basically political, waffling article pointed to, my own prejudice remains as it has been for some years:

A1: No. It's too complicated for the models to handle.

A2: No. There's a Solar Constant, and it's Constant.

A2.1: They don't bother, in view of A2.

B1: Piss poor.

B2: Completely wrong.

I'd be happy to be proven wrong, but don't expect it.

Regards,
Ric Locke

My personal favorite question comes from the 'Origin of Life' researches. (That is, Earth was once a lot hotter, wetter, and had a CO2 level orders of magnitude higher than the current level - even that state wasn't open-ended unstable.)

Picture a sealed system that is hot, wet, and has 'double normal' CO2 concentrations (held that way via a control system + tank) and monitored exactly how much more vegetation happens? A tropical rainforest is a lot better at making O2 from CO2 than Siberian taiga plants. Yes? Can someone point to this type of experiment?

Ric - the point is to come up with means of accurately handling uncertainty in climate modeling. The details of the model are a secondary consideration, to be worked out when the ability to handle uncertainties is solidly in the bag. That was the point of the article.

As to water vapor, I'm guessing that somehow figures into the little box including the words "clouds" and "rain" on the flow chart. Just a guess. The fact that a few years ago the role of water vapor was underestimated is nothing to hang your hat on. Climate modelling hasn't deen standing still in the interval between the early crappy models and now.

If you are really concerned about detailed dynamics of this particular model, there are links at the bottom of the page. It seems that you completely missed the point of the article, though, since it's about handling uncertainties rather than the output of a particular detailed model.

I guess I'm also missing the point (and I did read the article). I can understand a probabilistic model. An example might be: Based on such and such data with a well tested weather model, I can predict rain is 70% likely in two days at location X. I don't understand how you can quantify uncertainty without working out the model first and then show some reasonable test results. At the political level, opponents often don't agree on the fundamentals, so I don't see what this would add there.

VR - the aim is not to say that there's a X% chance that the result is Y, it's to generate complex models which can produce a distribution of outcomes, with associated probabilities. IOW, you want to know what the probability is that the result lies between A and B, between B and C, between C and D, etc. The modeling produces a distribution, not just a single number with a single associated uncertainty. The shape of the distribution is extremely important. We are most used to dealing with normal distributions, which have fairly intuitive behavior. If you look at the distribution in the article, you see that it's skewed over to one side, with a very long tail on the high side. That in itself is interesting, since it suggests that the mean value might not be the best one to use in planning.

Of course that plays directly into the arms of the people who reason that since the results change as the modelling improves we should simply ignore all climate modelling, but they are idiots.

What do you think of these hydroponic systems?