Michael Crichton: Aliens Cause Global Warming

Re:Michael Crichton: Aliens Cause Global Warming by yummyfajitas on Thursday November 06, 2008 @08:45AM (#355663)

But Dani, if the model is accurate enough to determine that warming is caused primarily by anthropogenic carbon, then it must also be accurate enough to correctly identify and estimate all other effects. After all, if we can't estimate them, how can we rule them out?

Perhaps the models are only valid in regions of phase space we have already visited, but in that case the model has little predictive power! We have a theory which fits some data from one experiment, and has no predictive power. Sounds like little more than a complex data fit to me.

Incidentally, sulfur has been tested for small magnitude changes; Pinatubo provides a natural experiment.

Anyway, I'm not trying to push any geoengineering scheme. I don't think we can accurately predict the effects of such things. But then again, I also don't think we can accurately predict what the earth's temperature would have been absent human intervention.

I'm just pointing out the fundamental point: if we understand climate well enough to determine the exact cause of a small change (which I don't believe we do), we also understand it well enough for geoengineering. But most people don't actually believe we can do geoengineering, even if they do believe AGW is proven.

As for Tim Ball, my bad. Didn't notice that.

[ Parent | on HN ]

Re:Michael Crichton: Aliens Cause Global Warming by DaniFong on Thursday November 06, 2008 @09:21AM (#355688)

I don't think it's true that the predictive accuracy of a model is always as accurate in one parameter as it is in another. I think that only holds if the model is believed almost for certain. In reality you have a distribution of models that you might believe. Suppose you impose some smoothness assumptions on the system: that the response isn't too wild anywhere that you're looking. Then you can make the successive approximations: first you assume that, in this domain, asymptotically, these are linear, separable PDE's. Then if you spend all your time jiggling one parameter, it gives you a bunch of information and constraints on the response for the one parameter. If you jiggle it still further, you might be able to ascertain some glimpse of the nonlinearity, and nonseparability of the system. So for example you might know a fair amount, might have constrained the space of possible models for CO2 - NO2 coupling, or CO2 - water vapor coupling, or CO2 - SO2 coupling. You might see some of the homogeneous CO2 nonlinearities. But you'll know way less about SO2 - acid rain coupling. That's much higher order.

Regardless, my point is that if we stay within the phase space that we know, it's a lot safer.

You need different levels of confidence depending on the choice you're considering. If the prediction is ecological disaster, 90% confidence is good enough to convince me not to try it. If the prediction is a 90% chance of it being okay, that has to be weighed against the much more confident prediction we have that if we gradually turn down the manmade forcing on the environment, things will stay pretty close to alright.

So it's less like your glider example than the reverse. Someone shows you a videotape of a large body, of mostly known shape, traveling through a fluid of unknown density and viscosity. We know that most things fall. We know that most gliders are unstable. Yet we are climbing this mountain and approaching this cliff ledge. And I think it's prudent to say, look. Ground is good. We had better be a lot more sure than this if we want to fly this thing. We shouldn't fly the climate recklessly.

[ps, I am going to sleep now.]

[ Parent | on HN ]