Do You Believe In Global Warming?

Earlier this evening, I got the following message from Ellis Wyatt:

"I'm curious-- in your research, have you come to the conclusion that climate change, at least at the hands of mankind, is provable? There is quite a lot of evidence to the contrary, and the State has done a great job of funding a lot of smokescreen research."

I wanted to address this on the main blog, rather than simply replying to it in some tucked away place where it would quickly be forgotten. So here goes nothing:

I think Stephen Gardiner put it best in his essay, "Ethics and Global Climate Change," when he wrote, "The skeptics are right…when they assert that the observational temperature record is a weak data set and that the long-term history of the climate is such that even if the data were more robust, we would be rash to conclude that humans are causing it solely on this basis. Still, it would be a mistake to infer too much from the truth of these claims. For it would be equally rash to dismiss the possibility of warming on these grounds. For, even though it might be true that the empirical evidence is consistent with there being no anthropogenic warming, it is also true that it provides just the kind of record we would expect if there were a real global warming problem."

There is very considerable evidence that climate change is being caused by human influences. Specifically, we have seen the most warming in the lower atmosphere, and a slight cooling in the upper atmosphere, which is consistent with a greenhouse gas driver. We have detected a clear warming trend during a period in which we would expect natural processes to have caused a cooling. And climate modelers are unable to reproduce historic climatic conditions without taking into account the influence of the roles of the factors which are believed to be responsible for current warming. Further, when modelers do incorporate what they believe to be the influence of anthropogenic factors, they are able to reproduce observed temperature fluctuations with a reasonable degree of skill.

That being said, it's certainly possible that the evidence is coincidental, and that modelers simply don't understand how the climate system works (though I'd say this possibility is relatively remote). I'd say that the IPCC was likely being fair to the available evidence when it said that it was very likely that humans have been responsible for most of the warming we've been seeing. That being said, it seems to me that considerable questions remain about the IPCC's predictions about future warming.

One danger is the possibility of model "tuning," which is where a model is designed in a way that makes it fit the data set, but doesn't accurately model the factors which caused the data set to have the characteristics that it has. It could be that rather than actually modeling the climate system, climate models are simply modeling the data sets, which would make them useless for predictions. Scientists are aware of this problem, and try to prevent if from having a significant influence, but given the enormous complexity of the climate system, the possibility is always there.

Another problem with models arises from the IPCC's practice of averaging model outputs. The idea is that by averaging multiple models, we should compensate for the shortcomings of each, and end up with a "best" prediction. Some will miss high, others low, and the result will probably end up closer than what any single model could have achieved on its own. The problem with doing this is that if none of the models work on their own, then it would seem that averaging them could only be providing the illusion of success. Again, this possibility is understood, and scientists try very hard to minimize it. But again, the possibility is there.

Another set of questions comes from an inability to accurately model small-scale phenomena when modeling larger climate systems. Climate models operate by dividing the climate system into "cells," where intracellular activity is only approximated, and intercellular activity is really what's being observed. If intracellular processes aren't being captured properly by the low-resolution approximations, then the result could be a butterfly effect in either direction. Small events which might have caused (or prevented) larger-scale phenomena would not be incorporated into the model, and the results could be game-changing. Modelers try to deal with this by running smaller-scale Regional Climate Models, but this problem is somewhat unavoidable given the limitations on computing power faced by climate modelers.

Yet another difficulty arises from the use of proscribed variables in models. Some processes are not well understood, and so they must be estimated outside the models and plugged in. As models become more complex, the proportion of internally calculated variables grows. But as long as we don't understand exactly how every piece of the puzzle works, this will continue to be a problem. Modelers often minimize the influence of this problem, though, by testing their models for their sensitivity to differences in proscribed values.

One last problem is that predictions are based on the assumption that the climate system will continue to behave as it has behaved in the past. Things like the the possible shutdown of thermohaline circulation or the collapse of the Greenland or Antarctic ice sheets would be game-changing, if they occurred, and would almost certainly add considerable uncertainty to our ability to make predictions. The roles of unquantified or undiscovered feedback mechanisms could also represent an instance of this problem.

All that being said, I don't think we can respond to uncertainty with inaction. I'm not sure who ought to bear the burden of uncertainty. But doing nothing places it all on the potential victims, which seems wrong in light of the fact that they'll never be able to obtain compensation, since those responsible will be long dead. Determining what we should do, though, is more complicated than I'd like to get into here. Hopefully that answer was helpful, and not too badly loaded with technical jargon. Feel free to ask for clarification!