Climate Models Don't Predict Climate

by Michael Tobis     Department: Earth

In recent years, climate models have had as their most prominent use the projection of climate for the next century or two under certain scenarios. These scenarios have two main parts: behavior of humans and behavior of polar ice sheets. The human behavior has three major parts; greenhouse gas concentrations, land use changes, and aerosol (dust) emissions. These in turn can be broken down further, but that's beside my point for now. And then there's the ice sheets, whose behavior is greatly hoped to be "nothing much", but is increasingly suspected to be something other than that.

The greenhouse gas concentrations themselves aren't quantities that even social and economic projections can give us. What economists try to offer is some sort of projection of emissions scenarios, that is, the amount of gas spewed into the air per year. There is some chemistry (including biochemistry) between those emissions and the actual concentrations (total amount in the air at any instant) over time. That chemistry is very poorly understood. We can identify most of the processes at work, we think, but we cannot quantify them very well even under current conditions.

So when we run the climate models into the future, there is a lot of guesswork about what conditions the climate system will be under. (For somewhat mathematical and somewhat historical reasons we tend to call these conditions "boundary conditions"; they're related to but not exactly the same as the boundary conditions you'll encounter if you study partial differential equations.) We can be pretty sure it isn't a realistic bet of the boundary conditions. Even so, when different models are run under identical scenarios, you'd hope they would get the same answer. (This is a useful sanity check if nothing else.) Well, what comes out is in some ways similar across models, but there are many differences. Models tend to agree that Arizona will get drier and Scotland wetter. They are literally all over the map when it comes to soil moisture in Texas, though.

I've recently come across a small group of people who seem to take great glee in finding technical problems with these models. I personally agree with some of their complaints though without much pleasure. I have to use the darned things myself.

Their main point, though, is that important decisions are based on these models. That point, though, isn't really true.

I think that the only important decision that climate models yet have any direct bearing on is whether greenhouse gases need to be curtailed, and even then the impact is more in the way of confirmation rather than motivation. We know the radiative properties of the atmosphere are changing as a result of human activity. We know that such changes are a primary factor in how the atmosphere behaves.  We have, therefore three choices (you can pick all in combination, or just one, but you can't pick none of them) 1) accelerating climate change 2) large cutback in emissions and 3) artificial mechanisms to remove greenhouse gases from the air. The climate models don't really have much to say about how much, when, or how, which are really the big questions.

So what are the models for? Should we improve the effort or should we give it up?

It turns out the press and the public are very confused about the role of the models, and the modeling centers are complicit in this confusion. The projections that give us the scary maps we often see in the press are pretty much of the right magnitude, but the details are still crude and the costs are pretty high.

See, the models aren't really for the policy sector. They are for the scientists. Their most prominent use is not their most important use. I'll explain more next time.

Tags: climate, computing, modeling

+ Add Comment