Simple Question, Simple Answer... Not

By Spencer R. Weart

The following article appeared originally in slightly different form on September 8, 2008 on the blog. It is being reproduced here, with the permission of The author has added a postscript.

I often get emails from scientifically trained people who are looking for a straightforward calculation of the global warming that greenhouse gas emissions will bring. What are the physics equations and data on gases that predict just how far the temperature will rise? A natural question, when public expositions of the greenhouse effect usually present it as a matter of elementary physics. These people get suspicious when experts seem to evade their question. Some try to work out the answer themselves (Lord Monckton for example) and complain that the experts dismiss their beautiful logic.

The demand that the case for dangerous global warming be proved with a page or so of equations does sound reasonable, and it has a long history. The history reveals how the nature of the climate system inevitably betrays a lover of straightforward answers.

The simplest approach to calculating the Earth’s surface temperature would be to treat the atmosphere as a single uniform slab, like a pane of glass suspended above the surface (much as we see in elementary explanations of the “greenhouse” effect). But the equations do not yield a number for global warming that is even remotely plausible. You can’t work with an average, squashing together the way heat radiation goes through the dense, warm, humid lower atmosphere with the way it goes through the thin, cold, dry upper atmosphere. Already in the 19th century, physicists moved on to a “one-dimensional” model. That is, they pretended that the atmosphere was the same everywhere around the planet, and studied how radiation was transmitted or absorbed as it went up or down through a column of air stretching from ground level to the top of the atmosphere. This is the study of “radiative transfer,” an elegant and difficult branch of theory. You would figure how sunlight passed through each layer of the atmosphere to the surface, and how the heat energy that was radiated back up from the surface heated up each layer, and was shuttled back and forth among the layers, or escaped into space.

When students learn physics, they are taught about many simple systems that bow to the power of a few laws, yielding wonderfully precise answers: a page or so of equations and you’re done. Teachers rarely point out that these systems are plucked from a far larger set of systems that are mostly nowhere near so tractable. The one-dimensional atmosphere model can’t be solved with a page of mathematics. You have to divide the column of air into a set of levels, get out your pencil or computer, and calculate what happens at each level. Worse, carbon dioxide and water vapor (the two main greenhouse gases) absorb and scatter differently at different wavelengths. So you have to make the same long set of calculations repeatedly, once for each section of the radiation spectrum.

It was not until the 1950s that scientists had both good data on the absorption of infrared radiation, and digital computers that could speed through the multitudinous calculations. Gilbert N. Plass used the data and computers to demonstrate that adding carbon dioxide to a column of air would raise the surface temperature. But nobody believed the precise number he calculated (2.5˚C of warming if the level of CO2 doubled). Critics pointed out that he had ignored a number of crucial effects. First of all, if global temperature started to rise, the atmosphere would contain more water vapor. Its own greenhouse effect would make for more warming. On the other hand, with more water vapor wouldn’t there be more clouds? And wouldn’t those shade the planet and make for less warming? Neither Plass nor anyone before him had tried to calculate changes in cloudiness.

Fritz Möller followed up with a pioneering computation that took into account the increase of absolute humidity with temperature. Oops... his results showed a monstrous feedback. As the humidity rose, the water vapor would add its greenhouse effect, and the temperature might soar. The model could give an almost arbitrarily high temperature! This weird result stimulated Syukuro Manabe to develop a more realistic one-dimensional model. He included in his column of air the way convective updrafts carry heat up from the surface, a basic process that nearly every earlier calculation had failed to take into account. It was no wonder Möller’s surface had heated up without limit: his model had not noticed that hot air would rise. Manabe also worked up a rough calculation for the effects of clouds. By 1967, in collaboration with Richard Wetherald, he was ready to see what might result from raising the level of CO2. Their model predicted that if the amount of CO2 doubled, global temperature would rise roughly two degrees C. This was probably the first paper to convince many scientists that they needed to think seriously about greenhouse warming. The computation was, so to speak, a “proof of principle.”

But it would do little good to present a copy of the Manabe-Wetherald paper to a technically trained person who demands proof that global warming is a problem. The paper gives only a sketch of complex and lengthy computations that take place, so to speak, offstage. And nobody at the time or since would trust the paper’s numbers as a precise prediction. There were still too many important factors that the model did not include. For example, it was only in the 1970s that scientists realized they had to take into account how smoke, dust and other aerosols from human activity interact with radiation, and how the aerosols affect cloudiness as well. And so on and so forth.

The greenhouse problem was not the first time climatologists hit this wall. Consider, for example, attempts to calculate the trade winds, a simple and important feature of the atmosphere. For generations, theorists wrote down the basic equations for fluid flow and heat transfer on the surface of a rotating sphere, aiming to produce a precise description of our planet’s structure of convective cells and winds in a few lines of equations... or a few pages... or a few dozen pages. They always failed. It was only with the advent of powerful digital computers in the 1960s that people were able to solve the problem through millions of numerical computations. If someone asks for an “explanation” of the trade winds, we can wave our hands and talk about tropical heating, the rotation of the earth and baroclinic instability. But if we are pressed for details with actual numbers, we can do no more than dump a truckload of printouts showing all the arithmetic computations.

I’m not saying we don’t understand the greenhouse effect. We understand the basic physics just fine, and can explain it in a minute to a curious non-scientist. (Like this: greenhouse gases let sunlight through to the Earth’s surface, which gets warm; the surface sends infrared radiation back up, which is absorbed by the gases at various levels and warms up the air; the air radiates some of this energy back to the surface, keeping it warmer than it would be without the gases.) For a scientist, you can give a technical explanation in a few paragraphs. But if you want to get reliable numbers—if you want to know whether raising the level of greenhouse gases will bring a trivial warming or a catastrophe—you have to figure in humidity, convection, aerosol pollution, and a pile of other features of the climate system, all fitted together in lengthy computer runs.

Physics is rich in phenomena that are simple in appearance but cannot be calculated in simple terms. Global warming is like that. People may yearn for a short, clear way to predict how much warming we are likely to face. Alas, no such simple calculation exists. The actual temperature rise is an emergent property resulting from interactions among hundreds of factors. People who refuse to acknowledge that complexity should not be surprised when their demands for an easy calculation go unanswered.

Postscript: Some bloggers have quoted from this essay to support their view that computer models can give no “proof” of global warming. That is true enough as they mean the term. So long as people disagree on what kind of “proof” is needed before we take action, we will never reach agreement on what to do, although we can hope to converge on where the disagreement is located. If we require a level of certainty equivalent to what governments use in deciding to intervene in markets or go to war, it is obvious that climate science has exceeded that level for a decade or more.

For history details and references see
On Monckton see

Spencer R. Weart
American Institute of Physics

This contribution has not been peer refereed. It represents solely the view(s) of the author(s) and not necessarily the views of APS.