December 8, 2008 | 54 Comments
If increasing temperatures are consistent with or are evidence of global warming, what theory is consistent with or evidence of falling temperatures? Global warming, too?
We have to ask this complicated question because it was just reported that this year’s global average temperature is on track to be the coldest in the last eight years. In other words, the temperature has dropped, and has been dropping for a couple of years.
So, do these falling temperatures mean that global warming has stopped or is false?
“Absolutely not,” said Dr Peter Stott, the manager of understanding and attributing climate change at the Met Office’s Hadley Centre.
We also hear from “a team of climate scientists at Kiel University” who
predicted that natural variation would mask the 0.3C warming predicted by the Intergovernment Panel on Climate Change over the next decade. They said that global temperatures would remain constant until 2015 but would then begin to accelerate.
This is going to be complicated, and there are not many ways to make this topic easy to understand, so hold on.
We have to explore what it means for evidence to be “consistent with” of “inconsistent with” a theory, what it means for a theory to be wrong or right, and what it means to be probably right or probably wrong.
Suppose, then, that the theory of anthropogenic global warming claims, among other things, that the temperature will certainly rise year by year. The temperature did not rise and in fact fell. This evidence is inconsistent with the theory and so the theory is false because it made a prediction that said that falling temperatures were impossible. If you like, we can say that this theory has been falsified (Popper’s term of very little utility; a topic to explore in full another day).
Our cartoon theory has been proven false. It can thus be dumped, forgotten, ignored. Skeptics may rejoice! Indeed, any theory that makes a prediction that says “X will certainly occur (in such and such circumstances)” and in fact “not X” occurs, has been falsified, which in plain English means that it is wrong. (“not X” means “Any circumstance that is certainly not X.”)
However, the actual theory of anthropogenic global warming (AGW) makes no such claim like “The year-by-year temperature will certainly rise.” Instead, it makes claims like this: “The year-by-year temperature will probably increase.” We can go further and say that nearly all the claims of that theory are statements of this kind, that is, probability statements.
There is only one claim of certainly in the AGW theory and it is that “Mankind influences climate.” That statement is trivially true. It is impossible for mankind not to influence climate. Every action you take, every breath you make, changes the climate. In fact, we can broaden the claim and say that “All organisms influence the climate” and it remains true. Since this is trivially true, the only interesting questions are (1) “How much does this or that organism influence climate?”, (2) “In what part are these influences helpful or harmful?”, and (3) “Can the helpful aspects be magnified and can the harmful effects be mitigated?”
Climatologists spend most of their time with (1), the first part of (2) is usually ignored, and “activists” make their business the last part of (3), the first part also being ignored. But we are getting away from our original goal. What do falling temperatures mean to the theory of AGW?
Let’s examine the claim: “The year-by-year temperature will probably rise.” What observations, if any, are inconsistent with that statement? None. There are no set of circumstances that can arise which directly contradict it. Temperatures may rise and they may fall; they may fall for each year from now until the year 3000 and that statement remains standing.
This is another way to state the the AGW theory is not falsifiable. But wait! I do not imply that this is a weakness of the theory. Any theory that makes probability statements cannot be falsified. Thus, it is a mistake—and one I often see—to attack the AGW theory by claiming it is not falsifiable. Of course it is not! How many theories are like this, that is, not falsifiable? Besides mathematical dictums, and not wanting to get to far into this, all theories that are of practical interest to mankind. Any theory that is forced to make probabilistic claims cannot ever be falsified. And, simply put, this is most theories that we ever work with (in real life). Since this is the case, we have to understand what a probability statement means, and what can we mean by data “consistent” and “inconsistent” with such statements.
Now, we are right to suspect a theory that makes a claim that “X will probably happen (in such and such circumstances)” and X does not in fact occur. We have not proven the theory is false, but we would be rational to give increased weight to the idea that it is false. Thus, it is not irrational to weaken our belief in the AGW theory in the face of falling temperatures. Further, Dr Stott is wrong—not probably mistaken, or likely wrong, but flatly wrong—if he says “Absolutely not” to the question “Given the falling temperature, is AGW less likely to be true?” But he would be right—exactly right—if he said “Absolutely not” to the question “Given the falling temperature, is AGW proven to be false?” Given the way newspapers report, we have no clear way of knowing which of these two questions Dr Stott was asked.
I told you this was going to be tricky! It’s now going to get worse, but stay with me.
What Claims Are Being Made?
It is very difficult to keep in mind the exact claims made by a complex theory, which is why when new evidence arises, we have so many different arguments about what this evidence implies. Proponents of controversial theories are usually happy to be vague about the claims they are making because of this. This way, no matter what evidence arises, arguments for the theory can be found within it. We discussed this the other day when we noted that precisely quantified predictions are hard to come by in global warming.
We might hear that “Temperatures will increase” but this is a vague statement because when temperatures do in fact fall in a location, the proponent can say “I meant it will go up elsewhere” or “I meant it will go up on average” or “I did not mean it will always go up.” This is how the Kiel University climatologists could claim that “natural variation would mask the 0.3C warming predicted by the Intergovernment Panel on Climate Change over the next decade” and that temperatures will start increasing in 2015. Their explanatory statement is impossibly vague (saying what is “natural variation” at least means you can exactly quantify the amount of change caused by the existence of mankind), but they have at least made a semi-quantifiable prediction starting six years from now. The IPCC, too, has made quantifiable predictions of global average temperatures, though they seek to introduce vagueness by calling their predictions “scenarios”, a weak attempt at evasion. They also do not explicitly attach probability to their predictions, which makes it harder to say how wrong or right their predictions were.
To judge how likely a theory is to be true requires explicit predictions. For example, suppose the AGW made one of these two statements: (A) “The global average temperature will likely increase”, and (B) “The global average temperature will probably increase.” Our new evidence is that temperatures have fallen. For which of these claims, A or B, would you have less certainty that AGW is true?
A difficult question! Let’s try to understand it. You can see that A and B differ only in the words “likely” and “probably”. Suppose that, to you, “likely” means “at least 90%” and “probably” means “at least 50%”. It’s clear that claim B is thus weaker than claim A. So that when contradictory evidence arises you would suspect that the theory that gave rise to A was more suspect than the theory that gave rise to B. The terms “likely”, “probably”, “might” and so forth, do not mean the same thing to every person. For example, I might say “likely” means “at least 50%” and “probably” means “at least 90%”, and so I would come to the opposite conclusion about the truth of the underlying theory as you did.
All this means is that the more precise a prediction is, the easier it is to judge it. The vaguer a claim is, the harder it is to dismiss or accept the theory that gives rise to it. Not every claim can be exactly quantified, but any prediction of importance can be made clear in terms of decisions or actions we would have to make given the claim is true. So it is no excuse to say “The problem is too hard to give an exact prediction.” Because if that were true, then the theory has no practical consequences, and so can be ignored. To be clear, AGW theory does make claim to have practical consequences, and so its proponents should be able to give us more exact predictions.
Like I said, precise claims from AGW are not always to be had—though of course they sometimes are—but let us suppose that the theory makes the claim that “There is a 90% chance that the year-by-year global average temperature will increase.” For the last several years the temperature has fallen. The theory is not falsified because the claim only mentioned a probability, one which is consistent with falling temperatures. We are rational, however, to decrease our belief in the truth of the theory.
Now suppose that a group of climatologists offer a rival theory called “Business as usual” (BAU) which makes the claim “There is a 50% chance that the year-by-year global average temperature will increase.” For the last several years the temperature has fallen. The theory is not falsified because the claim only mentioned a probability, one which is consistent with falling temperatures. We are rational, when offered a choice between the two theories, to increase our belief in the truth of the BAU theory over the AGW theory because the BAU’s predictions were closer to what actually happened. The temperatures fell, but saying there is a 50% chance of this happening each year is closer than saying their is only a 10% chance of this happening (10% = 100% – 90%).
The best theory, which is not on offer, is the one which predicted that the temperatures would fall for the last couple of years. So let us offer that theory—call it the Baby, it’s cold outside (BICO) theory—which makes the claim “There is a 10% chance that the year-by-year global average temperature will increase.” This is the best of the three theories in the sense that its probability statements were closest to what actually happened. The BICO theory says there is a 90% ( = 100% – 10%) chance that temperatures will fall.
It’s about to get tricky once more. Be sure you understand everything so far before reading more.
We do not just judge theories about how well they predict future data, but also by how well they explain already observed data. The BICO theory does not explain all the previous data we have very well, as you might guess. Proponents of the AGW theory say that their theory does. The BAU theory does a reasonable but imperfect job explaining historical data. Now, it is true that just because a theory explains past data well it does not mean that it will explain future data well. This is because it is always possible to create a theory that explains past data perfectly or as close to perfect as we want to be. Memorize this (especially if you read any paper which uses statistical results). Of course, any theory that does well on future data will also do well explaining historical data. Thus, given reasonable performance explaining historical data, the real test is always on how well a theory predicts data we have not yet seen (which is defined as data that was not used in any way to create the theory).
Realistically, the BICO model is out of contention because of its exceedingly poor performance on historical data. The two competitors left are AGW and BAU (I do not mean to imply that these are the only competitors in real life, just the they are the only two we are considering here). The contest is how well each model does on predicting future data.
There is some math that says that if we have two (or more) competing theories, the one that is calibrated is better than the other in the sense that anybody who acts on information from the calibrated theory would do better than had he acted on information from the other theory. Calibration means that if a theory says “There is a Y% chance that X will occur”, then on Y% of the times X could have occurred, it actually did. Neither the AGW nor the BAU theories are calibrated in this sense.
But there is the sense that the BAU theory is simpler than the AGW theory. The BAU theory requires no advanced “degrees” to understand, nor does it require suites of multi-million dollar computers, nor does it need panels of bureaucrats to meet yearly to discuss it. The AGW theory needs all these things and more. It is a sophisticated theory (I am not using this word sarcastically).
We would hope, given all the time, effort, and money that goes into the sophisticated AGW theory that it could beat the BAU theory in its predictions. If not—if the BAU theory routinely beats the AGW theory—then we would be rational to give more weight to the truth of the BAU theory.
Right now—as far as I am able to see—the BAU theory does beat the AGW theory in predictive ability. Actually, a modification of the BAU is what routinely wins. That modification states something like “There is a 90% chance that the global average temperature will do what it did last year.” This is technically called persistence (BAU(P)). When a sophisticated theory cannot beat either the BAU or BAU(P) theory, it is said not to have skill. If a theory is not skillful, it should not be used; that is, one should not base any decisions with respect to that theory. So far as I am able to see, the AGW theory is not skillful.
My caveat is “so far as I am able to see.” I am constrained by the inexact nature of the AGW theory’s predictions. The BAU and BAU(P) theories are certainly precise enough. It might be the case, for example, that I have mis-quantified the AGW’s predictions, or misunderstood exactly what physical variables the predictions are referring to, or that I have mischaracterized what the AGW predictions imply. For example, I have been taking them to mean that “There is a 90% chance that the year-by-year global average temperature will increase.” I welcome correction on my characterization. In fact, we all would welcome the correction and look forward to the issuance of precise statements and predictions.
Just What Are Falling Temperatures Evidence Of?
So how about it? Since temperatures have fallen, what are we to believe? It is not true that the AGW theory has been falsified, but neither have the BAU or BAU(P) theories. Given our characterization of the AGW theory, it is rational to say that our belief in it, given the contradictory observations, should be lessened. The BAU theories remain as true as ever—that is, we do not really increase nor decrease our beliefs in them based on this new evidence.
It might be, as I have admitted, that our characterization of the theory’s statement is wrong. Another characterization was offered by the Kiel University group who are probably claiming that temperatures will likely fall or remain constant until 2015, after which they will likely increase. I say “probably” because it’s not clear what their exact claim is. However, this is likely a fair summary of it.
Now, since predictions of “likely falling or remaining constant until 2015” are the same as the predictions made by the BAU and BAU(P) theories, it is obvious that the Kiel University (and similar) theories do not yet have skill. It is true to say that it might—but we won’t know until after 2015.
I for one am happy to wait before doing anything until then.
(To anticipate the counter to this conclusion, which we’ll discuss more fully at a later date—meaning I don’t expect this short reply to answer fully the counter: since you haven’t proven skillful, why should I do anything? If you say the consequences are too horrible if we do not, I ask you why you also refute Pascal’s argument.)