Climate change crystal ball clouds over

One of the main clubs with which the very vocal pro-greenhouse camp repeatedly beat their opponents is the assertion that there is a “consensus” of scientific opinion that temperatures are set to increase dramatically in coming decades.

This club is wielded again and again in response to almost every counter argument. Pro-greenhousers repeatedly state that “every scientist” agrees or that there is a “steady accumulation of scientific evidence” in favour of warming, and so on.

Some of this is due to public confusion over the object of the debate. There is no doubt that temperatures have increased by about a degree or so since 1860, but greenhousers seem to be arguing as if the sceptics are questioning that basic point. In fact, the sceptics are usually questioning the value of the temperature forecasts made by the Intergovernmental Panel on Climate Change, which is most certainly up for debate.

All that said, however, there is no doubt that many scientists, rightly or wrongly, support the IPCC forecasts: however, to claim there is a consensus about forecasts is plainly silly. It is like claiming a consensus for a long-range weather forecast. But there is at least a large and vocal minority who seem to regard them as a form of holy writ. Even before ABC television recently screened the program The Great Global Warming Swindle distinguished scientists with undoubted credentials in climate studies were organising seminars to debunk it. Plenty of others started tapping out articles at their computers.

We will discuss the program itself only in passing here. Instead this article will look at the question of whether this support among experts for the IPCC forecasts of global warming, constitutes a validation of those forecasts. In fact it is, at best, a very weak argument.

For we are not dealing with testable scientific propositions: we are dealing with a bunch of forecasts, issued earlier this year, that average temperatures will be anywhere between 1.1C to 6.4C higher than now in 100 years or so, with a best guess around the 3C mark. The IPCC scientists also state that they had a “very high confidence” (which translates to 90 per cent confident) that human activity would be responsible for anywhere between 0.6C to 2.4C of that increase, depending on what happens with emissions.

The first point to note is that all of this is completely unprecedented. It is difficult to think of any other forecasting exercise of this magnitude, or one apart from astronomy, that deals in such a long period. The limit for serious general forecasting in business and economics is five years, at a stretch. Forecasts for specific markets to be serviced by very large projects, such as power stations, may go out to 40 years. But the models used are almost trivial compared to the climate systems under discussion, and even then a single wrong assumption can make nonsense of the results.

As for trying to make usable forecasts - or say anything useful at all - about a complex system a full century out, anyone with knowledge of forecasting would refer those involved to the divination department at Hogwarts.

Forecasting has a long history which mostly involves failure, but by trial and error forecasters have learned what they can and cannot do; what works and what does not. Unfortunately, the IPCC seems to be repeating many basic mistakes.

This point was made with some force by the publication in late June of the paper Global Warming Forecasts by Scientists versus Scientific Forecasts (PDF 236KB). The authors are J.Scott Armstrong of The Wharton School, University of Pennsylvania and Kesten C.Green, of the business and economics forecasting Unit at Monash University in Melbourne.

Neither author is a climatologist - Armstrong is a professor of marketing - but then, as they point out in the paper, the IPCC climatologists are not forecasters. Readers can look at the paper, which is clearly written, for themselves. It makes a number of points in what amounts to an audit of chapter eight of the IPCC’s recent report which have already been made by myself and others, but with considerably more authority.

One such point involves the use of the technique of tweaking a computer model to fit historical data and then using the model to make forecasts: this is the IPCC approach. It is known to be of little use in forecasting. The paper also counts as the second independent review of aspects of the panel operations where the reviewers have been incontestably independent (the other was the Wegman report (PDF 1.41MB) on the hockey stick graph, involving the use of statistics).

The IPCC report has miserably failed both.

The one point we shall concentrate on here is the paper’s assertion that experts don’t matter much in unaided forecasting in their area of expertise. The paper cites several studies which deal largely with marketing and business in which experts fare no better in making forecasts concerning events in their field of expertise than non-exerts. In other words, their expertise counted for little in trying to forecast what would happen in their own fields.

As noted the surveys mainly concern marketing and business and pro-greenhouses may derive some comfort from that. They may declare that this time, because scientists are involved, it is different. Well, is it different? The paper points to individual examples of scientists making forecasts that turned out to be completely wrong - Einstein declaring that atomic energy is impossible and so on.

The classic, if rather dated, example not given in the paper is that of Lord Kelvin, a famous 19th century scientist who was mostly wrong in every pronouncement he every made about possible technological advancements.

A more recent example of how scientists can be collectively wrong, just like all other kinds of experts, is the medical profession’s collective insistence that viruses do not cause peptic ulcers - a point that doctors clung to, and note, clung to collectively for many years - in the face of pioneering work by two Australian scientists Robin Warren and Barry Marshall. The medical profession had earlier decided that stress and spicy food caused such ulcers. The scientific process eventually overcame the medical profession’s conservatism.

In other words there is no reason to believe that because a group of scientists collectively insist that they are right, that we should believe them - particularly when it comes to forecasts.

Armstrong and Green also note that the tendency of experts to be wrong is made worse, if they are working together or heed each other‘s work. The initial assumptions become dogma and difficult to overturn.

Nor are they alone in belitting the role of experts in forecasts. The point is made repeatedly in the book Megamistakes: Forecasting and the Myth of Rapid Technological Change, by Steven P.Schnaars published in 1989 but still the last word on the confident area of forecasting of the 1960s and 1970s. (The IPCC forecasts are, in essence, a throw back to that era.) As well as noting that forecasts by experts in certain fields seldom seemed to do better than lay people, Schnaars points out that conservative forecasts seemed to work best. That is, the forecasts are for little or no change in the system under study. This point is echoed by Armstrong and Green.

There may always be exceptions where the experts are right, and there are no other large scale predictive efforts by scientists on which to make a comparison, but there is very little in the panel’s output which gives independent observers any confidence. As noted, the IPCC has miserably failed at least two independent reviews of aspects of its work and, in any case, it freely admits that it does not know much about the bulk of the variables which it includes in its models. (See the table on page 4 of Climate Change 2007: The Physical Science Basis, Summary for policymakers (PDF 1.25MB). The relevant page is 16 in earlier versions.) 

Other arguments often put forward when the IPCC idols are shown to have feet of clay is that first, the risk is too great that they may be right, and second, that if we leave the decision making any longer it will be too late. The first part of that counter argument involves risk analysis which is a different discipline altogether and assumes that we know enough to assess risk properly. This is doubtful. The second objection naturally raises the question of too late for what? Are we going to have global cooling or global warming? As by now should be abundantly clear, forecasting is not about wishful thinking and it’s not about politics.

Waving a bunch of computers at a set of bad assumptions will not turn them into good forecasts, but the result can be bad policy.