Computer models in climate science

A new journal is out, Inference. International Review of Science, and its first issue carries a long, clearly written and most informative essay entitled ‘Physical Theories and Computer Simulations in Climate Science’, by William Kininmonth, who was for 12 years the head of the Australian Climate Centre, part of the Bureau of  Meteorology. Kininmonth was there when the UNFCCC was drafted, and knows about models, for they were part of his life. But he is not part of the orthodoxy. He has been writing books, articles in and letters to the press for some years now, but this is the longest essay of his that I have seen. It is very good indeed.

Its focus is on the use of computer models in the study of climate, and he says, almost at once, I argue that the relatively simple representation of the climate captured in computer models is inadequate for the purposes of prediction. I shall argue, in addition, that our rudimentary and incomplete understanding of natural variations in ocean and atmosphere fluids has made it difficult to interpret recent climate trends.

And he thinks that the scale of energy exchange processes associated with evaporation, precipitation and cloud formation (the hydrological cycle) are constraints on climate response to anthropogenic forcing. These processes are probably underestimated in climate models, leading to exaggerated projections of warming from carbon dioxide.

If you have only a vague sense of what the Global Circulation Models (GCM) are, then Kininmonth’s essay is a good primer. Climate models are a theoretical mathematical representation, in the form of a system of differential equations, of the earth’s climate system…  A solution to these equations projects an initial state of the system into the future. So far so good.

Models started as aids to numerical weather projection, back in the 1960s, and they refer to  ‘grid points’, of which the more the better, for obvious reasons. In fact, the earliest models had grids 400 km by 400 km, which is a pretty large point to be part of, and even today they are 100 km by 100 km. The trouble is that when you halve the size of the grid you increase the number of computations not by two, but by an order of magnitude. Hence the cry from within climate science for more and bigger computers. And even the current grid size is much too large to adequately resolve smaller scales of motion that are important for heat transport from the tropics to polar regions.

The GCMs come with three more serious problems, and Kininmonth deals with each of them. The first is you have to start with some assumptions in order to provide values for the equations. Some of the assumptions are well-founded, in radiative physics, for example. But some are just more-or-less reasonable estimates, for which no true values are known for each grid point. At once we have error, and unknown error, too. Modellers, for good and understandable reasons borrow and copy each others’ code, so the notion of an ‘independent’ GCM is moot.

The second is that we really don’t know a lot about much of the world, and there is a tendency for modellers to ignore that for which they have little information. A major gap is ocean variability, the circulation of warm and cold water. Another is the hydrological cycle, especially the role of clouds and water vapour. Without real knowledge of the internal variability of the climate system, and the models have not produced any, we simply don’t know how important the human contribution to global warming actually is.

The models looked for a time to be good at hindcasting (testing the models with known past climate data), and that did suggest for some that carbon dioxide accumulations were the real cause of a twenty-year rise in temperature at the end of the last century. But their capacity to forecast future warming, confidently predicted in the IPCC’s AR4 in 2007, has become less and less impressive.

The IPCC authors have argued that the ‘missing’ warming may have gone into the oceans, but the evidence for this is scanty, and in any case the GCMs are not equipped to demonstrate it. And if the GCMs can’t explain how it is that warming has stopped, for quite a long period now, it may well be the case that the apparent anthropogenic ’cause’ of warming in the late 20th century is not at all well supported.

Kininmonth finishes his thoughtful and reasoned analysis like this:  The real sensitivity of climate to anthropogenic activity must remain an open question. The complexity of the climate system and the importance of energy exchange processes on scales below that of the computational grids in use, with their necessary approximations and assumptions, mean that in their present state of development, GCMs are an inadequate tool to resolve the question of sensitivity or to project future climate states. 

My interest now is whether a rebuttal will appear in RealClimate or SkepticalScience, or this important paper is brushed aside, like so much good argument from the dissident camp.

[Extra: Another good essay on the inadequacy of models can be found here.]

Join the discussion 3 Comments

  • Mike says:

    A computer doesn’t help if you actually don’t understand how it works. That is if you don’t have a full understanding of how the climate works and all the factors that influence it in your toast. Adaption of other people’s code is one of the very big sources of computer error. There are many instances of disaster in this area. Added to that no clear concept of how this is to be achieved compounds this recipe for disaster.

  • BoyfromTottenham says:

    Hi Don,

    I keep reminding myself that the supporting evidence of the IPCC (report of 1997 I think) states that only 3% of atmospheric CO2 is man-made, the other 97% being natural (e.g. 60% from ocean outgassing, 20% from vegetation, etc.) I cannot understand why anyone with even average intelligence would think that a 30% increase in this 3% of CO2 was the sole cause of warming that will lead to CAGW. And yet a large percentage of the population (including a lot of pollies, academics, media folk, etc.) seems to be able to ignore this kind of basic evidence that should doom the whole CAGW scam. Or is it that I just don’t I understand cognitive dissonance?

    Meanwhile, keep up the good skeptical work!

Leave a Reply