A Real Climate of Misunderstanding

What Climate Science?
Are AGW climate scientists and science-prone skeptics talking about the same subject? I thought so, but am not sure of that any longer.

Having read Real Climate (RC)’s “Butterfly” blog and engaged in some commentary about it at that site, and having followed the AGW debate for the last five years, my impression is that:

  • the AGW climate scientists are just doing what they can, after heavily restricting their area of research
  • I and some other fellow science-minded skeptics are simply pointing out to the vast, unexplored regions outside of your average climate modeller’s understanding and computational ability.

Imagine if paleontologists had decided to concentrate on the skulls of Rift Valley hominids, treating with disdain (aka as “noise”) all of a find’s context, including other human bones, remains of other animals, local geography (and climate). And deliberatingly ignoring every other hominid find, anywhere else in the world.

That’d still be science, but within such a very focused line of research quite unlikely to add much knowledge or understanding, apart than about itself.

There Must Be Some…
If such a colossal misunderstanding is indeed in place, that would go a long way in explaining the extraordinary ill feelings surrounding the whole of climate science at the moment (and I am deliberately keeping politics outside of this), with one side treating skepticism itself as a dishonest scandal that should be stamped out of existence once and for all, and the other side dismissing years and years of research as pretty much irrelevant gibberish written by incompetent liars.

No wonder they (we) can’t see each other eye-to-eye…how could two judges agree at a canine show contest, if one of them were only interested in (and had built a whole theory of canine beauty about) the shape of the tails?

Climatology: An Abridged History
The story of how contemporary climatology has ended up like this is illuminating.

At first, basic laboratory experiments gave some indications on how atmospheric constituents could interact with one another, and with the incoming solar radiation. Notable among them, the study of CO2’s “greenhouse effect” by Arrhenius in 1896. But the real world of meteorology (including climate) is somewhat more complex than a lab’s setting.

For example, vast energy exchanges manifest in atmospheric cell circulation, oceanic heat exchanges, and whole climate-affecting cycles currently known as the Pacific Decadal Oscillation, the North Atlantic Oscillation, el Nino/la Nina, etc etc.

With no way of replicating that in controlled laboratory conditions, climatologists opted at one point to computational models of the atmosphere. This was of course possible only and after a minimum of computational power became available.

Computers of course understand only numbers and formulas/commands. In order to get to that, a momentous assumption was made: in an approach curiously reminiscent of the science of aeronautics, climate was taken as the response of the atmosphere to “forcings”, i.e. discernible components pushing and pulling the atmosphere in one or the other direction.

“Climate” is then the resulting overall effect of the action of each forcing, averaged over a certain lentgh of time.

In that context, “forcings” were purely operational, “digitizational” tools, providing some basis for computing the climate. By definition, in fact, forcings cannot be measured: all observations of the actual atmosphere will (obviously!) include the effect of them all. If “forcings” exist or not is therefore irrelevant. For all they were worth, forcings could have been substituted by Fourier analysis, or Principal Component Analysis, or whatever other technical tool that can transform a set of signals (and formulas) of any sort into computer-friendly figures (and procedures).

However, alongside a steady increase in available computational power, there came the a shift in focus, from real (observable) climate to forcings: in a first dichotomy with the real world, models became ways of investigating the (possible) effect of each forcing, instead of forcings being ways of investigating the (possible) evolution of the planet’s climate.

This change is less subtle than it appears. It entails throwing one’s hands up in the air about trying to understand the actual atmosphere, choosing instead to concentrate on known (pre-set) effects of known causes. Models in fact are far from independent from assumptions about forcings: they are made out of them. The effect of each forcing is already written in the code of each model, and model runs will show that effect at work. Even if results could vary for example modifying a model’s representation of geography, there is no way that model will be able to run contrary to its pre-assumed behaviour, for example in the case of increased CO2 concentration.

If I write a computer program that just adds one every time a white objects traverses a camera’s field of view, there is no way my program will ever count down, say to minus 20. And the fact that the counter always increases says nothing about how many white objects there are in the real world. It just shows how the counter works.

Nothing But Parameters
What can you do when all you have are models only useful to investigate what a particular forcing’s effect might be? You are left with playing with the parameters, modifying them to “fit” observations and “plausibility”. This is manifest for example in Hansen et al’s 2007 article, “Climate simulations for 1880–2003 with GISS modelE“, literally saddled with innumerable “estimations”, six of them explicitly “subjective” (little more than guesses, that is) but still able somehow to get published in a peer-reviewed scientific article.

Note that comparison to the real world is but a side issue in that paper. “Observations” (25+ years of averages) are useful to evaluate what the parameters are likely to be, i.e. the relative importance of each forcing. There is nothing important outside of them. In a second dichotomy with the real world, in such a vision of the world everything that is not included in the modelling is “noise”, in other words “irrelevant”.

There is no “going back to the lab” in contemporary mainstream forcings-based climate science, eg to learn anything new after finding unexpected observations, because those are “noise” (sometimes called, “weather”) and thus have to be ignored. And there is no meaningful effort to measure what if anything is going wrong: for example, comparisons between model results and observations are simply visual.

The good thing about this is that there are enormous avenues of research left open to future generations. The downside is that the reality of climate models is, at present, literally set in stone, whatever the real climate is out there.

Can climate models predict anything?
Skeptics and non-skeptics alike seem to agree that models cannot predict (i.e. make predictions that can be falsified, or confirmed, by observations) for timeframes shorter than around 25 years from the time of computation.

In fact, RealClimate seems to be willing to take a quarter of a century, more or less, as the minimum amount of time needed to get “averages” that can be called “climate” rather than mere “weather”. That is a second example of AGW climate scientists pigeonholing themselves: just as anything that cannot be modelled by forcings is “noise”, so anything that doesn’t cancel itself over 25 years is “noise” too.

So we started with “climate science” only to get stuck into “multi-decadal averaging to evaluate parameters to use in estimating the effect of forcings”.

Can anything ever disprove a forcings-based model?
No. Nothing at all ever will. Some AGWers are answering that with improbable claims about Popper being long dead, an eery reply one would expect only from inventors of perpetual-motion machines.

Actually, the prove/disprove question may simply be the wrong question. Models are only tools to investigate the possible effect of each forcing. Hansen et al talk about “using the model for simulations of future climate change”.

The key word there is of course “simulations”.

Models are not a weather-predicting tool (remember, they are about “climate”, not “weather”). And they are not a climate-predicting tool either, even if they are often abused as if they were. In its 2001 report the IPCC itself stated as much, in no uncertain terms: “In climate research and modelling, we should recognise that we are dealing with a coupled non-linear chaotic system, and therefore that the long-term prediction of future climate states is not possible” (from the IPCC TAR-WG1, 2001).

What models can do is simulate the effect of individual forcings in isolation, something that can never be observed anyway. They also simulate the cumulative effect of forcings, with added uncertainty as interactions must be modelled too. Such a cumulative effect is not necessarily expected to be observable either.

It must be stated that as far as I can remember RC has never claimed anything more. Good for them. Perhaps they could have been clearer before more clear and more often, but things are starting to move in the right direction, of late. As already quoted in a previous blog: “[…] The ensemble mean is monotonically increasing in the absence of large volcanoes, but this is the forced component of climate change, not a single realisation or anything that could happen in the real world. […]”

And Kevin Trenberth, a lead author of the IPCC-TAR report, recently wrote: “In fact there are no predictions by IPCC at all. And there never have been. The IPCC instead proffers ‘what if’ projections of future climate that correspond to certain emissions scenarios.”

Compounded Weaknesses
Don’t get me wrong: on its own, doing an estimation is part-and-parcel of conducting scientific research; computer modelling is a great tool for very complex situation; forcings are a good way to translate a system into a manageable model; and scenarios are the standard way to evaluate risk.

But with regards to forcings-based climate science, all of those combine together compounding their weaknesses rather than their strenghts: estimations are often subjective, computer models are used to study forcings rather than climate, forcings are taken as “real” even if they cannot be measured, and scenarios are interrogated not for current sensitivities but as forecasts.

They have become the basis for a large Intergovernmental organization, tens of international meetings, the collective action of thousands of people, one Oscar and one Nobel Peace Prize, all in the name of what every knowledgeable person knows it is impossible to predict.

What Kind of Science is Climate Science?
Restricted to “the computation of scenarios (the ‘what-ifs’ projections)”, climate modelling is a science (the “science of climate forcings”, in fact). And RealClimate is as good as it gets. The same applies to much of contemporary AGW scientific journalism and publications, including Scientific American, American Scientist, New Scientist, Nature, Science. And the BBC.

Just try, next time you read their reports, to imagine a world view (a “climate narrative“) where climatology, the most uncertain of exact sciences, is applied science, a policy-making tool where only forcings count and, among the forcings, only those of anthropogenic origin are relevant (as there is little to make policies about, for non-anthropogenic forcings).

That is too narrow a view to be useful for risk management, let alone to bring science forward. It may lead to worries wasting time worrying about possible future stronger hurricanes, rather than about certain concentrating on preventi present-day catastrophical levee failure for present-day storms.

Time to Expand the “Climate Narrative”
Models have been the cradle of climatology, Tsiolkovsky would have said, but we cannot live in the cradle forever. It is time to expand the “climate narrative”, by getting climate science of the models-forcings-scenarios hole.

Because “real” climate is much, much more than RealClimate.