The latest, greatest exercise in absurd climate-modeling hubris is the frankly preposterous project to create “Digital Twins of the Earth” inside the models, to overcome their hopelessly limited power to simulate the actual climate by, um, PR or something. Fans of the BBC comedy classic by Douglas Adams, The Hitchhiker’s Guide to the Galaxy, will recall the business where some aliens fed up with endless philosophical debate create a giant supercomputer, Deep Thought, to solve the ultimate question of life, the universe and everything only to wait 7.5 million years then get the answer “42” and have to build another even bigger computer to frame the question more precisely, one so vast and mighty that it actually is an entire planet, indeed the Earth. But now satire has become reality, as Andrea Saltelli explains in a guest post “Digital Twins of the Earth: Science or Pseudoscience?” on Roger Pielke Jr.’s “The Honest Broker” substack.
Saltelli “of the UPF Barcelona School of Management in Barcelona, Spain and the Centre for the Study of the Sciences and the Humanities at the University of Bergen, Norway”, a rather eclectic mix, discusses the presumptuousness of claiming to have modeled the whole planet inside a computer that fits into an office, even if:
“In the current ecosystem of science digital twins of the earth are currently unstoppable given their considerable appeal and the supporting advocacy.”
We ourselves have pointed out repeatedly that, judged by actual results, the models don’t do a very good job even of explaining known past data, let alone predicting the future. But they’re immensely prestigious partly because of a general obsession with mathematical analysis, especially linear algebra, starting in places like chemistry where it works but reaching out to areas like economic modeling, whose record is just as bad as that of climate modeling, and partly because those who now control the discussion, from activists to politicians to the gatekeeping editors of prestigious climate journals, know very well that the models will predictably give the answers they want on what is going to happen, however badly they flub the test of what actually does or did.
As Saltelli says:
“some scholars from different disciplines, including myself, have recently argued that this faith in the capacity of mathematical models to describe the real says more about the marketing skill of their proponents than about the planet or its climate.”
And he’s scathing on the resulting scandalous situation:
“Models exist in a state of exception, having appropriated the academic prestige of mathematics and physics while at the same time escaping the critical gaze of philosophers and social scientists, including to some extent that of the sociologists of quantification…. modellers have acquired a central position at the heart of the climate change discussion, and make use of this privileged state to increase their political standing and funding, putting themselves at the helm of the climate change narratives and making climate change itself into an all-encompassing meta-narrative, subsuming all ailments of humans and their planet, inclusive of wars, authoritarianism, migrations, and various forms of aggression to planetary ecosystems. The project of digital twins represents the pinnacle of this movement.”
And just possibly from that peak it will descend precipitously, as hubris is followed by nemesis. It’s one thing, and reasonable, to try to make a “digital twin” of, say, a wind turbine, and conduct tests inside a computer instead of out in a field. But an entire planet? A planet you can’t model precisely because your computers aren’t big and powerful enough? It really does smack of hucksterism.
At one point in the Hitchhiker’s Guide, when Deep Thought denies being omniscient, he’s asked whether he is not the mightiest computer in the universe, more powerful even than the “The Milliard Gargantu-Brain at Maximegalon, which can count all the atoms in a star in a millisecond” and responds “The Milliard Gargantu-Brain? A mere abacus. Mention it not.” And compared to the actual Earth, the massively transcomputably-complex Earth, the modelers’ computers that cannot even manage clouds are pretty much in that category.
So it’s astounding that the modelers, instead of exhibiting a bit of humility, are now claiming that they can escape the limitations of these feeble approximations using these feeble approximations, generating a full-fledged digital twin of the Earth on the same hardware that can’t even create a convincing simplification of it.
The only way to create a computer that really could model the Earth in all its non-linear complexity would be to make another full-size Earth. And the difficulty, apart from the fact that you obviously cannot do such a thing unless you are a Magrathean custom planet-builder (and anyone who wrongly thinks they are is better suited to a strait jacket than a lab coat), is that if you did, it would like the Earth be transcomputably complex and yield not crisp reliable linear projections but precisely the randomness the actual climate exhibits. You’d be right back where you started except, just possibly, a bit wiser.
As Saltelli writes:
“We advance doubts about the promised scientific advancements brought about by scaling mathematical models down to the scale of the kilometer: ‘The higher the resolution (i.e., the greater the localisation), the more non-physical feedbacks emerge as relevant, whether it be the micro-climatic effects of forest stands or the albedo micro-patterning of slush ponds on ice sheets. Scale change often brings about non-trivial shifts in complexity and governing principles, and finer-grained scales may reveal deterministically-driven chaotic behaviour. A different resolution may require different, perhaps yet unknown, process descriptions.’”
Chaotic behaviour. There’s the rub. The only reason the current models do not exhibit it is that they’re told not to, and are sufficiently simple, far simpler than the real climate, that it can be prevented by determined programming. If you really did create Earth Mark Two, all it would tell you is that climate is too complex to model.
P.S. In the Hitchhiker’s Guide, Earth Mark Two is cancelled. As it should be in real life.
" All models are wrong, but some are useful." George Box, in his paper for the Journal of the American Statistical Association, didn't evisage just what some people might consider useful.
Models,Schmodels!Mostly not to be trusted.GIGO!
It would be easier to figure out the rules of a game of CalvinBall than all of the rules governing climate in any particular locale…
Cui bono?
Models are just like a government adviser.
Government: How will carbon destroy our climate.
Adviser scientist: We have limited time to reduce carbon. We need to hurry. Even more than I said ten years ago. It is dire times.