The problem with climate modeling

2025-07-20 01:34 by Ian

Epistemological admonishments from a mathematical modeler

"The model is not reality."

The loudest voices about this topic have no idea what they are talking about. Many are conflicted, many more are simply NPCs and repeating an opinion that isn't their own (and doing so with religious zeal [an excellent indicator of an NPC]).

Disclaimer: I first read an IPCC model in 2013. Several of them. I read a few more in 2018. But have read none since then. My knowledge of their models at the time of this post is stale and degraded. Go read the models if you think I'm wrong, and email me. I will make a retraction if I am.

Pollution is not climate

Before the distraction happens, I want to state that I'm not talking about the great pacific garbage patch, nor the Hanford Nuclear Site, nor any such problem of pure pollution (that is, pollution that nobody wants). Those concerns may or may not be valid. But they are not germane here.

CO2 is not pollution

CO2 is produced intentionally at industrial scales for industrial purposes. It is used as a solvent, feedstock for reactions to make plastics, and also for feeding greenhouse atmospheres. Humans literally burn high-grade propane to produce CO2 for better plant growth.

If someone will pay for it, it has value.
If it is a conscious industrial product, it is not an externality.
If it is part of the carbon cycle (it is), it will find balance by way of Le Chatelier.
It is not poisonous, absent knowledge of the dose. Remember: Oxygen is also poisonous.
At the scales humanity is capable of forming it, CO2 is not pollution. So stop viewing it that way, as a first step.

"The" climate model, you say?

By my estimate, only 1-in-50 people who believe they have an opinion on climate change know that the IPCC "model" that they often talk about is actually a mash-up of at least 150 (last time I looked) submodels, each with its own (often mutually incompatible) free-parameters. IE, there are models that deal with ocean acidity. And others for cloud-cover. And yet others for atmospheric concentrations of various gasses. Many (most?) of these models cannot even predict the past. That is to say: when you feed them with historical data, they do not produce results that actually happened near the prediction date.

IIRC, the IPCC atmospheric temperature model used cubic cells more than 16km across, and didn't model convection at all.
Re-read that again if you need to.

The model assumes that volumes of atmosphere big enough to hold an entire storm has the same temperature everywhere. And they don't handle convection because modeling convection in a computer is damned expensive for any volume at decent resolution. Let alone a planet. So they didn't model the dominant mode of matter and energy circulation in the atmosphere.

So the "climate" models can't model "weather". Literally cannot without convection, as the weather forecasters do. The best response I've ever received to this criticism of climate modeling amounted to "climate and weather have nothing to do with each other", but doesn't sound credible to someone who just watched them gloss over the mixing effects of that weather by using a static free parameter where a proper convection-aware model should have stood.

I suspect this is part of the reason the zealots take so much issue with (valid) retorts having the flavor:
"When you can tell me a month in advance if it will rain, I will care what you think the temperature will be like next year."

Some of the IPCC component models come into mutual contention. That is: two (or more) models might disagree on several of their mutual variables. Ocean temperature model disagrees with the cloud cover model about air temperature (suppose). So the component models are tuned and finally passed through a "meta model" (I forget what the IPCC calls it) that is hand-weighted (more free parameters) to blend model disagreement into an epistemological mush that they believe looks plausible.

So there is no "climate model", as much as there is a very expensive suite of mathematical tinker-toys that are glued together by a hand-tuned biasing to produce a periodic report. That report is quote-mined for click-bait, and used as a prop-piece for grant money or policy favors. It is thoroughly tainted and has no predictive value.

Free parameters

If you don't already know what "free parameters" are, it means that you don't have an opinion on climate change. You have somebody else's that you may have even come to believe is your own.

When you talk to a mathematical modeler about "free parameters", it triggers a large group of concerns and questions about your model. The more free parameters a model has, the more "hard-coded" behaviors it has, and the more opportunities the modeler has to "get what he wants" out of the model, rather than "what may plausibly be". Sometimes modelers do this without realizing it. Others are outright frauds, and consciously make deceptive models for a living (profitability and stock market projections, to name a popular genre).
Sometimes, what appear to be sensible free-parameters are later discovered to be hideous over simplifications, or even deliberate fraud. In any case, specific values of free parameters are always a product of selection bias.

No matter how many free parameters a model has, if it is a system of differential equations, it also suffers from what we can regard as "granularity-induced noise". Granularity in both space and time. Typically, the more granular a model is, the better. But computers and observable reality put strict limits on how granular the model can reasonably be either in space or in time.

About those delta-epsilon values...

Imagine trying to predict where a hard rubber ball will come to rest when dropped on a rough (but level) dirt road. If the model considers the road as flat it will give correct answers approximately never. The real road has pebbles. And although you can precisely know the height of the drop, and its point of first impact, that pebble was enough to send even the first bounce off in the "wrong" direction, according your model.
So... what? Get a better map of the road if you want to better predict the ball, right?

Ok, sure... Spend a few thousand dollars on LiDAR scanners, map that road down to the mm, and cross your fingers that nobody but you notices that it isn't even remotely detailed enough, and you need to re-scan the area each time you drop, since the surface is gently re-arranged by the act of dropping the ball to an extent that it matters to the validity of your model. You find that no matter how precise you are, you can almost never predict the location of even the second bounce. The sixth bounce is clearly unattainable foreknowledge.

So what about time? That simulated ball isn't going to drop itself. And gravity isn't going to be the global average value of 9.8g like it was in your physics textbook (free parameter). And you aren't going to model every microsecond of an event that takes several seconds to complete.
Or perhaps you will, and it takes you a few $10,000 bills from your Kubernetes host before you discover why some places on your dirt road tend to reflect the ball very differently when you model only every third microsecond (or every 100ns).

If you aren't convinced by the difficulty of modeling that ball, it means you didn't actually try to do it. And the rubber ball is much easier than convection.

Are you certain you want to talk about sea-level?

First, a review of our situation....
You have lived your entire life on the hardened surface crust of an otherwise molten blob of (mostly) iron and silicate salts.
The blob is in the bottom of a gravity flask, and is being kept warm by latent heat, nuclear decay, sunshine, and strain from various sources (including its own weight).
Now... The oceans are all floating on that same crust. And although it is moving slowly, it is moving. And that will change the ocean's volume.

Sea level is not a simple thing to measure. If you do it by barometry, mean sea level might be 20 meters different in different oceans. Earth is not a sphere. It is clumpy and without uniform density (see the GRACE satellite missions).

The distance along the equator is longer than the distance around the Prime Meridian (That is, our daily rotation "squishes" the planet). It certainly does the same to the oceans.
Are you aware that the length of our sidereal day is always changing? The faster Earth spins, the more sea level at the equator will rise, and fall at the poles.

All of that said, mean sea level at any location will be at approximately the same distance from the Earth's barycenter. Since we are approximately spherical, that will be approximately the center (the position of which will drift over time).

If you can't give a well-reasoned explanation (right or not) for the sea level difference on either end of the Panama Canal, I can't even feign interest in hearing what you think you know about sea level, our ability to measure it with a noise floor comporting with an asserted signal, and certainly not predictions about it. You have a received opinion that you falsely believe to be your own.

Are you certain you want to talk about greenhouse gasses?

Most people who think they have an opinion on greenhouse gasses can't read a spectral absorption graph. If they could, I wouldn't so commonly encounter the expressions I do when I tell them that water vapor is a far more effective infrared absorber than both CO2 and methane.

But Greta cried on TV, you monster.

Bullsh!t sounds better when it is constant

It's 2025. Are we still on "climate change"? I'm old enough to remember when it was "global cooling".
Don't tell me it wasn't. Leonard Nimoy made a documentary about it.

Than, in the 90s, the modelers and their attendant NPCs were yelling about heat, melting glaciers, and other such hysteria. And that went on for a long time.
Don't tell me it didn't. Al Gore made a documentary about it.

Finally, sometime in the late 'oughts, the PR people apparently decided that it was hard to be credible when every specific prediction failed laughably, and they started calling their fake science "climate change" so they could definitionally never be wrong. Of course the climate is changing. Why wouldn't it be? Literally nothing stays the same after enough time passes.

But the idea that humans have any measurable impact on the change (or are even capable of predicting the change) is scientifically naive.
Hubris. Evidencing deficient standards for knowledge.
Probably a mix of all of those things.
But because the "science" now has a tautological label, the degree to which their models actually were wrong can be blamed on the fact that you ate a hamburger today.
Don't tell me it isn't. Leonardo DiCaprio made a documentary about it.

Celebrities are hired for PR campaigns. Which is what "climate change" usually is. And while we might poison the planet to the extent that we can no longer happily live on it, I doubt we could even do that if we tried.
Seriously... we are a fart in the wind on this planet. Get over yourselves, and spare me your concern trolling.

Previous:
Next: