I once met a developer who claimed he didn’t believe climate change was real because the software used for modelling isn’t used by many people and probably has lots of bugs. I don’t agree, but it was an interesting take.
Bugs and grad student software aside, ask anyone who thinks they can rely on a model to predict the future to implement a model of: a simple pendulum, a double pendulum, the three body problem, the stock market, and then finally the climate.
Those are roughly in order of difficulty, and if they fail at an earlier one, you shouldn't trust the later ones. The first one isn't even chaotic, and I wouldn't trust a model built from first principles alone to be in phase past a dozen cycles or so.
You can curve fit things after the fact (interpolation), but extrapolation is always on shaky ground.
Large scale climate models are more like modeling the possible energy distributions of such pendulums into the future. That can be done analytically for pendulums. You can do it analytically for very simple models of climate, too, but more complex models that include enough of the forces to be predictive require computers.
I've heard similar arguments before, but the details really matter. The amount of heat and CO2 is going to rely on things like albedo on the ground and from cloud cover, as well as plant mass and more. I think it's a mistake to ignore feedback on any of that, and it doesn't take too many moving parts with feedback to create a chaotic system.
If nothing interacted with each other, I think you could make reasonable energy-in / energy-out models. However even looking at big low-pass averages, plants/algae use CO2, and heat creates clouds, and clouds block sunlight, and so on. Unlike a double pendulum that bleeds a small amount of energy to friction, the climate bleeds a lot of energy into space, and the amount of energy it loses is a function of clouds, plant life, etc.
Sure, but we know it's near equilibrium, at least on human time scales, or the climate would not be stable enough for humans to evolve. So it's not chaotic like weather is chaotic.
I politely disagree, but my non-expert arguments shouldn't carry much weight. Here's a quote from a 2018 IPCC document instead:
"The climate system is a coupled non-linear chaotic system,
and therefore the long-term prediction of future climate states
is not possible. Rather the focus must be upon the prediction
of the probability distribution of the system’s future possible
states by the generation of ensembles of model solutions." [0]
I'd kind of like to dive in to the topic of whether an "ensemble of model solutions" is a fair and sufficient sampling of the problem space to trust the statistics, but I don't have enough details. However, I have done particle filters before, and when you have more than a few parameters to estimate, you need a shit-ton of particles before you can trust the statistics you get out. And that's with well behaved and fairly linear systems.
We can't predict what will happen when tipping points are tripped, but while we're near the current (dynamic) equilibrium, we can use perturbative methods to predict what will happen with relative small forces, such as doubling or trippling the amount of CO2 in the atmosphere, i.e. global average temperatures will rise.
Those are roughly in order of difficulty, and if they fail at an earlier one, you shouldn't trust the later ones. The first one isn't even chaotic, and I wouldn't trust a model built from first principles alone to be in phase past a dozen cycles or so.
You can curve fit things after the fact (interpolation), but extrapolation is always on shaky ground.