"There is a lot of stuff that we don't understand and don't account for in climate models" vs "There is a lot of stuff that we don't understand and don't account for in climate models... and so climate models should be ignored"
HN voting isn't always good at parsing nuance.
It's an interesting area I'm unfamiliar with. How are uncertainty bounds communicated in complex, compounding simulation models?
For simpler models you can do a sensitivity analysis by rerunning the model while systematically varying the uncertain parameters. Based on reading a handful of IPCC papers, a similar method is used in climate science for the most important parameters, but is severely limited by the resources needed to rerun the models and the sheer number of parameters.
Basically, any uncertainty due to complexity is mostly ignored and any skeptics who point this out are tarred as "deniers." Combine this with the fact that generally climate scientists do not produce any point predictions to verify models, and I think you can safely ignore many climate models. The papers that do produce explicit measurable predictions tend to find an interesting result by stretching the models to fit a narrative, eg: https://www.nature.com/articles/s41477-018-0263-1
How are uncertainty bounds communicated in complex, compounding simulation models?
Poorly to not at all.
Climate models have been diverging over time. As details are added (e.g. to try and handle clouds) the different teams making models get different results, and the gap between those results has been widening. We are told the science is certain but the reality is the science has got less certain. Journalists don't like to cover this because they see themselves as change agents, and discussing the actual levels of uncertainty would discourage change. This article in WIRED is an interesting exception, maybe it means something.
The gap between models is compounded by another problem. The individual models are often very buggy and suffer severe numerical instability. Even when told to assume no CO2 emissions at all, for which they are programmed to assume a steady state, they often end up predicting a massive ice age or that Earth becomes as hot as Venus. These runs are simply discarded and the issues are covered up.
Because these problems would cause ultra-wide CIs, climatologists prefer to just take an average and then present the predictions without accompanying measures of uncertainty. They also do this with the input data, for which there is also wide uncertainty.
They have little choice in the matter. Climatologists know their models aren't much good. Last year some of the most famous climatologists wrote an article in Nature warning scientists that there was a "hot model" problem, i.e. models were over-predicting actual warming, the very problem skeptics had been talking about for decades. It's paywalled but there's a copy here:
We are climate modellers and analysts who develop, distribute and use these projections. We know scientists must treat them with great care. Users beware: a subset of the newest generation of models are ‘too hot’2 and project climate warming in response to carbon dioxide emissions that might be larger than that supported by other evidence3–7. Some suggest that doubling atmospheric CO2 concentrations from pre-industrial levels will result in warming above 5 °C, for example. This was not the case in previous generations of simpler models.
"There is a lot of stuff that we don't understand and don't account for in climate models" vs "There is a lot of stuff that we don't understand and don't account for in climate models... and so climate models should be ignored"
HN voting isn't always good at parsing nuance.
It's an interesting area I'm unfamiliar with. How are uncertainty bounds communicated in complex, compounding simulation models?