[ooobit2]

The fact that we were just saying, as recently as 2016, that things hadn't become "catastrophic" yet, only to four years later make yet another correction that vastly accelerates past existing predictive models is why climate change continues to be seen as debatable. We need to either admit that we don't know or start acting consistently like we do know.

Pushing out an article with a headline like, "Major-global-feature is past the point of return," is not exactly selling the idea of a carbon tax. And the only argument against the critics has been, "Well, it could get even worse!" You mean in 60 days, when we announce something like "China will be underwater by February 2022?" Investments to curb climate change start to look like snake oil when we're saying something can be fixed by it but then announcing 30 days later that it no longer can.

[mhh__]

Which models are these? (i.e. [Citation needed])

The fact that people actually deny warming let alone the consequences of said warming does not bode well either - https://www.youtube.com/watch?v=kTk8Dhr15Kw

[magicalhippo]

> The fact that we were just saying, as recently as 2016, that things hadn't become "catastrophic" yet, only to four years later make yet another correction that vastly accelerates past existing predictive models is why climate change continues to be seen as debatable.

So they got the speed wrong, but it was still in the same direction.

Scientists have for years been worried about the state of the Greenland ice sheet, and have repeatedly said it's diminishing ever more quickly and that's not good.

It's not like they went 180 on this.

[ooobit2]

It's absurd I have to reply with this: Climate change continues to be seen as debatable because climate change models keep being corrected past the degree of correction most people understand invalidates similar models in other fields of science.

So, using a pharmaceutical as an example: If Pfizer published findings that stated an HBP pill permanently reduced blood pressure by a specific amount after 60 days, then four years later came out and corrected that to 10 days, most people would be concerned that the extra 50 days of treatment carried significant risk of lowering blood pressure too far, and would ask whether Pfizer did enough to prevent such a significant risk of lowering blood pressure too far. And almost everyone would be more cautious about taking other Pfizer medications. Makes sense, right?

That's what people do with climate science models in the general public. When someone asserts confidence in a model one year then publishes a significant correction to that model the next, skepticism over the integrity of all of their models is reasonable. If they messed up so bad here, we need to know whether it's isolated in scope.

That said, the entire model was about speed. No one inferred a general direction of decline. They asserted high confidence in a model that, given additional data, completely changed. There is a rule of logic in data that as new records are added, each record becomes an increasingly smaller proportion of the overall dataset, and save for extreme anomalies, the expected rate of change for the model given each additional record declines on a curve. That is why random sampling of a population is used instead of the population. After a certain number of records, the relevance of any N new records are insignificant to the output. If the addition of new data completely threw off their model, even as a time series, they're overfitting or underfitting, but either way, they are not going to be trusted as a source of reliable information.

What I'm saying is we need to stop doing that. If we talk about the models as just being marginally accurate, we're more likely to stem attempts to debate the entirety of climate science. Regardless of how smart or stupid you think those who debate climate change are, they do influence barriers to act.

[magicalhippo]

> Makes sense, right?

No, because it's an entirely different scenario. It's not at all similar to making some blood pressure medication.

[googthrowaway42]

This is like the 538 election models that swing wildly as more polling data comes in. A model that is so sensitive to new data is a pretty poor model since it should have roughly predicted that new data in the first place.

When you combine that with the fact that it is anathema to question Climate Science orthodoxy and this is wielded for massive political and financial ends, the only rational approach is to be extremely skeptical.

Central to this discussion is the point that climate science is not science in the same way that physics is. There are no repeatable experiments that can be run.

[ooobit2]

You point out a key fault in the approach people employ to derive inferential outputs vs. descriptive outputs. I feel like I could rant about the obvious wide MOE and high alphas that have to be employed for inferential models. But I also feel like it's a matter of convenience. Incredibly smart, well-learned data scientists are either subject to their own bias, to their employer's bias, to the audience's popular appeal, and a number of other skews to usable inferences.

Let's say we have 10 models of Greenland's ice sheet predicting non-linear declines in total ice at 10 different progressive rates. If those 10 models have an MOE in the top quartile, and the rate of change for the relevant variables meets high confidence, I'd say we can reasonably infer melt will progress similar to some of those models. I wouldn't shell out a publication declaring it an accurate model, or even one professionals should use to guide decisions. But, given additional precautions are taken, I would expect a sound decision to consider inferences from the 10 models.

I'm talking a whole different script and procedure for handling inferential outputs, communicating them to stakeholders, and setting fair expectations. It's an approach that people don't use, even though it is necessary. At the very least, it averts the kind of surprise conflicts in data that undermine productive, preventative planning and action.

Instead, like you point out, the approach and the bias inherent to the outputs necessitates extreme skepticism. I remind myself a lot of that line from Rick & Morty, when Rick says, "Sometimes science is more art than science." Data is the same kind of beast, especially as you attempt to work around rudimentary entropy. Physics isn't any less subject to the same context. We already have proof that the thermodynamics we see here on Earth are not uniform across the entire Universe. If you and I were standing at different ends of the Universe, I could as easily call you a quack for your declaration that a dropped ball's speed and trajectory can be predicted by your equation.

Anyway, I could keep going, but we both know the downvotes are self-righteous, self-aggrandizing people who have to be right to as deep an extreme as necessary. It's the only way they can sleep at night, so they can wake up and compound the mess they already compounded the previous day.