Scrutable

Here’s a really nice article about uncertainties in climate modelling, well worth a read:

https://e360.yale.edu/features/why-clou ... on-warming

Interesting but worrying read.

Can I suggest that that gets its own thread, as it's definitely off-topic in "Things wrongly attributed to climate change"?

(I'll do the report thing)

dyqik wrote: ↑

Thu Feb 06, 2020 2:15 pm

Can I suggest that that gets its own thread, as it's definitely off-topic in "Things wrongly attributed to climate change"?

(I'll do the report thing)

Thanks, new thread created.

Woodchopper wrote: ↑

Thu Feb 06, 2020 3:00 pm

dyqik wrote: ↑

Thu Feb 06, 2020 2:15 pm

Can I suggest that that gets its own thread, as it's definitely off-topic in "Things wrongly attributed to climate change"?

(I'll do the report thing)

Thanks, new thread created.

Ta.

Back on-topic: This lines up with what climate realists have been saying for a long time - yes, there's uncertainties and risks of errors in climate modelling, but those are two-sided probability distributions, and are at least as likely to be underestimating positive feedbacks that will destabilize the climate as underestimating negative feedbacks that stabilize the climate

dyqik wrote: ↑

Thu Feb 06, 2020 3:02 pm

Back on-topic: This lines up with what climate realists have been saying for a long time - yes, there's uncertainties and risks of errors in climate modelling, but those are two-sided probability distributions, and are at least as likely to be underestimating positive feedbacks that will destabilize the climate as underestimating negative feedbacks that stabilize the climate

Not disagreeing, just being nerdily pedantic - I'd quibble with "at least as likely".

Estimating uncertainty in model results is challenging <wiki>. Even if there are available reliable forms and parameters for the uncertainty distributions of the various model parameters and structures (and there aren't!), combining them to estimate uncertainty in results isn't easy. I agree that there is no reason to believe that errors in climate modelling lead to over-pessimistiic results rather than over-optimistic.

However commentators including the author will focus on the negative potential, which in some ways is fair enough, but we’re talking about some extremely expensive policy decisions here.

True, but the costs of underestimating versus overestimating the impacts are highly asymmetric.

KAJ wrote: ↑

Thu Feb 06, 2020 7:12 pm

dyqik wrote: ↑

Thu Feb 06, 2020 3:02 pm

Back on-topic: This lines up with what climate realists have been saying for a long time - yes, there's uncertainties and risks of errors in climate modelling, but those are two-sided probability distributions, and are at least as likely to be underestimating positive feedbacks that will destabilize the climate as underestimating negative feedbacks that stabilize the climate

Not disagreeing, just being nerdily pedantic - I'd quibble with "at least as likely".

Estimating uncertainty in model results is challenging <wiki>. Even if there are available reliable forms and parameters for the uncertainty distributions of the various model parameters and structures (and there aren't!), combining them to estimate uncertainty in results isn't easy. I agree that there is no reason to believe that errors in climate modelling lead to over-pessimistiic results rather than over-optimistic.

In principle, there's lots more way to be wrong that result in an unstable climate than ones that result in a stable climate not too far from where we are now. Of course, some of those can be ruled out because we haven't ended up there already. However, we know that other planets from our sample of three rocky planets with atmospheres have runaway.

dyqik wrote: ↑

Thu Feb 06, 2020 8:11 pm

KAJ wrote: ↑

Thu Feb 06, 2020 7:12 pm

dyqik wrote: ↑

Thu Feb 06, 2020 3:02 pm

Back on-topic: This lines up with what climate realists have been saying for a long time - yes, there's uncertainties and risks of errors in climate modelling, but those are two-sided probability distributions, and are at least as likely to be underestimating positive feedbacks that will destabilize the climate as underestimating negative feedbacks that stabilize the climate

Not disagreeing, just being nerdily pedantic - I'd quibble with "at least as likely".

Estimating uncertainty in model results is challenging <wiki>. Even if there are available reliable forms and parameters for the uncertainty distributions of the various model parameters and structures (and there aren't!), combining them to estimate uncertainty in results isn't easy. I agree that there is no reason to believe that errors in climate modelling lead to over-pessimistiic results rather than over-optimistic.

In principle, there's lots more way to be wrong that result in an unstable climate than ones that result in a stable climate not too far from where we are now. Of course, some of those can be ruled out because we haven't ended up there already. However, we know that other planets from our sample of three rocky planets with atmospheres have runaway.

Again, I don't think I'm disagreeing with the thrust of what you said in "Back on-topic:". I'm not criticising, other than pedantically, just suggesting a more careful wording. I'm not familiar with the various climate models, but I guess the results are something like maximum likelihood, and I've seen results quoted with confidence limits so there must be elements of probabilistic modelling. When I was involved in modelling I found that uncertainty in physical values was substantially [understatement!] easier to model than uncertainty in the nature and parameters of relationships. I agree that denialists are likely to concentrate on errors in one direction, and that climate realists (I like that phrase) should be ready to point out that errors can be in either direction. But "at least as likely to be underestimating positive ... as underestimating negative" quantitatively compares two things (likelihoods of positive and negative modelling errors), neither of which is quantitatively known with anything like confidence.

With respect to "lots more way(s) to be wrong that result in an unstable climate than ones that result in a stable climate", I don't really understand it, but I don't think you're using "ways to be wrong" to mean likely modelling errors, comprising modelling uncertainty. If you are then you seem to be saying that modelling errors in the direction of unstable climates are more likely than errors in the direction of stable climates. Your next two sentences seem to relate to the direction of model results, not the direction of errors in model results, so I deduce that this paragraph is about the relative likelihoods of stable and unstable climates. Although it would be entertaining to follow such handwaving arguments (and I'm not being critical) I have neither expertise or time, so I'll yield to the accepted models and say if we don't alter things quickly the state of the climate will [very, very likely] change to our detriment.

OK. Interesting, but things to do. Over and out.

I mean failing to model mechanisms that haven't been included yet, for example, feedbacks that only show up in the bizarre and unnatural climate scenario we find ourselves exploring.

There's modeling errors due to uncertainties in the values that get put into terms in the model equations, which are quantifiable, (known unknowns, if you will) and then there's modeling errors due to terms no one knew existed or thought to include, or assumptions no one really explored before (unknown unknowns).

dyqik wrote: ↑

Fri Feb 07, 2020 1:48 am

I mean failing to model mechanisms that haven't been included yet, for example, feedbacks that only show up in the bizarre and unnatural climate scenario we find ourselves exploring.

There's modeling errors due to uncertainties in the values that get put into terms in the model equations, which are quantifiable, (known unknowns, if you will) and then there's modeling errors due to terms no one knew existed or thought to include, or assumptions no one really explored before (unknown unknowns).

Yes. When I said:

I found that uncertainty in physical values was substantially [understatement!] easier to model than uncertainty in the nature and parameters of relationships

by

• "uncertainty in physical values* I meant what I think you mean by "uncertainties in the values that get put into terms in the model equations"
• "uncertainty in the nature and parameters of relationships" I meant what I think you mean by "failing to [correctly] model mechanisms"

In my (non-climatology) experience the latter is by far the greater, and extremely difficult to estimate quantitatively. One really has to trust (or not!) the experts to appropriately model appropriate relationships, and to appropriately assess and express their doubts, probably non-numerically.

I would hesitate to have this conversation with climate change doubters, fearing that they might seize on it to avoid action which I consider necessary.

Right now the conversation is with policy makers, who will be facing several extremely problematic alternatives, all of which will have economic impacts by disrupting business as usual - which is obviously why so little has happened to date. Scientists need to be very frank and honest about uncertainty in order to avoid genuine consequences in the here and now.

(for the avoidance of doubt, I agree we need to take action, and lots of it. But the more cuts to CO2 we make immediately, the harsher the consequences, so we’d better be damn sure that “extremely drastic “ is the right course of action)

This is cool, climatic impacts of contrails can be substantially reduced with adjustments in the altitude of a very small proportion of flights, with relatively tiny impacts on fuel use:

https://pubs.acs.org/doi/10.1021/acs.est.9b05608

note the abstract suggests that contrails have the equivalent impact as the rest of aviation emissions combined, something I didn't know.