Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Do climate models predict extreme weather?

backreaction.blogspot.com

Post reply on HN
Print view View original

61–70 of 223 posts

Re: Do climate models predict extreme weather?

#61

Post by labster »

contravariant wrote: post #25
timmg wrote: post #5

One of the things about "extreme weather attribution" that makes me uncomfortable is the asymmetry of it. To be clear: I'm in no way an expert (or even educated) in the area. To give an example. Let's say there's a 1% chance every year that there will be a catastrophic wildfire in some particular area. Now let's say climate change increases a chance of drought in that area. Then we get a wildfire one year and the att…

That's why in these discussions it's usually better to talk about specific statistics like 'the chance of a wildfire has gone up by 20%', or perhaps more clearly 'the average time between extreme floods has gone down from once every 100 years to once every 60' (which I find communicates the point a bit more clearly). That said personally I think that should be as far as the discussion goes, talking about a specific e…

As a climatologist by education and a dabbler in politics, I think you are radically misjudging the ability of most people to have rational discussions about unwelcome statistics. If you say severe weather will increase by 20% due to climate change, people will hear that there’s an 80% chance weather will stay the same. And they say I’m just being alarmist.

When I say that this storm may have been enhanced by climate change, but it could also could have happened anyway, statistically: people hear natural variability, and that I don’t know for sure that climate change impacted their daily lives. And if I don’t know for sure, there must not be much of a case for climate change.

Most of the world is not HN, where nuanced discussions work. Most of the world does not give a shit about climate change unless they will personally suffer in some way. So I lie. So that even if this storm or that hurricane didn’t hurt them, they will have empathy for those who did suffer from a climate-enhanced storm. Because if we wait for everyone to feel the impact in earnest, it will be far too late.

Re: Do climate models predict extreme weather?

#62

Post by avianlyric »

fasteddie31003 wrote: post #28
avianlyric wrote: post #26

Earlier quoted context omitted.

It’s complex and chaotic, sure. But as you increase time horizons, and area of simulation larger and larger, that complexity and chaos starts to average out. So you can’t take a singular datapoint from a climate model, and make any concrete statements, but you can take thousands to millions of them and start making probabilistic statements. Additionally it’s not like these models are untested. Their predictions are c…

Can you give me another example of a chaotic system that can be correctly forecasted years out?

Just about any basic physical system. If you zoom into our universe enough, down to the quantum level, you’ll find completely unpredictable chaos and complexity. But just because the quantum world is pretty much unpredictable doesn’t mean the macro world is unpredictable.

A climate model can’t you whether or not is going to rain in exactly 300 years time, but it can tell you how much rain will fall during an entire year, 300 years from now.

Climate modelling is macro scale modelling to weathers quantum scale modelling. You’re operating at entirely different scales, and further you zoom out, the more you can ignore the minutia and still produce useful predictions.

Additionally taking about absolute time periods is the wrong approach. The real question is now many time steps can your model take before it starts to significantly diverge from reality. A car collision simulation will be operating with time steps measured in milliseconds to nanoseconds, a climate model in time steps measured in days to weeks. Both of these models will probably be predicting a similar number of time steps into the future, and both will have a similar number of quantised simulation areas. For the car you’re simplifying cubes measured in millimetres, for climate, zones measured in km.

Again, climate models have been measured, tested and repeatedly proven their predictive power. If empirical results aren’t enough to convince you, then what is?

Re: Do climate models predict extreme weather?

#63

Post by credit_guy »

Looks to me Sabine made a fairly elementary mistake towards the end. She mentions 3 probability distribution curves, 2 model produced (blue and red) and one empirical (green). She forgot to mention a fourth one, the empirical-preindustrial. In any case, you'd like to be able to measure the ratio of empirical distributions, but let's say that's difficult because of data retention. So, instead of having the actual ratio x/y you have a ratio of proxies, x'/y'. This ratio could be both higher and lower than x/y. She claimed it can only be lower, and that's not quite true.

Re: Do climate models predict extreme weather?

#64

Post by Misdicorl »

fasteddie31003 wrote: post #7

I believe in climate change, but if you look into how the climate change computer models are written, it really makes me feel like it is contrived. Here's the problem with climate change modeling, you are modeling a complex chaotic system, you just cannot do it accurately. You have to be honest about it.

This sentiment is so frustrating to read. Fundamental science is hard. Working with extremely complex systems is hard. Putting them together is extremely hard.

That does not mean you cannot do it accurately. It means you have to continually refine and improve and throw out and restart and refine and....

Scientists as a group are by far the most honest group of people you can point at. No other profession comes even close. Alluding to some dishonesty in the work behind climate change is straight up denialism. Yes some of the science has not panned out. That's how science works. It's part of the bones.

Re: Do climate models predict extreme weather?

#65

Post by sanxiyn »

dpierce9 wrote: post #60
sanxiyn wrote: post #55

Earlier quoted context omitted.

Lung cancer risk was established with control group. We don't have control group for a planet so we use climate models. The article is about remembering that we use models because of lack of alternatives, not because they are as good as control group.

The parent was suggesting causal inference from general statistical evidence to particulars is impossible. You seem to be suggesting there can be no causal inference where there are no control groups. Both positions are wrong.

I am saying it is better to have control groups than not. Hopefully that is non-controversial.

Re: Do climate models predict extreme weather?

#66

Post by notatoad »

timmg wrote: post #5

One of the things about "extreme weather attribution" that makes me uncomfortable is the asymmetry of it. To be clear: I'm in no way an expert (or even educated) in the area. To give an example. Let's say there's a 1% chance every year that there will be a catastrophic wildfire in some particular area. Now let's say climate change increases a chance of drought in that area. Then we get a wildfire one year and the att…

"a fire this year" is a really vague and imprecise way to measure things, and leads to trying to conclude weird things like whether or not a quiet fire season was caused by climate change. that's why scientists don't measure things that way.

the more precise way to measure would be to increase the sample size by counting a precise metric like total acres burned per season, and measuring it over a longer period of time to determine whether or not forest fires are generally getting worse or not. and if they're getting worse, how much are they getting worse by, and how much of that trend can be attributed to climate change.

Re: Do climate models predict extreme weather?

#67

Post by magicalhippo »

refurb wrote: post #31
nabla9 wrote: post #14

Earlier quoted context omitted.

That's not true. Climate models model and simulate physical and chemical condition on Earth. Modellers don't know what they say beforehand. Offhandedly portraying scientists and researchers as dishonest and conspirators in every opportunity possible seems the default here.

Models are constantly tweaked to make them more accurate. How do you do that without looking at output and then going back to tweak it?

One way is to verify them. Run multiple variations on data from the 1950's and see how they compare to measured values up to today. Throw away the models that don't fit the measured values. Then you have some confidence that the remaining variations may be representative when run into the future.

This is also done in weather forecasting, though maybe they run the model in reverse, I can't recall. Same idea though, verify with existing data before predicting.

Re: Do climate models predict extreme weather?

#68

Post by avianlyric »

godelski wrote: post #58
8note wrote: post #37

Earlier quoted context omitted.

Generally the defining feature of non linear mechanisms is that the long term behaviour is non-intuitive. I'd expect that to include not having a predictable mean

Non intuitive doesn't mean not predictable. Nor does chaos. Chaos just means that a small perturbation in input parameters can result in a large change in the output. So to have good predictive power you need to pick good inputs with good accuracy. But chaos is also a spectrum so not all chaotic systems are as sensitive as others.

Chaos is also different at different scales. At certain point chaos can be simulated as noise with a known distribution. You can’t make predictions within the noise, but you can describe the shape of the noise and it’s impact on macro scale systems.

Perfect example of this is quantum effects. Zoom in far enough and quantum effects make the world entirely unpredictable, and yet I can still accurately predict where I throw is going to land with no knowledge of quantum effects. That doesn’t mean those effects don’t influence the ball, it just that noisy chaos of those influences is tightly bound enough that it doesn’t matter. Newtonian physics is enough to make a prediction, despite being an incomplete description of the physical system at play.

Re: Do climate models predict extreme weather?

#69

Post by sampo »

sampo wrote: post #59
freen wrote: post #18

Earlier quoted context omitted.

I am 100% sure that the Exxon model from 1982 was not what Exxon wanted. It also has been and continues to be tremendously accurate and predictive. https://xkcd.com/2500

> https://xkcd.com/2500 I am unable to find any 1982 Exxon prediction that would meander like the line in that graph. The only plot I was able to find is this [1] from this 1982 Exxon report [2], which is just a slightly curved upward rising sketch, not nearly as detailed as in the xkcd graph. > the Exxon model from 1982 I also failed to find any signs of there having been any "1982 Exxon model". What [2] is, is a 39…

Oh right, I misinterpreted the xkcd plot. It's only the single (x)-marked dot in the top right of the graph, that is the Exxon prediction. Not the graph itself.

Re: Do climate models predict extreme weather?

#70

Post by sanxiyn »

godelski wrote: post #51
fasteddie31003 wrote: post #7

I believe in climate change, but if you look into how the climate change computer models are written, it really makes me feel like it is contrived. Here's the problem with climate change modeling, you are modeling a complex chaotic system, you just cannot do it accurately. You have to be honest about it.

This is verifiably false. We can look at the predictions (inference) from older models and check how accurate they were at predicting the future. No back testing required. Those models have been very accurate. Current models have only improved and reduced error. We can test the accuracy, you have to be honest about it. https://www.science.org/content/article/even-50-year-old-cli...

Your point that we can check prediction against reality is well taken, but honestly I think we got lucky. Knowing what we know, 50-year-old model shouldn't have been accurate. It didn't model aerosol, which we now know to be important. It's just that errors happened to cancel out.
Post reply on HN