7 ms·

The most powerful supercomputers in the world are focused on weather analysis and they can only project a few days out. Yet, some people think they've got it f

The most powerful supercomputers in the world are focused on weather analysis and they can only project a few days out. Yet, some people think they've got it figured out.

justwalt 7y ago
That does put things into perspective, a bit.
- ceejayoz 7y ago
  Nah, it's a common talking point, but it's highly flawed. It's like claiming that because we can't predict if the stock market will go up or down tomorrow, we can't determine whether we're in a recession or a boom time. Short-term weather is fairly chaotic (and we've gotten a lot better at predicting it over the last 50 years). Long-term trends are different.
  - jessaustin 7y ago
    If you can figure out recessions and booms in real time you can make lots of money.
6d6b73 7y ago
The most powerful supercomputers in the world are not even close in processing power of a 4 year old. Isn't it possible that some people can see patterns that the supercomputers can't?
- chrisco255 7y ago
  Sure, but those supercomputers aren't autonomous, they're programmed and designed by thousands of human adult experts.
  - 6d6b73 7y ago
    And these people can make mistakes or be biased. If we can't predict what the weather in our hometowns will be how can we predict how will the climate change?
    - conception 7y ago
      You can't predict where a bubble will form when boiling water, but you can predict that the pot will get to over 100C.
      - 6d6b73 7y ago
        But doesn't that depend on how large is the pot and how much energy you're putting into boiling it? Not to mention that you also need to know the environment in the which you're trying to boil the water. If you're trying to boil the water in the middle of arctic winter, can you predict that the pot will get to over 100C if you're using just a small camping stove? :)
    - fwip 7y ago
      If you drop a glass full of water, I can't predict the exact splash pattern. But I can sure tell you the floor's gonna get wet.
  - Theodores 7y ago
    "Thousands" is true, but, in the specific domain of modelling the climate there are not that many people doing it. Weather forecasting is more about interpreting the results than writing the code for the model, and, realistically there are only the British and American weather supercomputers. Everyone else is borrowing the GRIB data and then doing their own interpretations. As for that interpretation, a surprising amount of forecasters operate much like how gamblers work. Think of a horse racing scene, the horses and the riders have form, i.e. how they have performed in the past. Then there is the course, this may or may not suit a given horse/rider combo. Weather, time of day and other factors come into it. Not to mention the prize money and the competition. A gambler looks at the form, the weather and works with their niche of specialist knowledge, similar with weather forecasting. The 'bet' might be what the weather is going to be in Dudley tomorrow or in Barnstaple, or anywhere else. The goal of the forecaster is to get it right. The observations are the key to it and the GRIB forecast data is not to be considered strictly accurate. Even if it is accurate it still needs interpretation. Even with the data and an eye on the metosat images it is incredibly hard to get the likelihood of rain right. Sometimes a step outside is needed too - with a local forecast. The resolution of the GRIB is not that good, so if you have a town at the foot of a mountain and next to a lake, it might not have a realistic prediction. The GRIB data might be for the top of the mountain or the middle of the lake. You need an expert to actually do an educated guess as to what temperature it will be there in the morning. Forecasters actually understand some things that are going on in the now, so a storm off Africa that goes across to the U.S seaboard as some tropical storm will come back across to the British Isles as rain. Forecasters don't tell you that there is a queue of storms heading across the Atlantic on the way to the U.S. and inevitably going to bring rain to the UK after that. But that information could inform your holiday plans for a few weeks ahead. The computer and the meteosat images might show all this but they won't interpret it for you. Then, when you get down to it, forecasting can often be done with horse betting mentality, like a gambling addiction. What would be cool is to have a consumer grade version of the software the weather forecasters use, where you can go forward and back in time layering up different types of data at different altitudes. But weather forecasting is like its own magic circle and it would be betrayal of sorts to share the tricks and the props with the audience.
- fromthestart 7y ago
  That's simply not relevant, nor true.* Modeling climate change is much more complex than the simple heuristic pattern matching of the human mind. You're talking about a huge, complex, and fundamentally chaotic system; the latter property precludes the usefulness of heuristic based predictions by a 4 year old (or a human of any age). There's a reason we run massive numerical simulations on multi billion dollar HPCs. *Depending of course on how you define processing power. Sure, in terms of raw flops it is estimated that human brains are indeed orders of magnitude more performant than supercomputers, but for the type of operations required to simulate something like climate, humans don't come anywhere close to the accuracy or precision of results you can obtain from a gridded 3D model fed into a purpose built simulation - without which you really can't answer any but the most basic of climate change questions.
  - pazimzadeh 7y ago
    I'm not sure what the use is of comparing human processing to "the type of operations required to simulate something like climate" since predicting the weather is not something our computers are particularly good at. Why is a 3D model even necessary? Personal anecdote: my grandmother is capable of predicting rain 1-2 days ahead of time based on smell and just "a feeling." Granted, we probably don't remember all of the times that she was wrong.
    - dlivingston 7y ago
      > Why is a 3D model even necessary? See [0]. PDEs can be solved discreetly in three dimensions on a three-dimensional grid (3D model, sort of). [0]: https://en.wikipedia.org/wiki/Finite_volume_method https://en.wikipedia.org/wiki/Finite_volume_method
      - lutorm 7y ago
        That doesn't really explain why you need a 3d model, though. If there was no important vertical process, you could get away with not resolving it. However, vertical heat transport is the entire game since the lower atmosphere is opaque to heat.
- masonic 7y ago
  not even close in processing power of a 4 year old Sure, but in most developed nations, you can't force 4 year olds to work even individually, let alone wire them into an array against their will.
tomrod 7y ago
Climate vs. weather. With many natural systems, we can understand qualitative elements of the system, at least in part. Of course, the long run general equilibrium may be quite drier.
TallGuyShort 7y ago
This is a bit like comparing the use of a supercomputer as a virtual wind tunnel to a prediction like "the airplane will fall to the ground when it's engines die".
trentnix 7y ago
Not to mention we have an extremely limited data set to work from when identifying what “normal” is.
- fwip 7y ago
  We're not talking about vague shifting definitions of "normal." We're talking about "the wettest 12 months ever recorded," in the 125 years we have data for.
  - trentnix 7y ago
    125 years is still a very small amount of time to assess climate, and the quality of the data 125 years ago isn’t very good. Considering all “solutions” to the “problem” of climate change involve a radical remaking of society, it’s not unreasonable or unserious to ask for more data.
    - fwip 7y ago
      You're having a totally different discussion than all the people you replied to. Nobody above you said "this is destroying the planet and we need to guillotine the oil barons." You introduced the idea of "we don't know what normal is" when it wasn't relevant. This is the wettest 12 months in the US on record. This is almost certainly because the global temperature has gone up significantly in the last century+. If you want to talk about this, talk about this - don't waste time attacking imaginary points you think other people are implying.
    - sbov 7y ago
      Reminds me of the an article a while back talking about how via studying tree rings and sediment, that scientists think California experienced back-to-back 200 year droughts starting around the year 850.
      - traeregan 7y ago
        Isn’t a back-to-back 200 year drought a 400 year drought?
viraptor 7y ago
Few days out projection is hard and influenced by lots of chaotic changes. Long term trends are a completely different thing. I can't tell you what temperature will be where I live tomorrow, but I can tell you it's going to be ~30°C higher in 6 months.
- ThomPete 7y ago
  But not whether its going to be 3 degrees hotter in 50 years and not whether its because of humans or natural patterns. This is the part of the story we somehow always forget.
  - hutzlibu 7y ago
    Yes we can. It is called science and modelling. We understand the greenhouse effect small scale very well. And we know very well, that we humans burned up many things in a short time that were underground before ... so we know there are much more greenhouse gases in the atmosphere. And the data matches the predictions so far. So no. Nobody knows for sure and nobody is claiming to know, what the temperature in 50 years will be. But we know that the current models predict bad things. And we know that we burned that oil. Regardles of the sun cycle, you are probably refering to.
    - ThomPete 7y ago
      Its called modeling not science and not science modeling in any shape or form even closely resembling ex physics. We dont know its going to be bad, we dont know if its us or natural variation and we dont have a single scientifically demonstrated conclusion of climate change we dont know how to deal with. Modelling is NOT science, its more akin to scenarioplanning which is why the results are all over the map. If it was science on the precision of physics we didnt need modelling. We understand greenhouse effect in a lab well, not in a chaotic non linear system like the climate.
      - viraptor 7y ago
        Statistics is a part of science. It's perfectly valid to use scientific methods to arrive at: this is the current trend, this increase started at year X, given all we know right now, there's a tiny chance that this increase is not caused by this specific change. From there we can extrapolate future changes with some likelihood brackets. Then feel free to discuss the brackets, the methodology, etc. but you can't really dispute this as a scientific method. This is literally how current drug research (well, validation) works. It would be great if we could model the entire body-drug interaction, but in the meantime the statistics we have are the best thing we have available.
btilly 7y ago
The widespread willful ignorance about the differences between modeling weather and climate was amusing to me 20 years ago. But the joke is long past its sell-by date. No, we can't predict whether it will rain on this day in August in Augusta Georgia. Detailed forecasts will probably never get a lot better a month out. But we can predict that December 2019 will be colder than August 2019 across most of the USA, and that 2020 will be warmer than 1980 was. And those forecasts are getting better over time. That's reality. Deal with it.
- trentnix 7y ago
  Not really. The delta in how data is collected in August and December is much smaller than 1980 to 2020. So such determinations aren’t as straightforward as you imply. Not to mention none of the catastrophic climate consequences predicted in 1980 came to pass, but somehow we are supposed to be so much more certain about the prognosticators of today.
  - viraptor 7y ago
    Which predicted catastrophes? The only one I remember is the ozone layer issue which we stopped with regulation and still pay the price (although it's getting better)
    - chrisco255 7y ago
      You ever see the movie Waterworld? Some people actually believe that to be a possibility.
      - NeedMoreTea 7y ago
        I'd take some convincing that anyone actually thinks a terrible movie that had a thousand metres, or whatever it was, of sea level rise is a real possibility. We're "only" looking at a rise under 100m even if we lose both ice caps. Which would take centuries to fully melt... Would still end up being catastrophic for the society we built.
      - 7y ago
ori_b 7y ago
I present to you a person who thinks that because we can't tell for certain that it's going to rain next Wednesday, we're have no idea if next summer will be hotter than next winter.
- zaroth 7y ago
  Not a very charitable interpretation of the comment.
  - SantalBlush 7y ago
    It was a false equivalence, and deserves to be treated as such.
  - ori_b 7y ago
    In which ways is it inaccurate?
    - zaroth 7y ago
      You can be accurate and still be uncharitable. Educate, don’t chastise or mock. And never make it personal.
      - ori_b 7y ago
        Normally, charity in an interpretation means that there's a way that it could be taken where it makes sense. I don't think that's the case for this comment. The most charitable interpretation I have is that the original poster is a paid shill -- in other words, they're bright enough to realize what they're saying is idiotic, but they're just doing a job to put food on the table. It'd take some evidence to sway me to that view.
        mirimir 7y ago
        Seriously? I don't agree with their comment. But after looking at their comment history, I see no evidence for them being "a paid shill". Contrarian and outspoken, sure. But not a shill.
noonespecial 7y ago
I can't predict the cards in my next hand of blackjack, but I'm sure if I play all night the house will win in the end.
- godshatter 7y ago
  That's a pretty good analogy, but the question is: will you lose your bankroll in 20 minutes, or an hour, or 12 hours, or a week? Climate models are doing the equivalent of trying to figure out when you're going to broke, which (although much easier than figuring what cards you will get) is still a tough question to answer.
  - newvoiceoldphne 7y ago
    Card decks are a much smaller problem space than the entirety of the Earth and factors affecting it's climate.
  - pqhwan 7y ago
    It is yet another (and much easier) question whether we should be gambling with the habitability of the one and only planet available to us, especially when technological advances are beginning to offer a way out.

The most powerful supercomputers in the world are focused on weather analysis and they can only project a few days out. Yet, some people think they've got it figured out.

justwalt 7y ago

That does put things into perspective, a bit.

ceejayoz 7y ago
Nah, it's a common talking point, but it's highly flawed. It's like claiming that because we can't predict if the stock market will go up or down tomorrow, we can't determine whether we're in a recession or a boom time. Short-term weather is fairly chaotic (and we've gotten a lot better at predicting it over the last 50 years). Long-term trends are different.
- jessaustin 7y ago
  If you can figure out recessions and booms in real time you can make lots of money.

6d6b73 7y ago

The most powerful supercomputers in the world are not even close in processing power of a 4 year old. Isn't it possible that some people can see patterns that the supercomputers can't?

chrisco255 7y ago
Sure, but those supercomputers aren't autonomous, they're programmed and designed by thousands of human adult experts.
- 6d6b73 7y ago
  And these people can make mistakes or be biased. If we can't predict what the weather in our hometowns will be how can we predict how will the climate change?
  - conception 7y ago
    You can't predict where a bubble will form when boiling water, but you can predict that the pot will get to over 100C.
    - 6d6b73 7y ago
      But doesn't that depend on how large is the pot and how much energy you're putting into boiling it? Not to mention that you also need to know the environment in the which you're trying to boil the water. If you're trying to boil the water in the middle of arctic winter, can you predict that the pot will get to over 100C if you're using just a small camping stove? :)
  - fwip 7y ago
    If you drop a glass full of water, I can't predict the exact splash pattern. But I can sure tell you the floor's gonna get wet.
- Theodores 7y ago
  "Thousands" is true, but, in the specific domain of modelling the climate there are not that many people doing it. Weather forecasting is more about interpreting the results than writing the code for the model, and, realistically there are only the British and American weather supercomputers. Everyone else is borrowing the GRIB data and then doing their own interpretations. As for that interpretation, a surprising amount of forecasters operate much like how gamblers work. Think of a horse racing scene, the horses and the riders have form, i.e. how they have performed in the past. Then there is the course, this may or may not suit a given horse/rider combo. Weather, time of day and other factors come into it. Not to mention the prize money and the competition. A gambler looks at the form, the weather and works with their niche of specialist knowledge, similar with weather forecasting. The 'bet' might be what the weather is going to be in Dudley tomorrow or in Barnstaple, or anywhere else. The goal of the forecaster is to get it right. The observations are the key to it and the GRIB forecast data is not to be considered strictly accurate. Even if it is accurate it still needs interpretation. Even with the data and an eye on the metosat images it is incredibly hard to get the likelihood of rain right. Sometimes a step outside is needed too - with a local forecast. The resolution of the GRIB is not that good, so if you have a town at the foot of a mountain and next to a lake, it might not have a realistic prediction. The GRIB data might be for the top of the mountain or the middle of the lake. You need an expert to actually do an educated guess as to what temperature it will be there in the morning. Forecasters actually understand some things that are going on in the now, so a storm off Africa that goes across to the U.S seaboard as some tropical storm will come back across to the British Isles as rain. Forecasters don't tell you that there is a queue of storms heading across the Atlantic on the way to the U.S. and inevitably going to bring rain to the UK after that. But that information could inform your holiday plans for a few weeks ahead. The computer and the meteosat images might show all this but they won't interpret it for you. Then, when you get down to it, forecasting can often be done with horse betting mentality, like a gambling addiction. What would be cool is to have a consumer grade version of the software the weather forecasters use, where you can go forward and back in time layering up different types of data at different altitudes. But weather forecasting is like its own magic circle and it would be betrayal of sorts to share the tricks and the props with the audience.
fromthestart 7y ago
That's simply not relevant, nor true.* Modeling climate change is much more complex than the simple heuristic pattern matching of the human mind. You're talking about a huge, complex, and fundamentally chaotic system; the latter property precludes the usefulness of heuristic based predictions by a 4 year old (or a human of any age). There's a reason we run massive numerical simulations on multi billion dollar HPCs. *Depending of course on how you define processing power. Sure, in terms of raw flops it is estimated that human brains are indeed orders of magnitude more performant than supercomputers, but for the type of operations required to simulate something like climate, humans don't come anywhere close to the accuracy or precision of results you can obtain from a gridded 3D model fed into a purpose built simulation - without which you really can't answer any but the most basic of climate change questions.
- pazimzadeh 7y ago
  I'm not sure what the use is of comparing human processing to "the type of operations required to simulate something like climate" since predicting the weather is not something our computers are particularly good at. Why is a 3D model even necessary? Personal anecdote: my grandmother is capable of predicting rain 1-2 days ahead of time based on smell and just "a feeling." Granted, we probably don't remember all of the times that she was wrong.
  - dlivingston 7y ago
    > Why is a 3D model even necessary? See [0]. PDEs can be solved discreetly in three dimensions on a three-dimensional grid (3D model, sort of). [0]: https://en.wikipedia.org/wiki/Finite_volume_method https://en.wikipedia.org/wiki/Finite_volume_method
    - lutorm 7y ago
      That doesn't really explain why you need a 3d model, though. If there was no important vertical process, you could get away with not resolving it. However, vertical heat transport is the entire game since the lower atmosphere is opaque to heat.
masonic 7y ago
not even close in processing power of a 4 year old Sure, but in most developed nations, you can't force 4 year olds to work even individually, let alone wire them into an array against their will.

tomrod 7y ago

Climate vs. weather. With many natural systems, we can understand qualitative elements of the system, at least in part. Of course, the long run general equilibrium may be quite drier.

TallGuyShort 7y ago

This is a bit like comparing the use of a supercomputer as a virtual wind tunnel to a prediction like "the airplane will fall to the ground when it's engines die".

trentnix 7y ago

Not to mention we have an extremely limited data set to work from when identifying what “normal” is.

fwip 7y ago
We're not talking about vague shifting definitions of "normal." We're talking about "the wettest 12 months ever recorded," in the 125 years we have data for.
- trentnix 7y ago
  125 years is still a very small amount of time to assess climate, and the quality of the data 125 years ago isn’t very good. Considering all “solutions” to the “problem” of climate change involve a radical remaking of society, it’s not unreasonable or unserious to ask for more data.
  - fwip 7y ago
    You're having a totally different discussion than all the people you replied to. Nobody above you said "this is destroying the planet and we need to guillotine the oil barons." You introduced the idea of "we don't know what normal is" when it wasn't relevant. This is the wettest 12 months in the US on record. This is almost certainly because the global temperature has gone up significantly in the last century+. If you want to talk about this, talk about this - don't waste time attacking imaginary points you think other people are implying.
  - sbov 7y ago
    Reminds me of the an article a while back talking about how via studying tree rings and sediment, that scientists think California experienced back-to-back 200 year droughts starting around the year 850.
    - traeregan 7y ago
      Isn’t a back-to-back 200 year drought a 400 year drought?

viraptor 7y ago

Few days out projection is hard and influenced by lots of chaotic changes. Long term trends are a completely different thing. I can't tell you what temperature will be where I live tomorrow, but I can tell you it's going to be ~30°C higher in 6 months.

ThomPete 7y ago
But not whether its going to be 3 degrees hotter in 50 years and not whether its because of humans or natural patterns. This is the part of the story we somehow always forget.
- hutzlibu 7y ago
  Yes we can. It is called science and modelling. We understand the greenhouse effect small scale very well. And we know very well, that we humans burned up many things in a short time that were underground before ... so we know there are much more greenhouse gases in the atmosphere. And the data matches the predictions so far. So no. Nobody knows for sure and nobody is claiming to know, what the temperature in 50 years will be. But we know that the current models predict bad things. And we know that we burned that oil. Regardles of the sun cycle, you are probably refering to.
  - ThomPete 7y ago
    Its called modeling not science and not science modeling in any shape or form even closely resembling ex physics. We dont know its going to be bad, we dont know if its us or natural variation and we dont have a single scientifically demonstrated conclusion of climate change we dont know how to deal with. Modelling is NOT science, its more akin to scenarioplanning which is why the results are all over the map. If it was science on the precision of physics we didnt need modelling. We understand greenhouse effect in a lab well, not in a chaotic non linear system like the climate.
    - viraptor 7y ago
      Statistics is a part of science. It's perfectly valid to use scientific methods to arrive at: this is the current trend, this increase started at year X, given all we know right now, there's a tiny chance that this increase is not caused by this specific change. From there we can extrapolate future changes with some likelihood brackets. Then feel free to discuss the brackets, the methodology, etc. but you can't really dispute this as a scientific method. This is literally how current drug research (well, validation) works. It would be great if we could model the entire body-drug interaction, but in the meantime the statistics we have are the best thing we have available.

btilly 7y ago

The widespread willful ignorance about the differences between modeling weather and climate was amusing to me 20 years ago. But the joke is long past its sell-by date. No, we can't predict whether it will rain on this day in August in Augusta Georgia. Detailed forecasts will probably never get a lot better a month out. But we can predict that December 2019 will be colder than August 2019 across most of the USA, and that 2020 will be warmer than 1980 was. And those forecasts are getting better over time. That's reality. Deal with it.

trentnix 7y ago
Not really. The delta in how data is collected in August and December is much smaller than 1980 to 2020. So such determinations aren’t as straightforward as you imply. Not to mention none of the catastrophic climate consequences predicted in 1980 came to pass, but somehow we are supposed to be so much more certain about the prognosticators of today.
- viraptor 7y ago
  Which predicted catastrophes? The only one I remember is the ozone layer issue which we stopped with regulation and still pay the price (although it's getting better)
  - chrisco255 7y ago
    You ever see the movie Waterworld? Some people actually believe that to be a possibility.
    - NeedMoreTea 7y ago
      I'd take some convincing that anyone actually thinks a terrible movie that had a thousand metres, or whatever it was, of sea level rise is a real possibility. We're "only" looking at a rise under 100m even if we lose both ice caps. Which would take centuries to fully melt... Would still end up being catastrophic for the society we built.
    - 7y ago

ori_b 7y ago

I present to you a person who thinks that because we can't tell for certain that it's going to rain next Wednesday, we're have no idea if next summer will be hotter than next winter.

zaroth 7y ago
Not a very charitable interpretation of the comment.
- SantalBlush 7y ago
  It was a false equivalence, and deserves to be treated as such.
- ori_b 7y ago
  In which ways is it inaccurate?
  - zaroth 7y ago
    You can be accurate and still be uncharitable. Educate, don’t chastise or mock. And never make it personal.
    - ori_b 7y ago
      Normally, charity in an interpretation means that there's a way that it could be taken where it makes sense. I don't think that's the case for this comment. The most charitable interpretation I have is that the original poster is a paid shill -- in other words, they're bright enough to realize what they're saying is idiotic, but they're just doing a job to put food on the table. It'd take some evidence to sway me to that view.
      - mirimir 7y ago
        Seriously? I don't agree with their comment. But after looking at their comment history, I see no evidence for them being "a paid shill". Contrarian and outspoken, sure. But not a shill.

noonespecial 7y ago

I can't predict the cards in my next hand of blackjack, but I'm sure if I play all night the house will win in the end.

godshatter 7y ago
That's a pretty good analogy, but the question is: will you lose your bankroll in 20 minutes, or an hour, or 12 hours, or a week? Climate models are doing the equivalent of trying to figure out when you're going to broke, which (although much easier than figuring what cards you will get) is still a tough question to answer.
- newvoiceoldphne 7y ago
  Card decks are a much smaller problem space than the entirety of the Earth and factors affecting it's climate.
- pqhwan 7y ago
  It is yet another (and much easier) question whether we should be gambling with the habitability of the one and only planet available to us, especially when technological advances are beginning to offer a way out.