Climate answers sought in supercomputers

Carl Franzen, for The Verge:
There's a dark cloud hanging over the science of climate change, quite literally. Scientists today have access to supercomputers capable of running advanced simulations of Earth's climate hundreds of years into the future, accounting for millions of tiny variables. But even with all that equipment and training, they still can't quite figure out how clouds work.


  1. GIGO.  If your basic assumptions are not correct it doesn’t matter how powerful the computer is or how perfect the model is.  This is especially true when trying to model a complex chaotic system 100 years into the future – an only slightly incorrect assumption in starting conditions can propagate massive errors throughout the entire model.  For at least this reason people should not be basing policy on a what a computer says the earth will be like in 50 or 100 years – and this is even assuming the model itself is 100% perfect, which it almost certainly is not.

    1. Instead, we should base policy on what Exxon-Mobil says the Earth will be like in 50-100 years.  (Hint: SUV’s.)  If scientific projections are less than 100% perfect, why not throw them out, and just do whatever generates the most revenue?

    2. Wait, are you using the fact that weather is hard to predict as a complex chaotic system where small changes can have large effects……to say we shouldn’t build policy around the idea that small changes can have large effects?

      I really wish that were a new level of incoherence on this subject.

      1. Science does not even know THE SIGN of the effect that clouds have on climate over the long term.  Let me repeat that: we don’t even know if more clouds means a hotter earth or cooler earth.  Further, we don’t know if more CO2 means more clouds or less clouds on a global scale.  Do you understand why I call GIGO on models that claim to predict future climate based on (among hundreds of other complex factors) assumptions about how CO2 affects clouds and how clouds affect climate, when these assumptions are basically wild ass guesses?

        Can we just state the truth here: You don’t care if climate models are accurate because ZOMG we have to DO SOMETHING NOW, even if we have no idea if what we do will have beneficial effect, harmful effect or no effect at all.

        1. Can we just state the truth here: You don’t care if climate models are accurate because ZOMG we have to DO SOMETHING NOW, even if we have no idea if what we do will have beneficial effect, harmful effect or no effect at all.

          Telling people what they think is a poor way to argue, because they usually know it better than you. In fact, I do care whether climate models are accurate. I care enough that I’ve noticed the many experts who study it actually put error bars on their results, compare them to measurements, and generally take steps to measure their uncertainty.

          Am I to believe that you genuinely have better reason to suppose clouds should completely override the shifts in the over-all energy balance they’ve been describing? Because right now, it doesn’t sound like you even clearly understand that having uncertainty in results is not at all the same thing as them being untrustworthy.

          Yes, I have an ulterior motive here, but it’s not ZOMG DO SOMETHING. It’s being sick of random jerks assuming the scientists who spend their lives studying this are uniformly morons, instead of listening to what they say.

        2. Bullshit.  You haven’t the faintest conception of how physics works, and somehow think you know more about it than every physicist in the world.   You might try reading this:

    3. That’s not how how modeling works.

      You don’t just make a model with a bunch of untested assumptions, stick in your data, get your output, and publish the result.

      Instead you continuously refine the model, working out if it can “predict” past results from earlier data, and changing it to make its predictions better and better. You have a completely testable model: given this ream of data, can it “predict” what actually did happen? If not, your model’s not good enough yet.

      But the general predictions aren’t based on these super-computer models, not yet anyway. The general predictions are based on the indisputable evidence that the world is indeed heating up, at a much faster pace than ever before.

      They’re also based on the indisputable evidence that the global temperature has correlated very strongly with the historical CO2 concentrations.

      We know the strong correlation between these two things, we know that CO2 is entering the atmosphere faster than ever before, and we know that the temperature is rising faster than ever before.

      We’re not sure exactly how clouds affect the dynamics yet, but it’s unlikely that they work differently than any other time in history, when high CO2 meant high global temperatures. All the evidence we have is more than strong enough to say that we need to be doing everything to reduce the CO2 in the atmosphere right now. From a policy perspective, that’s what’s important right now.

    4. With that comment you’ve shown you know absolutely nothing about climate modeling.  You are confusing numerical weather prediction with climate modeling.  The former predicts deterministically what the atmospheric conditions will be, based on current conditions, from a day to a couple of weeks in the future.  Get too far out in the future and, yes, the atmosphere becomes chaotic and you can no longer make a deterministic prediction.  This property of the atmosphere is well known, having been discovered by Edward Lorenz in 1963.

      However, climate models aren’t used to make deterministic predictions.  They are used to estimate how the mean and variance of the climate system changes under different boundary conditions (like doubling CO2, or after a large volcanic eruption).  For that you don’t need to know what the current conditions are, you just need to run the model until it reaches a quasi-equilibrium state.  Start with today’s conditions, or what the weather was like at the height of the last Ice Age – it doesn’t matter because the boundary conditions will drive the climate system to a state where there mean no longer drifts and the variations about that mean are relatively stable.

      This is climate modeling 101.  If you don’t understand these simple concepts then you really shouldn’t be spouting off about GIGO.

    5. Bullshit.  Can’t tell if you’re a troll or just so fucking dumb that you can’t see the irony that you’re using a computer that we wouldn’t have had the faintest chance of designing if statistical physics were not correct and extremely accurate, to blithely assert that we can’t use it to make much simpler predictions than those that go into designing semiconductor devices.   Any gas that is transparent to visual light but opaque to infared will cause an increase in ground level equilibrium temperature.  That is a trivial prediction of statistical physics, with ample evidence supporting it, and no remotely sane person would try to deny it.    The difficulty is in figuring out how fast it will reach equilibrium, and especially, how that initial effect will propagate.   That releasing greenhouse gasses into the atmosphere will lead to an increase has been known since before we actually started observing it. 

  2. “Carl Franzen (B.A., Magazine Journalism, Religious Studies) doesn’t understand how clouds work”? 

    “Scientists doesn’t understand every last detail of every possible type of cloud works”?

    Or “It’s impossible to map every sub-atomic particle in the formation and dissipation of cloulds even with supercomputers”?


  3. My kids know how clouds work.

    But they won’t tell me.

    I however know how the Easter Bunny delivers chocolate eggs.

    Check back in a week to see how this all rolls.

  4. That’s a terrible lede on an article that actually does a fairly good job of describing the current state of the art in cloud modeling. Of course we know how clouds work, and can simulate them very well. The problem is that you can’t do this on a global scale with current computing limitations, nor will we be able to in the foreseeable future. 

    It doesn’t really make any difference to the broad understanding of climate change though. At best, it adds some error bounds to long term future forecasts.

  5. We haven’t managed to have computers predict weather patterns 1 day into the future.  While it actually seems worthwhile, it would be far too easy for climate deniers to hide behind this until they get it working, which could easily be a hundred years more.  The climate is far too complex, (i.e. Chaotic), to make specific predictions about, but general predictions are a completely different story.

    We can’t say exactly which days will have hurricanes, but can predict that as the planet warms, there is be more hurricanes, and each hurricane will be on average stronger.

    This is just like the Hydrogen Fuel Cell.  Great idea in-of itself, but the climate deniers used it as an excuse to push back fully functional electric cars.  Now they’ve completely stopped talking about hydrogen cars, because the governments have completed stopped forcing the electric car issue.

Comments are closed.