back to article It's raining, then? Hallelujah. Big Blue super 'puter sharpens forecasts

IBM's embattled Weather Company subsidiary has said it is building a GPU-tastic supercomputer to model global weather conditions faster and more accurately. The firm – under fire in the US for allegedly not having the proper permissions to collect app location data – is aiming for 3km resolution and hourly updates for 40 per …

  1. Locky Silver badge
    Coat

    Useful

    But I'll stick to my more reliable method thanks

  2. Pascal Monett Silver badge

    3.5PB of Spectrum Scale parallel access file storage

    That is what is needed for a forecast precision of 3km and hourly updates. I wonder what will be needed for 1km precision and quarter-hour updates.

    I also wonder when they will be able to accurately tell me what time it is going to rain, because these days they pretend to know, but they're wrong 50% of the time which, as we all know here, means they know diddly squat.

    I'd also very much like to have a weather forecast that doesn't radically change between the time I go to bed and the time I wake up. Going to bed expecting rain and waking up being told rain is for tomorrow makes me feel that very much don't have a clue and are making it up as they go along.

    To forecast is to plan ahead. If you change your forecast every hour, you haven't forecast anything better than looking out the window would.

    1. Voland's right hand Silver badge

      Re: 3.5PB of Spectrum Scale parallel access file storage

      I also wonder when they will be able to accurately tell me what time it is going to rain,

      I can give you the answer my mom gives to this question. She is a retired met with 30 years of experience starting in the military and continuing with 20+ years of working in civilian aviation. The quote is actually about the UK Met office, but it is 100% valid here:

      "When they raise their eyes off the computer screen and look out of the f*cking window".

    2. tfb Silver badge

      Re: 3.5PB of Spectrum Scale parallel access file storage

      The amount of computation scales as the square of resolution in the horizontal plane, so probably at least 36 times the storage (a factor of 9 for spatial resolution and 4 for time resolution). But it may be more if they increase vertical resolution as well.

      Forecasts are in fact a lot better than they were (I think weekly forecasts are now as good as daily ones were 30 years ago or something) but weather is, rather famously, a system where long-term prediction (of specific events: it's much easier to predict averages of them) scales very badly indeed as there's chaos.

      In fact forecasts which tell you what is happening for the next hour or so in a apecific location are very interesting to people who go outside: do I need my raincoat when I go for my walk?

  3. Charlie Clark Silver badge
    Black Helicopters

    More detailed is not more accurate

    A 3km model shows much finer detail and will enable more accurate region and locality forecasts

    While the new setup might indeed be able to provide more accurate forecasts, it's not a given; it could continue to be wrong just in higher definition.

    My understanding is that the biggest deficit for forecasting is the lack of sensors, hence, the various ideas to use mobile devices and networks to gather more data. If this is the case, then can we have our forecasts without ads, as we're obviously adding value collecting all that data, which can then be sold at a profit to farmers, governments and spooks.

    1. ThatOne Silver badge
      Devil

      Re: More detailed is not more accurate

      > If this is the case, then can we have our forecasts without ads

      You have a refreshing sense of humor...

      1. jake Silver badge

        Re: More detailed is not more accurate

        "If this is the case, then can we have our forecasts without ads"

        What are these things you call "ads"?

        More seriously, I might unblock ads on weather.com if they ever get around to making the fscking thing work worth a shit. It's been utter crap for over a decade now. If you don't believe me, look at the page source ... Total amateur effort, at best. And hugely bloated, to boot.

        1. jake Silver badge

          Re: More detailed is not more accurate

          And now the fizz-heads have not one, but two pop-ups asking MeDearOldMum to turn off her ad-blocker. Now THERE'S a good way to win friends!

    2. tfb Silver badge

      Re: More detailed is not more accurate

      I would hope that if forecasting organisations start using data from phone sensors then the deal would be that you get forecasts without ads. But probably it will not be because the MBA pinheads have taken over the web stuff.

    3. tfb Silver badge

      Re: More detailed is not more accurate

      It could just continue to be wrong, but it's clear that low resolution forecasts can't be as good as high resolution ones: if there are phenomena which take place on a scale smaller than the resolution then nothing you can do, apart from increasing resolution, lets you model those phenomena, so low resolution models just miss that part of the system.

  4. Anonymous Coward
    Anonymous Coward

    5 seconds on Google verified my recollection that the NAM has a 4 km resolution model.

    https://www.weather.gov/akq/MapsModels

    What "US" is the article referring to that only has low res models? Perhaps it was written by Watson?

  5. fishman

    The variation in forecasts between the weather models (US, European, etc) is probably greater than the variation due to the resolution of the predictions.

  6. Gene Cash Silver badge

    Observations

    So here in Orlando, we have I-4, which is a major road cutting from Tampa in the SW, to Daytona in the NE. It's one of the reasons Disney is here, as a matter of fact.

    I ride a motorbike and we get lots of liquid sunshine in the summer, so I have a keen interest in the precipitation and have bought apps such as PYKL3 that even show the radar from the local international airport in addition to the usual Nat'l Weather Svc stuff.

    Anyway... a lot of time the rains come from the west, hit "the I-4 corridor" and evaporate. You can see where the roadway is the hard edge of the rainstorm.

    Makes me wonder if these models account for anything like that.

    1. tip pc

      Re: Observations

      if on IOS try rainalarm, doesnt cost much and shows rain radars etc.

    2. Charlie Clark Silver badge

      Re: Observations

      Makes me wonder if these models account for anything like that.

      Possibly, there are lots of such boundaries but, in general, more sensors rather than specific local models are a better way of handling them, which is why the state meteorological agencies are so keen on sea surface temperature sensors to help them track storm systems.

    3. tfb Silver badge

      Re: Observations

      I think people do take account of things like that in some models, yes.

  7. Anonymous Coward
    Anonymous Coward

    Global Weather Conspiracy

    The jury is out on whether global weather is a real thing. Science is divided. This is just another example of "scientists" jumping on the bandwagon to get research grants and justify their jobs.

    1. tfb Silver badge
      Alien

      Re: Global Weather Conspiracy

      I can't work out what this comment is trying to say. Rather trivially there is a global state of the weather since there is weather everywhere in the planet, and that is what a global weather model tries to predict.

      I rather suspect that this is some kind of attempt at global-warming denialism but from someone who does not understand the difference between climat and weather. But even then it's failing: clearly there is a climate, whether or not it is changing faster than it usually does.

      I'm tempted to think this might be some denialist bot which has seen some terms in the article which triggered it and then generated a hugely nonsensical comment since the article is not actually about climate change at all.

  8. Anonymous Coward
    Anonymous Coward

    My prediction

    Another brilliant success for Watson...

    To go along with all the other ways Watson has revolutionised our lives. Like....ummmm....being wheeled out by IBM for the latest gimmicky IT fad, having no impact and quietly moving on to the next IT fad in a few years while casually brushing the brown stains of failure off its shoulders.

    Slightly easier than predicting the weather.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019