Sign in / up

back to article Met Office wants better supercomputer to predict extreme weather

The UK's Met Office needs bigger and better supercomputers if it is to confidently and accurately predict the weather and give emergency services a longer lead time for extreme weather conditions, a government group said today. The Science and Technology committee of MPs advised in a report on the Met Office that, despite the …

COMMENTS

House rules Send corrections
This topic is closed for new posts.

Page:

      1. arrbee
        Reply Icon

        giraffe

        "as is done in the US" - they're having a laugh.

        The US television broadcasts use a basic conversion chart between phrases such as

        "slight chance of rain" -> " 25% chance of rain"

        "likely to rain" -> "75% chance of rain"

        etc

        There is no statistical basis behind the numbers, just smoke and mirrors.

        0 0
  2. Pete 2 Silver badge

    Sometimes the answer is "we just don't know"

    In Britain most of the weather we get only develops in the last 2 or 3 days before it hits us. While it may be possible (sometimes) to say "there's a storm coming our way", or "some places will be windy" predicting exactly WHERE will be hit is probably outside the bounds of the knowable.

    Even within 24 hours of an "event", the precise location - or worst hit spots - probably won't be evident until whatever it is actually starts to rain/blow/bake or blizzard all over it. if the Met Office was to buy some new sooper-dooper computers, would they really be able warn a particular village that it would be flooded, but the one 10 miles away wouldn't? Without that degree of certainty, we could end up in even worse bother: with the wrong emergency services being sent to the wrong place in advance of an imprecise prediction - instead of being held on "alert" until some calls for help actually came in.

    Maybe what Britain needs is a bit more awareness (such as not building houses, or critical services, on flood plains or near rivers) having a few more gritters and the will to use them and for someone to be in charge, rather than diluting responsibility to a mess of small and poorly organised local councils who don't really have any incentive to take precautions against once-in-10-year weather events.

    3 0
    1. Anonymous Coward
      Reply Icon
      Flame

      Re: Sometimes the answer is "we just don't know"

      What does a disastrous flood we know about and have forecast look like? a wet day.

      The councils know major rain is coming, have all the trucks ready (not in the garage being serviced), prioritize cleaning leaves from the gutters, lower the river water level, and life goes on, miserable and damp, but not flooded.

      What happens when we predict major rainfall, and empty a dam ahead of time so that it doesn't burst? another boring wet day.

      What happens when we predict icy roads properly? we don't need thousands of workers out every night over winter, gritting the road just in _case_ it might be icy later. We only do it on nights when its actually going to be icy.

      When we predict weather better, the weather doesn't get magically better, we just don't have as many disasters. We harvest more crops, knowing tomorrow and the next day will be dry and the day after will be pouring, so we can schedule collecting hay better, etc.

      Its easy not to notice the day to day improvements due to better forecasting.

      I've listened to too many rants from people complaining they were working in the garden at 11:30 and it started drizzling, when the forecast predicted rain in the afternoon. Look what it was like before folks: we're no longer losing dozens of ships in a single night because of storms we didn't predict!

      3 0
      1. Andrew Jones 2
        Reply Icon
        FAIL

        Re: Re: Sometimes the answer is "we just don't know"

        Yes - and now....... make it extremely clear that everything you just said only applies south of Newcastle. In Scotland - the Met Office like to issue AMBER or RED warnings (used to be Flash Warnings) a good 30-60 minutes after whatever they are warning about has actually started.

        As for the BBC.... 90mph winds across Scotland is "a bit windy" but in London it's going to gust UP TO 60mph and oh my god - it's the end of the world - "damaging gusts of wind" and "severe threat to life" - we get so fed up with this stereotypical view that because it's Scotland it is always windy/wet/snowy and therefore we expect it. What complete bollocks.

        2 1
      2. Pete 2 Silver badge
        Reply Icon

        I'd love to live in your world

        > What happens when we predict icy roads properly?

        Where I live the gritters go out and salt the 1 major road through my town (pop 20,000) and the bypass. However they never, ever grit any of the side roads. So the arterial routes are lovely and clear, but since nobody (the people who's council tax pays for the gritting) is able to get through the snow-blocked residential streets to use them, they remain clear - except for the occasional gritter wondering why they bother, since nobody is using the road.

        0 0
        1. cmp
          Reply Icon
          Thumb Up

          Re: I'd love to live in your world

          of course gritting costs per mile of road if you can go from gritting 30 times a year because "it might be icey" to 5 times a year when "it almost certainly will be icey" your local council could use the money save to either

          a) grit the side roads so you can get out or

          b) cut your council tax.

          So a nice new super computer for the met office will make yor life better \o/

          0 0
  3. Usually Right or Wrong
    Facepalm

    The likelihood of weather is...

    "In particular, we are keen to see broadcasters make greater use of probabilistic information in their weather forecasts,"

    In Scotland there is 150% chance of weather, that is 50% chance of rain in Glasgow, 50% chance of snow in the Cairngorms and 50% chance of sunshine in Aberdeen, or none of this may happen, so that would be 0% chance of weather.

    In England, the super-super-computer just said 42 and suggested that we release lots of white mice to see if they can get someone to work it out.

    1 0
  4. Barely registers
    Stop

    You want ANOTHER supercomputer?

    Bigger and better than the £33m IBM supercomputer that you installed in 2009?

    Yeah - you know - the one that consumes the same power as 2000 homes?

    My tip - write a better program.

    0 3
    1. Anonymous Coward
      Reply Icon
      Anonymous Coward

      Re: You want ANOTHER supercomputer?

      The 33 million in 2009 was the complete budget that covers the cost of both the supercomputers accepted in 2009, and the ones that are currently being built (the programme had two stages, and they are about half-way through).

      In the time since the current machines were installed, they've dropped from 73 and 74 on the June 2009 top500 to their current position of 478 and 479 on the November 2011 list. That's a fall of over 400 places, and there is a good chance that they will not appear at all on the next one! That's how fast the Supercomputing world moves.

      You really want to tell me that the Met Office can remain current and competitive, and a leading world forecaster without replacing them?

      Even the Indian Institute of Tropical Meteorology appear higher on the list (403)!

      (The new Met Office systems actually appear on the Nov. list at 62 and 63, which is interesting because they were not running at the time, and the spec's are wrong ; )

      2 0
    2. Anonymous Coward
      Reply Icon
      Anonymous Coward

      @Bareley

      Could you let us know what you think is an acceptable depreciation cycle for a supercomputer? How long should one live? Should the Met Office still be using their pair of Cray 1s? (Named Ronnie and Reggie, IIRC) or is a Cray 1 a tad old these days?

      1 0
    3. Fenwick
      Reply Icon
      Thumb Down

      Re: You want ANOTHER supercomputer?

      If you know how exactly to write a better program than the one we already have. I.e. a program that makes more accurate predictions for less computing power. Perhaps you should let the world know about your world beating understanding of atmospheric physics and numerical algorithms.

      Or perhaps you meant that you don't understand the point of knowing the weather in advance.

      0 0
  5. ukbabz

    Beyond 24 hours, its voodoo

    I try and keep an eye on what the forecast for the upcoming days are via the met website (not the sites that simply use the data given to them by the met). I look at the nearest weather station and the forecast for the next 24-48hours. More often than not its accurate enough (e.g. Wind picking up around 3pm, heavy rain around 6pm) to get by.

    If we want to predict things further in the future, with greater accuracy we need to ramp up the processing power to deal with this. No point taking 3 weeks to process a forecast for next weekend after all..

    0 0
    1. John G Imrie
      Reply Icon

      Re: Beyond 24 hours, its voodoo

      If we want to predict things further in the future, with greater accuracy we need to ramp up the processing power to deal with this.

      No we don't. you can't beet this simply by upping the processing power. Weather systems are Chaotic, which means even with infinite processing power you can't get it right, If you want to improve the forecast you need more data points. a grid of sensors at one .25m difference covering the globe from 15 km below see level to 15 km above might start to get the resolution you require

      1 0
      1. Lars Silver badge
        Reply Icon
        Coat

        Re: Re: Beyond 24 hours, its voodoo

        Voodoo perhaps, but if you want to predict the weather in one week you will have to compute what is happening with low and high pressure over say North America and Asia so the scale gets a lot bigger needing more power, also the amount of information gets bigger to compute.

        I cannot remember how long it took to fly a balloon around the world but that is about how long it is possible to compute anything about the weather, I think.

        Personally I don't give a damn about it. Things you cannot alter you just have to take like they are.

        But I understand there are those who really rely on and need that information badly.

        1 0
  6. Anonymous Cowherder
    Go

    It's seasonally average for the time of year.

    January to May, a bit iffy with the odd nice day, potential for some very bad weather.

    May - September, generally ok, will got some thundery storms but on the whole ok.

    October - December, On the whole cold, generally poor but some nice days.

    So, to summarise, keep an umbrella in your bag and during the summer months don't forgot to slap on the suncream if you are going to be exposed for some time.

    This is Britain, we have weather, none of it is too extreme, most of it is cloudy but we get some scorching days.

    Can I have the money that the Met office were going to spend on a supercomputer deposited in the joint account as it has just gone overdrawn as Npower have whacked me on the super high tarrif.

    1 0
  7. markw:

    Already ordered...

    I listened to a documentary on Radio 4 a few days ago and the Met bod said that a Power7 had already been ordered.

    1 0
  8. Giles Jones Gold badge

    Surely they could use "cloud" systems, it is surely better to rent than buy?

    0 0
  9. Wild Bill

    Obligatory

    http://www.youtube.com/watch?v=HMAt8ZXqtbc

    0 0
  10. Anonymous Coward
    Anonymous Coward

    A butterfly flaps it's wings...

    I'm sure greater processing power, wielded efficiently, will assist in better forecasting.

    This will just mean the probability in percentages will be higher, but it won't mean the long term forecast will be entirely accurate.

    There are so many variables involved and also not a massive amount of historical weather data to work with, when you consider just how long our planet has been whizzing around the sun.

    The more data that can be pumped in for processing, the better, so this request for faster hardware you have to take at face value as being valid.

    Weather geeks, from my limited experience, tend to also be computer geeks - I'm sure the Metoffice has some highly skilled individuals maintaining the software and hardware.

    I wonder whether they couldn't implement a similar project to SETI ? - donate idle cycles of tens of thousands of computers?

    0 0
  11. ich
    Holmes

    Met HPC

    I am studying the MSc. in HPC @ Edinburgh Univ. and distinctively remember a lecture on Exascale computing held by a Professor heavily involved in the HPC area within the UK and EU regarding the need for more and more computing power especially for weather forecasting. This is because as some have already described above, in order to have better medium and long term predictions, you need to compute bigger and bigger areas. With a normal distributed system, if you decompose millions of these sub areas across machines scattered around the world, the intensive halo swapping (neighbourhood borders) communications would absolutely grind the entire computation to a halt. This is why you need more cores and faster interconnects for being able to throw a bigger problem at the beast (Gustafson's Law) which in turn should hopefully mean a more accurate prediction in a faster manner.

    I wouldn't necessarily blame their models to be wrong, just their capabilities on how much data they can actually use them on which for this particular science seems to be a rather important point to take into account.

    0 0
  12. Anonymous Coward
    Anonymous Coward

    CPU Idle states

    I've got a four-core i5 here and it spends a great deal of its time clocked down to 1.2GHz and each core spending 90% of the time not clocked at all. http://code.google.com/p/i7z/ is a handy tool for looking at this sort of thing. Linux-only as far as I know.

    0 0
  13. nexsphil

    18 months given to produce a single business case eh?

    I wonder if they're hiring...

    0 0
  14. Anonymous Coward
    Anonymous Coward

    Met Office wants better supercomputer to predict EXTREME weather

    One word:

    Fuck Off!

    0 0
    1. Anonymous Coward
      Reply Icon
      Anonymous Coward

      Re: Met Office wants better supercomputer to predict EXTREME weather

      Thats two words.

      'Fuck off and die' - is more to the point.

      That'll be four words then.

      0 1
  15. Anonymous Coward
    Anonymous Coward

    Dear Politicians

    This is now the opportunity for us big corporations to show off our superior computing technology. And we sent our IT consultants to Downing Street last year, remember?

    Please don't think of taking this and similar projects as an opportunity to re-use existing super computing knowledge at CERN. And don't - eeeeuch - give contracts to local companies in the UK, provide apprenticeships for the unemployed young, and build up technological expertise in your country.

    No, don't do it like China, where the Tianhe supercomputers, Loongson processors and interconnects are developed and built in China. And don't do it like Japan with their K super, which has SPARC64 VIIIfx from Japan and a optimized parallel file system based on Lustre - developed in Japan - and the Tofu proprietary six-dimensional torus network - developed in Japan.

    No - don't get the idea of setting up a plan to develop 256 core CPUs and a 8TB storage chip and 1Tbits interconnects. Honestly, actively developing technology that can be exported to reduce the national trade deficit? Eeeeuch.

    I mean, we pay a hell a lot of money on generating hot air around small, incremental changes to our hardware and software. And we need to make more money from support contracts and additional software to patch up and gloss over our shortcomings. Remember, "it's not a fault, it's a feature".

    While we are on it, don't develop your own closed source but secure GovOS! Stay with GPS! Don't use Glonass and abandon Galileo! Don't revive anything similar to the Black Arrow carrier rocket....

    1 0
  16. Anonymous Coward
    Anonymous Coward

    Business case... which businesses are they talking about?

    Free (broadcast) weather forecasts are only a few hours and are useless for a few days out.

    Probabilistic forecasts going out several days and trending over the next month or so is a business in itself. Aviation, shipping, FMCG (especially the alcohol trade), MoD, big agriculture, wind energy operators, already buy the longer forecasts for essential planning. Whether they buy them from the Met Office or from elsewhere is the question.

    The business case should be examining how they can improve the accuracy and quality of service to those users, where decisions over when to harvest crops makes double digit differences to percentage yields, or changing the supplies of lager and wine in Tesco's according to the temperature.

    Broadcasters generating non-stories about extreme weather that isn't really significant should not be allowed to interfere with the real decisions that depend on weather forecasts, about how we make sure we operate the country's infrastructure and resources in the best ways we can.

    Let's stop being silly and re-frame their business case question for them...

    "Met Office wants a better supercomputer to make the forecasts worth paying for. Extreme weather forecasting is irrelevant unless we can make a business out of avoiding the costs of disruption."

    0 0
  17. Barry Lane 1
    WTF?

    One day, during summer 2010...

    ... I was sitting at the garden table having lunch in the sunshine. My laptop was opened at the BBC News page and I took a glance at the Met Office weather, which told me that the "current conditions" were heavy rain. The local weather observation site is around four miles away and a glance in that general direction proved pretty conclusively that the wall-to-wall blue sky extended to there and all the way to the horizon. I emailed the Met Office, who sent a reply a couple of days later that explained that weather in spot can be very different a little way down the road. If the Met Office spent more time adjusting the flawed modelling they use and less time on patronising PR, they'd do a better job. Oh, and a glance out the window might help them to divine "current conditions" with a tad more accuracy.

    0 0
  18. Iain Leadley

    Wouldnt trust them to speak their wheight

    I use the weather forcast for entertainment purposes.

    If you check the forcast for 5 days in the morning then again in the afternoon it has often totally changed. How can they get it that wrong.

    The only thing that they seem to predict with any real accuracy is snow.

    0 0

Page:

This topic is closed for new posts.

Other stories you might like