http://www.bbc.co.uk/news/science-environment-29789208

As well as running UK-wide and global forecasting models more frequently, the new technology will allow particularly important areas to receive much more detailed assessment.

For example, forecasts of wind speeds, fog and snow showers could be delivered for major airports, with a spatial resolution of 300m.

I think I’ll believe that one when I see it. The Met Office may have their weakenesses (Weather forecasting for example), but they are certainly good at PR

Met Office to build new super computer

…forecasts…could be delivered…with a spatial resolution of 300m.

all that really means is that their new toy can process a (much) larger number of data points.

just because it can process the calculations down to a resolution of 300m, doesn’t mean that they’ll be any more accurate.

Good news we are making this investment, hopefully to benefit as many UK companies as possible and them there West Country yokels. 🙂

Weather forecasting is very complex. They are actually pretty good in terms of what weather we will get its just he timing, ie when, which is difficult. they can predict heavy showers in the morning but if they fall a few hours earlier, ie at night, we don’t notice. I used to do a lot of offshore sailing and you make a point of downloading forecasts from multiple sources and comparing them. Plus you have all the live data coming in from weather stations/bouys/shipping, processing all this and upadating the models needs computer power.

Is this the big data molly was on about?

Today it’s sunny in Leicester with slight breeze; tomorrow will probly be similar. £97m saved right there.

I’m more amazed that Cray still exist. Didn’t they go bust over a decade ago?

Shame Sony stopped the ability to make PS3 clusters would be interesting to see how many it would take to match this bad boys processing power.

Then, as long as this guy stays perfectly still, we should be allright.

It’s all a bit Superman 3.

I saw a presentation from the project manager responsible for this a while back. Apparently Exeter uni approached him about having the old met office supercomputer until they realised the running costs were in the order of ~£10m/yr

And GIGO.

@Mr Woppit – nice reference 8)

The Met Office may have their weakenesses (Weather forecasting for example), but they are certainly good at PR

The Met Office are very very good at weather forecasting, but clearly crap at PR because everyone in the UK thinks they are no good at weather forecasting.

If you actually write down what they forecast and compare it with what actually happens, regularly (as people do) then you’ll find they are quite accurate. As long as you listen intelligently to what they say.

PS3 has about a quarter of a GB of RAM – so forgetting all the other dozens if not hundreds of advantages the Cray will have – just to match in in RAM terms you’d need about 40 thousand of them.

Cool, you could build one for about £4m then 8) I’ll spend the rest on a holiday in the sun

The Met Office are very very good at weather forecasting, but clearly crap at PR because everyone in the UK thinks they are no good at weather forecasting.

If you actually write down what they forecast and compare it with what actually happens, regularly (as people do) then you’ll find they are quite accurate. As long as you listen intelligently to what they say.
I use Meteogroup’s WeatherPro, which has consistently proved more accurate than the Met Office; I even get hourly forecasts, precip radar and satellite view.
On a number of occasions I’ve had rain within fifteen-twenty minutes of the forcast.

Many of these other websites use the metoffce prediction and just write a different forecast. The forecast depends a lot on the confidence interval you want to work with.

I suspect this is more about increased revenue generation rather than better forecasting. They rent out computing power to industry, especially aerospace, and a bigger, more powerful super computer means they can sell more capacity.

Is this the big data molly was on about?

It depends on the techniques used to partition up the data, process it, put the results back together, and then analysing it – as the is what the term ‘Big Data’ is generally meant to mean.

So probably not, as the ‘Big Data’ toolset is more focused on utilizing large amounts of commodity computing power to achieve results within reasonable timescales and costs, whereas basing something on Cray technology is focusing more on the other end of the scale.

Of course you’d need to knock up some kind of racking system, but that’ll be easier than keeping 40k play stations and their stupid in-board power supply design cool.

Simples

People have built ‘super computers’ by networking dozens of RaspberryPi’s the B+ model has twice the RAM of a PS3 forv £25! they use a lot less power than a PS3 too.

Seems to work for these guys though but maybe they should have used Raspberry Pis. I wasn’t actually being serious but 40,000 PS3 in a cluster would probably be quite a gaming set up 😀

The problem that the Met Office have is rubbish reporting in the media. The infamous “barbecue summer” forecast iirc actually said there was a one third chance of a heatwave, which the press then latched on to. So there was a two thirds chance it wouldn’t be a heatwave, which turned out to be correct.

For all their faults, I’d sooner rely on tbe Met Office than the publicity seeking loons that some of the tabloids use to give us dire warnings of floods, droughts, deserts and imminent snowmageddon.

the best thing the met office does is the rain radar map, best tool ever to get your local forecast. all the other stuff for me is pointless.

It’s a bit meh, only ranks at number 250 on the world supercomputer list.

The #1 entry is 😯

http://www.top500.org/system/177999

“Power: 17,808.00 kW”

17.8 MW ! Ouch.

Please please please tell me they are calling it Deep Thought.

Power: 17,808.00 kW

Didn’t Seymour Cray say that building a super computer was primarily an exercise in refrigeration?

Interesting graph regarding the accuracy of Met Office forecasts; five day forecasts are almost as accurate as one day forecasts used to be in 1980:

From here

The amount of improvement seems to be a fairly regular undulation, so maybe they’re buying a shiny new computer because it’s about time to start another peak…

MoreCashThanDash – Member

For all their faults, I’d sooner rely on tbe Met Office than the publicity seeking loons that some of the tabloids use to give us dire warnings of floods, droughts, deserts and imminent snowmageddon.

Do you mean Jonathan Powell, of vantage weather services? – he’s a comedy genius.

i’m sure ‘Johnathan Powell’ is really this guy:

The problem that the Met Office have is rubbish reporting in the media. The infamous “barbecue summer” forecast iirc actually said there was a one third chance of a heatwave, which the press then latched on to. So there was a two thirds chance it wouldn’t be a heatwave, which turned out to be correct.

For all their faults, I’d sooner rely on tbe Met Office than the publicity seeking loons that some of the tabloids use to give us dire warnings of floods, droughts, deserts and imminent snowmageddon.

Oh really?
http://www.metoffice.gov.uk/news/releases/archive/2009/summer2009

I suspect you have been misled by a bit of subsequent misinformation put out by the Met Office.

just because it can process the calculations down to a resolution of 300m, doesn’t mean that they’ll be any more accurate.

Not true. A lot of processes that affect the weather are small-scale. Wind blowing over hills etc. Finer resolution means you can model these directly.

300m will do a pretty good job of representing hills etc as well. Weather can be different one valley to next and this’ll have a chance of getting that right.

That looks like a press release rather than the proper forecast, but it does still make me look like a gullible idiot.

Again.

That looks like a press release rather than the proper forecast, but it does still make me look like a gullible idiot.

Never mind mate. Tell you what. Paypal Gift me a tenner and I’ll give a proper accurate forecast for 2015 😉

Didn’t Seymour Cray say that building a super computer was primarily an exercise in refrigeration?

It is, that’s why companies like Google etc are paying millions to hook up remote locations in the tundra to the internet, because the cold climate saves far more in aircon energy costs.

Oh really?

They are usually at pains to point out how unreliable their seasonal forecasts are. Why don’t you research how accurate they have been over the last few years?

From their own site:

Long-range predictions are unlike weather forecasts for the next few days. The nature of our atmosphere means it is not possible to predict the weather on a particular day months to years ahead. At this range we have to acknowledge that many outcomes remain possible, even though only one can eventually happen. Over the course of a whole season, year or decade, however, factors in the global weather system may act to make some outcomes more likely than others. It is because of this we can make a prediction , but we still need to show that a spread of outcomes is possible. To do this, we present the likelihood of each outcome by using a range or percentages. As a result, the forecast is useful to assess likelihood and risk, but not for warning of definite events.

As there is always a chance of one of the less-likely outcomes in the forecast happening, decisions based on a single forecast could be wrong. The full benefit of any long-range prediction is realised if they are used to guide decision-making over a longer period of time.

They are usually at pains to point out how unreliable their seasonal forecasts are. Why don’t you research how accurate they have been over the last few years?

They are a little more circumspect these days ever since they got their fingers burnt :-)over the barbecue summer.
FWIW I have spent the evening looking at a forecast of NW winds of 5kts, while the wind stays resolutely stuck at SE 5kts. Still 180 degrees out isn’t too bad.

Long-range predictions are unlike weather forecasts for the next few days. The nature of our atmosphere means it is not possible to predict the weather on a particular day months to years ahead. At this range we have to acknowledge that many outcomes remain possible, even though only one can eventually happen. Over the course of a whole season, year or decade, however, factors in the global weather system may act to make some outcomes more likely than others. It is because of this we can make a prediction , but we still need to show that a spread of outcomes is possible. To do this, we present the likelihood of each outcome by using a range or percentages. As a result, the forecast is useful to assess likelihood and risk, but not for warning of definite events.

TBH I think this is pretty good BS. The problem is, if they do a forecast and it says there is a 75% chance of x & y actually happens, you have no way of proving if this is one of the minority occasions or if they just got it wrong.

That’s all anyone can do. You just need to plan around those odds. What would you use the information for anyway?

P-Jay – Member

@RaveyDavey

People have built ‘super computers’ by networking dozens of RaspberryPi’s the B+ model has twice the RAM of a PS3 forv £25! they use a lot less power than a PS3 too.

Of course you’d need to knock up some kind of racking system, but that’ll be easier than keeping 40k play stations and their stupid in-board power supply design cool.

The problems with commodity computing clusters is the inter processor communication bottle neck. Depending on the problem this can be very important or not! Outside of cooling and power management that is where lots of the money gets spent infiniband networking fiber optic networking e.t.c.