In the UK these days, everyone loves to hate the Met Office, sometimes for getting long-range forecasts wrong - it predicted a barbecue summer and a warm winter for 2009 - but mostly because they don't make the weather right. Following the coldest winter for many years, the Met Office "appears to be in a losing battle with …
British? Or Daily Mail?
Over the weekend, the Mail had some "we're all going to be buried under ten feet of snow" screaming headline, for the third or fourth time this winter. My wife wondered if they were just rolling it out every fortnight until it came true, but perhaps they are simply trying to undermine confidence in weather forecasting.
I don't know if they do it on purpose, or whether it is an unconscious thing, but certain parts of the press do seem to be on a mission to undermine *anyone* who might have an informed opinion on any subject. That would, of course, clear the way for those with uninformed opinions, of which there are many but only those the loudest would actually be heard.
That's a joke, right?
"...writes the Daily Mail, the newspaper of choice for El Reg's British readers."
Please, just stop linking to them. You're only propogating their sheer stupidity and ineptitude. If I want my blood pressure to rise then I'll go and read the fail myself - I don't need the likes of El Reg pointing out their shite for me!
its gotta be said
Failing badly this winter
I don't know how the long range winter forecast worked out but the Met office managed to predict precisely the snow that we had in terms of times, amounts and when it stopped.
I tend to go to the met office direct because for some reason although the BBC has the same source data they seem to have a grumpy meteorologist who insists it's always going to be terrible.
Turns out the human interpretation of the data they have is still the most important bit.
Also, given the angle our soaraway Register likes to take, I'm pretty sure they have got confused between "Mail" and "Sport" for the daily paper of choice for UK readers...
I don't know where "breakfast" resides but here in Oxfordshire the Met office forecasts for the pre-Christmas weather were pretty far off, AccuWeather were much closer and I am now following them (as well as the Met office). I look forward to a third source so I can use a majority voting system.
does it matter
its not like its gonna change the weather.
i dont check the weather. and the worst bit is that i may need to scrape ice off the car window before i leave.
If its so bad i can drive its not going to change anything maybe knowing the day before.
e.g. this morning i was 3 min late because of ice.
The problem is their forecast for the current day is often wrong never mind them trying to piss about giving long term forecasts. Last week we were promised sleet but woke up to an inch of snow.
Maybe they should give a % chance and then we can measure them.
You'd probably still complain, and many people don't really understand percentage chance.
What does a 80% chance of rain mean to you? 80% of the region getting rain, it raining for 80% of the day, or it raining on 8 days out of 10 that they forecast an 80% chance of rain?
I can tell you that it is the latter, but I suspect that you had to read it twice to understand it. If you asked the readership of the Daily Mail this question, you would get most of them saying they don't know, and many of the rest just guessing one of the three.
And on the 2 out of 10 days that the chance says that it won't rain will still be wrong according to the media, even though the chances were correct. The media is fickle that way.
And if you remember your O Level or GCSC maths, you have to allow for short term anomalies . If you remember, when you toss a coin, over a larger number of tosses, you will get close to 50% heads and 50% tails, with some freak occurrences of it landing on it's edge. But that would not mean that you could not get 5 or even 10 heads in a row. It's unlikely, but it will happen sometimes. This is why it is never safe to bet on averages on a roulette wheel or any other game of chance.
The Met Office already do the analysis of predicted vs. actual weather. This is what they do to refine the model(s) to try to make them more accurate. But as I understand it, snow is particularly difficult to forecast, because very minor changes in the boundaries between the air layers can cause either rain, hail, sleet or snow.
The UK is caught between three major weather systems. You have the Arctic, that is cold air that attempts to push south, Europe that is fairly static, and at this time of year mostly cold, and the Atlantic which is very turbulent but quite wet and warm. As a result you have a three way battle, and I believe that it is one of the most difficult weather systems to predict in the world.
Precipitation is caused when warm moist air meets cold air in some way. The warm air moves up over the cold air, and as it gets higher and cooler, has to drop it's moisture. Depending on the temperature gradient, and the speed of air movement, and the turbulence at the air boundaries, the water droplets will coalesce and maybe freeze in different ways, leading to all of the possible outcomes.
Sometimes the forecast is easy, and sometimes it is not. Sometimes the warm air is kept south, leading to cold weather with no precipitation. Sometimes it just rolls straight off the Atlantic, leading to wet but warm weather spreading from west to east, and sometimes it diverts north, and then is carried back down south by the Arctic air, leading often to snow. And sometimes cold and warm air meet over the UK, and under these circumstances it is difficult to predict.
It is generally acknowledged that none of the current models give reliable results more than 10-14 days into the future. This means that the short term forecast will be based on modelling, and the seasonal forecast will be based on longer term cycles which can be identified by trend analysis of the of the past several years of actual weather rather than the air condition models. This makes a huge difference in the way that the forecasts should be used. You would not use the seasonal forecast to try to predict the weather for Christmas day, but you might use it to give an indication of what December in general may be like.
This is my schoolboy geography view of the weather, and no doubt somebody will pick holes in it. I do not claim to be a weather forecaster, but I believe what I've said is mostly correct.
I'm waiting for a maths or geography teacher to pull me up, so I've used the pedantic grammar alert icon!
"Last week we were promised sleet but woke up to an inch of snow."
Oh, how inaccurate of them. I imagine the difference needed to create one rather than t'other is pretty marginal. Reckon you could do it?
"...many people don't really understand percentage chance"
And never will if the Met continues to ignore it. In the US it is commonly used and understood by the vast majority as you described your third choice. It is never in the interest of the general public to hide information because the expert does not believe that it will be understood.
For heaven's sake
If you really want to try to find problems with the rock solid science behind global warming, you need to go to much more authoritative sources than the trashy UK press. They're great at reporting which sleb's just had a boob job or who's shagging who, or political intrigue, but science reporting in the UK press universally sucks golfballs through hosepipes. Just ignore it all (and yes, I'd say the same for the majority of stuff that's right, IMO; 99/100 times, they're right for the wrong reasons, and wouldn't know a Hadley Cell if it kneed them in the testes and nicked their iPhones.
Go to the source.
Read the journal articles.
Or admit you can't be arsed, and go with the scientific consensus. But don't claim it's all bollocks because of some blindingly obvious thing that Simon Heffer claims climatologists have missed out or not understood.
"rock solid science behind global warming"...
... yeah, right.
Mine's the one that doesn't have the University of East Anglia badge on it.
To be fair...
Not only is the British weather on the unpredictable side, there seems to be a lot of local variation as well which adds to the Met Office's problems. For example, I was driving back from Cambridge to the Midlands during the recent bad weather and I started off with clear weather. By the time I got to the A14/A1 interchange, it was snowing so heavily that I had to pull over at one point as visibility was literally zero. A few miles further on the A14 was clear and by the time I got home, I was faced with accusations of spending the evening down the pub.
Unless the Met Office can report on a 5-mile resolution, they have absolutely no chance of pleasing anyone. On this occassion the good people of Cambridge and Kettering would have applauded the Met Office for getting it right but I am sure the sheep and cattle buried under a foot of snow in the fields in between wouldn't.
Their forecast has a warm bias?
Hmm, I wonder where they got their climate model code from?
Very droll except that...
What the Mot Office's ACTUAL prediciton was that 2009 would be warmer than 2008, which despite the rain it was:
"Barbeque" is not to my knowledge an official metrological term or measurement of temperature, so sounds like this is a media communication issue not a scientific one.
After all, when you are trying to communicate with Daily mail readers, you do need to dumb down your message, especially when it doesn't involve immigrants etc
"What the Mot Office's ACTUAL prediciton was that 2009 would be warmer than 2008, which despite the rain it was"
Not sure what you're getting at? The met office spokesperson who came on the Today Programme on Radio 4 (on the day the 'bbq summer' press release was issued) predicted a hot summer for the UK weather and wasn't discussing the global temperature. I can't remember who used the word "barbecue" - interviewer or interviewee - but the met office bod was pretty confident about their prediction of a hot summer.
In the same "slot" on the same radio programme they also got another "sage" to make his prediction (a sort of druid-like weather predicter type person who probably lives on a farm somewhere in Sommerset). He predicted a bad summer. He said he based his predictions on some old English sayings (something along the lines of "if you get an early bit of good weather [which we had had in March] your summer will probably be crap" [the actual saying was more eloquent, of course]). Anyway as it turned out old father thyme (or whatever his name was) was right and the met office were wrong. So met office long range forecasting (at least on this occasion) wasn't up to the same standard as an old English 'rule-of-thumb' proverb - despite all their big super-puters.
It was the met office who used the phrase. See below:
Their predicition was for a warmer UK '09 summer not international temp - it was. In the press release you cite:
"Although the forecast is for a drier and warmer summer than average it does not rule out the chances of seeing some heavy downpours at times."
it might have been drier and warmer than 2008 by a few fractions of a degree but the 2008 UK summer was utterly utterly utterly crap (making a prediction of an improvement over 2008 almost a dead-cert). But they did predict a barbecue summer and got it wrong. I assume their confidence was boosted when we had a great period of weather during the early spring of 2009 (although I remember at the time of their "barbecue summer" press release thinking "didn't this pattern of weather occur early last year, just before the summer went to crap"?). Presumably their big grand computer didn't make this connection.
It will be interesting to see what they predict for this year. I doubt they'll put their neck on the line again with a prediction like last years. I expect if they do predict anything it will be along the lines of "2010 will be a mixture of sunshine, intermittent periods of hot weather, cold weather, storms and showers, with the possibility of long periods of rain". They'll probably also include a reference to climate change too - just so they can cover all bases and continue to push the idea that they're worthy of their generous funding.
Seems like it doesn't matter if the met office is right or you're wrong with their weather predictions - they can still expect to rake in the cash.
The Met Office get a large slice of its funding from...
...the taxpayer, purely to investigate Global Warming, ergo if they want more, Global Warming has to get worse.
Their predictions are farcical, the error margins huge due to their warming bias, nine of the last ten years have been cooler than the Met Office predicted! They are probably going to miss the next ice age with their long term forecasts.
They are no longer an impartial scientific body, they have become champions of a cause.
As for all the jokes about Daily Mail readers etc, reading some of the comments here, I am sure DM readers are intellectual powerhouses in comparison.
Are you using the same sources as the people mentioned in this article? Methinks you are... Apologies for not keeping up with the Mail's level of intellectual discourse.
It used to be
a 'best fit' methodology apparently. Look at the data for this year, map it to the best set of previous years' data and base your prediction on that. Maybe they don't use this method anymore, but I wouldn't be surprised and this would explain why they get it wrong for long range forecasts. In fact, I doubt they really get many right, it's just that we've had the best snow in 30 years, so the media pick up on it more.
A better use for that supercomputer
As the ultimate multi-user combat simulator c/w fully immersive graphics. How much would Reg readers fork up for a couple of hours on that? It would surely beat swimming with dolpins.
There's no visualization hardware on the cluster. It could do the maths, but would have to send the results to another system to drive the display hardware. And linking them together would be another story altogether. I doubt any display system could cope with the sheer raw compute power of one of these clusters.
Article wrong on so many counts
The European Centre for Medium-range Weather Forecasting (ECMWF) have two almost identical systems (Power 6 575's), except that they are each bigger (over twice the size, apparently), so I don't know where they got their facts about "Only Japan is known to have a bigger one". These systems are also installed in the UK, but not in Exeter. Look on the Top 500 Supercomputer list.
If you look at the IBM product website for a description of the Power6 575, you will find that each "filing cabinet-sized blocks you see in the picture" is actually a rack containing up to 12 Power6 575 systems, each of which has 32 Power6 processors. This is a possible total of 108 P6 575s if all of the racks are full, with 3456 Power6 speed-demon processors. The 'filing cabinets' are over two metres tall, and over a metre and a half deep according to the specs.
Because it is made up of separate systems, it will be configured as a cluster rather than a single system, using something like the IBM Cluster Systems Manager software (Xcat is not yet mature enough for such a premium customer), and using a cluster aware scheduler to control jobs that span multiple systems.
I think that there must be some more hardware somewhere (another row behind, perhaps), because the racks shown cannot contain all of the storage mentioned in the article.
If there is some hugely scalable storage, perhaps provided using another IBM technology such as GPFS, there are probably a number of storage nodes that also probably have Power6 processors.
One only has to wonder at the networking infrastructure necessary to glue it all together. According to the top 500, it uses Infiniband, but even using this I would suspect that they would need multiple fabrics to allow all of the nodes to communicate fast enough. Judging from the photograph, I would suspect 8 separate fabrics (it looks like 8 medium sized switches in the two communication racks). The cabling must be a nightmare.
Still, I would imagine it looks very impressive. Or maybe I don't have to imagine, as they do sometimes allow visitors in to see them.
One wonders if there are plans are to install Power7 systems as a mid-life performance kicker? Something like Blue Waters but no as big maybe?
Global warming is a poor term, climate change is better since the weather seems to be strange just about every place. In Toronto we have had almost no snow, I have not used the snow blower once. It's been warmer then normal too. Washington has had WAY more now then normal, the Vancouver Olympics don't have enough snow...
"Chief weapon against British cynicism"
The *chief* weapon would be a *published* detailed description of the model they use along with the data that drives it and the assumptions (and there will be quite a few I suspect) used.
I'd *like* to think the present model already covers the *obvious* things like a non-spherical Earth with irregular gravity (check the model in the CRC Physics & Chem handbook) and the wildly varying solar activity (It's *roughly* 11 years a cycle, and for a 70 year period did not seem to happen at all). IMHO trying to split the surface of a sphere (or anything like a sphere) into squares is already asking for trouble as this implies an infinitely short top edge when you get to the poles.
But I don't know. I *do* know that putting *everything* on the table *would* show which areas the data is strong and well understood (IE that have conclusions that are defensible) and which ares are in doubt.
My gut feeling (and a skimming of the "readme" file in the climategate leak) is that Knuth was right. Premature optimisation *is* the root of *all* evil. Let's get *all* the factors included on a grid that recognises we live on near spherical Earth. If it takes 1 week to get a 99% accurate forecast 1 day in advance SFW. You would *now* have to speed it up (by at least a factor of 14) but you would *know* its findings are trustworthy. Speeding it up is an optimisation excercise. Historical efforts in this regard (EG n-body simulations of whole galaxies) have managed to compress run time from 1 year to 1 day (IE 365:1)
You can guess what' s in my (large) pocket.
try investing in glass
Living 6 miles from the Met Office in Exeter they should try looking out the window before providing today's forecast. I know the peninsula has very different weather north to south and east to west but they invariably WRONG for the locals.
You would think the smokers who go out for a ciggie would think occasionally "I didn't forecast this".
Try it yourself.
I suggest you try to produce a 90 second encapsulation of the weather for an entire region, and see whether you can do it in sufficient detail to be completely correct across the entire south-west for an entire day.
FYI, they do a local forecast for Exeter every week for internal consumption in an open briefing to staff, and each week reflect on how accurate the last week's forecast was. On the ones I have seen, they are quite candid about when it is wrong, and also when it was right. More often than not, it is right within reasonable bounds (a degree or two min and max, and within a reasonable time discrepancy for significant events). All forecasts are done with a margin of error, represented by a percentage chance. This means that it is expected to deviate from the forecast sometimes. 60% certainty means that they expect that the forecast will not match 40% of the time.
Oh, by the way. The building has lots of windows, but the secret is not reporting what the weather is doing now, but predicting what it will be doing in the future. This is what forecasting means!
An Obvious Error
The problems with the recent Met Offices 'long-term' forecasting(BBQ Summer, and Mild Winter), is that they have had to include the 'debunct' and erroneous data provided by the IPPC - because this lamentable government has been totally sucked into their claims. More and more of their claims are proving to be at best error-strewn, at worst corrupted(intentionally). Until the world wakes up to the fact that they have to be ignored, and the Models used by the Met Office have this IPPC data removed, the mid to long-term forecasting will continue to be fulle of error.
Propaganda not forecasting
The reason that people are fed up with the MET is that just like the CRU and the BBC they have thrown their objectuivity away, in order to be prophets for the great global warming apocalypse.#
Now that the vacuity of the IPCC report has been exposed, their anger is turning on the con-men; and I'm sure that's the description I want to use for the senior MET people who've taken huge bonuses while pushing propoganda.....
- Product Round-up Smartwatch face off: Pebble, MetaWatch and new hi-tech timepieces
- Geek's Guide to Britain The bunker at the end of the world - in Essex
- FLABBER-JASTED: It's 'jif', NOT '.gif', says man who should know
- If you've bought DRM'd film files from Acetrax, here's the bad news
- VIDEO Herschel Space Observatory spots galaxies merging