WUWT readers may recall the story by the Daily Mail about the new supercomputer.
The Met Office has caused a storm of controversy after it was revealed their £30million supercomputer designed to predict climate change is one of Britain’s worst polluters.
The massive machine – the UK’s most powerful computer with a whopping 15 million megabytes of memory – was installed in the Met Office’s headquarters in Exeter, Devon.
With a total peak performance approaching 1 PetaFlop — equivalent to over 100,000 PCs and over 30 times more powerful than what is in place today. It is capable of 1,000 billion calculations every second to feed data to 400 scientists and uses 1.2 megawatts of energy to run – enough to power more than 1,000 homes.
With all that power, surely it must produce some quality digital reckoning.
Bishop Hill has located the “supposedly secret” winter forecast sent to the British government. The details of the forecast produced are nothing short of astounding.
When the kerfuffle over the Met Office’s winter forecast blew up, I wrote to the Quarmby team to see if they had actually received a copy of the Met Office’s cold-winter forecast, which was apparently sent to the Cabinet Office. It is alleged that the forecast should have provided sufficient warning to the government machine to ensure that everyone was ready for what happened in December.
Today, rather later than I expected, the Quarmby team have responded and have helpfully provided a copy of the forecast:
Met Office Initial Assessment of Risk for Winter 2010/11
This covers the months of November, December and January 2010/11, this will be updated monthly through the winter and so probabilities will change.
Temperature
3 in 10 chance of a mild start
3 in 10 chance of an average start
4 in 10 chance of a cold start
Precipitation
3 in 10 chance of a wet start
3 in 10 chance of an average start
4 in 10 chance of a dry start
Summary: There is an increased risk for a cold and wintry start to the winter season.
Looking further ahead beyond this assessment there are some indications of an increased risk of a mild end to the winter season.
Yes that seems clear, doesn’t it? Seeing the numbers produced, personally, I think this less expensive computer, using Digital Advanced Reckoning Technology (DART) can do the job of making odds equally well, using less power, less space, and less money:
I really love this one:
Looking further ahead beyond this assessment there are some indications of an increased risk of a mild end to the winter season.
I think its been done, something about “March coming in like a lion and out like a lamb” IIRC. But really, I never thought that a “mild end to winter” could be categorized as a “risk”.
But this forecast for the start of winter still doesn’t square with the Met Office map output.
Here’s the Met Office supercomputer enhanced model output forecast from October 2010: 
See the story about that controversy here and here
Gary Mount says (January 21, 2011 at 6:59 pm): “Ummm, 15 million megabytes is only 15 gigabytes.”
Actually 15 gigabytes is 15,000 megabytes. So the MET computer with 15,000,000 megabytes has 15 terabytes of memory.
Hu McCulloch says (January 21, 2011 at 3:19 pm): “With a total peak performance approaching 1 PetaFlop — equivalent to over 100,000 PCs and over 30 times more powerful than what is in place today. It is capable of 1,000 billion calculations every second.
A petaflop would be 1 quadrillion flops, or 1,000,000 billion calculations per second. 1,000 billion flops would be just a teraflop.”
I wonder if the article was using the older british meaning of “billion”, i.e. a million million, instead of the newer one of a thousand million, as in standard Americanese?
BTW, kudos to Craig Moore for his computing METric: the “CowFlop”. 🙂
LazyTeenager says (January 21, 2011 at 9:05 pm): “Well the agenda is to discredit the met office and this is followed by a lot of predictable, breathless commentary.”
Actually, that’s one thing the met office does pretty well all by itself.
doesn’t matter if they have 5,00o of these supercomputers they’ll get whatever “results” or predictions they want just by tweaking the software…and the sad thing is they won’t get any real answers because there are TOO MANY VARIABLES…
They would have been better off buying the machine that goes PING. Or better still, sell it and lease it back from the company they sold it to and that way it would come under the monthly current budget and not the capital account. Then people at WUWT wouldnt be able to poke fun at the amount it all cost.
Schrodinger’s Cat says:
January 21, 2011 at 10:31 am
Julia Slingo of the Met Office is complaining that they don’t have nearly enough computing power. Four times more would be good…
http://www.nature.com/news/2010/101230/full/news.2010.685.html
Unbelievable!”
Wow, four times greater. Maybe instead of all this 3 in 10 stuff they could then say a 12 in 40 chance of ….
LazyTeenager says:
January 21, 2011 at 9:05 pm
Well the agenda is to discredit the met office and this is followed by a lot of predictable, breathless commentary.
But it is all being done in an information vacuum.
The verbal forecast supplied here has been simplified, I guess, to make it easy for bureaucrats to skim? but it is near enough to useless for decision making. I would make a wild guess and suggest that this forecast is not to be used by the people whose real job it is to prepare for winter.
Weather is a matter of probabilities and probabilities do allow 2 sixes to be thrown every now and again. People who claim they knew that 2 sixes were going to come up, after the dice are thrown are just attention seekers.
Sheesh lazy, we are having a little fun here. BTW your post is reflective, referring to your self you say “I guess”,…”wild guess”…and referring to the MET; “the dice are thrown”
Sounds like you and the met work alike.
“Gary Hladik says:
January 21, 2011 at 10:17 pm
Actually 15 gigabytes is 15,000 megabytes. So the MET computer with 15,000,000 megabytes has 15 terabytes of memory.”
I see where I went wrong. I am used to the modern notation of using Gigabytes instead of megabytes, the old archaic amount of memory they used to have in late 20th century computers.
Why they could not have used the more modern term Terabyte of ram I do not know.
As an aside, I used to buy memory for about $40 a megabyte.
Vicky Pope says that her computer is not big enough and wants to update it. It will still produce poor forecasts whatever is spent.
Bill Jamison says:
January 21, 2011 at 11:00 am
Looking at those numbers another way, the Met forecast predicted a 40% chance of a cold dry start to winter.
Actually Bill, it’s a 4/10 chance for cold multiplied by a 4/10 chance for dry. This equates to the probability of it being cold and dry as 16/100 or 16%. Very poor performance by the MET.
I’d give them a score for this test as 16% as well. A serious failure in my book.
Elsewhere (?) a certain James Sexton commented, “While all of the connections haven’t been made, yet, I think its fairly obvious to any casual observer that drywet is a product of warmcold.”
Sounds like the kind of forecasting the SuperDuperPooperScooperPuter could deal with!
P.S. While the Met ‘puter can out-flop mine, my ancient P4 system is maxed out with a Gigabyte of RAM. That “massive 15 megabytes of memory” sounds pretty wimpy.
A mild end to winter eh? Methinks the cuckoo’s are here already!
It is difficult to understand why they use the word “risk” when they are talking of probabilities.
It would be nice to see Christopher Booker run with this story in his Sunday column. Could be quite a laugh and it would bring to a wider audiance the quality of forecasting from the Met Office and would make more people question uts worth.
I have to make one correction here. The performance of this computer is not measured in PetaFlops but MetoFlops.
It’s late fall and the Indians on a remote reservation in British Columbia asked their new chief if the coming winter was going to be cold or mild. Since he was a chief in a modern society, he had never been taught the old secrets. When he looked at the sky, he couldn’t tell what the winter was going to be like.
Nevertheless, to be on the safe side, he told his tribal community that the winter was indeed going to be cold and that the members of the village should collect firewood to be prepared. But, being a practical leader, after several days, he got an idea. He went to the phone booth, called the Environment Canada National Weather Service and asked, ‘Is the coming winter going to be cold?’
‘It looks like this winter is going to be quite cold,’ the meteorologist at the weather service responded. So the chief went back to his people and told them to collect even more firewood in order to be prepared.
A week later, he called Environment Canada’s National Weather Service again. ‘Does it still look like it is going to be a very cold winter?’ ‘Yes,’ the man at National Weather Service again replied, ‘it’s going to be a very cold winter.’
The chief again went back to his people and ordered them to collect every scrap of firewood they could find. Two weeks later, the chief called Environment Canada’s National Weather Service again. ‘Are you absolutely sure that the winter is going to be very cold?’ ‘Absolutely,’ the man replied. ‘It’s looking more and more like it is going to be one of the coldest winters we’ve ever seen.’
‘How can you be so sure?’ the chief asked. The weatherman replied, ‘The Indians are collecting firewood like crazy.’
Remember this whenever you get advice from a government official!
I thought I would post this again, seems apt for this thread.
http://s446.photobucket.com/albums/qq187/bobclive/?action=view¤t=Spike2.mp4
Is the Met supercomputer the one they used in Emerald City over the rainbow?
This is really funny, but seriously isn’t it possible for a meteorological condition to exist where it is not possible to predict an outcome as being more likely than others? On this blog you hear “chaotic”, “ever-changing”, etc. used often in describing climate systems. Maybe the Met computer is smart enough to know it’s limits.
Just The Facts says:
January 21, 2011 at 7:58 pm
R. Gates says: January 21, 2011 at 4:42 pm
“I have no doubt that all the GCM’s have major flaws, and that the real root cause is trying to model a system at the edge of chaos.”
Agreed, Earth’s Climate System is unbelievably complex and continuously evolving, and GCM’s development is still nascent. But why are we investing a tremendous amount of time and resources reacting to alarming prognostications from these obviously flawed models?
_________
I agree with much of what you say, but here’s probably a current difference between us at the moment. While some people might skoff a the notion that I’m a 75/25 warmist vs. skeptic, that’s about where I place the odds that some level of AGW warming is occurring. But more to your point about models and chaos. I do have faith that the models are giving me general information about trends, though not about the exact timing and particulars. I go back to the analogy of the snowstorm. Imagine you were fairly confident it was going to snow at your house overnight, and wanted to develop a model that would predict exactly when and where the first snowflake would fall on you front sidewalk and exactly how much snow was going to fall. Such a model would be impossible to develop as there are too many interacting and chaotic variables to take into account. That does not mean however, that you can’t be pretty certain, that if the NWS issues a winter storm warning for your area, that you won’t wake up and find some level of white stuff on the ground.
Such is the case for GCM’s. I think in fact they might be decent at giving us pretty good clues about general trends…i.e. the polar ice will be declining overall leading to a seasonalably ice-free arctic sometime later this century, the hydrological cycle will be acclerating leading to heavier downpours in areas that are prone to downpours, the oceans will be warming, Greenland and Antarctica will be melting, etc. I just happen to think that the chaos in the system from the many excellent other influences such as you’ve pointed add way to much unknown, such that no matter how big the Met office computer gets, and how refined the GCM’s get, they are still inherently dealing with a chaotic system, and as such, they might get better at telling more about the trends, but never about the specifics, and certainly will be horribly wrong about the timing and of course, will never be able to predict those inevitable “tipping” points that exist with any system undergoing chaotic changes.
It does not take a fancy computer to see that UK winters and in particular the beginning of winters namely the month of December were getting colder.
WINTER TEMPERATURES [CENTRAL ENGLAND]
2007 6.4C
2008 5.6C
2009 3.5C
2010 2.4 C
2011 NA[ estimate 2.2C to mid january]
WINTER TEMPERATURES HAVE BEEN DROPPING [4 YEARS ?] IN ROW
DECEMBER TEMPERATURES[CENTRAL ENGLAND]
2006 6.5C
2007 4.9C
2008 3.5C
2009 3.1C
2010 -0.7C
DECEMBER TEMPERATURES HAVE BEEN DROPPING 4 YEARS IN A ROW
I’m going to install a Digital Advanced Reckoning Technology board immediately at home for improved weather prediction
The problem with the GCMs is that because most if not all use the ‘two-stream’ approximation to the optical physics of aerosols introduced by Sagan** to predict cloud albedo, they wrongly predict aerosol pollution increases albedo of thicker clouds. This allows much higher climate sensitivity for CO2 than is really the case.
The classic example of the effect of this basic mistake is the ‘warming bias’ of the Hadley model. As each year passes, assumed higher CO2-AGW distorts the prediction upwards. It’s time they got this bit of physics right.
The clue is to look at rain clouds, dark underneath, higher albedo. Yet they have larger droplets, lower optical depth, so the models predict lower albedo. You correct matters by direct backscattering at upper cloud surfaces. Mie theory predicts this – the second interaction, also reduction of droplet size switches the effect off, making another AGW. This is a game changer: there is now no proof of any net CO2-AGW.
I suspect the inadequacy of this physics was known to insiders well before AR4 was published. After it was learnt there was no experimental proof of ‘cloud albedo effect’ cooling, NASA ‘commissioned work’ to find out why. It concluded there was less water in polluted clouds, the inference being no problem with the physics.
Twomey, who proved the albedo of thin clouds was increased by pollution but warned this should not be extrapolated to thicker clouds, was given a prize. A NASA web publication replaced Twomey’s correct physics with a plausible ‘surface reflection’ argument; there’s no such physics: http://geo.arc.nasa.gov/sgg/singh/winners4.html
This false premise is widely believed in climate science: it appears to be taught. Until the models are corrected they will always be wrong. Unfortunately, the fix requires the IPCC’s prediction of high-feedback CO2-AGW be rejected as plain wrong..
**It assumes just internal diffuse scattering, biased in the backscattering direction by constant ‘Mie asymmetry factor’. No-one appears to have realised that Mie’s solution of Maxwell’s equations for the interaction of light with a dielectric sphere assumed a plane wave. The resultant equation appears to fit experimental albedo-tau data but because there’s a second optical process, direct backscattering at upper cloud boundaries, it’s essentially a curve fit with no predictive capability.
Video from a recent service upgrade of the Met Office supercomputer
That’s outrageous!
We Brits should use seaweed and a pocket Casio, to forecast the weather, like you guys in the States.
Just The Facts says:
January 21, 2011 at 7:58 pm
Thankyou for posting all the references . . . that must have taken some time and thought . . . I look forward to going through them . . . again thank you.