Global climate change: World's most accurate computer for weather predictions

[jay gw]

Earth Simulator will allow British experts to predict climate change

By Michael McCarthy
17 January 2005

The Greeks went to the oracle at Delphi. The Romans looked into the entrails of slaughtered chickens. But our society has a different way of foretelling the future: the climate model.

For the past 20 years, scientists around the world have been using huge mathematical models of Earth's climate system, run on computers, to predict the coming course of one of the gravest threats the planet has known: global warming.

Now British researchers have built the biggest such model. It is so ambitious it may be able to warn of "surprises" - sudden, potentially disastrous leaps in climate change which have not so far been predicted but which could overwhelm any defensive preparations for global warming.

It will almost certainly make the most accurate predictions yet about how the atmosphere will heat up during the coming century, and how the climate may change, with more extreme weather events such as hurricanes and violent rain.

But it is so big it cannot be run on any computer in Britain; it can be operated only on a giant supercomputer in Yokohama called the Earth Simulator. Everything about the Earth Simulator, the biggest super-computer in the world when it was built two years ago, and still the fastest machine of its type anywhere, is awesome.

It needs a floor area the size of four tennis courts to house its 5,120 processors, and it has a speed of 36 teraflops, or 36 trillion - yes, that is 36,000,000,000,000 - floating-point operations (or calculations) per second.

http://www.belfasttelegraph.co.uk/news/story.jsp?story=601573

The EU took global climate change seriously and the US didn't. Now they are way ahead of everyone on this issue.

[Nikk]

Originally posted by jay gw
Earth Simulator will allow British experts to predict climate change


It will almost certainly make the most accurate predictions yet about how the atmosphere will heat up during the coming century, and how the climate may change, with more extreme weather events such as hurricanes and violent rain.



But can it accurately predict the weather here 14 days ahead?

[LucyR]

Originally posted by jay gw
Everything about the Earth Simulator, the biggest super-computer in the world when it was built two years ago, and still the fastest machine of its type anywhere, is awesome.


Actually the Blue Gene/L is now the fastest. Currently nearly twice as fast as the Earth Simulator. When completed (it's being built at Lawrence Livermore National Lab) it will be capable of more than a third of a petaflop.

http://www.infoworld.com/article/04/11/08/HNibmsupercomputers_1.html

Btw, I think the word "accurate" in your title is a misnomer. Doesn't matter how fast it is, if you put ******** in it you'll get ******** out.

[Grammatron]

Originally posted by jay gw

The EU took global climate change seriously and the US didn't. Now they are way ahead of everyone on this issue.

As evident by what? Seems to me they are as much in the dark as anybody.

[a_unique_person]

Originally posted by Nikk
But can it accurately predict the weather here 14 days ahead?

Climate prediction and weather prediction are two completely different tasks.

[LucyR]

Originally posted by a_unique_person
Climate prediction and weather prediction are two completely different tasks.

This is very true and you would think well known, but someone always makes the same witty remark.

[a_unique_person]

Originally posted by LucyR
Actually the Blue Gene/L is now the fastest. Currently nearly twice as fast as the Earth Simulator. When completed (it's being built at Lawrence Livermore National Lab) it will be capable of more than a third of a petaflop.

http://www.infoworld.com/article/04/11/08/HNibmsupercomputers_1.html

Btw, I think the word "accurate" in your title is a misnomer. Doesn't matter how fast it is, if you put ******** in it you'll get ******** out.

The climate can be modelled at a macro level. This has been done with existing records to demonstrate the validity of the concept.

[username]

Originally posted by LucyR
This is very true and you would think well known, but someone always makes the same witty remark.

Perhaps someone with some understanding in this area could clarify the difference for me.

If I say it is raining, I am describing the weather. If I say it is not raining, I am describing the weather. If I say a given area only gets about 2 inches of rain per year I am describing the climate as arid. If I say an area will have 20F temps tonite I am describing the weather. If I say the area has an average low temp of 10F and an average high temp of 90F I am describing climate. At least I think all the above is correct, but I am open to correction if I am using the terms incorrectly.

So the thing is, climate appears to be weather trends. Perhaps it isn't, but that is why I am asking for clarity here.

Given the inability to accurately predict weather in the short term I find it difficult to have confidence in climate models looking decades out. I believe there is evidence of warming such as melting polar ice caps and such, but I don't know that the models we have of climate change are accurate. That said I don't see how a more powerful computer would make the models more accurate.

[BobK]

The title of the thread is very misleading. Either a computer is accurate or it is broken.

It's the software that determines the accuracy of the simulation.

Remember. With computers the acronym GIGO applies. Garbage In, Garbage Out.

Anyone know of a complicated piece of software that is totally bug free? Or do they all simply have unintended and undocumented features?

[LucyR]

Originally posted by a_unique_person
The climate can be modelled at a macro level. This has been done with existing records to demonstrate the validity of the concept.

I'm not claiming anthing about the science on which the code is based. All I'm saying is that it doesn't mean anything to refer to a computer as the most "accurate".

[a_unique_person]

Originally posted by LucyR
I'm not claiming anthing about the science on which the code is based. All I'm saying is that it doesn't mean anything to refer to a computer as the most "accurate".

Understood.