WORLD WEATHER modelling and predictions
 

Dear colleagues!

I am thinking about global world weather modelling and prediction center.

If we shall take the hole Earth's atmosphere, cover it with computational grid till 50 km high and shall try to model the world's weather.

Can anyone suggest, how much RAM and clusters it will be neccessary to perform such task?

Thanks in advance

Michail

Re: WORLD WEATHER modelling and predictions
 

The climate modeling work at the Oak Ridge National Lab uses most of the available time on the most advanced parallel processors that Cray, IBM, and other US computer manufacturers can provide. You can probably learn more about those computers at www.ORNL.gov.

The guys doing that work are adament that they are not doing weather forecasting, which doesn't require the entire earth in one model. As I understand it, forcasters use actual data for initial and boundary conditions and look only a few days ahead. So they only need to model part of the earth's surface at once.

So the ORNL work is evidently not directed at your question directly, but it's a place to get an idea of the size of the problem. And of the computing power needed ...

Re: WORLD WEATHER modelling and predictions
 

I started setting up a problem like this for fun awhile ago in Fluent - I assume you are just doing it for a hobby.

My system only has few Gigs of RAM so I was going with something quite crude. Then I realized I would have to map out the continents, so I looked for some digital data on the web. Then, should land be included in 3D, that is, should mountains be included?

At this point, I started thinking about other things that could be modeled for fun and left the world modeling to the experts.

WORLD WEATHER the most ambitious project
 

Dear Jim Park

Dear Allan Walsh

Thank You very much for Your responces. About "hobby". I am just started my CFD carreer, so I am looking for interesting task. One can be the modeling of swirling reacting flows in very complex geometry. Other - this.

I've made some calculations. Earth surface area is 510 mln. square km. (71% oceans and 29% ground). With 50 km altitude it will be:

25500 * 10^6 cubic km.

I suppose that avarage scale should be about 1500-500 m.

To this we should add

- radiation and heating modeling,

- evaporation modeling

- turbulence modeling

- cooling modeling

- gravitation modeling

- phases and particles modeling

- pollutions modeling

------------------------------------------------------

ESA's CryoSat cost was $165 million

If we shall add to this project Latin America, USA & Canada, EU, Russia, China, India, Japan, Australia - we shall have these money. Center may be placed in the Africa and use the Sun energy.

Re: WORLD WEATHER the most ambitious project
 

Michail wrote: Center may be placed in the Africa and use the Sun energy.

----------- You can put it at my Mom's house, if you like... :) She lives in Africa.

diaw...

Re: WORLD WEATHER the most ambitious project
 

Why not? Into the heart of Sachara desert. It will be the best and safiest place.

The summer season will give enough energy to produce H_2 and O_2 (in order to use it when will be cold or unsunny season).

Moreover - it's a good investition. I even dare to suppose that it will return money during 7-10 years after opening.

To that - it will be a new Los-Alamos, just International. And bigger.

Its development will help to the poorest African countries.

Its appearence can help to coordinate work of all climatoligsts.

Re: WORLD WEATHER the most ambitious project
 

... because most of the equipment would probably be stolen, or damaged by local wars. Your security cost would be enormous...

Re: WORLD WEATHER modelling and predictions
 

As a loose estimate you will have a grid of atleast (to compete with current lower model resolutions) 432x325x38 which corresponds to about 41Mb per 3D field. multiply this by the number of required fields - 3 velocity, pressure, density, temperature, humidity, + (>)10 chemical species etc, and you are talking about 1-2Gb of memory.

This is just the estimate of the memory required to run the model once an initial start dump has been created. The memory cost/computing time required by a 4Dvar data assimilation on the whole globe is around (at the very least) an order of magnitude greater than this.

All the major weather centres perform these calculations daily along with nested higher resolution regional simulations in order to give the weather forecast on the news.

A European centre for weather forecasting already exists!

Re: WORLD WEATHER modelling and predictions
 

Yes? But why noone could predict the Katrina? Why NASA provide the sattelite information? - Because these pridictions - are not reliable.

Why was the flooding of EU without emergency signal?

Re: WORLD WEATHER the most ambitious project
 

Firstly I haven't heard about local wars in Sachara.

Secondly - the density of population there is very low.

Re: WORLD WEATHER modelling and predictions
 

I think you'll find that Katerina was predicted - it depends upon whose forecast you look at and which one was used by the Americans.

You need to learn a great deal more about Meteorology/NWP and what is currently state of the art before making suggestions like this; i.e. the funding of the individual (big) weather centres easily exceeds your $165 million and all of them have the latest (multi-million $) supercomputers and access to global satelite and Earth based observations.

You also have to keep in mind most weather centres are not only in (friendly through the WMO) competition with one another but are Government funded organizations with defense related responsibilies.

Re: WORLD WEATHER modelling and predictions
 

Dear Tom

Thank You very much for Your answer. I've got a lot from it. BUT.

They operated not so well as You believe.

If Katerina would be predicted (and its trajectory and its power would be sent to the US Goverment), as it should be done, I suppose the Washington Post and New-York Times would not wrote about failures and human losses.

It was flooding in France recently. And what??? Any emergency signal.

I am thinking about World Climate and Nature Center. It will be completely International organisation, where one could investigate all climate and weather problems applying to the whole Earth.

When some groups work on the local solutions - they will reach nothing. I suppose it would be better to create the global Center.

I've seen some articles about this in QNET-CFD. I am sure - they can not afford to theirselves such computations. And noone in the World separately can afford the same. And it's clear that we should consider the whole Earth for adequate modelling. List of models which should be implemented I already listed (please see above).

Re: WORLD WEATHER modelling and predictions
 

-------------------------------------------------

Thank You very much for Your answer. I've got a lot from it. BUT.

They operated not so well as You believe.

If Katerina would be predicted (and its trajectory and its power would be sent to the US Goverment), as it should be done, I suppose the Washington Post and New-York Times would not wrote about failures and human losses.

-----------------------------------------------------------

You are missing the point here (politics has a lot to do with this). A counter example to what you are saying is the storm (hurricane) that hit the UK in the late eighties which was predicted by one forecating centre and not another. The forecast that went out in the UK went with the one that said there would be no hurricane! The forecasters had access to both forecasts.

And, as I've already said, nearly all forecasts are global and all major centres have access to the fastest computers presently available (and also share data through the WMO). A new centre is not going to be able to perform bigger and better simulations simply because they will have to use the same type of supercomputers that are currently used in weather forecasting.

Also to make one big centre you will have to divert funds from the current Government funded centres which is not foing to happen.

A question : how many people do you see working at this new centre?

I'm curious about this because I don't think you realise just how big the major weather centres are - a single one is bigger (in terms of manpower, computer resources and finance) than all the commercial CFD companies put together.

Re: WORLD WEATHER modelling and predictions
 

Michail,

Please check out the following websites for information on weather modelling.

http://vortex.atgteam.com

http://www.emc.ncep.noaa.gov/modelinfo/

http://www.ecmwf.int/

Regards, Nashat

Re: WORLD WEATHER modelling and predictions
 

Dear Tom.

Thank You very much for Your question

I've found WMO

http://www.wmo.ch/index-en.html

Then I descovered European Centre for Medium-Range Weather Forecasts

And its computational facilities:

http://www.ecmwf.int/services/comput...m_cluster.html

and

http://www.ecmwf.int/services/comput...ahandling.html

---------------------------------------------------

The equipment is an IBM Cluster 1600 system, based on pSeries 690+ servers interconnected by a low latency high speed network . Each separate cluster comprises 70 pSeries 690+ servers, each of which has 32 1.9 Ghz Power 4 CPUs. Most of the pSeries 690+ servers have 32 GB of memory, however 7 in each cluser have 128 GB, and 2 of these are logically partioned ("LPAR-ed") into 2 16-cpu logical nodes with 64 GB each. This allows for greater flexibility, especially for programs that require large amounts of memory, but which cannot be converted to parallel programs that could use the memory systems of multiple nodes.

Funding £26,394,200 was contributed to the budget for 2004 by the Member States and Co-operating States. The States contribute pro rata their GNI.

---------------------------------------------------------

What does prevent us to build more powerful system???

TEN (i.e. x10) times powerful ???

As I've already said - we shall develop the global Earth's system. And we shall develop it both for the Brazil and Japan, as well as for Russia and Australia. Of course USA Canada and EU.

It will allow us to simulate the breathing of the our planet completely. Noone exept us will save us from the destruction. Creating and development of such Earth's Center can face many enemies such as powerful corporations, because it will be clearly seen, who and how affect on the environment.

I dare to suggest so hipotesis about the Katerine. Someone was very interested in what had happened. Oil prices also have its lobbists.

Re: WORLD WEATHER the most ambitious project
 

"Most ambitious" maybe, but still not adequate. If you're really aiming for global long term weather prediction with minimum use of measurements and accurate resolution of relatively small scale and highly dynamic phenomena (hurricanes, blizzards, even floods, hey... how about earthquakes (let's do earth-atmosphere fluid-structure interaction)?), you may as well bury your money in the Sahara sand. Think again about the required spatial and temporal resolution and computing power and then come up with a more realistic plan... and don't take the current practice as a guideline because that won't be good enough for your ambitions... not even close...

how many CPUs would be required... what should be the CPU speed... how much RAM... how much disk space... what network speed...?

chances are, you might not find a suitable machine in your local computer store... who is supposed to build (and maintain) such a machine for the peanuts that you're offering?

now, assuming we can build huge scale circuit of solar panels, efficient enough to power your machine... and self-cleaning so they won't get buried in Sahara sand within minutes... how are you going to cool the computer which will create enormous amounts of heat by itself, and additionally has to stand the heat of it's desert environment... and how are you going to protect the whole setup against the drastically changing temperatures (day vs. night) and against the finest sand grains that tend to find the most unlikely ways to get into every corner of your machine... do you think to maintain all this on a daily basis is going to be cheap?

no, I think I need some more detailed and more professional information before I start investing into your project...

...and I haven't even mentioned the biggest of all practical problems you are going to deal with, if you're thinking of an international collaboration: Building political consensus.

well, for a dinner table discussion your idea is good enough, but a far cry from being realistic...

Re: WORLD WEATHER modelling and predictions
 

Michail,

I can only say one thing after reading your postings.

"YOU ARE NUTS"..

Re: WORLD WEATHER modelling and predictions
 

Thank You.... :))

You are not the first who said this. :))

Many things in this world have begun from dinner coversations, often such conversations seemed impossible and crazy.

I've already have an engineer's degree and served in desert.

When I'll get my degree in CFD I'll do this.

any offence.

But I suppose You have a lack of imagination. As for me I try to think without brakes.

Re: WORLD WEATHER modelling and predictions
 

A few points as my last word on this subject - you appear to selectively (mis)reading the information and drawing your own incorrect conclusions.

(1) ECMWF exists under the agreement that it does not compete with any of the member states. This actually means that it has free access to all the data used by the member states and as such does not need to budget for obtaining ground and space based observations. This is a huge saving in expenses. Hence the low figure you quote which is less than 1/10th of the budget of any one of the larger members. If ECMWF were in competion this goodwill would be removed and the budgetry requirements would become enourmous.

(2) The reason why you can't have a machine 10 times faster is that if one existed they would already have it. Most NWP models run at close to the scalabilty limit of the supercomputer and so doubling the number of cpus in the simulation will actually slow the code down. Also if, as you appear to think, bought a computer with 10 times the number of CPUs it would cost close to half a billion dollars! (i.e. just scale up the cost of the computer by 10).

(3) As Mani above mentioned the Sahara is probably a little too hot to have a functioning supercomputer - you'd be better off in Antartica. Interesting fact here - most of Cray's patents were on coolant systems and not computer designs!

(4) As I've said in all my previous posts - All the large meteorological centres already perform global weather forecasts at close to your ideal resolution. It is also believed, and there is a great deal of supporting evidence for this, that slight improvements to the resolution do not have a significant effect upon the forecast skill. Current thinking is to run an ensemble with slightly different initial conditions and to look at the spread of the predictions.

Re: WORLD WEATHER modelling and predictions
 

Dear Tom

Thank You very much.

I'll think about all this, as well as about what respectful Mani said.

I've read somewhere recently about possibility of developing meteorological and geological warfare. May be it's only rumors and nonsence, but....

Creating and Developing of such Global International Earth's center would be very helpful for all.

There won't be any politics, only scientists, engineers and technitians.

Michail