BOINC Weather Projects... BBC Climate Change vs Climate Prediction

[CupCak3]

What are the differences between these two projects? I'm traditionally a F@H guy but I would like to branch out to some other science facets.

I'll be putting together an x2 machine soon and want to run BOINC on at least one of the cores.

 

[BlackMountainCow]

The BBC project only studies about 5 years of change in realtion to the recent dramatic weather phenomena in the UK (IIRC 2001 to 2005) while the CPDN project simulates a much longer period, with a calibration period starting in 1945 IIRC. Also, a BBC model takes about 14 days to crunch on a X2 3800+ core while a CPDN model after 25 days of crunching is only at 23% complete. Last but not least, the BBC project is limited to some 1000 models, while the CPDN is more or less open ended.

:beer:

 

[Geomagick]

I have to say that the BBC project is slightly more detailed than this. After over 300 hours of time it has only got to 14% on my X2 4400.

The BBC model also starts in 1920.

 

[BlackMountainCow]

Ouch, right I mixed that up with the Seasonal Attribution Project, sorry.

Basically BBC and CPDN are the same, but the Seasonal Attibution Project is the shorter one.

 

[petrusbroder]

Well, the BBC project uses the data from 1920 to 2005 to find a model (or: to calibrate the models) which fits the data best. Then from 2006 until 2046 (IIRC) they make a prediction and the outcome of the different models will be analysed.

 

[Rattledagger]

Well, currently there's 3 different CPDN-projects:
1; Seasonal Attribution, uses 430 MB memory, runs high-res 1-year model looking on a specific problem. A model takes roughly 4 weeks on a 2 GHz-computer, less on faster computer.
2; BBC Climate Change Experiment; Runs Coupled Model, 80 year hindcast + 80 year forecast, takes roughly 5 months on 2 GHz-computer, less on faster computers, more on slower...
The BBC-experiment is a "limited" experiment, to get "normal" users to run a climate-model in combination with some programs BBC is showing in UK. If not mistaken, the plan was 7 months or maybe a couple months longer, and afterwards the web-site will likely point to the "normal" CPDN-experiment so downloaded model will run to the end. Also, users (and work done) will likely be moved to CPDN...
3; Climateprediction.net, CPDN, the "main" project. Currently runs the same Coupled Model as BBC, so for the science, run-times and credits, there's currently no difference between running BBC or CPDN...

For future plans, the latest released application apparently also supports "short" models, making it possible to make 80-year or 40-year-runs instead of the current 160-year-runs... But, how long till some shorter models is actually distributed is another matter...

Also, planned released later this year or early next year is a 64bit-model, but haven't got any real system-requirements (appart for obviously 64-bit-cpu) on this yet...

 

[Buzzman151]

CP is the to go...

 

[CupCak3]

thanks for all the help guys!