New CPDN experiment

Desti

Joined: 20 Aug 05

Posts: 117

Credit: 23762214

RAC: 0

26 Aug 2005 23:47:54 UTC

Topic 189771

(moderation:

)

sulphur cycle experiment launched
Fri, 26 Aug 2005

From today, anyone downloading a new experiment from climateprediction.net will download an extended version of the experiment, which adds a fully interactive sulphur cycle to the model, to investigate the so-called 'global dimming' effect. People whose computer spec. (available memory or speed) is not sufficient for this experiment will continue to run the standard, 3 phase experiment. We expect this experiment to continue until early next year.

http://www.climateprediction.net/science/s-cycle.php

Desti

Joined: 20 Aug 05

Posts: 117

Credit: 23762214

RAC: 0

New CPDN experiment

31 Dec 2005 1:03:41 UTC

Message 15935

(moderation:

)

Quote:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=3772#18779

just some more "preview" of the transient coupled model:

1) we will have it out in beta test in about two weeks, and it will be available from the "regular" (i.e. this) CPDN site as well as a separate site BBC viewers will come in from (but they will be a separate "pool" of users).

2) it's quicker than the slab & sulphur as far as timesteps per second, but as Les mentioned it's 160 years (+ 6 days, for 4147632 total timesteps). It will be from the years 1920 to 2080 -- so will span 80 years "historical" (hindcast) and 80 years in the future (forecast) -- which they tell me is a common range for studies, IPCC reports etc.

3) we were going to split it up into two 50-year jobs, then two 80-year as we figured that would best match the literature. But there's a lot of trouble with say, somebody doing workunit 0j50_0004322 on a Linux 64-bit box, and then for the final 80 years a Win 32-bit box gets it, so it's not a well "controlled" thing. So having the "full monty" 160-year seemed the best

4) it runs about as fast as a spinup, so say the fastest machines, i.e. P-IV 3.8GHz will do the full run in 2.5 months.

5) I didn't think it was more unstable than the others, but maybe it is!

6) I am moving from the old "CPDN hogs your whole hard disk" paradigm. In the old days we fancied everyone running CPDN would do perhaps 1 model, and it would be nice to keep your data around locally for advanced visualization etc. But I don't think that fits in well with BOINC, as we ended up using too much space, people end up doing 10 models a year etc. So I will have the model "clean up after itself!" The trickles will provide plenty of information at monthly global & regional scales.

7) the trickles will be smaller than the spinups but more than the slab runs, probably about 100KB per day (since a meg per day is too much for the general public!)

8) there will be 16 uploads (intermediate uploads every 10 years) -- so even if you crash in the middle or towards the end we will have a lot of valuable information. Each upload will be about 3-5MB Max (so cumulatively it will be a lot, i.e. 50-80MB). Which means I hope the BBC or someone buys us more upload servers!