CPDN/BBC

Message
Author
User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

CPDN/BBC

#1 Post by rebirther » 27.06.2006, 23:16

Schon wieder ein neues beta Projekt:
http://climateapps1.oucs.ox.ac.uk/cm3beta/
Last edited by rebirther on 06.09.2006, 12:50, edited 1 time in total.

Cheeky@Boinc
Guest
Posts: 177
Joined: 19.12.2005, 23:29
Location: Warstein

#2 Post by Cheeky@Boinc » 28.06.2006, 20:48

This one will test different length workunits (e.g. 1/10/20/40/80/120/160 years)
Jetzt schon bis 160 Jahre :wink: Klasse

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#3 Post by rebirther » 06.09.2006, 12:50

BBC hat wohl die Credits jetzt erhöht:
226,80 auf 259.20

User avatar
grmbl
Admin
Posts: 274
Joined: 18.12.2005, 21:57
Location: Potsdam
Contact:

#4 Post by grmbl » 06.09.2006, 21:08

Haste Beweise? Wenn das stimmt, müsste ich ja den BOINCcalculator abändern.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#5 Post by rebirther » 07.09.2006, 09:56

lt Forum-Admin bleibt das so.

Cheeky@Boinc
Guest
Posts: 177
Joined: 19.12.2005, 23:29
Location: Warstein

#6 Post by Cheeky@Boinc » 08.09.2006, 04:45

Wielange braucht CPDN für einen Checkpoint?

Ich hab BOINC ja nicht ständig an :wink:

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#7 Post by rebirther » 08.09.2006, 09:12

Cheeky@Boinc wrote:Wielange braucht CPDN für einen Checkpoint?

Ich hab BOINC ja nicht ständig an :wink:
Das hängt ganz vom Rechner ab (s/TS). Mit meinem P4 3,2Ghz aller 20min Speicherpunkt, das senden eines Trickles dauert so um die 20-24h. Ansonsten für weitere Infos hier im Forum BBC Guide :!:

Cheeky@Boinc
Guest
Posts: 177
Joined: 19.12.2005, 23:29
Location: Warstein

#8 Post by Cheeky@Boinc » 09.09.2006, 00:16

aller 20min Speicherpunkt,
Damit könnte ich leben :wink:
Ansonsten für weitere Infos hier im Forum BBC Guide
Das Projekt BBC ist ja begrenz darum interessiert es mich nicht!

Wenn dann CPDN

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#9 Post by rebirther » 09.09.2006, 09:28

Das Projekt BBC ist ja begrenz darum interessiert es mich nicht!

Wenn dann CPDN
Wie begrenzt? CPDN gibt auch nur die BBC WUs aus. Hadsm und Sulphur sind schon wieder Geschichte.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#10 Post by rebirther » 09.05.2007, 10:39

"aborted by user,,
CPDN Monitor - Quit request from BOINC...
Suspended CPDN Monitor - Abort request from BOINC..."
Bevor einer die Fehler-Meldung bekommt, folgende Punkte beachten, betrifft kurz vor PC runterfahren:

1. Modell/WU anhalten
2. BOINC beenden, kurz warten, ob auch alles aus dem Speicher raus ist
3. Computer runterfahren

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#11 Post by rebirther » 23.06.2007, 22:34

Database maintenance
Wed, 20 Jun 2007

Due to some unscheduled maintenance, the database server will be down over the next day or two. We hope to have things back up as soon as possible.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#12 Post by rebirther » 11.09.2007, 17:28

Neuigkeiten:
In den Prefs kann man jetzt entscheiden, ob lange, kurze oder beide Modelle:
* The HadSM3 slab ocean model consists of 45 years and is the shortest model CPDN has offered for a long time. This is the most suitable model for slower computers and often also for computers running several projects.

If you do not select either type of model in your preferences you will receive either of the above or a HadCM3 full 160-year model (ie the same as we have been running since Feb 2006).
Image
Image

Je nach CPU läuft die WU nur ein paar Tage. Im 2. Bild kann man sehen, das die %-Anzeige irgendwie aus dem Ruder läuft, knapp 50% erreicht aber nur 13% angezeigt.

Update:
Lt. Admin im Forum hat das Modell 3 Phasen, das erklärt dann auch den kleinen %-Wert, pro Phase also knapp 33%.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#13 Post by rebirther » 22.09.2007, 20:52

Leider sind die UK Slab Modelle verbuggt, folgende Fehler sind zu beobachten:

- Modell friert ein und bekommt eine blaue Farbe, Zeit läuft ewig langsam weiter
- BOINC 5.10.20 gibt eine Crashmeldung (log) aus, das Modell kann leider nicht mehr reaktiviert werden auf Normalzustand

Eine neue Anwendung gibt es leider nicht, das Problem ist aber bekannt.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#14 Post by rebirther » 05.02.2008, 09:57

Mal ein großes im letzten Jahr veröffentlichtes Update aus dem Forum:
We have just recently bid goodbye and good luck to PARDEEP PALL, who is moving on to take up a position at ETH, Zurich. Pardeep successfully completed his DPhil early this year looking at attributing the UK floods of Autumn 2000 to anthropogenic influences. He has since been working at CPDN as a post-doc. His experimental set-up - the Seasonal Attribution Project - will already be very familiar to many users! At ETH he will continue working on understanding and attributing/predicting extreme rainfall events.

HIRO YAMAZAKI, whose research background includes oceanography, GFD and planetary modelling and who joined CPDN originally on the computing side, is now working on the EU funded project MILLENNIUM investigating European climate of the last millennium. He is using the 'FAMOUS' model, a coarse resolution version of HadCM3 that enables about 1000 years of simulation to be performed in approximately the same time as one 160-year BBC simulation. Via CPDN he hopes to acquire a large ensemble of runs that begin in 800 AD and run for 1200 years, and will investigate changes not only in temperature but also in other quantities for which there are proxy data in Europe. Later on in the project he will also make future predictions of climate change under a variety of different emissions scenarios.

KUNIKO YAMAZAKI is in full writing up mode and is close to finishing her DPhil thesis. Kuniko is a physical oceanographer and has been studying some of the processes occurring in the ocean of the coupled model experiments. She is interested in how ocean heat uptake relates to the state of the ocean and how feedbacks are thereby formed. She has helped to develop a simple 2-level model, based on that of Jonathan Gregory at Reading University, and comparing it with the 20 ocean levels in the HadCM3 coupled model she finds it does very well. Her application for funding to continue as a post-doc was successful - congratulations Kuniko! - and in the New Year she plans to distribute experiments whereby she can investigate possible changes in the meridional overturning circulation in the North Atlantic resulting from idealised greenhouse gas change scenarios.

Third year DPhil student HELENE MURI is investigating how well CPDN models fare in simulating paleoclimates. She is looking particularly at the mid-Holocene period - 6000 years before present. Her work is important in helping us decide which models could most reliably be used for climate prediction. Her models will be tested against a variety of paleo-observations; as part of this she will use an offline vegetation model to determine which models produce a realistic shift of the Arctic treeline. She is very close to being able to distribute her slab experiments so look out for those!

HELEN HANLON, who is also a third year DPhil student, studies the summer '03 heatwave in southern Europe and is investigating whether and how this may be attributable to anthropogenic greenhouse gas increases. One aspect she is looking at is how soil moisture feedbacks may have contributed. Technically she is not actually involved in CPDN, instead using ECMWF's IFS (Integrated Forecasting System) running on their supercomputer, but it is just possible she might decide to involve herself with CPDN at some point. Watch this space.

RUTH CEREZO-MOTA, again a third year DPhil student, works with the high resolution regional model PRECIS. She is currently running this with NCEP and ERA-40 data providing boundary conditions. She is particularly interested in monsoon rainfall over Mexico and the importance of atmosphere/ocean coupling to the processes involved in this. She has a background in physical oceanography and has previously studied hurricanes using a simpler model. She does not plan to distribute any experiments via CPDN - rather her work contributes towards the general set-up and testing of the PRECIS model.

NEIL MASSEY originally joined CPDN as a research assistant but began a DPhil two years ago, during which time he has been developing a storm tracking algorithm which he will use to investigate possible changes to the location of storm tracks under various greenhouse gas emission scenarios. He is now taking a year 'out' to be employed on a contract among the Met. Office, CEH Wallingford and Oxford University, during which time he will calibrate his model against ERA-40 reanalyses. He is using a relatively high resolution atmosphere model (HadAM3P) forced with observed SSTs, similar to that used in Pardeep's Seasonal Attribution Project, but with a few modifications. Once the behaviour of this model has been validated he will then perform his storm track investigations.

DAN ROWLANDS has just started as a DPhil student having successfully completed an MPhys project with CPDN earlier this year. In his MPhys work he tested Clausius-Clapeyron predictions of changes in extreme precipitation over Northern Europe, using data from Pardeep's Seasonal Attribution Project ensemble. For his DPhil he is now investigating the issue of deciding which parameter combinations are likely to give the most reliable climate forecasts. Starting with the slab model data, he is initially calculating which combinations give the best agreement with observations; he will then use statistical techniques to devise new combinations which could potentially give an even better fit.

Earlier this year ROSALIND WEST completed her fourth year MPhys project in which she used CPDN data to investigate using the seasonal cycle in temperature to constrain climate sensitivity. She found quite different results for the slab and the coupled experiments and is in the process of writing this up for publication in the peer-reviewed literature.

Two new MPhys projects are planned to take place between now and March:
CHLOE SHARROCKS will be looking at relationships between temperature and precipitation changes in the slab model data, and
BENJAMIN GRANDEY will be investigating the effect of sulphate aerosol forcing in the coupled model experiments.

Scientists WILLIAM INGRAM, DAITHI STONE and DAVE FRAME continue to be very productively involved with the Climate Dynamics group in a large variety of ways - update to follow in the New Year!

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#15 Post by rebirther » 14.10.2008, 09:15

Sieht so aus das die Linux Anwendung 2x schneller ist als im Windows, zumindest teste ich gerade die neue hadsm3, der derzeitige s/TS liegt bei 0,54, das Modell hat 4 Phasen.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#16 Post by rebirther » 22.10.2008, 10:20

The work on the database is 80% complete and we expect it to be finished tomorrow at the latest. Apologies for the unexpectedly long outage, and thanks for your patience.
Na hoffentlich geht die Seite heute wieder, mein Modell ist kurz vor Ende und die anderen Trickles warten und warten und...

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#17 Post by rebirther » 07.03.2009, 22:17

CPDN main project

HADAM3P NEW! SHORT!

Many thanks to Tolu who expects to release this new model on the main CPDN project soon. In some ways it is quite different from previous models we have run, so this advance description should help members decide whether to try it.

* We expect Tolu to release it for Windows, Linux and Mac.

* It's much shorter than any previous models and will probably take from 65 to 200 hours to complete, depending on computer speed.

* Each model covers a period of 25 months. The time slices of different models overlap.

* Previous HADAM models could only be downloaded to computers with at least 1.5GB RAM. This restriction does not apply to HADAM3P models.

* These models appear to move a lot of data between CPU cache and main RAM. They are not recommended for older, slow computers. On multicore computers several HADAM3Ps running side-by-side will usually slow each other down. It would be better to run them alongside HADCM, HADSM or HADSM MH models which are less demanding in this respect.

* The initial download for the first model is very large, over 125MB. The downloads for subsequent models are smaller, but still large. The final and only file upload is about 43MB. These models are probably not suitable for members with a dialup connection.

* There are model graphs as well as globe graphics, as usual.

* Each completed model gives 1982 credits.

* Neil Massey of the Oxford CPDN research team describes in detail this interesting and important validation and attribution experiment on the new website here .

* We can select the model types we want to run in the climateprediction.net preferences section of our accounts. (I can only see my own there, not yours!)

* If you definitely don't want one of these models, for example because you have dialup, in the Projects tab of BOINC Manager set climateprediction.net to No New Tasks now. (It must say 'Won't get new tasks' in the Status column.) When HADAM3P appears in your account's model preferences, deselect it. Update climateprediction.net in BOINC Manager to tell the server about your preferences. Then allow new tasks for CPDN.

This post may be edited later if the CPDN version of this model turns out to be significantly different from the version we have beta-tested.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#18 Post by rebirther » 01.08.2009, 18:50

On July 27 Ian Hay posted on the CPDN forum News threads:
================================================================================

BBC Climate Change Experiment

The BBC CCE project is being shut down. All models which are still running
will be sent a killer trickle next time they send a message to the scheduler
and you should not attempt to restore from backup. The scheduler and upload
servers will be disabled at the end of this week and you will not be able to
return any more trickles or upload files after that happens.

The results which have been returned will be used as part of a submission to
the IPCC AR5 (Intergovernmental Panel on Climate Change Fifth Assessment
Report). Carl Christensen is working part-time for CPDN on this and is
adapting the model to collect some further data for the submission. This
will be released on the main CPDN project.

Any BBC CCE users who have yet to sign up for the main project should follow
the instructions at
http://www.climateprediction.net/board/ ... 235#p56235 .

================================================================================

The BBC CCE, run by CPDN as a separate project, is now therefore a retired
project. Most of the very few BBC models still sending in trickles and files
would not have completed in a usable timescale. One BBC model, for example,
recently sent in its first trickle of 160 more than two and a half years
after the model was downloaded. It is not an efficient use of manpower or
server and home computer resources to maintain such a situation
indefinitely. Carl is therefore freezing credits to remove this overhead
from the database. The basic website functions and forum remain available
and the forum is still regularly visited by moderators who provide
continuing support.

We may recollect that the BBC project attracted a large number of members of
the general public both in the UK and worldwide as well as exisitng CPDN
members and multi-project Boinc crunchers. While the attrition rate among
the general public was high, thousands of very long 160-year climate models
were completed, contributing to the BBC television programme that presented
the results and the splendid BBC results website
http://www.bbc.co.uk/sn/climateexperiment/. The project was an example of
the sort of public outreach that Boinc wishes to demonstrate in its
forthcoming application to the NSF for continued funding.

Many BBC members later transferred to CPDN. My subjective impression is that
the membership retention and model completion rates were highest among those
who joined teams and used the forums.

BBC climate model results, like all CPDN results, have been made freely
available to climate researchers worldwide through Milo's results website.

The Boinc programmers contributed to the CCE by providing a special version
of Boinc. Many thanks to all the subscribers to this mailing list who
contributed to the BBC project.

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#19 Post by rebirther » 08.11.2009, 16:59

Mal was zum Status von CPDN Modellen gefunden:
Myles Allen, CPDN's chief researcher, has given an update on current experiments to Milo. Here it is:

Dear all,

Here is a long-promised update on the experiments we are writing up, running or preparing for release, with apologies for not having kept everyone in the loop. First, a status update on the experiments we have already released, in more-or-less chronological order of release:

1) The original slab experiment. This continues to provide grist for doctoral theses and detailed studies of how the atmosphere in the Met Office model responds to parameter perturbations. An ongoing debate is whether it is possible to generate the very high responses to doubling carbon dioxide that were reported in Stainforth et al without rendering the climate of the model demonstrably unrealistic. The original “11oC” models were, with the benefit of considerable hindsight and additional scrutiny, relatively unrealistic, but plenty of models have now been found that generate these high sensitivities through different mechanisms (indeed, the main result of that whole experiment was how easy it turns out to be to increase the sensitivity of this model to alarmingly high levels), and we are sending out targeted work-units to explore these in more detail.

2) The BBC Climate Change Experiment / climateprediction.net transient ensemble. Dave [Frame] has the main results of this experiment now written up, and it’s on my desk for editing (sorry, Dave), so we hope to submit that paper soon. It’s taken a while, because we’ve been having to develop a whole new approach to analyzing coupled model ensembles as we go along. It’ll have been worth the wait. There are still gaps in parameter space that we want to fill in on this experiment, so we are still sending out these coupled work-units as these are identified.

3) Pardeep Pall’s seasonal attribution experiment is written up and on my desk for editing (sorry, Pardeep).

4) Neil Massey’s study of extreme weather events has completed its first stage (simulation of 1950-2000), which he is writing up for his thesis, and will shortly be distributing work-units for 2000-2008 – this one feeds directly into Jara Imbers’ experiment (see below).

5) Kate Ricke’s geo-engineering experiment. This is now written up and submitted for publication, and she has some ideas for follow-up runs.

6) Falk Niederhorst’s study of climate predictability, based on the slab experiment, is in progress.

Those are the experiments that we have released through the main climateprediction.net site. We now get onto the experiments we have in beta release or still in preparation.

7) William Ingram’s transient ensemble for the IPCC Fifth Assessment. Lots of people are contributing to this experiment, but we’ve asked William to coordinate it, so I’m using that title as short-hand. This follows essentially the same format as the BBC Climate Change Experiment but with an updated model, allowing a wider range of perturbations to be explored, updated climate drivers (projected greenhouse gas levels and the like) to be consistent with the drivers used by the major modeling groups in the so-called “CMIP-5” model inter-comparison, and an updated design: we are now running from 1880-2120 split up into 40-year work-units rather than single 1920-2080 160-year work-units, which none of you liked because they took so long to finish. It has been a fair amount of work for Tolu making this possible, but now he has done it, we will be able to target work-units much more flexibly in the future. This is a multi-stage experiment:
a. Ocean spin-ups, completed last year through the beta site.
b. Atmosphere spin-ups (40-year work-units to compute the surface fluxes associated with perturbed atmospheres), currently under distribution through the beta site.
c. Coupled control work-units, which have the same drivers as the spin-ups (a simulation of average climate conditions in the late 19th century), run to check the model’s stability. We aim to start distributing these soon through the beta site.
d. Coupled transient work-units, which contain the drivers for simulating 1880-2120, which we hope to be distributing by the end of the year (we are checking the driver files very carefully).

We will, of course, be using results from the existing BBC Climate Change Experiment to target these runs, so we should be able to explore both parameter space and climate driver uncertainty more comprehensively this time. One of the main criticisms often leveled at the CMIP model inter-comparisons is that, to allow all the modeling groups to run the simulations, they only explore a very restricted ranges of scenarios for future emissions. We can explore a much wider range, to put the CMIP-5 results into context.

8 ) Jara Imbers’ extension of Neil Massey’s experiment. Neil has been running an atmosphere-only model driven with prescribed sea-surface-temperatures which allows higher resolution and shorter work-units. Jara is extending at least a few decades into the future now that we have a range of simulated surface temperature changes from the various coupled experiments (both our own and other peoples) to draw on. This isn’t entirely straightforward, to put it mildly, mostly because of complications with how to make the sea-ice consistent with the surface temperatures (in the coupled models, sea-ice is computed, so while it might be wrong, it is at least self-consistent), but it does provide a way of getting substantially higher atmospheric resolution than is possible in the coupled experiments, which is valuable for impact studies. The software for this experiment already exists (it is identical to Neil’s), but we are working on the driver files.

9) Hiro Yamazaki’s MILLENNIUM experiment. This follows the same format as the BBC Climate Change Experiment but uses a faster model (with a lower-resolution atmosphere) to simulate climate change over the past millennium. The longer time-scales should provide a much stronger constraint on climate change over the next couple of centuries. Hiro is aiming to get this on general release by the end of the year.

Finally, our most ambitious experiments under preparation involve the use of the embedded regional model to give substantially enhanced resolution in selected regions for climate impact studies. These will provide a unique way of simulating realistic weather for a wide range of possible climate changes, which will be invaluable for planning adaptation strategies. This is also a perfect application of distributed computing, because of the large ensembles required to generate adequate weather statistics. There are plenty of other groups around the world looking at this problem, but they are all hampered by not being able to run ensembles as large as they would like to, and hence have to resort to complicated statistical post-processing to try and estimate what they would have found if they had been able to. We feel (and plenty of people agree with us) that it is much better as a general rule to provide climate impact assessments with actual climate model output rather than simulated output from a statistical model, because this ensures that the fields provided are internally consistent (that is, satisfy the laws of physics). We are taking two complementary approaches based on the experiments we have already performed, both funded largely by Microsoft Technical Computing:

10) The coupled regional climate modeling experiment, led by Richard Jones at the Met Office. In this experiment we embed the regional model HadRM3 directly into the coupled atmosphere-ocean model HadCM3L, to simulate regional climate change directly for selected 5-year periods within the coupled transient ensemble. This builds on an 11-member ensemble performed with this set-up for the UKCP09 Climate Projections project. So far, we have been testing a domain located over Southern Africa at 50-km resolution and a domain located over the Pacific Northwest at 25-km resolution. As many of the beta-testers will know, the Pacific Northwest domain has been giving trouble, which may simply result from that specific choice of domain and resolution, in which case fixing it will simply be a (tedious but straightforward) matter of varying the domain until we find something that works. Or it may also indicate a bug, which clearly we would want to track down before releasing this experiment. Given these problems, we have decided to release another experiment in parallel.

11) The atmosphere-only regional climate modeling experiment. This is a straightforward synthesis of Neil and Jara’s atmosphere-only experiments with the coupled regional modeling experiment, in which we embed the regional model into the higher-resolution HadAM3P model rather than into HadCM3L. The lower contrast in resolution between the driver atmospheric model and the regional model may help make this setup more stable (it has certainly been used very successfully under the Met Office PRECIS project), but time will tell. This is still under preparation, but given it is substantially simpler than the coupled regional modeling experiment, it may end up getting released first. It remains an open question which is the better experimental design: running an atmosphere-only model in so-called “time-slice” mode allows for higher global atmospheric resolution, which improves the simulation of the circulation driving the regional model, while running the regional model directly embedded in the coupled model allows atmosphere-ocean interactions to be simulated more faithfully. When a question is open like this, the best option is to try both approaches which, with your help, we will.

Once again, none of these experiments would be possible without our participants’ support: thank you all again, and please keep modeling.

Myles

User avatar
rebirther
Admin
Posts: 2902
Joined: 19.12.2005, 00:59
Location: Germany

#20 Post by rebirther » 29.03.2010, 16:41

Die FAMOUS 6.02 im Windows ist zwar schnell aber völlig instabil, viele WUs gehen schon nach einiger Zeit mit Fehler raus. Im Beta-Projekt wird gerade die 6.03 getestet, die s/ts Zeiten sind doppelt so hoch, dafür sollte es stabiler laufen.

Post Reply