Getting started with climateprediction.net
From Unofficial BOINC Wiki
[edit] How do I get started with Climateprediction.net?
Before starting it may be sensible to read Rules and Policies especially if you are not familiar with BOINC. There is also System Requirements if you are not sure if your computer is capable enough. The Climateprediction FAQ also has a section on what you may want to find out before starting.
[edit] Overview installing BOINC
1. Download BOINC (if you have already got version 5.2.13 or later running skip to point 3) from here and run the .exe file to install BOINC.
2. Run boincmgr.exe to start the program.
More detailed instructions on installing BOINC are available here.
[edit] Overview attaching to a project
3. On starting boincmgr for the first time, it may automatically start the attach to project wizzard. If it doesn't or you are already running another project, you can start that wizzard by choosing Projects, Attach to project.
4. Enter the project URL http://climateprediction.net/
Wait while this communicating message is displayed.
5. Enter account information of email address and password. Take care to avoid typing mistakes like in this example.
Wait while this communicating message is displayed.
If successful this 'Attached to project' message should display. Click finish to end the attach to project wizzard.
6. Check to make sure your firewall allows the BOINC Client Software full browsing capability on port 80.
That should be it! The BOINC Client Software should look after downloading the Climateprediction.net (CPDN) Science Application, the details of your particular Climateprediction.net Model and all the other communications necessary. If this does not happen, see BOINC FAQ: Troubleshooting on attaching to a project or Climateprediction FAQ - Issues with Getting started or BOINC FAQ.
For more on what the BOINC System does to get you started see Getting Started.
A more detailed version of the attach to project guidance including dealing with proxy configurations is available here.
[edit] What is CPDN?
Climate Prediction Dot Net (or Climateprediction.net (CPDN)) is a Distributed Computing project which aims to refine the accuracy of climate modeling. While there are simple models, running a full global atmosphere ocean coupled model has always been a very computationally intense process and, before this project started, had only been run on 64 bit computers which largely restricted it to supercomputers.
The aim of the project is to run hundreds of thousands of different models using the idle time of the Participant's Computers, thereby leading to a better understanding of how models are affected by small changes in the many Parameters known to influence the global climate.
The vast majority of Participants are running Climateprediction.net (CPDN) as a Science Application under BOINC. The project can also still be run using the Climateprediction.net (CPDN) classic program which is used in an Open University short course on weather and climate modeling. This might be useful for those using Windows 98 and others that find their computer does not like running the BOINC version. Note, however, that the output from the classic version is no longer of any use scientifically to the project, and is supplied for educational purposes only.
[edit] How is it done?
The project has previously run three equilibrium experiments which were aimed at finding out more about the model by imposing large Forcings on the Climateprediction.net Model. These models had simple oceans which allowed equilibrium to be reached quickly. This enables the Climateprediction Team to know more about the effects of different Parameter combinations and perhaps filter out some combinations that will not result in a suitably stable climate.
The next stage which was launched in February 2006 uses a Coupled Model with a dynamic ocean that can now start to try to simulate realistic climate. The filtered Parameter sets will be use to perform a Hindcast of the climate from 1920 to 2000 with realistic Forcings. The models will be scored according to their accuracy. After a model completes the Hindcast, it will move on to a Forecast of the climate from 2000 to 2080. A variety of assumptions of future Forcings (or scenarios) will be used. The scores from the Hindcast will be used to weight the results from the Forecast.
Further details are available on the strategy page.
[edit] The Experiments
For a basic understanding of what the project is trying to accomplish please read "An Introduction To The Science Behind Climateprediction.net".
The different models that Climateprediction.net (CPDN) has and will distribute are detailed below in time order. Therefore if you have joined recently you are likely to be running the Sulphur Cycle, or, if you started after February 2006 you will probably be running the Transient Coupled Model.
- Classic Slab Model - The original experiment not under the BOINC System. This is a 3 Phase 45 year model. First Phase measures fluxes necessary to keep the model stable. Second control Phase tests whether the measured fluxes do keep the model stable. The third Phase includes a forcing of doubled CO2 to see how the climate reacts to such a change. This model is used in the OU short course but can still be used by those unable to run the BOINC versions.
- BOINC Slab Model - this is the same as the classic Slab Model just released under the BOINC System.
- THC - an investigation of how the climate might change in the event of a decrease in the strength of the ThermoHaline Circulation. This experiment has now been closed to new Participants as they have sufficient Results. It was a four phase model totalling 60 model years. The first three phases were identical to the above slab models. The fourth phase imposed the effects of a 50% slowdown in the Thermohaline circulation. The Climateprediction.net (CPDN) page is here.
- Sulphur Cycle - an investigation of the effect of sulphate aerosols on the climate. This commenced on 26th August 2005. See Sulphur page. This is a pre-requirement for the Hindcast. It is a 5 phase model totalling 75 model years. Timesteps are around 70% longer making it around 2.8 times longer than the initial slab model. The Climateprediction.net (CPDN) page is here.
- Coupled Spin-Up Model - inclusion of oceanic influences into the basic model in a more dynamic and realistic way than the initial Slab Model. This is a pre-requirement for the Hindcast. This is in progress but it is not planned to be publically released. The fastest 200 - 500 computers have been invited to join because it is a 200 year model and Results are needed by February 2006 for the Transient Coupled Model launch.
- Transient Coupled Model. This comprises an 80 year Hindcast and an 80 year Forecast. The Hindcast is to test how well the models perform at recreating the climate of 1920 to 2000. This is also described as "experiment 2" on the strategy page. Estimated launch February 2006. The forecast of what will happen in 2000 to 2080 follows the Hindcast and this is described as experiment 3.
- High Resolution Model - this will be a Slab Model as normal followed by a period calculating at high resolution. It is unclear whether/when this will be launched.
[edit] What are the system requirements
The BOINC version runs under Microsoft Windows®, Apple's Apple Macintosh OS-X and Linux, with further details on the BOINC tech FAQ page.
The Climateprediction.net (CPDN) classic program will only run under Microsoft Windows®, with the requirements detailed in the Classic tech FAQ page.
I am mentioning Climateprediction.net (CPDN) classic in the BOINC documentation as Windows 98 may not cope with Climateprediction.net (CPDN) under the BOINC System.
It is also worth noting that the Climateprediction.net Models are very long. So if you have an old computer that is only switched on for a few hours a week, it may take too long for these models to complete particularly if you try to run more than one project.
[edit] Can I run more than 1 model on my computer?
Yes, but only with the BOINC version. If you have a multi-processor or hyper-threaded Pentium 4 system you can run as many models as you have physical and/or virtual processors.
Also see BOINC FAQ: Running more than one project or Work Unit simultaneously
[edit] How long does a model take to complete?
Don't expect your model to finish quickly - we're talking about weeks or months. But the results are trickled back perhaps as frequently as four times a day. The very comprehensive statistics are updated daily.
[edit] What about storage requirements?
Each running model will require around 750MB (2.7GB for Sulphur Cycle) of disk space. At the end of a run this will be compressed down to 330MB (1GB for Sulphur Cycle) , but only 7MB (20MB for Sulphur Cycle) of this will be returned to the project servers. The data for coupled models will be automatically deleted after a successful run, since it will have largely been uploaded (via the yearly 'trickles'). If a model crashes, it cannot clear the data from the disk and this needs to be done manually.
For Sulphur Cycle models, it is a good idea to hang on to the compressed Result Data Files - if your run is of particular interest it is possible that the project team may want to retrieve more of the results or download another program to perform further analysis. If you've deleted the Result Data Files the model will have to be run again, and even on the Cray supercomputers it was originally developed for that's going to take over a week!
Note: If you've deleted the result data files recently it's possible you can recover them with software. I'd recommend Recuva, which is free and very effective.
The more recent coupled model uploads the climate data every year in the trickles, and therefore there is no requirement to retain the data after the model has successfully completed.
If you are short of disk space you can always back up the Result Data Files using the procedure "Saving Or Deleting CPDN Result Data Files".
[edit] What is the support like?
Excellent.
- The website is very comprehensive, and well worth further investigation.
- The two forums (CPDN/Boinc forums and shared CPDN forums) are full of useful information, with plenty of experienced members willing to provide solutions to problems. The Project Team members are also active participants. The shared CPDN forums are outside the Boinc system and therefore require registration, but are shared between all the climate projects.
- The response to email support queries is usually very prompt, but do bear in mind how busy the Project Team members might be doing other things. Seeking support from other Participants via either of the forums is almost always a better option.
You may also want to look at the Climateprediction FAQ.
