ClimatePrediction.Net gateway

Home

Taking part in CPDN

About

	Contact - main
	FAQ

Frequently asked questions about the experiment

Here are a few questions we have been asked by the people who wished to take part in the climateprediction.net experiment. We hope they would be also useful to you. However, if you want to ask something not covered below, feel free to post your question or suggestion via our Discussion Forums

We also have some very useful answers to technical questions provided by the participant UK_Nick. These were orginially posted on the forum, but have been grouped together and can be read here.

The participants' frequently asked questions are grouped into five categories:

Registering as a participant
Running climate model
Learning more about climate & interactive user area
Additional technical questions
User generated questions

«» Registering as a participant

What data do I give you to register?
Will you sell my email address or other data?
Climate dynamics is an evolving science; will you upgrade the models/clients during the experiment?
I've registered, but I couldn't get an experiment. What should I do?
I seem to be having problems with my firewall. (Or the 'Error Code 0 From Test Connection' message). What should I do?
Sounds great, what do I do now?
Your registration form doesn't include the XYZ platform!

What data do I give you to register?

In order to take part in the experiment, we will ask you to register at our climateprediction.net User Portal. Just supply your name, email address, username and password. Other information (such as your location, age, etc.) is optional. The username, email and password will identify you for the purposes of downloading your specific climate model (client) as well as for the purposes of interacting with the portal and other users.

Will you sell my email address or other data?

Never! We're interested in your email address only so we can contact you when the climateprediction.net client is ready for distribution and to keep you informed of the project's progress. We shall use your username only to customise the information on our website, User Portal and in various statistics.

Climate dynamics is an evolving science; will you upgrade the models/clients during the experiment?

Yes. The purpose of the climateprediction.net project is to assess the uncertainty in a variety of state-of-the-art climate models. In the first instance, we are releasing a series of Hadley Centre models. In the future we will be incorporating models from other climate modelling groups around the world.

I've registered, but I can't get an experiment. What should I do?

Note that it is likely that you have already created an account (username/email and password), so you will want to check the box that says "Are you already taking part...". Then you can enter your username or email address, and the password that was emailed to you (or the password you made at the user stats pages if you have already changed it.

Here are some more tips to help you over various registration problems we've heard about:

Received Email With Username and Password / Already Registered / Account Exists / Username or Email Exists
On a server connection failure (either due to our server having a problem, or your firewall settings), it's possible that you were still able to create an account for your username.

If you got a password email (or if you didn't, but the server says "name / email exists," it means you are in our system already. So when registering for a new experiment, you must now click on the "check-box" that will allow you to return as "Existing User" (the box at the top that says "Are you taking part...").

Then you will enter your already existing username or email address, and the password from the email you hopefully received. If you do not have your password, leave the field blank and the server will resend your password.

I seem to be having problems with my firewall. (Or the 'Error Code 0 From Test Connection' message). What should I do?

http://proxy-address:proxyport

The DNS of our registration server is climateapps.oucs.ox.ac.uk which resolves to either 163.1.13.14 or 163.1.13.15. On an upload you will need Port 80 access to a different IP address for a server that is allocated to you at that stage.

You possibly had "partly" registered before when the HTML post or our server failed, therefore you may already have an account with a username & email address. You can go to the user page http://www.climateprediction.net/users and verify this with your username/emailaddress & the password that would have been emailed to you (or leave password field blank to resend the email). You can change your email address for your registered client by the menu option on the client.

Note for WinNT/2000/XP Users: Some firewall software requires you to restart your "Internet Sharing Service" which you should find in "Control Panel / Administrative Tools / Services" then click "Stop" and then "Start."

Sounds great, what do I do now?

Thanks for deciding to take part in the experiment. Basically, the best way to start is to read through technical details to make sure that the model will be operational on your computer. The next few steps include downloading the model, installing it, registering your details, etc. In order to see the registration details, follow the How to join link. Once your model is running, it may occasionally ask you to connect to Internet in order to send some of the results back to us. Also, at the end, when all the modelling is finished the client will attempt to send some data to us through Internet. These data are the output of your climate model run, and will help us complete scientific aims of the climateprediction.net project.

Your registration form doesn't include the XYZ platform!

There are too many platforms out there to include each of them in our registration form. We have attempted to include the most prevalent platforms, but have no doubt missed many. We obviously won't be able to produce a client for all platforms, so we need input from you to decide which platforms to support. Feel free to contact us if you feel we have made a serious oversight and we will attempt to include your suggestions.

«» Running the climate model/client

I have an old 286 computer with 4 Mb of memory and Windows 3.1, will there be a client I can run?
How long will the modelling take?
Is your model running as a screensaver or a background client?
What should I do if I have a dual-processor machine?
Will running climate model interfere with other processes?
If there's a power failure or my Windows freezes up, how much data have I lost?
Are you going to port to the XYZ platform?
Will climateprediction.net be Open Source?
I can't bring up the client, but the taskbar icon shows up How can I get the client up?
I noticed the model displays "February 30th". What's going on?
How do I Install / Uninstall the CPDN Service (for Win NT4/2000/XP)?
Is there an easy way to stop / start the CPDN Service (Windows NT4/2000/XP)?
Can I schedule the client to run at certain times (for Win NT4/2000/XP)?
My model is uploading data before it has got to the end of the third phase Why?
I'm continuously getting a "model.exe" crash. Is there anything I can do?
How can I change my registered email address?
I'm having strange visualisation problems. What can I do?
Do you support Win 2003 Server?
I occasionally seem to lose timesteps. Why?
Can I keep running other distributed computing projects in parallel with climateprediction.net?
Does My Computer / Internet Connection Always Need to Be On?
I've tried launching the visualisation from the client menu, but it won't come up. Is there anything I can do?
How can I stop the model when running as a Service?
I want to move my run to a new computer. Can I do this?
My screensaver seems to conflict with climateprediction.net. What should I do?
What will the software do to my machine?

I have an old 286 computer with 4 Mb of memory and Windows 3.1, will there be a client I can run?

We're afraid not. The climateprediction.net client requires at least 128MB of memory, and 500MB of free disk space. In order for results to be produced in on a reasonable time-scale, one will need at least a 800MHz processor. This is a scientific research project based on a full-scale climate model, representing decades of research in super-computing centres. Although it would be a great idea to have a simpler model available for anyone to run on any old PC, and we don't like being CPU-elitist, our primary goal is the science, for which we need to use a state-of-the-art model.

How long will the modelling take?

Quite a while. We are anticipating at least four weeks of CPU time on fast machines, perhaps three or four months on slower machines. Serious climate science requires serious computer time. The flipside of this is that the fact we can conduct this sort of experiment at all is one of the minor miracles of the computer age: general circulation climate models are among the most sophisticated computer models in the world. They have always been run on supercomputers because of their complexity and the intensity of the number-crunching involved. It's yet another (if more were needed) testament to the phenomenal development of personal computing that you can run such a model on the machine you have in front of you.

Is your model running as a screensaver or a background client?

Screensavers only run when a computer has been idle for a period of time, are resource-hungry and place a limit on the platforms that can be supported. A background client will run whenever there is spare processing power, can be made more efficient than a screensaver and will support many more platforms. This is why we've designed the climateprediction.net client as a backgrounds process.

What should I do if I have a dual-processor machine?

At the moment we do not support dual processor machines. Parallelising the model across both processors would require a significant rewrite of much of the client code, while running two models simultaneously is not possible because of shared memory conflicts - both models would try to call the same memory handles and it would get very ugly. To avoid this we would need to rewrite quite a bit of code. It would be nice to do, but we have a few other priorities first.

Will running the climate model interfere with other processes?

Almost certainly not. In developing the package we have made every effort to fit in with other sorts of Windows-based software, such as Powerpoint, Word, Excel, and even games. Those of you who use other systems (such as reasonably tuned Linux-box) don't need to worry at all.

If there's a power failure or my Window freezes up, how much data have I lost?

We ensure the model state is saved to disk at least once per day (real time) so as little information as possible is lost in the event of an interruption -- if all goes to plan, you should be able to copy the re-start file over to a new PC if you decide to upgrade while running the experiment. More frequent dumps would simply slow the integration down and eventually wear out your hard disk.

Are you going to port to the XYZ platform?

Out industry partners Tessella Support Services plc have begun work on a Linux/UNIX port. See the news centre for details. We'll keep you posted on its status. We also have plans to include a Mac OSX version of the client, and we'll let you know how these develop.

Will climateprediction.net be Open Source?

The source code of any model used in climate research is accessible to the academic community -- we would never contemplate using a model for which the authors were not prepared to disclose the source. On the other hand, some of the most advance climate models are "dual use", meaning they are used for operational weather forecasting as well as for research. As a result, the source code has considerable commercial value, so anyone who wants access to it (to port it to a different platform, for example) would need to sign a specific agreement with the model's developers stating the purpose for which it was to be used and guaranteeing that it would not be passed it on to a third party.

We recognise this is not fully Open Source, but since anyone who wishes to verify that the model is doing what we claim it is doing could access the source in this way (free of charge, of course) it's close -- not all Sources are Open, but some are more Open than others...

I can't bring up the client, but the taskbar icon shows up How can I get the client up?

Some users are reporting that a reboot is required to show the client window on the desktop when you double-click on the system tray icon.

I noticed the model displays "February 30th". What's going on?

The model is based on a 360 day calendar (with 12 months, each of 30 days). This is basically for convenience and makes no difference to the model's performance.

How do I Install / Uninstall the CPDN Service (for Win NT4/2000/XP)?

"c:\program files\climate prediction\execs\cpdnservice.exe" /Service

Note the use of quotes around the program name, and check the path, I used the default install path above. If you already have Administrative Tools/Services up, you will need to do a "Refresh" to see your new "CPDNService."

Then you can control your start/stop of the model (outside of the Halt/Pause options on the running client) via the Administrative Tools/Services list (e.g. Start / Stop, set to Automatic service or Manual etc, defaults to Automatic).

Conversely, if you get tired of it running as a service and would rather go back to a "start via desktop icon" or Startup Folder, simply uninstall the service by "Start/Run":

"c:\program files\climate prediction\execs\cpdnservice.exe" /UnRegServer

Is there an easy way to stop / start the CPDN Service (Windows NT4/2000/XP)?

1) Open Notepad then copy & paste the following

net stop CPDNservice
echo off
cls
echo . CPDNservice paused.
echo .
echo . Click anywhere outside this box before using your keyboard.
echo .
echo . To restart the CPDNservice, click on the title bar then.....
echo .
pause
echo on
cls
net start CPDNservice

Save as "CPDNpause.bat" wherever you want - I used C:\Windows\Program Files\Climate Prediction

2) Navigate to the CPDNpause.bat batch file using Windows Explorer - Right click on CPDNpause.bat and 'Send to...' Desktop (as shortcut)

3) Grab the new icon on your desktop and dump it on your 'Quick Start' bar then trash the extra icon that will remain on your desktop.

Anytime you want the CPDN service halted whilst doing something especially I/O or CPU intensive on your computer then simply click the Quickstart icon titled "CPDNpause". This way you'll have a reminder that the CPDNservice is effectively paused from the Commandline box on your desktop or in your taskbar if you minimise it - only problem with that is CPDN won't resume if another user kills the Commandline box before it completes running CPDNpause.bat but that's not my endaba. If the main user clicks the "CPDNpause" taskbar icon again and follows the instructions onscreen then it will restart anyway.

(Ed. note: Thanks to CPDN user 'uk_nick')

Can I schedule the client to run at certain times (for Win NT4/2000/XP)?

If you want a rudimentary scheduler and are running NT4/2000/XP, you can setup an "at" job and issue a "net stop cpdnservice" and "net start cpdnservice" to cleanly stop & start the client.

The easiest way would be to make a small batch file that you can call from within the "at" job, or even use from a command-line or Start/Run prompt. Simply open up "notepad" or your favorite text editor and type:

net %1 cpdnservice

Save this file to "cpdn.bat" somewhere convenient; preferably in a directory that is in your PATH, such as "c:\winnt" or "c:\windows" (depending on your installation). Now to start the client you can just type (or call from the "at" command): "cpdn start" and of course "cpdn stop" will stop it.

To use this as a scheduler via the "at" command, you would want to setup the "CPDNService" if you haven't already (see previous section), and probably go to Administrator Tools/Services and set it to "Manual" instead of "Automatic" startup(unless you really do want it running upon reboot, whether you are there or not, or whether it's during your preferred start times or not).

So using this and the "at" command in NT/2000/XP you can setup a basic scheduler to stop/start the client. Say you want it to stop during your "working hours" of 9AM-5PM M-F, but it's OK on the weekends. You can simply issue (from an Administrator account):

at 9:00am /EVERY:M,T,W,Th,F c:\winnt\cpdn.bat stop
at 5:00pm /EVERY:M,T,W,Th,F c:\winnt\cpdn.bat start

This should keep it running non-stop over the weekend also, and shut down for you Monday morning when you're ready to go to work.

To delete a job, you can just say

at /delete

to get rid of them all, or to just get rid of one, put in the Job ID # of it, e.g.

at /delete 1

will get rid of job #1. If you just issue "at" by itself it lists the currently scheduled jobs and their #'s.

There are many other options for using the "at" command to setup the climateprediction.net client to run at certain times & dates, please refer to the Microsoft documentation appropriate to your version of Windows. Other scheduling package add-ons should work similarly, as long as you use the "net stop cpdnservice" and "net start cpdnservice" syntax.

My model is uploading data before it has got to the end of the third phase Why?

If your computer is uploading before the end of Phase 3, it could mean that your computer is incompatible with our software. It's possible that you just had a bad experiment downloaded (since it's hard to predict all the parameters that would put the model into an unstable state), and when it is done the early upload you will download a new model.

I'm continuously getting a "model.exe" crash. Is there anything I can do?

Unfortunately due to the heavy requirements imposed by the climate modelling software (>500,000 lines of code), a machine that is slower than a Pentium III/600MHz is not recommended.

How can I change my registered email address?

Simply double-click the CPDN icon in your system tray, go to the "Settings" menu and select "Change My Registered Email Address" to input your new email address..

I'm having strange visualisation problems. What can I do?

The 3-D visualisation uses the COIN 3-D graphics library which is based on OpenGL. You must have a graphics card that is fully OpenGL compliant via hardware or software drivers. Visiting your card manufacturer's website and downloading the latest drivers for your card and Windows version seems to help in most cases.

Do you support Win 2003 Server?

We hope to support this platform very soon.

I occasionally seem to lose timesteps. Why?

The client saves every 144 timesteps as a "restart checkpoint." So it is possible that you will "lose" timesteps when you restart. [144 timesteps was chosen as a good point to save because any faster would require too much disk I/O and detract from the performance of the calculations (144 timesteps = 20 minutes on a typical computer).] If your model becomes unstable for any reason, it might crash. if it does, it goes back and tries again (from the previous checkpoint). If it still crashes, it returns to the start of the month and tries again from there. If it still crashes after that, it goes back to the beginning of the year and tries one last time. If it still crashes after all that, it gives up, sends us what data it can, and picks another run to try.

Can I keep running other distributed computing projects in parallel with climateprediction.net?

You should be able to run multiple distributed computing projects; however it may not be a good idea. Different programs use different means of detecting idle CPU time & priorities, therefore the "threshhold" for "Folding Proteins" is different from "SETI" which is different from "climateprediction.net" -- and one DC app may "hog" more of the idle CPU cycles away from the others. You can try it out and see how it goes by monitoring the rate that each DC application is performing its tasks.

Does My Computer / Internet Connection Always Need to Be On?

The computer doesn't have to be left "on" (either physically powered on or connected to your ISP) continually for any particular time. If it notices that you are online, it sends a "ping" to our servers about once a day just to tell us you're "alive" and where you are at in the model, which you can track on your user pages.

I've tried launching the visualisation from the client menu, but it won't come up. Is there anything I can do?

It may be that the DLL didn't register properly, can you go to Start/Run and type (or cut-and-paste): regsvr32 "c:\program files\climate prediction\execs\cpdnshmem.dll" (replace the above path with the correct one to your CPDN install/execs directory, if different).

How can I stop the model when running as a Service?

Similarly, if you want to stop the model running as a service, change the Startup Type from "Automatic" to "Manual" or "Disabled." You can still run the model as a shortcut to "client interface.exe" (via your Startup folder automatically, or by clicking on the icon).

I want to move my run to a new computer. Can I do this?

If you are copying from a CD/DVD backup, make sure to take off the "read-only" attributes that may have copied over from the CD/DVD media.

Make sure that only one computer is ultimately running this archive, because only the first computer to upload will be allowed to finish the run!

My screensaver seems to conflict with climateprediction.net. What should I do?

If your default screensaver is "high-powered", e.g. CPU intensive or uses OpenGL (e.g. the "Windows Pipes" screensaver), it is probably "competing" with the CPDN model. The best options are a simple blank screen / monitor off kind of screensaver. These are also the best for saving power as well. If you really want a graphical screensaver, go for a non-OpenGL/3-D-style screensaver, such as a simple picture viewer, etc.

What will the software do to my machine?

As long as you leave the computer on (after having closed other applications for the night, for instance) the model will run. You don't even need to close the model down when you start new processes (such as Word or Powerpoint or Explorer) because the climateprediction.net software is set to be a background process - all other processes will automatically be given priority over the model, so it won't slow up your other work.

However, as stated above, if you are going to use your company's computer, please DO ask for permission to do so. We cannot accept any liability or responsibility for any breaches in your company's code of conduct and/or Internet usage!

«» Learning more about climate & our interactive user area

Won't all those computers left on for 24 hours a day have a detrimental impact on the climate system?
Okay, so I'm running the model. How can I find out what is going on in my model run?
My model turns into an ice-ball/mass of boiling oceans, yet you guys reckon this is a reasonable representation of 21st century climate. Should I be alarmed or is your model just rubbish?
How do you choose which parameters/model physics schemes to perturb?
Why do the dates go from 1810-1840, and then jump to 2050?
Isn't the sudden doubling of CO2 in the model extreme/arbitrary/unrealistic?
Do these models introduce stochastic variability or are they deterministic?
Is the probabilistic information found from all these simulations useful if at best you get a few that are right but for the wrong reasons?
What physical processes are the model?
How long will the project last?
How many model runs do you need?
I want to particpate in the climateprediction.net community. How do I go about it?
What peer to peer services do you provide to the user?

Won't all those computers left on for 24 hours a day have a detrimental impact on the climate system?: Assume a computer running 24hrs/day requires, on average, 50W of power. If 100,000 computers join the climateprediction.net project, the project will require 5,000kW of power. There are 24 hours in a day, so each day the project will consume 120,000kW-hrs, or 432,000,000kJ of electrical energy.

That's a big number, so let's try and put it in perspective by calculating how much energy is necessary to boil water for a cup of tea. Let's use a tiny bit of physics to do it. Assuming a specific heat of water of 4.19 kJ/(kg-K), 0.237kg/cup of water, a necessary temperature rise from 20 degrees Celsius to 100 degrees Celsius, and that only one cup of water is boiled for each cup of tea, then about 80kJ/cup of energy are necessary (assuming our kettle is 100% efficient). This means that running the climateprediction.net project for one day is equivalent to boiling water for 5,400,000 cups of tea!

Is five and a half million cups of tea a lot? According to the Tea Council, some 37 million people in the United Kingdom drink, on average, 3.4 cups of tea per day. That's nearly 126 million cups of tea per day in the UK alone!!!

Each day, about 23 times more energy will be spent boiling water for tea in the United Kingdom than would be used by the computers involved in the climateprediction.net project. More seriously, a rough calculation suggests that 100,000 computers running 24hrs/day for one year at a power consumption of 50W will contribute approximately 0.0001% of the total amount of CO₂ generated in one year. This is not an insignificant amount, but seems (to us) a worthwhile investment to better understand the climate system.

Assuming you are convinced this experiment needs to be done, there are basically two options: to buy a hangar-full of PCs and run it ourselves (not even an option right now, since the climate research community doesn't have the resources); or to recycle spare CPU out in the community, as we propose to do under the climateprediction.net experiment. Since the main environmental impact of a PC is in manufacture and disposal, not the power consumed in running it (never mind the air-conditioning costs and visual impact of that hangar on some innocent rural community), environmentalists will, we hope, approve of our strategy.

Nevertheless, this would be an interesting topic for further discussion, and we warmly invite you to go to the interactive parts of this portal and make your voice and/or opinion heard in the climateprediction.net community. See also the question on joining the user community further down.

Okay, so I'm running the model. How can I find out what is going on in my model run?: The client interface displays a variety of information about your model. It tells you how far through the experiment you are, including the timestep number and model date, and gives you information about what sort of processing is going on (eg., it warns you if it needs to do some file processing between different phases of the experiment). Menus on the client interface allow you to open your browser at the climateprediction.net help pages or user profile pages, as well as giving you the option of opening the visualisation package, through which you can see various fields (temperature, cloud/snow/ice fields, etc) which are read directly from the model as it runs.

My model turns into an ice-ball/mass of boiling oceans, yet you guys reckon this is a reasonable representation of 21st century climate. Should I be alarmed or is your model just rubbish?: No need to worry. We're making changes to the physics of the model, and we expect (and want!) to probe a range of climates beyond that which we currently observe. So we fully expect some strange climates (and a few very strange climates). These are actually pretty interesting because although they probably can't tell us much about the earth, they can tell us a lot about the model.

How do you choose which parameters/model physics schemes to perturb?: We consult the folks who built the model (the Met Office in this case, and various other modelling groups when we start to port their models over). We ask them which bits of the model are observationally poorly constrained and (in their view) contributing to uncertainty in climate forecasts. Cloud schemes are generally pretty ropey in most climate models, and we expect water vapour and clouds play a big role in climate (and many parts of cloud schemes are poorly constrained by observations) so that has been a sensible place for us to start in this experiment. We are starting with (and generally concentrating on) atmospheric perturbations, but in the future we will also be lookng to make some perturbations to the ocean, too. [And beyond that we want to run the entire experiment on completely different models.]

Why do the dates go from 1810-1840, and then jump to 2050?: The first two phases have pre-industrial levels of CO2, and the third phase has twice that. The first fifteen years has quite firm bounds on the sea surface temperature and salinity (saltiness). In the second phase we take some fields (diagnosed from the first phase) and use them to keep the model in balance. We take the firm bounds on sea surface salinity and temperature off, and let the model evolve on its own (this is the "control" phase). Then in the third phase we have double the previous level of CO2 and, like phase 2, we let the model evolve under its own steam. Details can be found on our experiment strategy page.
The experiment is testing climate sensitivity, or the effect of a doubling of CO2. We're interested (among other things) in the degree to which you can have a family of models that have very similar base climates (control phases) but very different responses to changes in CO2.

Isn't the sudden doubling of CO2 in the model extreme/arbitrary/unrealistic?: There's a quantity known as "climate sensitivity" that is defined as the equilibrium response of the climate to a doubling of CO2 (see the IPCC TAR). This is a useful quantity because it lets us know how much a given model responds to changes we're quite likely to see in the next century (hence the "sensitivity" bit). The idea of doubling CO2 and seeing what happens is very common in the literature. However, nobody sensible pretends it's the whole story. It's a very handy way for us to identify "interesting" models from our very large ensemble of slightly different models, but it's not sufficient to make a forecast from.
The next step is to take the interesting models we find (identified above) and then see how they do at actually trying to simulate the period 1950-2000 (using historical climate forcings). This is called a "hindcast" and is basically a quality control step that tells us which models are "good" (in that they seem to be reasonably realistic climate models). We may have some very interesting models that are completely hopless at simulating real climates (which would be, as you say, empirically useless). So we'll drop those, and just use the ones that are both interesting and good. From these we'll make a forecast (spanning an appropriate range of possible futures). See our strategy pages for more on the experimental strategy.

Do these models introduce stochastic variability or are they deterministic?: The models are deterministic in the sense that if you feed it exactly the same starting conditions it will do the same thing. If the initial conditions differ even slightly then chaotic processes will lead to changes fairly quickly (in a couple of weeks of model time). The parameterisations are deterministic in this model, but there is research looking in to the development of stochastic parameterisations for climate models. The advantage of stochastic parameterisations is that you can do a much better job of (a) sampling observed distributions of the things you want to parameterise and (b) reproducing observed natural variability in the system. [And of course as long as you keep track of the random seed you can still do bit reproducible experiments.]

Is the probabilistic information found from all these simulations useful if at best you get a few that are right but for the wrong reasons?: There are three distinct aims here, each approached with similar means.
1) Determining the sensitivity of the model to various changes in parameter (or parameterization).
2) Comparison of an ensemble of historical runs (say, 1950-2000), each with reasonable initial conditions and parameter values, to the single realization of the real world we have access to.
3) Interpreting an ensemble of simulations run into the future as a forecast, both for understanding the model, the real world and for policy guidance.

For (1) there isn't really all that much probabilistic info; the runs help us to define where the model is sensitive and, ideally, what the "reasonable" parameter values are.
For (2) we are again learning more about the model than the future. This phase will help us identify initial conditions and parameter values for this model better still; it will also give us a glimpse at what the model cannot do, as opposed to what it simply did not do. this is the most interesting for me, since we learn a bit about how to use this model and how to build a better one.
For (3) we really want probabilistic information. There are many open questions on how to turn simulations into pdfs, much depends on the results from (2). But we can deal with issues of deterministic chaos or stochastic dynamics. It is the simple fact that the model is wrong that muddies the water most. Are the limitations merely due to things like finite resolution? Can the model shadow reality? And if so which aspects?
i'd suggest that a major goal of the experiment is to learn if the model is wrong for the right reasons.

What physical processes are the model?: The model you are running at the moment (HadSM3) has an atmosphere, as well as a land surface that can get wet or dry out, and a "slab" ocean that has a limited ability to respond to changes in termperature (it can respond thermodynamically (ie heat up or cool down) but not dynamically (the thermohaline circulation can't shut down in a slab ocean). [In the coupled release next year we'll be replacing this simple slab with a "proper" fully dynamic ocean (and if you get one with the THC shutting down we'll be very interested!).] So the model you have at the moment is mostly an atmospheric model (and this is fair enough because the parameters we're perturbing at the moment are mainly cloud parameters...).
The atmosphere has two main bits: dynamics and radiation. There are five dynamical timesteps and then a radiation timestep (the radiation code is quite demanding - it's the reason the model flows quickly through the five dynamics timesteps and then chugs through the radiation one).
The following is a list of the physical processes represented in the atmospheric model:

Cloud Scheme
Radiation
Boundary Layer
Sulphur Cycle
Precipitation
Convection
Land Surface and Vegetation
Gravity Wave Drag

This is in addition to the dynamics. The dynamics involves utilising the equations of motion (3 equations for the full 3d wind field), the conservation of mass or continuity equation, an energy equation and an equation of state. It's a total of six equations: for the three wind components, temperature, pressure and density. The e quations of motion include Coriolis terms due to rotation of the Earth. (Each of these equations is evaluated at each point in the model [96 longitudes by 73 latitudes by 19 vertical levels] every half hour timestep.)

How long will the project last?: The climateprediction.net project will last several years. You can read about the experimental strategy here. Initially we will be using one family of models, the Hadley Centre's model suite. In the coming years we intend to port in new models and test them in the same sort of way. All told, this will take several years to complete.

How many model runs do you need?: Lots. In the first part of the experiment, we are determining the sensitivity of the model to various changes in parameters (or parameterizations). This helps us work out where the model is sensitive and what the "reasonable" parameter values are. There are hundreds of parameters we could look at in the model, and, when you couple them all together to explore all possible combinations, this leads to an enormous amount of space to cover. Naturally, some combinations of parameters will make very little difference, while others will have a more dramatic impact. [And some may even turn out to be improvements upon the standard model...] It makes sense to start with those regions of paramater space that we believe to comprise the most significant sources of uncertainty, so we're starting by looking at cloud schemes. Basically, the sheer number of parameters in the model means that there is the potential to run many thousands of millions of runs to check, if we were to do so one at a time. So we intend to scale the experiment to fit the number of participants: if we have a very large number, we can explore more of parameter space. If we have a smaller number, we can explore a smaller area. So we can alter the size of the experiment to fit user uptake. Naturally, we'd prefer a large number of participants. (So please join in!)

I want to particpate in the climateprediction.net community. How do I go about it?: Over the next few months we're going to develop our user community/interactive area. You'll be able to personalise your own user profile and see how your model fits into the experiment as a whole. You'll also be able to participate in various teams: friends and neighbours, countries and regions, and neighbours in parameter space.

What peer to peer services do you provide to the user?: Basically, the services provided by our User Portal fall into two categories. First, community-wide services such as newsletter, discussion forums and similarly, where you are going to interact with many participants in this experiment. Second category contains so-called peer to peer services, by which we mean those tools that enable a contact between limited number of individual participants - usually friends, colleagues, fellow students, etc.

We start with a provision of a specially enhanced instant messaging (IM) tool called BuddySpace that puts you in touch with your mates using the leading IM networks (e.g. MSN, ICQ, Jabber, Yahoo!), and in addition provides map-based interface that will later be embedded directly into our climate model, so you can have all the benefits at your fingertips. Later, we intend to extend peer to peer services to include collaborative visualisation of the results, result comparison between the peers, and similarly.

«» Additional technical questions

(03 Feb 2004) Problems With New Microsoft Internet Explorer Security Patch

Please make sure you the latest upgrade/install package for CPDN (currently version 2.2.28). The latest package is available here. Versions prior to 2.2.28 do not work with Windows 2003 Server and the MS IE Patch, however 2.2.28 works.

Errors Uploading / Unfinished Upload

First, please make sure you the latest upgrade/install package for CPDN (currently version 2.2.28). The latest package is at available here

If that does not fix the problem, we now have a manual upload page if your run did not finish an upload. It should be easy to use via the instructions on the page, and the page is linked off of your user stats page (i.e. where you login to see stats etc). You can get to the page directly by clicking on this link:

http://cpdn.comlab.ox.ac.uk/user/preupl.php

"Checksum Verification Failed" Error Message

There is an upgrade patch (v 2.2.28) to fix the "Checksum Verification Failed" error which is available here.

Received Email With Username and Password / Already Registered / Account Exists / Username or Email Exists

On a server connection failure (either due to our server having a problem, or your firewall settings), it's possible that you were still able to create an account for your username.

If you got a password email (or if you didn't, but the server says "name / email exists," it means you are in our system already. So when registering for a new experiment, you must now click on the "check-box" that will allow you to return as "Existing User" (the box at the top that says "Are you taking part...").

Then you will enter your already existing username or email address, and the password from the email you hopefully received. If you do not have your password, leave the field blank and the server will resend your password.

An image follows of this screen and your selection if you already have an email address/username and password registered:

Firewall Issues / Error Code 0 From 'Test Connection'

Setting up your firewall to allow climateprediction.net If you are running firewall software, please make sure that the program "client interface.exe" can send on ports 80 & 443 (the basic HTTP/S ports your web browser would use). If you do not use a proxy server you can leave that setting blank, if you do use a proxy server and it doesn't work with your default IE settings, please enter in the client as follows:

http://proxy-address:proxyport

The DNS of our registration server is climateapps.oucs.ox.ac.uk which resolves to either 163.1.13.14 or 163.1.13.15. On an upload you will need Port 80 access to a different IP address for a server that is allocated to you at that stage.

You possibly had "partly" registered before when the HTML post or our server failed, therefore you may already have an account with a username & email address. You can go to the user page http://www.climateprediction.net/users and verify this with your username/emailaddress & the password that would have been emailed to you (or leave password field blank to resend the email). You can change your email address for your registered client by the menu option on the client.

Note for WinNT/2000/XP Users:

Some firewall software requires you to restart your "Internet Sharing Service" which you should find in "Control Panel / Administrative Tools / Services" then click "Stop" and then "Start."

Missing Client But Taskbar Icon Shows up

Some users are reporting that a reboot is required to show the client window on the desktop when you double-click on the system tray icon.

Leap Year / 30th February etc

The model is based on a 30-day month (360 day year), so that is why you see peculiarities such as 30-Feb-1811 or even 29-Feb-1811.

Installing and Uninstalling the CPDN Service (for Win NT4/2000/XP)

Simply run the following (in "Start/Run") and it should install it in your Administrator Tools / Services list as an "Automatic" service which you can manually click "Start" (or reboot and it will come up automatically and be running by the time you login):

"c:\program files\climate prediction\execs\cpdnservice.exe" /Service

Note the use of quotes around the program name, and check the path, I used the default install path above. If you already have Administrative Tools/Services up, you will need to do a "Refresh" to see your new "CPDNService."

Then you can control your start/stop of the model (outside of the Halt/Pause options on the running client) via the Administrative Tools/Services list (e.g. Start / Stop, set to Automatic service or Manual etc, defaults to Automatic).

Conversely, if you get tired of it running as a service and would rather go back to a "start via desktop icon" or Startup Folder, simply uninstall the service by "Start/Run":

"c:\program files\climate prediction\execs\cpdnservice.exe" /UnRegServer

Easy Way To Stop / Start the CPDN Service (Windows NT4/2000/XP)

Open Notepad then copy & paste the following
net stop CPDNservice
echo off
cls
echo . CPDNservice paused.
echo .
echo . Click anywhere outside this box before using your keyboard.
echo .
echo . To restart the CPDNservice, click on the title bar then.....
echo .
pause
echo on
cls
net start CPDNservice

Save as "CPDNpause.bat" wherever you want - I used C:\Windows\Program Files\Climate Prediction
Navigate to the CPDNpause.bat batch file using Windows Explorer - Right click on CPDNpause.bat and 'Send to...' Desktop (as shortcut)
Grab the new icon on your desktop and dump it on your 'Quick Start' bar then trash the extra icon that will remain on your desktop.

Scheduling the Client To Run At Certain Times (for Win NT4/2000/XP)

If you want a rudimentary scheduler and are running NT4/2000/XP, you can setup an "at" job and issue a "net stop cpdnservice" and "net start cpdnservice" to cleanly stop & start the client.

The easiest way would be to make a small batch file that you can call from within the "at" job, or even use from a command-line or Start/Run prompt. Simply open up "notepad" or your favorite text editor and type:

net %1 cpdnservice

Save this file to "cpdn.bat" somewhere convenient; preferably in a directory that is in your PATH, such as "c:\winnt" or "c:\windows" (depending on your installation). Now to start the client you can just type (or call from the "at" command): "cpdn start" and of course "cpdn stop" will stop it.

To use this as a scheduler via the "at" command, you would want to setup the "CPDNService" if you haven't already (see previous section), and probably go to Administrator Tools/Services and set it to "Manual" instead of "Automatic" startup(unless you really do want it running upon reboot, whether you are there or not, or whether it's during your preferred start times or not).

So using this and the "at" command in NT/2000/XP you can setup a basic scheduler to stop/start the client. Say you want it to stop during your "working hours" of 9AM-5PM M-F, but it's OK on the weekends. You can simply issue (from an Administrator account):

at 9:00am /EVERY:M,T,W,Th,F c:\winnt\cpdn.bat stop
at 5:00pm /EVERY:M,T,W,Th,F c:\winnt\cpdn.bat start

This should keep it running non-stop over the weekend also, and shut down for you Monday morning when you're ready to go to work.

To delete a job, you can just say

at /delete

to get rid of them all, or to just get rid of one, put in the Job ID # of it, e.g.

at /delete 1

will get rid of job #1. If you just issue "at" by itself it lists the currently scheduled jobs and their #'s.

There are many other options for using the "at" command to setup the climateprediction.net client to run at certain times & dates, please refer to the Microsoft documentation appropriate to your version of Windows. Other scheduling package add-ons should work similarly, as long as you use the "net stop cpdnservice" and "net start cpdnservice" syntax.

Early Uploading (before end of Phase 3)

Standby/Hibernation

Question : Can I run the application software in Standby or Hibernation mode?
Answer : NO
They're basically means of conserving energy and temporarily storing the last state of your machine. Standby is ideal when you PC is idle for short periods while Hibernation is better suited for longer periods if idle time. All programs are suspended in either standby or hibernation mode
so its impossible to expect the ( or any ) software to run during this period.
Standby or Sleep mode effectively turns off the visual display, any internal devices and stops the hard-disk drive, so that the power consumption of the PC is greatly reduced. When normal operation resumes , the desktop is restored exactly as it was prior to entering Standby mode. Hibernation mode conserves power by copying system data to a reserved area on the hard drive and then completely turning off the computer. So this operation takes longer. When you exit hibernate mode, the programs and files that were open before you activated hibernate mode are still open. If the computer loses power while in Standby mode, all data that was being worked on is lost while data is saved and recoverable in hibernation state.

Slow Computer / Model.Exe Crash

Unfortunately due to the heavy requirements imposed by the climate modelling software (>500,000 lines of code), a machine that is slower than a Pentium III/600MHz is not recommended, and can lead to crashes & instabilities.

Changing Your Email Address

It is easy to change your email address -- simply double-click the CPDN icon in your system tray, go to the "Settings" menu and select "Change My Registered Email Address" to input your new email address.

Visualisation Problems (Strange Graphics/Artifacts Appear)

The 3-D visualisation uses the COIN 3-D graphics library which is based on OpenGL. You must have a graphics card that is fully OpenGL compliant via hardware or software drivers. You may want to go to your card manufacturer's website to download the latest drivers for your card and Windows version.

Windows 2003 Server

The latest package (version 2.2.28) supports Windows 2003 Server and you can download it here.

'Lost' Timesteps

The client saves every 144 timesteps as a "restart checkpoint." So it is possible that you will "lose" timesteps when you restart. 144 timesteps was chosen as a good point to save because any faster would require too much disk I/O and detract from the performance of the calculations (144 timesteps = 20 minutes on a typical computer).

Running Other DC Projects

Does My Computer / Internet Connection Always Need to Be On?

Unstable Model / Planet Turned Into An Ice-Ball etc

It sounds like your client has downloaded an experiment with "interesting" parameters that has veered off into an unstable state. The client attempts to "rewind" to recalculate timesteps just to be safe, but if it fails after three times it will contact our upload server, send what it can, and automatically download a new (and hopefully more stable) experiment for you to run.

Win32 Errors / Bad Downloaded CPDN Package

Please verify the download file size/checksum, and perhaps try a new download from one of our mirror sites at the bottom of the download page

Uninstall

You can uninstall by simply going to the "Climate Prediction" folder and clicking on the "Uninstall" option. Or by going to "Control Panel : Add/Remove Programs" and then selecting "climateprediction.net." If you wish to notify us that you have uninstalled, please "check off" the "active" checkbox on your Edit Your Profile page.

Visualisation Not Starting

It sounds like the DLL didn't register properly, can you go to Start/Run and type (or cut-and-paste):
regsvr32 "c:\program files\climate prediction\execs\cpdnshmem.dll"
(replace the above path with the correct one to your CPDN install/execs directory, if different).

Stopping the Model When Running As A Service

If you have installed the model as a service, you will probably want to stop it for such tasks as virus scanning, defragmenting your hard drive, scandisk etc. Simply go to Control Panel / Administrative Tools / Service and select the "CPDNService," then hit the "Stop" button (or double-click to get the "Stop" Button). The model will restart on a reboot or you can manually "Start."

Similarly, if you want to stop the model running as a service, change the Startup Type from "Automatic" to "Manual" or "Disabled." You can still run the model as a shortcut to "client interface.exe" (via your Startup folder automatically, or by clicking on the icon).

Screensavers

Please note that some screensavers take up CPU time so that it may be best to use a "low-powered" screensaver (blank screen, monitor power off etc) so as to leave the most idle CPU time for your climate model.

Moving Your Run To Another/New Computer

If you need to move your run to another/new computer, simply copy the "Climate Prediction" folder (either via a network or CD/DVD-R/RW). Then you will want to run the "Repair" option from the original installation package to register these files.

If you are copying from a CD/DVD backup, make sure to take off the "read-only" attributes that may have copied over from the CD/DVD media.

Make sure that only one computer is ultimately running this archive, because only the first computer to upload will be allowed to finish the run!

Conflicts With Screensaver / Screensaver Slows Down CPDN

Why Isn't CPDN Open Source / Running on a Mac / Linux /etc

First, let us say that we are an eager part of the distributed computing community and we have nothing but respect for all of the projects out there in the world. We are especially grateful to David Anderson and our friends at SETI who have been invaluable in helping us get off the ground and are kind enough to host our download package as a mirror site.

We are definitely not trying to compete and turn this into a "horse race", and hope that everyone runs the distributed computing application that they are most interested in and can get most involved in; because that is what will give you the incentive to stick with it.

We get a lot of emails from people who rightly would love to run the climate model on a Mac or Linux, or want us to make it "open source" so that they and others can help the port etc. We realize that right now we have unfortunately left a lot of users behind, but would like to explain our side of it.

The programming and resources involved in running a full-scale climate model is tremendous, as well as the peer-review required over years of academics using something like the UK Met Office model. The UK Met Office model is not "open source" -- it is an extremely large, complicated system (something like 500,000 lines of Fortran; two miles of continuous paper if printed out and laid end-to-end). The UK Met Office has been superb in allowing us to bring it over to run on a Windows platform and distribute to the world. They are certainly not getting any money out of it.

So it is not simply a matter of (as many emails suggest) "just send me the C program and I'll cross-compile it" etc.

There are compiler and machine architecture issues involved that have to be applied all throughout this enormous climate model, rigorous (and long) tests to be done to make sure the same input parameters have reproducible results across the different platforms, etc.

So we appreciate the concern and interest from the DC community, but it is unfortunately not feasible to make this "open source" nor to port around to various operating systems at the current time, although we are working on the latter!

user page login problems

Symptoms: Unable to log in to the user pages
Cause: Changed default Internet settings.
Remedy :
IE version 6.0.28 or above: Navigate to the "tools" item on the menu bar and select "Internet options". Click on the "privacy" tab then select the "Advanced" settings. enable "override automatic cookie handling" check box and finally enable "always allow session cookies". Click OK and make sure you "Apply" the setting as well so it actually takes effect. default IE setting permit session cookies unless otherwise configured.

Mozilla: Navigate to the "Edit" item on the menu bar and select "Preferences". Expand the Privacy & Security trees and then highlight "Cookies". check "Enable cookies for the originating web site only". Click OK and that's it (see screen grab).

Opera: Click on "File" , "Preferences" then highlight "Privacy". Ensure Enable cookies is turned on and add "cpdn.comlab.ox.ac.uk" to the server category.

Firewall Configuration

With Norton Internet Security 2003:

Double click on the " Norton Internet Security 2003" program icon usually on the bottom right hand corner or you screen.
This should bring up the console shown below
Click "Personal Firewall" then click "configure"
Hit the program "control tab"
At the bottom select "Add"
From the next window, navigate to the installation directory and select "Client Interface.exe". Usually in the default installation folder
C:\Program Files\Climate Prediction\execs\Client Interface.exe
In the Next dialog, i.e "What do you want the program to do?" Select "Permit". Click ok and you can choose General category in the next window.