Start with a simple factor which you haven't mentioned involving the Earth's major greenhouse gas: water vapor.
Increased temperature obviously encourages evaporation of water. Will that water stay as a gas, or will it cause greater cloud formation? Will those clouds be flat or tall? Look at today's weather satellite pictures -- are clouds an unusual event? What will cloud changes do to climate?
Which is exactly why the first half dozen or so parameters we checked out were cloud-related. Most climate change modellers acknowledge that there are large uncertainties around cloud-climate feedbacks, and these form a central focus of the experimental design.
We're in the process of moving the software to BOINC, which will make us much more platform neutral. We reckon this ought to be done in a few weeks (it's been quite a big job). We'll be having a public beta test, so if you want to get involved (on your Mac or linux box) keep an eye on http://www.climateprediction.net
Exactly. Although in a chaotic system predictability due to initial conditions washes out over time (in the atmosphere initial condition predictability washes out over about 2 weeks) predictability due to changes in the boundary conditions of the system emerges over time.
Imagine a choppy and complicated lake which is fed by a river. The river's flow is getting bigger and bigger (perhaps due to some earthworks in the upstream catchment area). You take a snapshot of the lake, and use your model + initial conditions to predict the surface in thirty second's time. You do okay. But (say) you do a lousy job of predicting the surface in a day's time. BUT, you might do an okay job of predicting the *average level* of the surface in a month's time, not by knowing the initial conditions very well, but by knowing the rate of change of the river's flow. So though we can't predict the exact state (the weather) on longer timescales, we can (we hope, models and data permitting) do a reasonable job of predicting the average state (the climate) of the system on longer timescales.
Dave Frame
climateprediction.net coordinator
We're not running a parallelised model across lots of computers, we're farming out a run to each of several thousand machines. And the purpose of the experiment is precisely to look into the feedback processes that govern how climate changes. You say: "what's holding back the state of the art right now is the quality of the algorithms we're using" and this - on climate timescales - is what we're looking to explore.
Basically, the models that we have these days (IPCC TAR, for instance) lack any sort of quantitative measure of uncertainty. We're looking to find "error bars" for these sorts of predictions. See http://www.climateprediction.net/science/strategy_ adv.php for details of the experimental strategy. [We (& friends overseas) have submitted bids in recent EU Framework 6 and NSF rounds, to try to do something similar with very different models. This will help us conduct a convergence/verification process.]
We have recently submitted a "first results" paper and are awaiting the reviewers' comments. So far, things seem to be going pretty well (though we'd love some more participants!).
Cheers,
Dave Frame
climateprediction.net coordinator
Hi,
Thanks very much for the kind words, and I hope you really enjoy participating in climateprediction.net. It's got off to a pretty good start (the odd teething trouble excepted) and we're hoping it'll roll along nicely (we've got 27000+ participants already, but we want loads more...)
Dave
These are all fair questions.
We hope the experiment is worthwhile - it has been about 4 years since it started and in that time we have been at pains to ensure the basic integrity of the experiment. We've ported the model over to windows, checked that the model works reliably and gives sensible climates, and have set it up in an open way (all our funding has been through peer-review processes, so it's been through the same sort of hoops as any other bona fide science project). There are design papers and papers detailing different aspects of the experiment at http://www.climateprediction.net/science/publicati ons.php), so please feel free to browse these.
Cheers,
Dave Frame
climateprediction.net coordinator
Just to clear up any misunderstanding here - climateprediction.net is NOT a commercial project. We are funded by the UK Natural Environment Research Council (http://www.nerc.ac.uk) and the UK Department of Trade and Industry's eScience programme (http://www.escience-grid.org.uk/). The dataset generated by the project will be a public domain dataset, and will be released to the public following standard prcatice in our field (ie like a satellite project or like a climate model intercomparison project).
Dave Frame
climateprediction.net coordinator
Slashdot Mirror
Which is exactly why the first half dozen or so parameters we checked out were cloud-related. Most climate change modellers acknowledge that there are large uncertainties around cloud-climate feedbacks, and these form a central focus of the experimental design.
Dave Frame
climateprediction.net coordinator
University of Oxford
Cheers
Dave Frame
climateprediction.net coordinator
Exactly. Although in a chaotic system predictability due to initial conditions washes out over time (in the atmosphere initial condition predictability washes out over about 2 weeks) predictability due to changes in the boundary conditions of the system emerges over time. Imagine a choppy and complicated lake which is fed by a river. The river's flow is getting bigger and bigger (perhaps due to some earthworks in the upstream catchment area). You take a snapshot of the lake, and use your model + initial conditions to predict the surface in thirty second's time. You do okay. But (say) you do a lousy job of predicting the surface in a day's time. BUT, you might do an okay job of predicting the *average level* of the surface in a month's time, not by knowing the initial conditions very well, but by knowing the rate of change of the river's flow. So though we can't predict the exact state (the weather) on longer timescales, we can (we hope, models and data permitting) do a reasonable job of predicting the average state (the climate) of the system on longer timescales. Dave Frame climateprediction.net coordinator
We're not running a parallelised model across lots of computers, we're farming out a run to each of several thousand machines. And the purpose of the experiment is precisely to look into the feedback processes that govern how climate changes. You say: "what's holding back the state of the art right now is the quality of the algorithms we're using" and this - on climate timescales - is what we're looking to explore. Basically, the models that we have these days (IPCC TAR, for instance) lack any sort of quantitative measure of uncertainty. We're looking to find "error bars" for these sorts of predictions. See http://www.climateprediction.net/science/strategy_ adv.php for details of the experimental strategy. [We (& friends overseas) have submitted bids in recent EU Framework 6 and NSF rounds, to try to do something similar with very different models. This will help us conduct a convergence/verification process.]
We have recently submitted a "first results" paper and are awaiting the reviewers' comments. So far, things seem to be going pretty well (though we'd love some more participants!).
Cheers,
Dave Frame
climateprediction.net coordinator
Hi, Thanks very much for the kind words, and I hope you really enjoy participating in climateprediction.net. It's got off to a pretty good start (the odd teething trouble excepted) and we're hoping it'll roll along nicely (we've got 27000+ participants already, but we want loads more...) Dave
These are all fair questions. We hope the experiment is worthwhile - it has been about 4 years since it started and in that time we have been at pains to ensure the basic integrity of the experiment. We've ported the model over to windows, checked that the model works reliably and gives sensible climates, and have set it up in an open way (all our funding has been through peer-review processes, so it's been through the same sort of hoops as any other bona fide science project). There are design papers and papers detailing different aspects of the experiment at http://www.climateprediction.net/science/publicati ons.php), so please feel free to browse these.
Cheers,
Dave Frame
climateprediction.net coordinator
Just to clear up any misunderstanding here - climateprediction.net is NOT a commercial project. We are funded by the UK Natural Environment Research Council (http://www.nerc.ac.uk) and the UK Department of Trade and Industry's eScience programme (http://www.escience-grid.org.uk/). The dataset generated by the project will be a public domain dataset, and will be released to the public following standard prcatice in our field (ie like a satellite project or like a climate model intercomparison project). Dave Frame climateprediction.net coordinator