Home, Archive, Stuff, Random thoughts, London, My Rigs, Pictures, Dreams, Links, About, Contact, Search

spikegifted - climateprediction


Planet Earth is a very small space. When we look at our home compare with the vastness of space, there's no other way of describing it. So far, Planet Earth is the only place in the Solar System that supports complicated intelligent life forms like us. Yet, due to our 'progress' in industrialization, mechanization of farming, continual exploitation of the planet's natural resources, we may be inadvertently getting ourselves into deeper and deeper trouble. Scientists have long been warning us of the dangers and implications of global warming. However, despite much good will in the international community to cut down green house gases, we are still exploiting this home of ours, with the two infamous crimes being deforestation and irresponsible CO2 emission.

Before we dive deep into the world of climateprediction.net, let me borrow its introduction: "The climateprediction.net experiment has been developed to allow a state-of-the-art climate prediction model to be run on home/ school/ work computers. By getting data from thousands of climate models, we will generate the world's largest climate prediction experiment."

What is climateprediction.net?

The background to climateprediction.net is all about simulations. Not just one, but hundreds and thousands. The scientist behind climateprediction.net build a model (or even dozens of models) which simulates the Earth's climate using a range of initial conditions and a wide array of boundary conditions. Of course such models are full of assumptions and approximations - it is simply impossible to accurately simulate every parameter that can influence a planet's long-term climate even if we know all the initial and boundary conditions. The idea is that based on a range of educated guesses, certain parameters in the models are fixed and others are changed for each simulation run. Over a large sample, a statistical model can be built up using the simulation results that will have some resemblance to the real world. All of that is based on history - given what we know have happened in the past, the models apply slight changes to some or all of the parameters and let the models do their work.

This is a little like what economic statisticians do - collect past data, define the present and simulate the future. If you know what are the assumptions you’re making, you come up with a ‘control’ and then vary those assumptions one by one or in combination. With sufficient number of runs, you'd build up a statistical picture of the economic system (or the world) that you try to predict. Then you either take the results from the simulations or re-run the models with more changes to the parameters so see if the changes you’ve made make any difference and if they do to what extent. That’s what science is - trial and error. Unfortunately, planetary climate, much like economic systems, is a collection of very complex systems which interact in unpredictable ways. On the other hand, it is far less dangerous to run models to figure out what may or may not happen than to allow a crazy experiment ‘go live’... In economics, such experiment was allow to ‘go live’! That was called communism. Did you not observe the consequences?

What does climateprediction.net hope to achieve?

Due to the large array of conditions and parameters that can and will fact the models, there is a great deal of uncertainty regarding the interpretation of the results and often the results themselves. The problem with modeling is that in the event that we take the wrong set of results or taken the wrong interpretation of the results, we'd ended up with wasting huge amount of resources trying to do something that is inappropriate for the situation. For example, if the models resulted in an overestimation of the speed and the scale of climate change, we may ended up spending large amount of money to correct a problem that may not affect us at all or in the worst case scenario creating a bigger problem by over-correct a small one. On the other hand, the wrong set of results may lead us to not devoting sufficient resources to correct the problem and allow the problem to get out of hand.

The objectives of climateprediction.net is to:
- improve the models themselves and improve the range of parameters, the variables within the parameters, the range of data and the scope of the data; and
- massive runs and re-runs of the models with a range of parameters and variables and home in on those that achieve within a pre-defined confidence level.

In achieving these objectives, climateprediction.net is hoping to help "improve methods to quantify uncertainties of climate projections and scenarios, including long-term ensemble simulations using complex models", as outlined in the Intergovernmental Panel on Climate Change in 2001.

How to take part in climateprediction.net?

In a fashion similar to other distributive computing projects, anyone can participate in climateprediction.net by downloading the client. You'd need to have your internet connection active to download the client the models. Each model takes around 4 to 6 weeks to complete as each of them models 45 years of climate. The client also has some rather neat graphic display tools which demonstrate, on screen, the results of your current model.

Official climateprediction.net website
spike's BOINC climateprediction.net official statistics
spike's BOINC climateprediction.net profile
2CPU.com BOINC climateprediction.net Team official statistics
2CPU.com BOINC climateprediction.net Team Statistics from SETI Synergy
2CPU.com BOINC climateprediction.net Team Statistics from BOINC.dk
spike's BOINC climateprediction.net Personal Statistics from SETI Synergy
spike's BOINC climateprediction.net Personal Statistics from BOINC.dk
UK BOINC climateprediction.net User Statistics from SETI Synergy