Whatever you're using as a screensaver right now is nowhere near as nifty as what you could be using- take part in the Climate Change Prediction Project, hosted by the
BBC, and you can lend your computer's processing power to running climate model experiments in its downtime. Essentially, this works just like a screensaver; as soon as you need your machine again, the model pauses until your computer goes unused for a few minutes again.
Climate models are sophisticated and very expensive to run because of the processing power required, so by dividing the energy amongst many lower-powered machines climate scientists are able to run more iterations of the model. The technique is known as distributed computing, and 10,000 people with ordinary machines can give the BBC's modelers more speed and power than the world's fastest supercomputer.
Why model at all? Climate models are run against predictions made from paleoecological data to see if the model can predict climate change over time that matches up with what we know happened from different environmental indicators in the past. Once the model has been tested against the past, we can start to use it to predict the future, like getting a handle on how rainfall patterns might change in the next twenty years with global warming, or how different biomes will respond to increased temperatures. This is practical knowledge that is essential to scientists making any kind of predictions that will be used to guide future policy decisions.
It also looks pretty spiffy, too (and you can select from a variet of displayed climate variables, like precipitation, temperature, and air pressure). The Geography department at UW Madison, where I am a grad student, just installed it on all the machines in the computer lab. These computers spend more time sitting in the lab than they do actually being used, and now that extra downtime is being put towards a good cause. All you need is an internet connection (though dial-up isn't recommended) and either a Widows XP or 2000 operating system or Linux (there is sadly no Mac version at the moment).