Phrak said:
How good a science is global warming?
Yes, I agree that it isn't good at all. For one, climate modeling is extremely complex. It isn't just because the variables are unknown; actually doing the modeling involves a huge amount of computing power. Since it obviously isn't possible to include every single water and air molecule, scientists have to split the atmosphere and ocean into different "chunks" and assume that the chunk is a single, monolithic entity. The smaller the chunks, the more accurate the modeling, but the more intensive the required computing power. Since the parameters--like how much rain forms given a certain temperature, humidity, and aerosol concentration--are not well-known, different sets of parameters have to be tested. To see which set is more reasonable, simulations are run on past climate to see how well it predicts the current climate in a process called "hindcasting". It's a complicated process, and one that isn't too accurate.
I don't know if any of you have heard of Climateprediction.net, but if not, read this:
http://climateprediction.net/
It's a distributed computing project that involves predicting the climate of the 21st century. They've a lot of computing power at their disposal, and have produced some interesting results: http://climateprediction.net/content/scientific-papers
Basically, there's a huge variation in predictions of climate sensitivity, ranging from 2-11 degrees. Much of the higher estimates have been ruled out by hindcasting, but none of the models predict a sensitivity of less than 2 degrees, and certainly none of them say the climate will remain unchanged if we keep on burning fossil fuels. One thing to note is that the scientists running this experiment are well aware of the inherent uncertainties and dedicated a paper to calculating them. Climate modeling isn't ESP research; errors are estimated just as they are in every experiment, and you won't hear a scientist proclaiming global warming as real if the forecasted temperature rise isn't significantly larger than the error.