'The forecaster is like an entrepreneur,' says Roman Frydman. 'He uses quantitative methods, but he also studies history, and relies on intuition and judgment. He is not a scientist.'
The quote refers specifically to economics forecasters, but it applies just as well to any forecaster of complex systems, such as those who forecast climate changes.
Predictions are not science. No way, no how. And unfortunately, bad predictions (which are more common than good ones) come with little cost. As a result, they're dreadfully oversupplied. This doesn't mean that predictions should be ignored, but nor should they be elevated to anything more than they are: guesses--and often very crude ones at that.
On a semi-related note, one way to increase the accuracy of a forecast is through the use of prediction markets. More on that later, perhaps.