This post on RealClimate provided an enlightening insight into computer modelling.

I've not much to add really, except to say that I think shows clearly the difficulties and dangers of modelling something as complex as climate (or, indeed carbonates) in a computer. It's not as easy as you might at first think...