Although I often criticize numerical models, since they often don't manage to reproduce correctly the observations I am working on, I have to recognize the following statement by Mark Lynas:
"Nevertheless, many skeptics base their objections on the suspicion that models are somehow fiddled in advance to come up with the "right answers" by scientists eager for the next global warming grant - "you get out what you put in", as the old adage goes. But climate models do have an important grounding: they are based not on subjective judgments by their constructors but on the fundamental laws of physics. These observable physical laws, governing everything from convection within clouds to the reflectivity of sea ice, cannot be changed by anyone, whatever their politics. After all, models don't do anything magical. All they do is solve physical equations. All the processes of HadCM3, for instance, could theoretically be worked out by hand - except that it would then take centuries of human labor to complete one "model run". What computers do is speed up the process, just as pocket calculators speed up mathematics lessons in school.
No one, however, suggests that models are perfect. They all tend to come out with slightly different answers to the same question, a reflection of their varying design. The reason here is that some of the physical laws that underpin them are not known precisely. How clouds interact with the wider atmosphere is a big uncertainty, for example, so some cloud model parameters are best guesses. Nor is it known exactly how far sulphate "aerosols" - tiny particles of pollution blamed for "global dimming" - cool things down. But models are a useful tool and give a valuable insight into likely future conditions on this planet - something humanity has never had access to before. Unlike the oracles consulted by the ancients, models offer a way of divining the future based not on the miraculous visions of some unseen prophetess but on observable physical data."
Mark Lynas, Six degrees - our future on a hotter planet, chapter "three degrees".
The present task for climate scientists is to improve our understanding of the physical relationships between different processes and parts of the earth system that are still not known precisely, and not well represented in climate models, in order to reduce the uncertainties in the forecasts of our future climate, and check that some neglected processes may not hold big surprises! For example, the IPCC's 2007 report forecasts between 18 and 59 centimeters of sea level rise by 2100, but acknowledges that uncertainties about ice-sheet response time to global warming were not taken into account because the physical processes involved had not yet been studied enough to allow for a reliable assessment of their potential effects. Since the Greenland and West Antarctic ice sheets contain enough ice for a multi-meter sea level rise, understanding the physical processes controlling ice sheet stability is an urgent scientific question to address, with primordial societal relevance.