The obvious starting point in this chapter is to consider what we mean by predictability. A helpful approach is to say that it is the extent to which it is possible to predict the climate with a theoretically complete knowledge of the physical laws governing it. Using this definition of predictability, it follows from the discussion of computer models of the climate in Chapter 10 that predicting how the climate may change over the next century or so faces four basic challenges. The first is model error and how much we can improve the mathematical representation of the workings of the climate. The second is the closely related question of reducing the initial-condition error to ensure that the model starts with as accurate a representation of the state of the climate system as possible. The third is how do we measure the results of our studies. Finally, there is the issue of the potential for generalising the concepts and results of the model in framing policy.

The final concept of the generalisation of the results of modelling studies leads us into the economic and social consequences of any future climate change. Here the prediction is even more difficult. Because economic models lack the well-defined physical laws that underpin climate models the uncertainties are much greater. In particular, these uncertainties hinge on public behaviour in the light of any emerging evidence of climate change.