Towards a statistical paradigm for climate change

It is argued that an acceptable 'paradigm' needs to be devised to clarlfy climate change as a statistical concept. A relatively simple statistical model for climate change is proposed as a start towards producing such a paradigm. Climate change is envisioned to involve changes in both the location and scale parameters of the probability distribution of a climate variable, whereas the shape of the distribution remains the same. Modifications of the model are treated to circumvent some apparent limitations of this approach, for instance, to allow for changes in the shape of the distribution of total precipitation. An indirect validation of the model is attempted through consideration of 2 analogues for global climate change: (1) a 'spatial analogue', involving patterns of variation in climate across space; and (2) the urban heat island, involving real temporal changes in local climate. These analogues indicate that any statistical model for climate change would need to be at least as complex as the one proposed here. Related issues include the need to preserve internal consistency when generating scenarios of climate change, as well as the need for information about how the variability of climate would change.