91福利

The science of climate change aims to understand the processes which drive changes in climate and their consequences for many aspects of the earth system. The greatest attention is on the 20th and 21st centuries so the subject is not only of scientific interest but also provides an important input to national and international policy. Policy makers look to the science to provide predictions on which they can build both mitigation policy, which is intrinsically international in nature, and adaptation policy, which is intrinsically regional in nature.

Complicated three-dimensional global circulation models are a key element of climate change research and source of information to wider society. Yet our ability to explore uncertainty within these models, as well as their resolution and complexity, is constrained by computational limits. Furthermore, since they are being used to extrapolate to a new state of the system there is no possibility to verify either the models themselves or their post-processed results.

These aspects of the problem raise significant statistical, philosophical and practical challenges for the production of relevant information both within scientific research and for use in policymaking. Here I will explore and illustrate some of these problems and present some ways forward in the use of data to guide adaptation decision making. Barriers to the relevance of traditional statistical methods will be illustrated with a 40,000 member grand ensemble from the climateprediction.net project. Results from a smaller, initial condition, ensemble will be used to explore linearity within a climate model, and to illustrate the need for large ensembles of this type to understand behaviour within even a single model. Finally a dataset of observations from across Europe will be used to demonstrate a non-model based approach to guiding climate change adaptation.