1.2 Methods of detection and attribution of observed changes

In the TAR (Mitchell et al., 2001), detection of climate change is the process of demonstrating that an observed change is significantly different (in a statistical sense) from what can be explained by natural variability. The detection of a change, however, does not necessarily imply that its causes are understood. Similarly, attribution of climate change to anthropogenic causes involves statistical analysis and the assessment of multiple lines of evidence to demonstrate, within a pre-specified margin of error, that the observed changes are (1) unlikely to be due entirely to natural internal climate variability; (2) consistent with estimated or modelled responses to the given combination of anthropogenic and natural forcing; and (3) not consistent with alternative, physically plausible explanations of recent climate change.

Extending detection and attribution analysis to observed changes in natural and managed systems is more complex. Detection and attribution of observed changes and responses in systems to anthropogenic forcing is usually a two-stage process (IPCC, 2003). First, the observed changes in a system must be demonstrated to be associated with an observed regional climate change within a specified degree of confidence. Second, a measurable portion of the observed regional climate change, or the associated observed change in the system, must be attributed to anthropogenic causes with a similar degree of confidence.

Joint attribution involves both attribution of observed changes to regional climate change and attribution of a measurable proportion of either regional climate change or the associated observed changes in the system to anthropogenic causes, beyond natural variability. This process involves statistically linking climate change simulations from climate models with the observed responses in the natural or managed system. Confidence in joint attribution statements must be lower than the confidence in either of the individual attribution steps alone, due to the combination of two separate statistical assessments.