IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group I: The Physical Science Basis

9.1.2 What are Climate Change Detection and Attribution?

The concepts of climate change ‘detection’ and ‘attribution’ used in this chapter remain as they were defined in the TAR (IPCC, 2001; Mitchell et al., 2001). ‘Detection’ is the process of demonstrating that climate has changed in some defined statistical sense, without providing a reason for that change (see Glossary). In this chapter, the methods used to identify change in observations are based on the expected responses to external forcing (Section 9.1.1), either from physical understanding or as simulated by climate models. An identified change is ‘detected’ in observations if its likelihood of occurrence by chance due to internal variability alone is determined to be small. A failure to detect a particular response might occur for a number of reasons, including the possibility that the response is weak relative to internal variability, or that the metric used to measure change is insensitive to the expected change. For example, the annual global mean precipitation may not be a sensitive indicator of the influence of increasing greenhouse concentrations given the expectation that greenhouse forcing would result in moistening at some latitudes that is partially offset by drying elsewhere (Chapter 10; see also Section Furthermore, because detection studies are statistical in nature, there is always some small possibility of spurious detection. The risk of such a possibility is reduced when corroborating lines of evidence provide a physically consistent view of the likely cause for the detected changes and render them less consistent with internal variability (see, for example, Section 9.7).

Many studies use climate models to predict the expected responses to external forcing, and these predictions are usually represented as patterns of variation in space, time or both (see Chapter 8 for model evaluation). Such patterns, or ‘fingerprints’, are usually derived from changes simulated by a climate model in response to forcing. Physical understanding can also be used to develop conceptual models of the anticipated pattern of response to external forcing and the consistency between responses in different variables and different parts of the climate system. For example, precipitation and temperature are ordinarily inversely correlated in some regions, with increases in temperature corresponding to drying conditions. Thus, a warming trend in such a region that is not associated with rainfall change may indicate an external influence on the climate of that region (Nicholls et al., 2005; Section Purely diagnostic approaches can also be used. For example, Schneider and Held (2001) use a technique that discriminates between slow changes in climate and shorter time-scale variability to identify in observations a pattern of surface temperature change that is consistent with the expected pattern of change from anthropogenic forcing.

The spatial and temporal scales used to analyse climate change are carefully chosen so as to focus on the spatio-temporal scale of the response, filter out as much internal variability as possible (often by using a metric that reduces the influence of internal variability, see Appendix 9.A) and enable the separation of the responses to different forcings. For example, it is expected that greenhouse gas forcing would cause a large-scale pattern of warming that evolves slowly over time, and thus analysts often smooth data to remove small-scale variations. Similarly, when fingerprints from Atmosphere-Ocean General Circulation Models (AOGCMs) are used, averaging over an ensemble of coupled model simulations helps separate the model’s response to forcing from its simulated internal variability.

Detection does not imply attribution of the detected change to the assumed cause. ‘Attribution’ of causes of climate change is the process of establishing the most likely causes for the detected change with some defined level of confidence (see Glossary). As noted in the SAR (IPCC, 1996) and the TAR (IPCC, 2001), unequivocal attribution would require controlled experimentation with the climate system. Since that is not possible, in practice attribution of anthropogenic climate change is understood to mean demonstration that a detected change is ‘consistent with the estimated responses to the given combination of anthropogenic and natural forcing’ and ‘not consistent with alternative, physically plausible explanations of recent climate change that exclude important elements of the given combination of forcings’ (IPCC, 2001).

The consistency between an observed change and the estimated response to a hypothesised forcing is often determined by estimating the amplitude of the hypothesised pattern of change from observations and then assessing whether this estimate is statistically consistent with the expected amplitude of the pattern. Attribution studies additionally assess whether the response to a key forcing, such as greenhouse gas increases, is distinguishable from that due to other forcings (Appendix 9.A). These questions are typically investigated using a multiple regression of observations onto several fingerprints representing climate responses to different forcings that, ideally, are clearly distinct from each other (i.e., as distinct spatial patterns or distinct evolutions over time; see Section 9.2.2). If the response to this key forcing can be distinguished, and if even rescaled combinations of the responses to other forcings do not sufficiently explain the observed climate change, then the evidence for a causal connection is substantially increased. For example, the attribution of recent warming to greenhouse gas forcing becomes more reliable if the influences of other external forcings, for example solar forcing, are explicitly accounted for in the analysis. This is an area of research with considerable challenges because different forcing factors may lead to similar large-scale spatial patterns of response (Section 9.2.2). Note that another key element in attribution studies is the consideration of the physical consistency of multiple lines of evidence.

Both detection and attribution require knowledge of the internal climate variability on the time scales considered, usually decades or longer. The residual variability that remains in instrumental observations after the estimated effects of external forcing have been removed is sometimes used to estimate internal variability. However, these estimates are uncertain because the instrumental record is too short to give a well-constrained estimate of internal variability, and because of uncertainties in the forcings and the estimated responses. Thus, internal climate variability is usually estimated from long control simulations from coupled climate models. Subsequently, an assessment is usually made of the consistency between the residual variability referred to above and the model-based estimates of internal variability; analyses that yield implausibly large residuals are not considered credible (for example, this might happen if an important forcing is missing, or if the internal variability from the model is too small). Confidence is further increased by systematic intercomparison of the ability of models to simulate the various modes of observed variability (Chapter 8), by comparisons between variability in observations and climate model data (Section 9.4) and by comparisons between proxy reconstructions and climate simulations of the last millennium (Chapter 6 and Section 9.3).

Studies where the estimated pattern amplitude is substantially different from that simulated by models can still provide some understanding of climate change but need to be treated with caution (examples are given in Section 9.5). If this occurs for variables where confidence in the climate models is limited, such a result may simply reflect weaknesses in models. On the other hand, if this occurs for variables where confidence in the models is higher, it may raise questions about the forcings, such as whether all important forcings have been included or whether they have the correct amplitude, or questions about uncertainty in the observations.

Model and forcing uncertainties are important considerations in attribution research. Ideally, the assessment of model uncertainty should include uncertainties in model parameters (e.g., as explored by multi-model ensembles), and in the representation of physical processes in models (structural uncertainty). Such a complete assessment is not yet available, although model intercomparison studies (Chapter 8) improve the understanding of these uncertainties. The effects of forcing uncertainties, which can be considerable for some forcing agents such as solar and aerosol forcing (Section 9.2), also remain difficult to evaluate despite advances in research. Detection and attribution results based on several models or several forcing histories do provide information on the effects of model and forcing uncertainty. Such studies suggest that while model uncertainty is important, key results, such as attribution of a human influence on temperature change during the latter half of the 20th century, are robust.

Detection of anthropogenic influence is not yet possible for all climate variables for a variety of reasons. Some variables respond less strongly to external forcing, or are less reliably modelled or observed. In these cases, research that describes observed changes and offers physical explanations, for example, by demonstrating links to sea surface temperature changes, contributes substantially to the understanding of climate change and is therefore discussed in this chapter.

The approaches used in detection and attribution research described above cannot fully account for all uncertainties, and thus ultimately expert judgement is required to give a calibrated assessment of whether a specific cause is responsible for a given climate change. The assessment approach used in this chapter is to consider results from multiple studies using a variety of observational data sets, models, forcings and analysis techniques. The assessment based on these results typically takes into account the number of studies, the extent to which there is consensus among studies on the significance of detection results, the extent to which there is consensus on the consistency between the observed change and the change expected from forcing, the degree of consistency with other types of evidence, the extent to which known uncertainties are accounted for in and between studies, and whether there might be other physically plausible explanations for the given climate change. Having determined a particular likelihood assessment, this was then further downweighted to take into account any remaining uncertainties, such as, for example, structural uncertainties or a limited exploration of possible forcing histories of uncertain forcings. The overall assessment also considers whether several independent lines of evidence strengthen a result.

While the approach used in most detection studies assessed in this chapter is to determine whether observations exhibit the expected response to external forcing, for many decision makers a question posed in a different way may be more relevant. For instance, they may ask, ‘Are the continuing drier-than-normal conditions in the Sahel due to human causes?’ Such questions are difficult to respond to because of a statistical phenomenon known as ‘selection bias’. The fact that the questions are ‘self selected’ from the observations (only large observed climate anomalies in a historical context would be likely to be the subject of such a question) makes it difficult to assess their statistical significance from the same observations (see, e.g., von Storch and Zwiers, 1999). Nevertheless, there is a need for answers to such questions, and examples of studies that attempt to do so are discussed in this chapter (e.g., see Section