By studying the climate of the past millennium in great detail, it is possible to analyze the impact of human activities during the past few centuries compared with natural driving forces, such as volcanic eruptions and solar variations, and internal variations related only to the complex dynamics of the climate system itself. An optimal analysis is obtained when proxy data and model results are combined. Proxy data provide the "ground truth" (e.g. Mann et al. 1998; Jones and Mann 2004; Luterbacher et al. 2004; Moberg et al. 2005), whereas models can be used to interpret the observed changes, in particular to analyze the mechanisms responsible for the observed response of the climate system to changes in external forcing (e.g. Crowley 2000; Shindell et al. 2001; Stott et al. 2001; Bertrand et al. 2002; Bauer et al. 2003; Goosse et al. 2005b). The majority of past studies have been heavily biased towards one or other of these two particular ways of analyzing past changes, while only briefly discussing (if at all) the other. Our goal here is to show that very interesting insights and synergies could be gained from a combined analysis of proxy records and model results.

Model results and proxy data can be complementary in several ways. First, proxy data can be used to evaluate the validity of model results (e.g. Shindell et al. 2001; Bertrand et al. 2002; Bauer et al. 2003; Goosse et al. 2005a). This is a necessary step before any analysis of model results, but it is not a trivial task, as any discrepancy between model results and data could be due not only to model deficiencies but also to several other processes. The uncertainties in the forcing history could have an impact on model results. Proxy data record a climatic signal but also include a component of nonclimatic noise that is difficult to estimate. Furthermore, proxy records could be influenced by regional or local changes whereas model results provide information at a spatial scale of hundreds of kilometers. Second, model results could be used to provide information on the climatic signal that is recorded by the proxy data, disentangling the contribution of various climatic variables. For instance, changes in ground surface temperature are influenced by variations in surface temperature, snow cover, and changes in surface vegetation. Model results include an estimate of variations of all such variables and could thus help in the interpretation of the recorded variations in proxy data (e.g. Gonzalez-Rouco et al. 2003; Mann and Schmidt 2003). Additionally, models provide a physically consistent and complete set of data that could be used to test the method applied to reconstruct past changes from proxy data with incomplete temporal and spatial coverage (e.g. Rutherford et al. 2003; von Storch et al. 2004; Mann et al. 2005b). Finally, evidence deduced from proxy data and models can be combined to describe and analyze the causes of past changes. All these elements will be illustrated below.

Reviews have extensively described recent work on the past millennium climate (e.g. Jones and Mann 2004), and we therefore provide only some general information about proxy records, reconstructions, and model simulation below to give an introduction to the subject. In subsequent sections, reconstruction methods are tested using synthetic time-series derived from model results; reconstructions and model results are used to detect the role of various forcings at the hemispheric scale and changes at regional scale are discussed; and some considerations are presented about promising techniques that could allow the direct assimilation of proxy records in models, providing a single estimate of past changes and of the mechanisms responsible for those changes based on all the available information.

0 0

Post a comment