Abstract/Summary

The warmings associated with Dansgaard-Oeschger cycles (D-Os) during the last glacial were as fast as and of a similar magnitude to expected warming over the 21st century. Identifying these warmings in pollen records objectively may assist understanding of current and future climate change impacts. Identification of D-Os by age alone is hampered by dating uncertainties in both Greenland and pollen core age-depth models, so additional methods of identification are sought. Several statistical techniques applied directly to pollen series from the circum-Mediterranean area do not show D-Os clearly and consistently. An alternative approach using quantitative climate reconstructions made using WA-PLS and fxTWA-PLS is explored. First, the robustness of these methods is evaluated. A training set should sample a climate space which includes the target climate as comprehensively as possible. Actual distributions of taxon abundances along climate gradients often differ from the symmetrical unimodal shape assumed by WA-PLS and fxTWA-PLS, so many taxa are unsuited to a mean-based method, and many are weakly evidenced or even misleading. Calibrations made including and excluding such suspect taxa are indistinguishable, yet provide different reconstructions. Large differences between the balance of abundances in the training set and a fossil core lead to poorly evidenced reconstructions. A proposed alternative reconstruction method avoids a problem identified in WA-PLS (but fxTWA-PLS is unaffected). Potential D-Os are found in reconstructions by their similarity to the signature asymmetrical shape of D-Os warmings in Greenland (‘pattern matching’). By combining the age uncertainties of the Greenland GIs and of the core age models, updated to IntCal20, many younger candidate D-Os can be assigned with high plausibility to specific GIs, and some rejected. Many older candidates are highly plausibly D-Os, being similar in number to the expected GIs, but cannot be firmly assigned.