Weighted averaging partial least squares regression (WA-PLS): an improved method for reconstructing environmental variables from species assemblages

Dissimilarity coefficients measure the difference between multivariate samples and provide a quantitative aid to the identification of modern analogs for fossil pollen samples. How eight coefficients responded to differences among modern pollen samples from eastern North America was tested. These coefficients represent three different classes: (1) unweighted coefficients that are most strongly influenced by large-valued pollen types, (2) equal-weight coefficients that weight all pollen types equally but can be too sensitive to variations among rare types, and (3) signal-to-noise coefficients that are intermediate in their weighting of pollen types. The studies with modern pollen allowed definition of critical values for each coefficient, which, when not exceeded, indicate that two pollen samples originate from the same vegetation region. Dissimilarity coefficients were used to compare modern and fossil pollen samples, and modern samples so similar to fossil samples were found that most of three late Quaternary pollen diagrams could be “reconstructed” by substituting modern samples for fossil samples. When the coefficients indicated that the fossil spectra had no modern analogs, then the reconstructed diagrams did not match all aspects of the originals. No modern analogs existed for samples from before 9300 yr B.P. at Kirchner Marsh, Minnesota, and from before 11,000 yr B.P. at Wintergreen Lake, Michigan, but modern analogs existed for almost all Holocene samples from these two sites and Brandreth Bog, New York.