Comparison of taxon calibrations,modern analogue techniques,and forest-stand simulation models for the quantitative reconstruction of past vegetation

We compared results of three quantitative methods for reconstructing past vegetation, including (1) calibrations of per cent dominance vs. per cent pollen for individual tree taxa; (2) coefficients of dissimilarity between fossil and modern pollen assemblages; and (3) simulation modelling of long-term forest succession. For calibration, we used 1684 modern pollen samples and 1742 forest-inventory summaries from eastern North America. We applied the calibrations at 1 000-year intervals to the 19000-year pollen record from Anderson Pond, Tennessee. Dominance values reconstructed using taxon calibrations typically paralleled the changes in percentages of arboreal pollen. Dissimilarity values between fossil and modern pollen assemblages were lowest from 19 000 yr BP to 15 000 yr BP and from 9 000 yr BP to the present, indicating good modern analogues for both full-glacial and Holocene time periods. During times of good analogue, forest composition estimated from forest inventories at the locations of best pollen analogues was similar to that reconstructed using taxon calibrations. Relatively poor modern analogues exist for late glacial and early Holocene pollen spectra from Anderson Pond. During times of poor analogues, forest composition reconstructed from taxon calibrations differed from that derived from analogue methods. FORET-model simulations of forest-stand biomass differed markedly from palaeovegetation estimates made from the other two methods. This is primarily because the FORET model is scaled to forest succession on 1/12-ha forest plots rather than forest dynamics integrated over a broad landscape mosaic. Because they are based upon fundamental relationships of pollen production and dispersal, taxon calibrations provide the best available means for reconstructing forest history both during times of good and of poor analogues.