Predicting tree frequencies from pollen frequency: an attempt to validate the R value method

SUMMARY Pollen-tree correction factors are presented using regression analysis of pre-settlement tree and pollen frequency data from 32 small lakes (1–10 ha) in Wisconsin. Regression results using vegetation samples from various sized areas around each lake for Fagus grandifolia (Ehrh.) and Tsuga canadensis (Carr.), are consistent with my initial design of using vegetation data for the single township in which a lake was found to compute the regression results. Pollen-tree correction factors from pre-settlement and modern data were compared for all major pollen types. Notable differences were found between pre-settlement data and modern data sets for Wisconsin in Betula spp, T. canadensis, and Quercus spp. Little difference was observed in the fit of the regressions for these data sets. Estimations of pre-settlement relative tree frequencies, derived from the application of modern correction factors on pre-settlement pollen frequencies, were compared to measurements of pre-settlement vegetation. The reconstruction of pre-settlement vegetation was incongruous with measurements of pre-settlement tree frequencies. It is unclear, given minor differences in the data sources for modern and pre-settlement data, whether the observed differences in regression results are artifacts of the data sets, or reflect differences in pollen-tree representation between the two time periods sampled. Regardless, the error associated with prediction of tree frequencies from pollen frequencies is shown here to preclude any assurance of accuracy with normal pollen count levels.