Pollen productivity estimates strongly depend on assumed pollen dispersal

Past plant abundance may be reconstructed from pollen data if dispersal distances of pollen and pollen productivities of each taxon are known. Using surface sediment samples from small and medium sized, closed and near circular lakes from lowland Central Europe, we tested the validity of three pollen dispersal models by comparing empirical pollen data from each lake with simulated pollen data derived from applying various pollen dispersal models to vegetation data from rings situated up to 100 km from each site. Pollen assemblages simulated with a Lagrangian stochastic (LS) model best fit real pollen assemblages, simulations with the commonly used Prentice model on pollen dispersal underestimated the amount of pollen arriving from distances larger than 10 km and overestimated the differences in dispersal distances between lighter ( Pinus) and heavier ( Fagus, Picea) pollen grains. The LS model appeared to provide more appropriate simulations. Pollen productivity estimates (PPEs) calculated for the data set showed that the choice of the dispersal model has great impact on the results. If derived with the Prentice model, PPEs for Fagus and Picea are three times higher than with the LS model. Studies on pollen productivities thus need to consider the apparent limitations of the Prentice model. We suggest an alternative approach, which uses simulations instead of the extended R-value model, to calculate PPEs. The approach is flexible in the use of dispersal functions and produced consistent results for two independent data sets from small and medium sized lakes.