Sensing Vegetational Patterns with Pollen Data: Choosing the Data

Abstract

Comparisons between modern pollen and tree—inventory data from northeastern North America show how the clear portrayal of vegetational patterns in pollen data depends upon several choices made in selecting and preparing these data. These choices concern the size and location of the area studied, the types of pollen examined, the quantitative measure of pollen used, the types of sediments sampled, the number of analysts contributing samples to the data set, and the degree of smoothing applied to the data either by adding samples together or by combining pollen types into groups. Our results show that these choices can control not only what scale of vegetational pattern will appear on maps of pollen data but also whether these maps portray vegetational patterns at all. Of the choices examined, the most important were those concerning the size of the area studied, the types of pollen examined, and the amount of smoothing used. Before presenting the results, we introduce an expression for a signal—to—noise ratio that contrasts the degree of vegetational variation affecting the pollen data with the size of the uncertainties and nonvegetational variation (e.g., changes in pollen dispersal) that also influence these data. The effects of the different choices in data selection are discussed in light of their effect on this ratio.