NUMERICAL-ANALYSIS OF MODERN AND FOSSIL POLLEN SPECTRA AS A TOOL FOR ELUCIDATING THE NATURE OF FINE-SCALE HUMAN ACTIVITIES IN BOREAL AREAS

Modern pollen assemblages from the major vegetation units (both natura l and anthropogenic) on the island of Hailuoto, Finland are studied fr om 29 surface moss samples. A total of 59 pollen and spore taxa are re corded. The pollen data-set is related by redundancy analysis (RDA) to six external synthetic variables that characterize in general terms t he different major ecological situations relevant to the land-use on t he island (sand, humus, forest, field, deforested, treeless). Weighted average (WA) optima are calculated to identify which pollen taxa are most indicative of four of these external variables. Two fossil pollen diagrams from Hailuoto are re-interpreted by positioning their sample s on the RDA plot of the modern data-set and by classifying the modern and fossil spectra together in a minimum-variance cluster analysis. T he RDA plot shows that the strongest features in the modern pollen ass emblages are the contrast between spectra from dry sandy soils and tho se from damper soils with more organic content, and between assemblage s from forested and unforested areas. For anthropogenic situations the numerical analyses detect a distinction in the pollen assemblages cha racterizing fields from those representing farms and trackways but is less successful in separating the latter two. The WA optima indicate t hat most pollen taxa considered as indicator types primarily represent a mixed field/farm category of land-use. The RDA and minimum-variance cluster analysis confirm that the two fossil sites have had different vegetational histories. In neither is there any phase equivalent to t he modern farm situation on Hailuoto. The initial phase in each diagra m represents shore meadows and fields but because Poaceae pollen is id entifiable only to family level it is not possible to separate these t wo habitat types.