Search

This paper presents a probabilistic method of pollen spectra analysis. The method relies on a pollen taxon characterization using biotic and abiotic plant attribute modes, and their occurrence in a given pollen spectrum at a specific site. This type of analysis can provide an interpretation, which can lead to the reconstruction of the biome and, to an extent, of the abiotic conditions at the site.The analysis has been carried out at the European scale using data provided by the European Pollen Database for about 1000 sites. This dataset contains about 50,000 pollen spectra from the last 21 ka. In these spectra, each pollen taxon has been characterized by a set of 10 chosen attributes. These have been selected with regard to their relevance in biome reconstruction, but also on the basis of available literature. By using the probability of occurrence of each taxon in a given pollen spectrum, it is possible to calculate an affinity index for the spectrum to the attribute considered. To overcome difficulties caused by pollen identification in low diversified pollen spectra, a co-occurrence concept has been used to give more information.The method has been validated on a set of 1327 modern surface samples by comparing the results to the major climatic and environmental variables that control the distribution of the vegetation. A reconstruction exercise on various characteristics of the plants was then carried out on a 6-ka dataset. This confirmed previous studies by showing a strong dominance of deciduous forest over most of Europe, related to a milder climate than at present in the north and a wetter and colder climate than at present in the south. By analysing the change in pollen/seed dispersion strategies and the light requirement, we show that the history of vegetation dynamics in relation to human influences can be assessed using this method.Our results show that the probabilistic method is an objective tool for pollen assemblage analysis. It allows reconstruction of various characteristics of the vegetation at the continental and global scale for periods and sites with significantly different climate conditions. This method can also be used to compare maps of vegetation attributes for the validation of the new generalized dynamic ecosystems models. [ABSTRACT FROM AUTHOR]