A probabilist model of the relations between pollen and climate and its application to 80 European annual spectra

The relation between pollen and climate is commonly computed by regressing the climatical factor. The disadvantage of this method is that it does not respect the ordinal and intermittent nature of field data. This paper overc omes the artefacts created by this problem by using a probabilist calibrati on, that quantifies the ecological linkage between a taxon T and a factor w ith a general parameter, the probability PROX for an abundance A being conf ined near the rank F of the factor. Confining simulates the effect of the f actor upon the concentration of presences and ordering of abundances, and c alibrates the climatical behaviour of a taxon with the set of PROX for all possible pairs (A,F). It summarizes a behaviour with the probable position of each abundance A in the range of the factor. Calibration was applied to 130 pollen taxa observed in a network of 80 standardized annual aeropollini c spectra. Spectra were mostly from France, the rest being from a transect stretching from Sweden to Algeria. Spectra were characterized by the values of 10 climatic factors, as well as the presence and abundance of 130 polle n taxa. The influence of geographical climate differences upon pollen conte nt in the atmosphere was quantified by comparing the spectra. Pairs from di fferent localities but the same year were compared. The reliability of indi cator taxa was tested by estimating the climate in the 80 spectra using cal ibration. For all the taxa observed in a spectrum, the envelope of confinin gs generally followed an unimodal gradient, whose mode was the probable pos ition of the spectrum. Reliability of the estimate was measured by its accu racy, being the agreement between estimates and measures; and by its stabil ity, being the agreement between two estimates from the same climate accord ing to two different flora (the two halves of a spectrum for instance). Ave rage accuracy was 72%, and average stability 87%.