Information content of zero pollen counts in Holocene profiles

Abstract

Percentage-based pollen diagrams display the relative proportions of various taxa in relation to a stated pollen sum. A zero count for a given palynological taxon, as recorded for a given sample, does not mean that the taxon was absent at the time and place represented by the sample, but only that the concentration of palynomorphs is, at most, below the detection limit of the method. Here we apply the statistical concept of the confidence interval to approach this problem. In general, observing zero counts, one can state that there is a 95% probability that the true average for the given sample is simply below 3.0. However, if the sample has a neighboring non-zero sample, the best-fit value for the expected number of counts is 0.6–0.7, as estimated in the statistical analysis of a data base of 39 pollen diagrams from the Carpathian mountains, with a combined total of 4304 pollen curves. We suggest applying the confidence intervals (for a zero as well any positive number of counts) in drawing pollen diagrams, in order to clearly indicate the measurement uncertainties connected with pollen counting.