Pollen counts statistics and its relevance to precision

In the day to day management of pollen counts from aerobiological samples of national networks, only a small proportion (usually from 12 to 15%) of the daily microscope slide is read. It is generally believed that, otherwise, too much time will be spent reading slides for a minimal increase in precision. Different networks use different slide sampling methods (longitudinal, transverse or at random) and a different number of lines are routinely read. However, the topic of the precision of the different pollen count strategies has seldom been the object of serious investigation. In this study, the precision of different sampling methods of 12 pollen types was investigated by: a) counting pollen grains over the whole slide (3 slides per taxa), b) spatially (i.e. microscope field per microscope field) recording over the 3120 fields found at 400x the location of each pollen grain, c) sub- sampling, by macro procedures, this population by selecting a number of transverse (1 to 48) or longitudinal (1 to 20) lines, or a number of random fields (90 to 2340), so that between 0.96 to 46.15% (transverse), 3 to 66.6% (longitudinal) or 3 to 75% (random) of the whole slide was artificially counted. Between nine and twelve procedures were built per reading strategy. The error found is much higher than what is normally believed, and it was significantly correlated with the abundance of a pollen taxa on the sampled slide. It is only with a total count over 1000 (corresponding to a concentration of above 500 m-3) that the mean error of 4 longitudinal lines (or 13.3% of the slide), the standard protocol of both the Italian Association of Aerobiology (AIA) and the Spanish Aerobiology Network (REA), was always below 30%.