Taxonomic aggregation does not alleviate the lack of consistency in analysing diversity in long-term phytoplankton monitoring data : a rejoinder to Pomati et al. (2015)

Abstract

Long-term phytoplankton monitoring provides an important resource for studying the effects of environmental change on communities and testing ecological hypotheses. However, because of identification difficulties, maintaining consistency in the data over long periods is extremely difficult. It is usually assumed that consistency is improved when only one taxonomist is responsible throughout, and/or when data are aggregated to a coarser taxonomic level. Neither assumption has been critically tested. We address the comment of Pomati et al. (2015) on our earlier Opinion paper (Straile et al , 2013) and test these assumptions with the long-term data from Lake Zurich.2. We show that aggregation to coarser taxonomic levels does not improve data set consistency because: (i) the proportional effect of misclassification is unlikely to be reduced by lumping taxa since the fewer misclassifications affect the dynamics of an overall lower number of taxa, that is the proportional effect is constant, and (ii) because changes in detection limits will affect all taxonomic levels proportionally.3. We also show that, although a single taxonomist supervised phytoplankton recordings, data consistency is undermined by: (i) learning via exchange with other taxonomists and participation in taxonomic workshops, and (ii) a reduction in detection limits of species, presumably due to an increase in the number of taxonomists (allowing an increased processing time per sample).4. As a consequence of (i) a reduction in detection limits, (ii) the confirmed taxonomic learning and (iii) the failure of taxonomic aggregation to improve consistency, our new evidence strengthens the view that there are consistency problems in the Lake Zurich data set, and the need for a critical review of the conclusions of Pomati et al. (2012) and Matthews & Pomati (2012).