Subsampling error in soft algae counts: Results and the possible implications for biomonitoring programs

Repeatability is critical to producing valid results in bioassessment and can be significantly affected by the use of subsampling in the processing of biological samples in the laboratory. While developing protocols and standards for internal Quality Control of phytoplankton identifications on a large project for the US EPA, we became interested in the effect of subsampling error on total disagreement calculations. Initially, taxonomists with a minimum of 10 years counting and identification experience followed a protocol of identifying 300 natural units in a phytoplankton sample to lowest practical taxonomic level. 30 samples were then randomly selected for re-identification by the original taxonomist to provide more consistent identifications and better isolate the effects of subsampling. We then calculated similarities using the Sørensen index for taxa present and percent similarity based on relative abundance. Preliminary calculations show a mean value of 72 on the Sørensen (minimum=32, maximum=100) while the mean for relative abundance was 73 (minimum=40, maximum=95). We are currently recalculating these indices at different levels of identification (i.e. genus and family) as well as comparing the results of the two data sets on different bioassessment indices and will be presenting these results at NABS.