Seven community similarity indices were compared using a manipulated data set with known changes and an experimental data set of phytoplankton community data. Community similarities were calculated based on (1) strict number of individuals per species and (2) corrected for species size with an estimate of total biomass. The similarity techniques used in this study were heavily affected by the abundant species, and changes in the minor taxa were frequently undetectable. Two of the seven statistics were previously unpublished modifications of the Percent Similarity method of Whittaker , which were specifically modified to enhance the sensitivity to changes in low frequency species and proportional changes.
While complex numerical techniques are commonplace in ecological literature, there is no generally accepted mathematical definition of community similarity. Ideally community comparison techniques should be sensitive to the loss or gain of a species and to changes in abundance of low and high frequency taxa. They should also detect proportional changes. None of the community similarity techniques used in this study fulfilled all the above requirements completely. All the similarity techniques assigned similarity values based on the presence of individuals. Perfect replication of cell counts for scarce species added little or nothing to the similarity values. At this time, a combination of statistical techniques using both numbers per taxa and an estimate of total biomass per taxa is recommended for the analysis of phytoplankton community data.