The influence of data transformations on biological monitoring studies using macroinvertebrates

The relative merits of untransformed, √-transformed, √√-transformed and binary data for the discrimination of sites of varying water quality was investigated using macroinvertebrate data from the river Trent system in England. The influence of these four data treatments was investigated using the CLUSTER, ANOSIM and SIMPER programmes of the PRIMER computer software package. In general, the results showed that relative abundance is important in distinguishing between the more heavily polluted sites, which have poor faunas. In contrast, the down-weighting of abundant taxa, through the use of data transformations or binary data allows for a more effective discrimination of richer sites. The √√-transformation can be recommended over the other data treatments as it offers a good compromise between untransformed and binary data, and as it allows for the effective discrimination of sites over a wide range of water qualities.