Analysing data hom post-impact studies using asymmetrical analyses of variance: A case study of epibiota on marinas

Several recent methods have been developed for detecting anthropogenic perturbations. Most have been analyses of data collected before and after some anthropogenic disturbance. It may, however, be more common that data can be collected only after a disturbance has occurred. In such situations, the only appropriate sampling design to use will often be an asymmetrical design because it will avoid problems of spatial confounding. Here, I describe in detail the steps involved in constructing asymmetrical analyses of variance using a case study of subtidal epibiota around marinas as an example. Differences between the marina and control locations were detected for a number of taxa, but this was often only possible after post-hoc pooling of non-significant terms. Marina and control locations varied greatly from estuary to estuary and consequently it was not possible to identify suites of species that were typical of either type of location. This result highlighted the need for multiple control locations near each marina to allow a reliable estimate of the variability among controls. Large variability among controls would mean that if differences existed between disturbed and control locations they would rarely be detected. These and other problems associated with analysing 'after data' are discussed in addition to the precautions to take when designing environmental sampling regimes.