Jackson, J.K., and V.H. Resh. 1988. Canadian Journal of Fisheries and Aquatic Sciences 45(2):280–286.
Sequential decision plans provide a statistical approach that can reduce the number of benthic sample units needed to classify a site as impacted or unimpacted, thus reducing the cost of using benthic macroinvertebrates in water quality assessment programs. These plans require information about unimpacted and impacted conditions, the mathematical distribution of the data, and acceptable risks of classification error. A large benthic data set (n = 55) was used for simulations that created and tested sequential decision plans. Using 10–60% reductions in species richness, mayfly (Cinygmula) population density, and species diversity as definitions of impact in the simulations, the average number of sample units processed for identification of the unimpacted reference site was reduced (compared with fixed sample-size methods that are commonly used) by 50–64% for species richness, 59–79% for density estimates, and 51–55% for species diversity. Unimpacted data sets were initially classified as representing impacted conditions in 0–5% of the cases. If classifications are to be interpreted properly, sampling error and spatial and temporal variation in biological parameters must be considered when sequential decision plans are created.