Implications of Sampling Frequency for Detecting Temporal Patterns during Environmental Effects Monitoring

To estimate the effects of sampling frequency on detecting temporal patterns during environmental effects monitoring, we used multivariate analyses and data subsampling to investigate long-term (spanning 20 years) patterns in benthic invertebrate community structure downriver of a large pulp mill in southern British Columbia. Patterns in invertebrate abundance sampled yearly were related to long-term patterns in several physicochemical variables measured in the river using multidimensional scaling ordination. The only available physicochemical variables that were significantly correlated with invertebrate community structure over the 20-year period were the mill outputs of total phosphorus and suspended solids, and these were associated with increased abundances of five families of mayflies, stoneflies and caddisflies. To evaluate the implications of sampling on a more coarse (than yearly) time scale, the full data set spanning 20 years was subsampled to produce a series of smaller data sets, each simulating a sampling frequency of once every three years. Ordination of the subsample data sets showed that an average of 71% of the important taxa and 50% of the important physicochemical variables highlighted in the full analysis were missed in the subset analyses. These results underscore the importance of ensuring adequate temporal replication of sampling effort when a major goal is to directly measure or test for temporal patterns of stressor impacts.