Watershed Symposium 2019

Accounting for Local Variation: Bugs in the Biomonitoring System

Summary:
Aquatic Invertebrates are often collected as environmental monitors; but the design recommended by regulatory agencies is usually ineffective for most monitoring goals. Our results are especially useful to watershed groups, or foundations hoping to describe change (good or bad) in benthic community structure.

Full Abstract:
Macroinvertebrate studies have become commonplace since the acceptance and promotion of rapid bioassessment protocols. Certain assumptions are implicit to the use of these protocols whenever they are used to detect change. We examined four inter-related aspects of bioassessment sampling designs that may preclude their use (senu stricto) to describe change: (1) aspects of design, (2) assumption of variance homogeneity, (3) assumption of taxonomic completeness, (4) assumption of standard unit effort. Due to time limitations, this presentation will focus primarily on the underlying assumptions (aspects of design) and within-site variation (variance homogeneity). One of the underlying premises of the bioassessment design is that a single, large sample can sufficiently represent all the variation in a riffle, or in a reach. Therefore, assessments are conducted without true replication in the field. However, this assumption is only valid if the variation is zero—or sufficiently close to zero that the error is negligible. That is, if you were to sample the same site multiple times, can you get the same (or similar) result? Although many people using bioassessment designs promote the idea that large samples eliminate variation, this was never actually tested in bioassessment development. We examined within-site variation of a bioassessment design by collecting five replicate composite samples from five individual sites in Sublette County Wyoming. The method of collection for WY DEQ samples is to collect eight randomly placed Surber (sample area 1sq. ft.) samplers from a riffle. Thus, to collect five replicates from a riffle, 40 individual, 1-square foot samples, had to be collected and apportioned among five large 8-surber-composites samples. This requires a herculean field effort to collect a modest number of replicates. Therefore, the advantages of collecting these samples should be substantial. We found significant variation in all metrics. Moreover, the variation in taxa richness, using the most-resolute taxonomy, was sufficiently large to cause problems with both multimetric indices and predictive models (e.g., rivpacs-style monitoring). In some cases, the variation in the observed taxa richness exceeded the impairment threshold defined for rivpacs models. We did not find a significant reduction in variation through compositing. Furthermore, compositing samples actually prevents the use of more rigorous statistical designs (e.g., modeling, and covariance). Finally, there is reason to believe that these observations have widespread application (e.g., Townsend 1989) Most watershed groups, conservation groups, foundations, and water quality monitors are interested in detecting spatial/temporal changes in community structure. Our results suggest this cannot be adequately evaluated using standard bioassessment designs. Discussion will briefly include strategies to use variation to detect community changes.