| |
Presented at the NABS Annual meeting, Pittsburgh, Pennsylvania, 2002
in Bioassessment: Predictive Models
A RIVPACS-INSPIRED BIOASSESSMENT MODEL FOR ILLINOIS: EFFECTS OF TAXON RESOLUTION AND FREQUENCY.
M.J. Paul1, J. Gerritsen1, E. Leppo1, B. Ettinger2, and M. Joseph2. 1Tetra Tech, Inc., 10045 Red Run Boulevard, Suite 110, Owings Mills, Maryland 21117, 2Bureau of Water, Illinois Environmental Protection Agency, Springfield, Illinois 62702
Site-specific taxa prediction bioassessment models (e.g. RIVPACS, AUSRIVAS) have gained considerable attention in the USA. As with most bioassessment approaches, there are many factors that constrain the precision and discriminatory power of these models. We developed these bioassessment models using macroinvertebrate data from the State of Illinois to investigate two factors: taxonomic resolution and capture probability (a measure of taxon frequency). We also compared four different impairment designation approaches to assess discrimination between a priori reference and disturbed sites. Lastly, we compared discrimination using the taxa prediction models to that using multimetric bioassessment models (IBI). We found that genera level taxonomic resolution and capture probabilities greater than 0.5 satisfactorily discriminated between disturbed and reference sites. It appeared that the frequency of sensitive taxa determined which capture probability level was sufficient for identifying disturbance. We also found that impairment threshold approaches varied, but using quartiles of the reference database provided the most even and clearest impairment classification. Taxa prediction bioassessment scores were significantly correlated with multimetric bioassessment scores, however, the multivariate models exhibited 9 to 30 percent greater discrimination between reference and disturbed sites. Taxonomic resolution and taxon frequency affect these models and have to be considered carefully during model construction.
|
|
