| |
Communication at the NABS Annual meeting, Keystone, 2000
(471) BUILDING A BETTER MOUSETRAP: HOW TAXONOMIC RESOLUTION AND SAMPLING METHOD AFFECT RIVPACS-TYPE MODELS IN THE UNITED STATES.
V.B. DeWaard, C.P. Hawkins, and J.D. Ostermiller. Department of Fisheries and Wildlife, Utah State University, Logan, UT 84322-5210
RIVPACS-type models have not been thoroughly evaluated in the U.S. as a method for biological assessment, and few studies have evaluated their use for taxa other than invertebrates. In this study, we used data from Maine (macroinvertebrates, artificial substrates), Ohio (macroinvertebrates, artificial substrates and fish, natural habitats), North Carolina (macroinvertebrates, multiple natural habitats), and Oregon/Washington (macroinvertebrates, single natural habitats) to examine 3 questions: (1) how does level of taxonomic resolution influence model performance, (2) how does sampling method affect performance, and (3) can models be used with fish assemblages. We evaluated model performance by examining accuracy (model bias) and precision (model error) and are currently examining sensitivity (departure from model predictions as a function of stress). All models were unbiased. Models constructed with genus-level data and with data from natural habitats were generally more precise than models based on species, families, and artificial substrates. Error for the fish model was within the range of model errors observed for invertebrate models. Our ongoing research is designed to further refine our understanding of how natural conditions, sampling methods, assemblage used, and data treatment influence the performance of predictive models and to elucidate the underlying causes for the patterns we have observed.
Presented at 1:00 PM on Wednesday, May 31, 2000 in Bioassessment: Techniques
|
|
