PT03 Landscape-Scale Ecological Risk Assessment
Exhibit Hall
8:00 AM - Tuesday, 11 November 2003
(PT043) Ohio 1: Attributing Impact of Mixtures and Habitat Degradation via Predictive Models.
Dyer, S.D.1, Zwart, D.2, Posthuma, L.2, Hawkins, C.P.3, 1 The Procter & Gamble Company, Cincinnati, Ohio, USA2 RIVM - Lab. for Ecological Risk Assessment, Bilthoven, The Netherlands3 Utah State University, Logan, Utah, USA
ABSTRACT- A dataset consisting of nearly 700 sampling sites with fish, invertebrate, habitat and chemistry data has been compiled for Ohio, USA, using state and federal agency data sources. The objective was to produce a comprehensive, eco-epidemiological description of this dataset in terms of habitat degradation and the role of chemical mixtures at site, regional and statewide scales. To accomplish this, we linked an ecologically predictive model (RIVPACS, River InVertebrate Prediction and Classification System) with ecotoxicologically-based methods (e.g., the ms-PAF, multi-substance-Potentially Affected Fraction of species in the ecosystem) via general linear models. The RIVPACS approach was used to provide a numerical expectation on the number and types of taxa expected in a location given a set of geo-chemical and habitat descriptors. The proportion of fish taxa not observed, but expected (e.g., low O/E score), provided the magnitude of impact that might be explained by the cumulative risks of toxic compounds and/or degraded habitat conditions. The potential for adverse ecological effects of chemical mixtures was estimated as the joint risk of compound mixtures using Species Sensitivity Distributions, mixture principles and by taking account of the influence of local factors such that affect availability. On a per taxa basis, models describing presence and/or abundance vs. measures of habitat degradation were developed via the Ohio dataset. Probabilities of cumulative effects from chemical mixtures and habitat degradation were expressed in simple pie-diagrams with each slice corresponding to factors statistically attributable to the magnitude of impact. Grey slices show the unexplained variance per site. The models derived from the basic training set are being confronted with newly collected data as model validation attempt.
Key words: RIVPACS, SSD, Habitat, Multiple stressors