| |
Presented at the NABS Annual meeting, Vancouver, British Columbia, 2004
in Bioassessment 5
Investigating the relationships between environmental stressors and stream health using belief network models
J.D. Allan1 and L.L. Yuan2. 1School of Natural Resources & Environment, University of Michigan, Ann Arbor, MI 48109, 2National Center for Environmental Assessment, U.S. Environmental Protection Agency, Washington, DC 20460
A rapidly growing literature addresses the impacts of changing land use on lotic systems by coupling land-use data to stream bioassessments. However, these statistical associations provide limited insight into the pathways by which land use influences stream condition. Diagnosing the causes of impairment and restoration activities for impaired streams would benefit from a more mechanistic understanding of the pathways by which specific response variables are influenced by land use. We model the influence of agricultural and urban land use on stream condition using Bayesian Belief Networks (BBN), a method to depict logical or causal relations among factors that influence the likelihood of outcome states such as ecological condition or species viability. BBNs offer several benefits: relationships can be expressed probabilistically using both empirical data and expert judgment, presumed causal relationships are formalized, uncertainty is considered, and models can be refined through addition of new cases. Information regarding the states of individual variables (e.g., extent of riparian forest, presence of woody debris) and their probabilistic relationships are derived from literature review, professional judgment, and theoretical expectations. The resulting models aid in identifying the relative influence of specific pathways, and in diagnosing cause of impairment for an observed outcome.
|
|
