A Bayesian Approach to Landscape Ecological Risk Assessment Applied to the Upper Grande Ronde Watershed, Oregon

We present a Bayesian network model based on the ecological risk assessment framework to evaluate potential impacts to habitats and resources resulting from wildfire, grazing, forest management activities, and insect outbreaks in a forested landscape in northeastern Oregon. The Bayesian network structure consisted of three tiers of nodes: landscape disturbances, habitats, and the ecological resources or endpoints of interest to land managers. Nodes at each tier were linked to lower nodes if ecological and spatial relationships existed between them. All parameters had four potential discrete states: zero, low, medium, and high. Our model reliably predicted probable risk to habitats and endpoints from natural and anthropogenic disturbances. The disturbances most likely to transform habitats and effect ecological resources were forest management and wildfire. Of the six habitats, moist forest (characterized by Douglas fir and grand fir) was found to be at greatest risk of ecological impacts. The management endpoint with the highest likelihood of impact was historical range of variability (HRV) for salmon habitat, followed by recreation (hunting native ungulates) and HRV wildfire. We found that the Bayesian approach to ecological risk assessment was a useful method to assess potential impacts to ecological resources resulting from forest management and natural disturbances.