You are here:
Combining Bayesian networks and conceptual models for Superfund remediation support
Citation:
Carriger, J. AND R. Parker. Combining Bayesian networks and conceptual models for Superfund remediation support. SETAC North America 39th Annual Meeting, Sacramento, CA, November 04 - 08, 2018.
Impact/Purpose:
This presentation will explore the application of Bayesian networks combined with Superfund conceptual models for remediation evaluation and support. The initial work discussed here could be further developed, tested, and incorporated into Superfund site assessment guidance. A broad spectrum of listeners involved in developing or using Superfund monitoring and technical data might be interested including analysts, environmental scientists, decision makers, and risk managers.
Description:
Conceptual models play a central role in Superfund remediation. They are used as a communication tool and a platform for quantitative models. Bayesian networks offer many opportunities for supporting conceptual and quantitative modeling in Superfund remediation. However, their application to Superfund risk assessments and management is still rare. A framework will be presented that combines Bayesian network knowledge engineering and development with requirements and recommendations for Superfund conceptual models. We call the resulting products conceptual Bayesian networks (CBNs). A development process for Superfund CBNs will be introduced that can support all phases of site investigation and cleanup. Learning is incorporated into the process for updating knowledge of the model structural components and connections as new data and understanding becomes available. Inferences to support decision making with a CBN will be presented including causal pathway analysis from sources of stressors to receptors of concern and capabilities for examining the causal influence of remediation interventions in breaking exposure pathways. The CBN process is designed for adaptability to multiple sites and stressor-types as well as comprehensiveness for adequately incorporating required legal and scientific guidance for site assessments. Components and connections that are broadly transferrable will be shown. The focus of the presentation is on the qualitative side of Bayesian network development and inferences, which is necessary for subsequently developing quantitative models. The potential value added of a CBN for quantitative risk assessments will be demonstrated. The flexibility of CBNs allows adaptation across all phases of Superfund assessments and can support management decision making, uncertainty evaluation, and communication from the structural and quantitative capabilities of Bayesian networks.