Increased Use of Bayesian Network Models Has Improved Environmental Risk Assessments

Area of Science:

• Environmental Science
• Ecological Risk Assessment
• Probabilistic Modeling

Background:

• Environmental and ecological risk assessments evaluate potential environmental impacts from stressors.
• Traditional risk assessments often use nonprobabilistic methods, like risk quotients.
• Bayesian network (BN) models provide a probabilistic and causal modeling approach increasingly adopted in environmental science.

Purpose of the Study:

• To highlight the growing application and advancements of Bayesian network models in environmental risk assessment and management.
• To showcase a range of BN applications from cellular to national scales.
• To demonstrate how BNs can integrate diverse data and adapt to existing risk assessment frameworks.

Main Methods:

• Utilizing Bayesian networks, which are directed acyclic graphs quantifying causal relationships and uncertainty via conditional probability tables.
• Integrating various information types, including expert elicitation.
• Applying BNs to diverse environmental systems and scales, incorporating spatial (GIS-based) and temporal (dynamic) modeling.

Main Results:

• A steady increase in BN applications for environmental risk assessment and management over the past two decades.
• Presentations at major scientific meetings (SETAC, EGU) highlight new theoretical developments and applications.
• BNs have been successfully adapted to established frameworks like adverse outcome pathways and relative risk models.

Conclusions:

• Bayesian network models inherently incorporate uncertainty and can integrate diverse data sources.
• Recent advancements include spatial and temporal BN modeling.
• The increased use of Bayesian network models is predicted to significantly improve environmental risk assessments.