Modeling an Uncertain Risk: Using Belief Networks to Assess the Environmental Implications of NanoTechnology
Monday, March 15, 2010
Eric Money, Ph.D., Post-Doctoral Associate, CEINT/NSOE
Abstract: Understanding the potential environmental risks of nanomaterials requires extensive knowledge of particle characteristics and their transport, fate, uptake, and induced effects in environmental systems. These are complex mechanisms that are affected by existing environmental conditions and the specific characteristics of at risk populations, all of which are highly uncertain given our present scientific knowledge. Even though this uncertainty exists, decision-makers and other stakeholders (including the public) are increasingly aware of nanomaterials and are already suggesting regulatory and other measures to control their use and release into the environment.
Therefore, there is a need for a risk framework that can be informative in light of scientific uncertainty and which can be rigorously updated as our knowledge increases through the experimental work of CEINT and others. The framework presented here relies on the use of Bayesian Networks (BayesNets) to model the potential environmental implications of nanomaterials, with a current focus on nano-silver in aquatic environments. A Bayesnet is a powerful tool for formally integrating expert knowledge and experimental data into a probabilistic framework that can be updated with new information, thereby incorporating existing uncertainty and expressing the likelihood of possible effects given the current state of knowledge.
Here, we briefly introduce the concepts of BayesNets and show how we are developing a complex model that links all of the mechanisms important for assessing the ecological outcomes of nanomaterial presence in the environment. In addition, a complete sub-model related to particle aggregation and particle fate is discussed in detail to show the diagnostic and predictive capabilities of the Nano-BayesNet. We conclude with a discussion of some of the challenges in applying this approach and how we can use these models to not only assess risk, but to also identify the sources of uncertainty and the information we need to reduce that uncertainty.