Modeling and risk forecasting for engineered nanomaterials


We have proposed a process for risk forecasting that includes the following key features: 1) the ability to generate forecasts and associated levels of uncertainty for questions of immediate concern, 2) a consideration of all pertinent sources of nanomaterials, 3) an inclusive consideration of the impacts of activities stemming from nanomaterial use and production that extends beyond the boundaries of toxicology and include full life cycle impacts, 4) the ability to adapt and update risk forecasts as new information becomes available, 5) feedback to improve information gathering and, 6) feedback to improve nanomaterial design. A paper detailing this framework was published in the Proceedings of the French Academy of Science. The following section detail some of the progress made over the last year towards realizing such risk forecasting tools.

Estimating the Inventory of Nanomaterial Production

The first step to enable any other interpretation of exposure potential must be a quantification of potential release; therefore an inventory of U.S. production quantities was carried out for nano-Ag, nano-TiO¬2, nano-CeO2, fullerene C60, and carbon nanotubes (CNTs).  This work estimates ranges for the 2009 production in tons per year within the United States.  In addition to the ranges of values, the data quality is reported in the form of pie charts that indicate the source of production information or assumptions, to highlight the limitations of obtaining proprietary data that to date has no reporting requirements. The highest production level was estimated for nano-TiO2, followed by CNTs, nano-CeO2, C¬60, and finally nano-Ag. The resulting manuscript “Estimating production data for five engineered nanomaterials as a basis for exposure assessment” was published in ES&T. In addition to this work, we have completed an inventory for nano ZnO production.

Monte Carlo Simulations of nanomaterial removal in wastewater treatment

Products containing nanomaterials may enter municipal wastewater (WWTP). This subsystem was modeled as a complete mixed reactor based on key plant operation parameters, the quantity of U.S. production, the fraction released, and a partitioning coefficient describing the nanomaterial affinity for sludge vs. effluent.  Distributions were assigned to each input parameter based on literature values, the previously mentioned inventory research, and on lab experiments carried out to determine partitioning coefficient values.  Monte Carlo methods were applied to estimate probability functions for nanomaterial concentrations in the effluent of the treatment plant and in the sludge. Our simulations show that significant differences in the removal of silver nanoparticles (Ag NPs) can be expected based on the type of engineered coatings used to stabilize these materials in suspension. At current production estimates, 95% of the estimated effluent concentrations of the Ag NPs considered to be least well-removed by the average wastewater treatment plant, are calculated to fall below 0.12 mg/L, while 95% of the estimated sludge concentrations of Ag NPs with coatings that increase their likelihood of being present in biosolids, fall below 0.35 mg/L.

Modeling exposure from land applications of wastewater sludges

From the WWTP described above, estimates are obtained for the concentrations of  nanomaterials in digested in biosolids that have the potential for land application. The WWTP model has been refined to allow for changes in the distribution coefficient during treatment.  In a collaboration with RTI, we are porting models for land-transport of nanomateials that will use the WWTP model as input and evaluate long-term accumulation and runoff associated with biosolid applications to fields.

Mechanistic sub-models in forecasting nanomaterial exposure potential

Nanomaterial exposure assessment requires relationships between the amounts of nanomaterials entering various aquatic systems (wastewater treatment plants, mesocosms, actual wetlands, lakes, experimental systems, etc.) The objective of this project is to predict where nanoparticles go into the environment once they enter an aquatic compartment, and how fast they get there. This effort involves the develop of numerical code for describing the aggregation of nanoparticles  with each other (autoaggregation) aggregation with other suspended surfaces (bateria, algae, clays, etc.) referred to as heteroaggregation, and transport (including sedimentation).  This in turn requires the incorporation of significant detail regarding heteroaggregation and the hydrodynamic properties of the resulting aggregates (structure and drag coefficients). The core of our aggregation model performs a numerical integration of an extended form of the Smoluchowski equations, and follows the concentrations of purely natural aggregates, purely nano-aggregates, and mixed aggregates.

The model suggests that as long as there is a significant probability of attachment between nanoparticles and background particles, the distribution of natural particles would be almost entirely contaminated by the 25ppm Ag NP spike within a day, or even within seconds in the case of attachment efficiencies higher than ~10%. Upon validation, the model will allow to evaluate the extent to which a spike of nanoparticles will contaminate an existing distribution of natural aggregates, or the actual exposure of bacteria to toxic nanoparticles, for examples.

Bayesian network modeling

We have successfully created the structure for a Bayesian network model that links particle characteristics/behavior, with environmental conditions, potential exposure, hazard, and ultimately a measure of risk. The overall objective of this work has been to create a probabilistic model (Bayesian Network) to forecast the potential risks of nanomaterials to the environment, using Nano-Ag as an initial example.    This modeling project consists of four major phases as follows:  model structural development, parameterization, testing and analysis, and Bayesian updating.  The model structure and baseline parameterization using expert elicitation has been completed, resulting in the FINE (Forecasting the Implications of Nanomaterials in the Environment) Bayesian network, and a publication detailing the model development is in the submission phase. We have completed a project where the baseline FINE model from Phase I has been updated with literature data centered on the behavior and impacts of Nano-Ag in aquatic environments.  Subsequently, the model has been validated using the first round of the CEINT mesocom experiments, with an overall model accuracy of nearly 70%, meaning that by providing FINE with the environmental conditions and particle characteristics in the mesocosms, it accurately predicted the Nano-Ag concentration in the water column 70% of the time, which is a significant level of accuracy given the complexity of the model.  The newly updated and validated model was then used to conduct an exposure analysis for Nano-Ag in a real-world case study involving wastewater treatment and a North Carolina river basin. The results of this scenario analysis suggest that Nano-Ag risk may be highly localized and the exposure concentration will be highly dependent on localized environmental conditions, and may not pose a significant risk, even under extreme scenarios.