Shanna Shaked

Assessing regional intake fractions in North America

This paper develops the IMPACT North America model, a spatially resolved multimedia, multi-pathway, fate, exposure and effect model that includes indoor and urban compartments. IMPACT North America allows geographic differentiation of population exposure of toxic emissions for comparative risk assessment and life cycle impact assessment within U.S. and Canada. It looks at air, water, soil, sediment and vegetation media, and divides North America into several hundred zones. It is nested within a single world box to account for emissions leaving North America. It is a multi-scale model, covering three different spatial scales--indoor, urban and regional--in all zones in North America. Model results are evaluated against monitored emissions and concentrations of benzo(a)pyrene, 2,3,7,8-TCDD and mercury. Most of the chemical concentrations predicted by the model fall within two orders of magnitude of the monitored data. The model shows that urban intake fractions are one order of magnitude higher than rural intake fractions. The model application and importance is demonstrated by a case study on spatially-distributed emissions over the life cycle of diesel fuel. Depending on population densities and agricultural intensities, intake fractions can vary by eight orders of magnitudes, and even limited indoor emissions can lead to intakes comparable to those from outdoor emissions. To accurately assess these variations in intake fraction, we require the essential three original features described in the present paper: i) inclusion of the continental model within a world box for persistent pollutants, ii) addition of an urban box for short- and medium-lived substances (for grid size larger than 100 km), and iii) assess indoor emissions. This model can therefore be used to screen chemicals and assess regionalized intake fractions within North America for population-based human exposure assessment, life cycle impact assessment, and comparative risk assessment. The model can be downloaded at http://www.impactmodeling.org.

Spatial analysis of toxic emissions in LCA: A sub-continental nested USEtox model with freshwater archetypes

This paper develops continent-specific factors for the USEtox model and analyses the accuracy of different model architectures, spatial scales and archetypes in evaluating toxic impacts, with a focus on freshwater pathways. Inter-continental variation is analysed by comparing chemical fate and intake fractions between sub-continental zones of two life cycle impact assessment models: (1) the nested USEtox model parameterized with sub-continental zones and (2) the spatially differentiated IMPACTWorld model with 17 interconnected sub-continental regions. Substance residence time in water varies by up to two orders of magnitude among the 17 zones assessed with IMPACTWorld and USEtox, and intake fraction varies by up to three orders of magnitude. Despite this variation, the nested USEtox model succeeds in mimicking the results of the spatially differentiated model, with the exception of very persistent volatile pollutants that can be transported to polar regions. Intra-continental variation is analysed by comparing fate and intake fractions modelled with the a-spatial (one box) IMPACT Europe continental model vs. the spatially differentiated version of the same model. Results show that the one box model might overestimate chemical fate and characterisation factors for freshwater eco-toxicity of persistent pollutants by up to three orders of magnitude for point source emissions. Subdividing Europe into three archetypes, based on freshwater residence time (how long it takes water to reach the sea), improves the prediction of fate and intake fractions for point source emissions, bringing them within a factor five compared to the spatial model. We demonstrated that a sub-continental nested model such as USEtox, with continent-specific parameterization complemented with freshwater archetypes, can thus represent inter- and intra-continental spatial variations, whilst minimizing model complexity.