Local to Global Multi-Scale Multimedia Modeling of Chemical Fate and Population Exposure

Abstract

To assess environmental and human exposure to chemical emissions, two types of approaches are available: 1. intermediate- to high-resolution, substance/location-specific analyses, and 2. lower resolution, less specific analyses aiming for broad coverage. The first category is time/resource intensive, which limits its utility, while the second is less accurate but allows for evaluation of large numbers of substances/situations. None is well suited for analyzing local to global population exposure. We need a multi-scale approach of intermediate complexity that bridges the advantages of both approaches: high resolution when relevant, the ability to evaluate large numbers of substances, and a level of accuracy that is “useful” (for decision-makers). This thesis aims to 1. develop a multi-scale, multimedia fate and transport, and multi-pathway population exposure modeling framework, 2. evaluate it using large-scale inventories of emissions and measured environmental concentrations, 3. evaluate local to global population exposure associated with large sets of point sources covering a wide variety of local contexts (e.g. up/down-wind/stream from large populations, important water bodies or agricultural resources), and 4. simulate a large national inventory of emissions and perform multi-media source apportionment. Coupling a geographic information system and a computation engine, we develop the Pangea framework, which offers a unique ability to discretize the globe using three-dimensional multi-scale grids, to overlay Eulerian fate and transport multimedia models, and to compute multi-pathway population exposure. We first apply this framework to predict the fate and transport of home and personal care chemicals in all of Asia. This study provides a large-scale high-resolution spatial inventory of emissions and a large data set of ~1,600 monitoring values. We compare predicted environmental concentrations (PECs) and measurements and find good agreement for the long-lived triclosan in fresh water (Pearson r=0.82), moderate agreement for shorter-lived substances, and a large discrepancy specifically for parabens in sediments. This study highlights the limitation of the present underlying gridded hydrological data set (WWDRII) when comparison with measurements at monitoring sites is required, which prompts the evaluation of a finer, catchment-based hydrological data set (HydroBASINS). We then focus on human exposure and the evolution of the population intake fraction with the distance from the source. We simulate emissions from 126 point sources (stacks of solid waste treatment plants) in France, and compute radial distributions of population intake fractions through inhalation and ingestion. We determine that a substantial fraction of emissions may be taken in by the population farther than 100 km away from point sources (78.5% of the inhaled benzene and 54.1% of the ingested 2,3,7,8-TCDD). We demonstrate the feasibility of simulating large numbers of emission scenarios by extending the study to 10,000 point sources. We finally extend the previous emitter-oriented studies with receptor-oriented analyses (source apportionment). We simulate 43 substances emitted from 4,101 point sources defined by the Australian National Pollutant Inventory for 2014-2015. We compute population exposure and severity (DALY). Formaldehyde, benzene, and styrene are the three top contributors in terms of DALYs. We demonstrate the technical feasibility of multimedia, large-scale source apportionment. This research opens new perspectives in spatial, local to large-scale fate and exposure modeling. The flexibility of Pangea allows to build project-specific model geometries and to re-analyze projects following the evolution of data availability. Major limitations come from the underlying first-order fate and transport models and from a limited availability of global spatial data sets.