Simplifying complex models : Application of modeling tools in exposure assessment of organic pollutants

Sammanfattning: Thousands of chemicals are used in society, but the exposure to humans and other organisms has been measured only for a small number of compounds. Modeling tools constitute low-cost and effective alternatives to measurements for the assessment of exposure. In this thesis, the prerequisites for the application of modeling tools in environmental exposure assessment of organic pollutants were explored. The first aspect discussed was emission estimates, which are crucial for any quantitative modeling study. In Paper I, the only currently existing high throughput tool for ranking emissions was evaluated and found to have limited predictive power, suggesting that further research is necessary to enable exposure based screening. The second aspect was the model’s treatment of dynamic processes. A strategy for deciding on the temporal resolution required for the description of dynamic processes was proposed in Paper II, which involved identification of major transport routes and time to approach steady state. The third aspect was prediction of partition coefficients for use in bioaccumulation models. The traditional single parameter regressions (spLFER) employed for this purpose were compared to the more mechanistically sound ppLFER equations in Paper III. The two methods had a similar accuracy when compared to measured data, implying that the choice of approach should be based on other factors than methodology (e.g. availability of accurate input data). The fourth aspect was the influence of system characteristics on human exposure. The susceptibilities of several ecosystems with diverging characteristics to exposure to organic chemicals were compared in Paper IV. The strong variation in exposure susceptibilities found suggests that the choice of model system can be relevant for exposure assessment and that models may have to be tailored to the ecosystem of interest. In the broader context, this work provides methodologies for handling model complexity in exposure modeling.