Identifying the long-term effects of airborne pollutants requires the computation of the spatial and temporal variability of their concentration in air to estimate the exposure of the population. To estimate exposure levels of individuals in a breast cancer case control study nested in a national prospective cohort, we determine here the differential impact of a variety of cadmium and dioxin sources on urban air over a large urban area and over a period of almost 20years. To that end, we couple an emission model, to estimate dioxin and cadmium atmospheric annual releases, with an urban dispersion model in order to compute pollutant concentration fields at a fine temporal (1h) and spatial (25m) resolution. The reliability of the modelling chain is compared to two types of measurement: i) localized industrial emissions and ii) dioxin and cadmium air monitoring data (from 2007 to 2008), collected at a fixed station, placed in the city centre, as well as at three mobile short-term dioxin monitoring stations, located in the suburbs (the latter providing dioxin data, only). Comparisons between measured and estimated emissions show non-negligible difference, with a correlations for dioxin (rs=0.42) and cadmium (rs=0.41). Despite this, mean values between estimated emissions and emission measurements are close to each other, in particular for cadmium. Weekly average modelled concentrations show an overall good agreement with weekly average measured concentrations in spring and summer but are generally lower than monitored data in winter due to peak concentrations from diffuse sources representing an important proportion of emissions in 2007/2008. The model provides better results for cadmium than for dioxin. Despite the relevant errors in the model predictions, the model meets the validation criteria, defined by Chang and Hanna for an urban dispersion model. Simulation scenarios of air pollutant concentrations, reconstructed over the last 20years, show the effects of the variability of the pollutant sources over time with decreasing levels of dioxin and cadmium concentrations in air. This is primarily due to the reduction in localized industrial releases, which results in a general trend of homogenization of the exposure of the population. The model further allows us to dissociate the contribution of different types of pollutant sources on the population exposure. The impact on local concentrations due to industrial emissions, which were originally responsible for the major impact on air quality, is shown to drop over the years by 99% and 92% for dioxin and cadmium, respectively. Today, the major contributions are due to diffuse miscellaneous sources in the case of dioxin and to traffic-related emissions for cadmium. Average modelled concentrations at the study subjects' residential locations range from 10.2 to 82.1fg-TEQ/m3 for dioxin and 0.10 to 1.6ng/m3 for cadmium and are comparable with data from the literature. The study results will be essential to increase the accuracy of the assessment of long-term airborne dioxin and cadmium exposure and improve the results of epidemiological studies.
Copyright © 2017 Elsevier Ltd. All rights reserved.