This work deals with current problematic issues of environment-health interactions. This thesis describes the origin of mining and smelting polymetallic pollution (Ag, Au, Cd, Cu, Pb, Sb, Sn, Zn) and its impact on environment and health in the mining city of Oruro, Bolivia. Samples of tap water, the hydrographic network, soils, dusts, aerosols (PM2,5 et PM10) and bioindicators of atmospheric contamination (Tillandsias capillaris) were geochemically (ICP-MS) and mineralogically (DRX, SEM) characterized. All the compartments of the urban area studied ("water ", "soil" and "atmosphere") were contaminated by trace elements, especially near smelters and mines (reaching 16000 µg/g Pb in dust and 200 ng/m3 As in PM10 close to the smelting area). Statistical studies and spatial interpolations helped in the identification of atmospheric contamination sources and their scope. Soil and dust studies also indicated other more local sources, such as old ore stockage areas. Granulometric separation of dusts showed that trace elements were more highly concentrated in fine fractions, but not necessarily in the clay (< 2µm), which is generally considered to be the most contaminated fraction. Oral bioaccessibility of dust fractions (UBM method), is higher overall in the smelting area than in the mining one. Bioaccessibility greatly varies depending on the granulometric fraction or elements considered (up to 90% for As and less than 5% for Sn and Sb). Exposure to trace elements in children depending on their place of residence was calculated considering dust and aerosol inhalation and ingestion. Taking into account granulometry and bioaccessibility can considerably change the absorbed dose calculated. Estimated exposure and trace element content in children's hair previously measured prove to be correlated. In this study, Pb showed a particular behaviour, with soil Pb concentrations were very high when compared to weak aerosol concentrations. Child exposure is thus greatly linked to type and location of the child's activities, which can explain the variability observed in a previous epidemiological study
