Source apportionment studies of particulate matter (PM) link chemical composition to emission sources, while health risk analyses link health outcomes and chemical composition. There are limited studies to link emission sources and health risks from ambient measurements. We show such an attempt for particulate trace elements. Elements in PM2.5 were measured in wintertime Beijing, and the total concentrations of 14 trace elements were 1.3-7.3 times higher during severe pollution days than during low pollution days. Fe, Zn, and Pb were the most abundant elements independent of the PM pollution levels. Chemical fractionation shows that Pb, Mn, Cd, As, Sr, Co, V, Cu, and Ni were present mainly in the bioavailable fraction. Positive matrix factorization was used to resolve the sources of particulate trace elements into dust, oil combustion, coal combustion, and traffic-related emissions. Traffic-related emission contributed 65% of total mass of the measured elements during low pollution days. However, coal combustion dominated (58%) during severe pollution days. By combining element-specific health risk analyses and source apportionment results, we conclude that traffic-related emission dominates the health risks by particulate trace elements during low pollution days, while coal combustion becomes equally or even more important during moderate and severe pollution days.