There are large uncertainties in the estimation of sources of reactive mercury in current photochemical models, with many models suggesting that three quarters or more of reactive mercury in the atmosphere is due to secondary oxidation of the gaseous elemental mercury present in the global background. An inverse model is used to estimate the sources of mercury at an urban site in Wisconsin based on a year long time series of hourly measurements. The model combines high resolution backtrajectories simulated with WRF and WRF-FLEXPART, forward simulations of passive tracers using a transport model (CAMx), hourly time series of atmospheric pollutant concentrations and time series from a chemical box model for oxidation of elemental mercury by ozone, the hydroxyl radical and bromine. The hybrid formulation provides an estimate of the mercury concentrations on a polar grid surrounding the site along with emission scaling factors for emissions from forest fires and lake surfaces. In addition, the model estimates the impact of oxidation of gaseous elemental mercury from the three pathways. The inverse model identified direct emissions of reactive mercury, defined as the sum of reactive gaseous mercury and particle-bound mercury, that are associated with regional sources and with forest fires. The results suggest that oxidation by ozone is adequately characterized in existing chemical mechanisms, but that oxidation by the hydroxyl radical and by bromine may be underestimated. The results suggest that between 50 and 70 percent of the reactive mercury at the measurement site is due to direct emissions and hence suggest the importance of developing emission inventories for reactive mercury species. © 2013 Elsevier Ltd.