Mercury (Hg) is a critical environmental concern. Although an important component of its biogeochemical cycle, large uncertainties still exist in the estimates of surface fluxes of mercury. Three projects presented in this thesis improve our understanding of mercury surface fluxes at different spatial scales by combining atmospheric observations and models. First, a global scale inverse model study uses observations at multiple ground-based stations and simulations from a three-dimensional chemical transport model (GEOS-Chem) to obtain a total mercury emission of about 5.8 Gg yr⁰ (gaseous elemental mercury). The optimized Asian anthropogenic emissions (0.7-1.8 Gg yr⁰. The inversion also suggests that the legacy mercury releases tend to reside in the terrestrial system rather than in the ocean. Second, the comparison of nested grid GEOS-Chem model simulations with aircraft observations support results from the global inversion, and further suggests that the Northwest Atlantic Ocean is a net source of Hg⁰, with high evasion fluxes in summer (related to the high precipitation rates and deposition fluxes of oxidized mercury), whereas the terrestrial ecosystem in the eastern United States is likely a net sink of Hg⁰ during summer. Third, a one-dimensional chemical transport model is built and used to simulate the mercury diurnal variabilities observed at Dome Concordia on the Antarctic plateau. The model simulation best reproducing the Hg⁰ observations shows that in summer mercury is rapidly cycled between the shallow atmospheric boundary layer and the surface snowpack. A two-step bromine initiated scheme oxidizes Hg⁰. Oxidized mercury is deposited, photoreduced in the surface snow, and reemitted as Hg⁰ back into the atmosphere.