At polar sunrise, gaseous elemental mercury (GEM) undergoes an exceptional dynamic exchange in the air and at the snow surface during which GEM can be rapidly removed from the atmosphere (the so-called atmospheric mercury depletion events (AMDEs)) as well as re-emitted from the snow within a few hours to days in the Polar Regions. Although high concentrations of total mercury in snow following AMDEs is well documented, there is very little data available on the redox transformation processes of mercury in the snow and the fluxes of mercury at the air/snow interface. Therefore, the net gain of mercury in the Polar Regions as a result of AMDEs is still an open question. We developed a new version of the global mercury model, GRAHM, which includes for the first time bidirectional surface exchange of GEM in Polar Regions in spring and summer by developing schemes for mercury halogen oxidation, deposition, and re-emission. Also for the first time, GOME satellite data-derived boundary layer concentrations of BrO have been used in a global mercury model for representation of halogen mercury chemistry. Comparison of model simulated and measured atmospheric concentrations of GEM at Alert, Canada, for 3 years (2002-2004) shows the model's capability in simulating the rapid cycling of mercury during and after AMDEs. Brooks et al. (1) measured mercury deposition, reemission, and net surface gain fluxes of mercury at Barrow, AK, during an intensive measurement campaign for a 2 week period in spring (March 25 to April 7, 2003). They reported 1.7, 1.0 +/- 0.2, and 0.7 +/- 0.2 microg m(-2) deposition, re-emission, and net surface gain, respectively. Using the optimal configuration of the model, we estimated 1.8 microg m(-2) deposition, 1.0 microg m(-2) re-emission, and 0.8 microg m(-2) net surface gain of mercury for the same time period at Barrow. The estimated net annual accumulation of mercury within the Arctic Circle north of 66.5 degrees is approximately 174 t with +/-7 t of interannual variability for 2002-2004 using the optimal configuration. We estimated the uncertainty of the model results to the Hg/Br reaction rate coefficient to be approximately 6%. Springtime is clearly demonstrated as the most active period of mercury exchanges and net surface gain (approximately 46% of annual accumulation) in the Arctic.