Mercury deposition in snow near an industrial emission source in the western U.S. and comparison to ISC3 model predictions

Mercury (total and methyl) was evaluated in snow samples collected near a major mercury emission source on the Idaho National Engineering and Environmental Laboratory (INEEL) insoutheastern Idaho and 160 km downwind in Teton Range in westernWyoming. The sampling was done to assess near-field (<12 km)deposition rates around the source, compare them to those measured in a relatively remote, pristine downwind location, andto use the measurements to develop improved, site-specific modelinput parameters for precipitation scavenging coefficient and thefraction of Hg emissions deposited locally. Measured snow waterconcentrations (ng L^-1) were converted to deposition (ugm^-2) using the sample location snow water equivalent. Thedeposition was then compared to that predicted using the ISC3 airdispersion/deposition model which was run with a range ofparticle and vapor scavenging coefficient input values. Acceptedmodel statistical performance measures (fractional bias andnormalized mean square error) were calculated for the differentmodeling runs, and the best model performance was selected. Measured concentrations close to the source (average = 5.3 ngL^-1) were about twice those measured in the Teton Range(average = 2.7 ng L^-1) which were within the expected rangeof values for remote background areas. For most of the samplinglocations, the ISC3 model predicted within a factor of two of theobserved deposition. The best modeling performance was obtainedusing a scavenging coefficient value for 0.25 μm diameterparticulate and the assumption that all of the mercury isreactive Hg(II) and subject to local deposition. A 0.1 μm particle assumption provided conservative overprediction of thedata, while a vapor assumption resulted in highly variable predictions. Partitioning a fraction of the Hg emissions to elemental Hg(0) (a U.S. EPA default assumption for combustion facility risk assessments) would have underpredicted the observed fallout.