Scientific uncertainties in atmospheric mercury models II: Sensitivity analysis in the CONUS domain

In this study, we present the response of model results to different scientific treatments in an effort to quantify the uncertainties caused by the incomplete understanding of mercury science and by model assumptions in atmospheric mercury models. Two sets of sensitivity simulations were performed to assess the uncertainties using modified versions of CMAQ-Hg in a 36-km Continental United States domain. From Set 1 Experiments, it is found that the simulated mercury dry deposition is most sensitive to the gaseous elemental mercury (GEM) oxidation product assignment, and to the implemented dry deposition scheme for GEM and reactive gaseous mercury (RGM). The simulated wet deposition is sensitive to the aqueous Hg(II) sorption scheme, and to the GEM oxidation product assignment. The inclusion of natural mercury emission causes a small increase in GEM concentration but has little impact on deposition. From Set 2 Experiments, it is found that both dry and wet depositions are sensitive to mercury chemistry. Change in model mercury chemistry has a greater impact on simulated wet deposition than on dry deposition. The kinetic uncertainty of GEM oxidation by O3 and mechanistic uncertainty of Hg(II) reduction by aqueous HO2 pose the greatest impact. Using the upper-limit kinetics of GEM–O3 reaction or eliminating aqueous Hg(II)–HO2 reaction results in unreasonably high deposition and depletion of gaseous mercury in the domain. Removing GEM–OH reaction is not sufficient to balance the excessive mercury removal caused by eliminating the HO2 mechanism. Field measurements of mercury dry deposition, better quantification of mercury air-surface exchange and further investigation of mercury redox chemistry are needed for reducing model uncertainties and for improving the performance of atmospheric mercury models. r 2007 Elsevier Ltd. All rights reserved.