Elsevier

Atmospheric Environment

Volume 42, Issue 8, March 2008, Pages 1828-1845
Atmospheric Environment

Scientific uncertainties in atmospheric mercury models III: Boundary and initial conditions, model grid resolution, and Hg(II) reduction mechanism

https://doi.org/10.1016/j.atmosenv.2007.11.020Get rights and content

Abstract

In this study, the model response in terms of simulated mercury concentration and deposition to boundary condition (BC), initial condition (IC), model grid resolution (12 km versus 36 km), and two alternative Hg(II) reduction mechanisms, was investigated. The model response to the change of gaseous elemental mercury (GEM) concentration from 0 to 2 ng m−3 in IC/BC is found to be very linear (r2>0.99) based on the results of sensitivity simulations in July 2001. An increase of 1 ng m−3 of GEM in BC resulted in an increase of 0.81 ng m−3 in the monthly average of total mercury concentration, and 1270 ng m−2 in the monthly total deposition. IC has similar but weaker effects compared to those of BC. An increase of 1 ng m−3 of GEM in IC resulted in an increase of 0.14 ng m−3 in the monthly average of total mercury concentration, and 250 ng m−2 in the monthly total deposition. Varying reactive gaseous mercury (RGM) or particulate mercury (PHg) in BC/IC has much less significant impact. Simulation results at different grid resolutions show good agreement (slope=0.950–1.026, r=0.816–0.973) in mercury concentration, dry deposition, and total deposition. The agreement in wet deposition is somewhat weaker (slope=0.770–0.794, r=0.685–0.892) due to the difference in emission dilution and simulated precipitation that subsequently change reaction rates in the aqueous phase. Replacing the aqueous Hg(II)-HO2 reduction by either RGM reduction by CO (5×10−18 cm3 molecule−1 s−1) or photoreduction of RGM (1×10−5 s−1) gives significantly better model agreement with the wet deposition measured by Mercury Deposition Network (MDN). Possible ranges of the reduction rates are estimated based on model sensitivity results. The kinetic estimate requires further verification by laboratory studies.

Introduction

The mercury model of Community Multiscale Air Quality modeling system (CMAQ-Hg) has been used extensively as a modeling tool for atmospheric mercury studies in North America (Bullock and Brehme, 2002; Gbor et al., 2006, Gbor et al., 2007; Lin and Tao, 2003; Sillman et al., 2007), and for intercontinental transport (Lin et al., 2006b). Recently, Lin et al., 2006a, Lin et al., 2007 assessed quantitatively the model uncertainties in mercury emission processing, gaseous and aqueous chemistry, aqueous mercury speciation, dry deposition, and wet deposition through sensitivity simulations using CMAQ-Hg. In this study, the impacts of BC, IC, model grid resolution, and potentially missing chemical reactions in the models on simulated mercury concentration and deposition are investigated systematically.

Model sensitivity to BC has been studied by Pai et al. (1999) using a 3-D model, which reported a negligible effect on simulated wet deposition by varying RGM from 0 to 160 pg m−3. However, since mercury deposition is primarily driven by chemistry (Lin and Pehkonen, 1999; Lin et al., 2007), it is necessary to understand the impact of GEM in BC, which is investigated in this study. Model sensitivity to IC is also included because of its importance to regional modeling. Special attention was devoted to understanding the impact caused by IC in terms of its magnitude and time period corresponding to a domain size.

On the effect of grid resolution, Pai et al. (2000) noted that short-term wet deposition results (daily or weekly variation) are more sensitive to grid resolution as opposed to long-term simulations (seasonal and annual averages). They also reported a two-fold increase in total concentration and dry deposition near major point sources when increasing the spatial resolution from 100 to 20 km. Seigneur et al. (2003) reported a poor correlation between the simulated wet deposition at a 100 km and at a 20 km resolution. The use of the finer resolution improved the model performance upwind of major emission sources, but the performance is compromised downwind (overestimation of wet deposition). They hypothesized that some key mercury chemical transformations are likely missing in the model (Seigneur et al., 2003), which leads to a greater degree of oxidation of the emitted mercury at the finer resolution. In this study, the effect of grid resolution is further investigated in two domains using 36 and 12 km. The results at 12 km resolution have not been reported in mercury modeling studies previously.

One of the uncertainties in mercury models is the treatment of Hg(II) reduction mechanism (Lin et al., 2006a). The gas-phase reduction of RGM to GEM has been reported in power plant plumes using measured concentration downwind and model estimation (Lohman et al., 2006). They proposed two possible reduction pathways of RGM: a pseudo-first-order decay with a rate of 0.3 h−1 and a second-order reduction by SO2 with a rate of 0.007 ppb−1 h−1 (8×10−18 cm3 molecule−1 s−1). Seigneur et al. (2006) adopted those two pathways in a global mercury model. They tested a first-order reduction rate of 0.15 h−1 and found that it caused unreasonably high GEM concentrations, i.e., approximately four times greater than observations. They also replaced the aqueous Hg(II) reduction HO2 (Pehkonen and Lin, 1998; Gårdfeldt and Jonsson, 2003) with the reduction of RGM by SO2 using a reaction rate of 8×10−18 cm3 molecule−1 s−1 and found that this mechanism alone was not sufficient for producing realistic GEM concentration in the model. Other alternative reduction pathways have not been tested in regional models.

The objective of this study is to quantitatively assess the uncertainties of atmospheric mercury models caused by different grid resolutions, BC, IC, and chemical reduction mechanisms. One approach to address the uncertainty issue is to perform sensitivity simulations for quantifying the variation caused by different science implementations in models. In this study, sensitivity simulations were performed using identical emission inventory for the 12 and 36 km domains to demonstrate the effect of model grid resolution. The effect of varying GEM, RGM and PHg concentrations in BC and IC on the simulated concentration and deposition is shown. Two alternative gas-phase reduction mechanisms of RGM replacing the aqueous Hg(II) reduction by HO2 are also tested using CMAQ-Hg. The goal is to verify whether or not implementing new reduction mechanisms would improve model performance in regional-scale modeling.

Section snippets

Model domains and grid resolutions

Two model domains were used in this study: (1) the continental United States (CONUS) domain, and (2) the Northeast, Midwest, and South (NMAS) regional US domain, which is nested inside the CONUS domain. Both domains use a Lambert conformal projection centered at 40°N and 97°W, as shown in Fig. 1. The CONUS domain has 148×112 horizontal cells and a resolution of 36 km with 14 vertical layers (see Table 1). The NMAS has 279×240 horizontal cells with a 12 km horizontal resolution and uses the same

Model grid resolution

The simulation results at 36 and 12 km grid resolutions in July 2001 are shown in Fig. 2. The spatial distribution at both resolutions is similar for the monthly average surface mercury concentration (Fig. 2a) and the monthly mercury dry deposition (Fig. 2b). The 12 km results give a smoother texture. The distribution of mercury wet deposition (Fig. 2c) and total (dry+wet) mercury deposition (Fig. 2d) from the two resolutions are slightly different, which is caused by the difference in emission

Conclusions

In this study, the model response to the boundary and initial conditions of mercury species, model grid resolution (12 km versus 36 km), and alternative Hg(II) reduction mechanisms, i.e., RGM reduction by CO and photoreduction, was investigated. Although model results using the two different resolutions are similar, using the finer spatial resolution better resolves the simulated concentration and deposition, especially near the major emission sources. The model response to the change of GEM

Acknowledgments

The authors would like to acknowledge the funding support from Texas Commission on Environmental Quality (TCEQ work order number: 64582-06-15), the USEPA Office of Air Quality Planning and Standards (RTI subcontract number: 6-321-0210288) and Texas Air Research Center (TARC project number: 077LUB0976). A portion of the research presented here was performed under the Memorandum of Understanding between the US Environmental Protection Agency (EPA) and the US Department of Commerce's National

References (37)

  • C.J. Lin et al.

    Scientific uncertainties in atmospheric mercury models II: sensitivity analysis in the CONUS Domain

    Atmospheric Environment

    (2007)
  • P. Pai et al.

    On artificial dilution of point source mercury emissions in a regional atmospheric model

    Science of the Total Environment

    (2000)
  • C. Seigneur et al.

    On the effect of spatial resolution on atmospheric mercury modeling

    Science of the Total Environment

    (2003)
  • C. Temme et al.

    Trend, seasonal and multivariate analysis study of total gaseous mercury data from the Canadian atmospheric mercury measurement network (CAMNet)

    Atmospheric Environment

    (2007)
  • C. Walcek et al.

    Preparation of mercury emissions inventory for eastern North America

    Environmental Pollution

    (2003)
  • M.L. Abbott et al.

    Atmospheric mercury near Salmon Falls Creek Reservoir in Southern Idaho

    Applied Geochemistry

    (2008)
  • Aldrich Chemical Company, 2006. Handbook of Fine Chemicals. Sigma-Aldrich Corporation, Milwaukee,...
  • A.O. Beckman et al.

    Microdetermination of carbon monoxide in air

    Analytical Chemistry

    (1948)
  • Cited by (0)

    1

    Now in Graeagle, CA 96103, USA.

    2

    In partnership with the US Environmental Protection Agency.

    View full text