Sensitivity model study of regional mercury dispersion in the atmosphere

Atmospheric deposition is the most important pathway by which Hg reaches marine ecosystems, where it can be methylated and enter the base of food chain. The deposition, transport and chemical interactions of atmospheric Hg have been simulated over Europe for the year 2013 in the framework of the Global Mercury Observation System (GMOS) project, performing 14 different model sensitivity tests using two high-resolution three-dimensional chemical transport models (CTMs), varying the anthropogenic emission datasets, atmospheric Br input fields, Hg oxidation schemes and modelling domain boundary condition input. Sensitivity simulation results were compared with observations from 28 monitoring sites in Europe to assess model performance and particularly to analyse the influence of anthropogenic emission speciation and the Hg(g) atmospheric oxidation mechanism. The contribution of anthropogenic Hg emissions, their speciation and vertical distribution are crucial to the simulated concentration and deposition fields, as is also the choice of Hg(g) oxidation pathway. The areas most sensitive to changes in Hg emission speciation and the emission vertical distribution are those near major sources, but also the Aegean and the Black seas, the English Channel, the Skagerrak Strait and the northern German coast. Considerable influence was found also evident over the Mediterranean, the North Sea and Baltic Sea and some influence is seen over continental Europe, while this difference is least over the north-western part of the modelling domain, which includes the Norwegian Sea and Iceland. The Br oxidation pathway produces more HgII (g) in the lower model levels, but overall wet deposition is lower in comparison to the simulations which employ an O3 / OH oxidation mechanism. The necessity to perform continuous measurements of speciated Hg and to investigate the local impacts of Hg emissions and deposition, as well as interactions dependent on land use and vegetation, forests, peat bogs, etc., is highlighted in this study.

-Every option for emissions inventories, chemistry schemes, etc. are presented as equally valid but it would be helpful to know what discussion has already been performed of these inventories/chemical mechanisms etc. to indicate the strengths and weaknesses of each and how this paper's results fit in. I think it is misleading to present Br and OH/O3 oxidation mechanisms on equal footing, for example, but then the paper shows that Br oxidation alone best represents TGM results which adds further support to what previous studies have suggested.
-I think the paper would benefit from more clearly outlining the purpose of each of the sensitivity studies. It would also be beneficial to have the explanation of most sensitivity studies be more centralized -since some of the sensitivity studies are untrue hypotheticals (e.g., NOCHEM and NOANT) while others are testing different hypothesized physical processes (e.g., BRCHEM1 vs. BRCHEM2), but both types are valuable. There is some of this scattered throughout the results section, but I think it can be more explicit; e.g., 'the ANTSPEC experiment, which assumes all anthropogenic emissions are as GEM, would represent a lower bound on deposition from local anthropogenic sources and an upper bound on long-range transport of anthropogenic emissions because GEM has a much longer lifetime against deposition than RGM/PBM' or something to that effect.
-Some background information is missing -e.g., the lifetime of GEM vs. RGM/PBM is not explained in the paper, or the sinks of GEM vs. RGM/PBM (dry deposition vs. wet deposition affinities), making the results from the ANTSPEC simulations and others less easily understood. Some aspects of atmospheric mercury are missing from the discussion (e.g., reduction of Hg(II), uncertainty in present oxidized Hg measurement capabilities that would be relevant for the 2nd to last sentence of the paper).

Specific comments:
Section 2.1: There are a lot of details given here about the models but it is not clear how the model representations of Hg are different or how their differences would affect Hg C2 results. It would be helpful to rewrite this a bit and make it more explicit whether/how the differences would affect Hg (e.g., if CMAQ is offline while WRF is online meteorology, the transport of Hg will be different to some degree between the models.) Page 4 line 2: it would be helpful to explain briefly here whether there are multiple heights in the EDGAR inventory, as it is written it is unclear what the difference is and this comes up later in the results section. Section 2.3: It is unclear from the way the paper is written what the base model oxidation chemistry schemes include -is it OH, O3 and Br oxidation at the same time in both CMAQ and WRF-Chem Hg? This should be made clear.
Page 4 Lines 23-25: some of this would be helpful to mention in the model description section 2.1. Table 3: add a total deposition column, just makes it easier to follow Figure 3: It is hard to compare the results between CMAQ and WRF-Chem Hg because the same model setups do not line up. I think it is worth reordering the individual panels to lining up the ANTSPEC and ANTSPEC_C vertically, or NOCHEM with NOCHEM_C below it, etc. to make more obvious the comparisons made in Section 3 of the paper. Figure 4: This is total dry deposition of Hg(II)+Hg(0), correct? Just making sure since wet deposition is of course just Hg(II). I think this is worth mentioning as it also helps explain the changes in dry vs. wet deposition seen in the sensitivity studies. Figure 5a: there are two legend labels with "BASE". I assume the green square is BASE2? Also, I would explain that the order of the sites is by the magnitude of the mod/obs ratio of the BASE simulation as it is not immediately obvious.
Page 5 Lines 20-23. It is counterintuitive to me how an overestimate relative to observations for the ANTSPEC sensitivity study where all anthropogenic emissions of RGM & PBM are as GEM means long-range transport is less important than regional emissions. I think more explanation would be helpful -so when PBM and RGM are C3 emitted normally as in the BASE simulation, the model is no longer overestimating the observations because the Hg(II) is deposited fast enough before it reaches the CZ03 site? I suppose part of this is defining what is local vs. "regional" vs. long-range more clearly.
Page 5 last paragraph (beginning line 29): do you have any hypotheses why TGM at ES08 is so low relative to the models?
Page 6 Lines 1-2: The discussion of OHCHEM and NOANT experiments comes as a surprise as it is not in Figures 5, 6, or 7; more connection to the rest of the section is needed.
Page 7 Lines 15-16: it is also not immediately obvious why dry deposition decreases so much more than wet deposition, since GEM is not wet deposited. Is it because even though there is higher GEM in the ANTSPEC experiment, it is dry deposited so much more slowly than the RGM/PBM species, or is it something about how RGM and PBM contribute different amounts to wet vs. dry dep of the oxidized species? Setting up the background on this in the intro or methods would be helpful.
Page 7 lines 21-23: given that a no anthropogenic emissions scenario is currently untrue, I think something more insightful can be said about the results -e.g., something about how anthro. emissions contribute to 2/3 of total deposition (not counting the fact that "natural" emissions from soil/ocean as they are tuned in models also implicitly are impacted by legacy anthropogenic sources), or how a hypothetical policy scenario of shutting off emissions could have huge local benefits? Section 3.3 paragraph 2 (pg 7 lines 30-32): I would emphasize that this shows a significant proportion (exact percent varying on the model) of total Hg deposition to ecosystems is coming from the oxidation of GEM which can be transported from far distances as opposed to the Hg(II) locally emitted. This is an interesting result with policy implications and could be highlighted more.

C4
Page 8 lines 3-4: "A number of studies have shown the importance of O3, and the OH radical, and also reactive halogen compounds. . ..": I understand that there have been review papers discussing the intricacies of this and you don't want to repeat that here, but it is overly simple to group all three oxidants together and not mention that studies have found that the homogeneous gas-phase oxidation of Hg(0) by O3 and OH are thermodynamically and/or kinetically impossible (e.g., Hynes et al. (2009), Goodsite et al. (2004, Calvert and Lindberg, 2005). I think it is still interesting to compare the results from the three species, but it needs to be introduced with a bit more nuance. Moreover, as Theodore Dibble posted in his comment, there are additional HgBr+X second-step oxidation reactions that can greatly increase the total Hg(II) production and deposition through the Br-initiated pathway. Somewhere in the paper there should be a discussion of how this would affect the results presented -e.g., Hg deposition in the BRCHEM1 and 2 sensitivity simulations would be increased and TGM would be decreased.
Page 9 Lines 1-8: it is hard to understand the differences in the two Br concentration fields from the description given -would it be possible to show (perhaps in the supplemental) a difference plot of the Br concentrations over the Europe domain? (e.g., zonal mean latitude on x-axis vs. altitude on y-axis or something like that). Most GEM oxidation is not occurring in the PBL but in the free and upper troposphere (because of Br distribution and the temperature dependence of the oxidation reactions), so it is not surprising that the huge differences in Br in the PBL between the two Br fields doesn't impact on Hg(II) deposition; I am more interested in the differences at higher altitudes.
Page 9 Lines 13-16: I think it is essential to connect this to available observations of Br. The Shah et al. study tripled bromine concentrations of the GEOS-Chem Parrella et al. 2012 model which was consistent with observations of BrO during the NOMADSS field campaign (Gratz et al., 2015). Parrella et al. (2012) showed previously that BrO was underestimated in GEOS-Chem by 30% in the global mean against satellite observations. So it is not just a model exercise but shows that higher Br in GEOS-Chem (closer C5