April 17, 2019

Administrator Andrew Wheeler US Environmental Protection Agency 1200 Pennsylvania Avenue, NW Washington, D.C. 20460

Docket No. EPA-HQ-OAR-2018-0794

Subject: Comments on “National Emission Standard for Hazardous Air Pollutants: Coal- and Oil-Fired Electric Utility Steam Generating Units—Reconsideration of Supplemental Finding and Residual Risk and Technology Review”

As academic researchers studying fate, transport, and impacts, we appreciate the opportunity to comment on the Reconsideration of Supplemental Finding and Residual Risk Technology Review. We write to encourage the Environmental Protection Agency to withdraw its Proposed Finding and Residual Risk Review, given limitations in its assessment methodologies and developments in scientific research on mercury processes and impacts since 2011. This assessment is based on our own research (Giang and Selin 2016; Perlinger et al. 2018; attached; Angot et al. 2018), as well as our review of relevant scientific literature. In the below, we focus on mercury as one important component of Hazardous Air Pollutants (HAPs). Our comments are in response to solicitations for comments C-2 and C-24.

(C-2) The Regulatory Impact Analysis (RIA) from 2011 quantifies a very limited range of mercury- related benefits of regulation. Recent evidence indicates that benefits that were unquantified in the RIA may be substantial, and larger than the subset that were monetized.

In Giang and Selin (2016), we estimated the benefits of the Mercury and Air Toxics Standards (MATS) policy for US populations. Annualized, our estimates out to 2050 are $3.7 billion/year, using a similar methodology to the RIA. Although methodological differences between our analysis and the RIA prevent direct comparison of numerical results, our research suggests that including a larger subset of health endpoints (IQ and heart attacks) and affected populations (consumers of self-caught freshwater fish and consumers of commercial marine and estuarine fish in the US market) may lead to mercury-related benefits estimates that are orders of magnitude larger than those reported in the RIA. Considering these additional mercury-related benefits could result in benefits that are of comparable magnitude to estimated costs (Rice et al. 2010; Grandjean and Bellanger 2017).

Other recent research supports the conclusion that the subset of mercury-related benefits monetized in the RIA represent an incomplete picture of mercury impacts—both in terms of health and wellness impacts considered, and in the populations considered. There are several health benefits from reducing mercury emissions that could not be quantified in the RIA: cardiovascular effects, neurobehavioral effects aside from IQ loss, and immune endpoints (Karagas et al. 2012; Roman et al. 2011; Genchi et al. 2017). While efforts to develop dose-response relationships for use in regulatory benefit-cost analysis are underway, these potentially large health benefits should be taken into account in a holistic assessment of benefits and costs. Other benefits to individual and community health and well-being that are not easily quantifiable should also be included in a holistic assessment of benefits and costs, such as socio-cultural integrity (Ranco et al. 2011; O’Neill 2004). New evidence has also affirmed the importance of considering additional exposure pathways to self-caught freshwater fish: US Electricity Generating Units (EGU) also contribute to mercury pollution in coastal fisheries, and marine fish make up the vast majority of seafood diet for the US population as a whole (Sunderland et al. 2018, 2016).

In the Proposed Finding, the EPA affirms that benefits described above that were not quantified in the RIA “are relevant to any comparison of the benefits and costs of a regulation” (p. 2678). If the EPA chooses to focus only on benefits associated with HAPs in its “appropriate and necessary” determination—an approach that does not consider the full public health benefits of MATS—it should take into account recent research on the full scope and potential magnitude of mercury-related impacts.

(C-24) The residual risk analysis methodology underestimates non-cancer risks associated with dietary exposure to mercury through fish for most-exposed populations. Recent research using state-of-the-science atmospheric and aquatic modelling suggests that fish-dependent communities may be exposed to levels exceeding EPA’s reference dose—even with policies like MATS in place.

The multi-pathway exposure and risk screening assessment described in the Proposed Finding, and based on the National Emissions Standards for Hazardous Air Pollutants: Benzene Emissions from Maleic Anhydride Plants, Ethylbenzene/Styrene Plants, Benzene Storage Vessels, Benzene Equipment Leaks, and Coke By-Product Recovery Plants (Benzene NESHAP), may not be appropriate for methylmercury risk assessment. The assessment focuses on local (<50 km) impacts of single-facilities, rather than aggregate and cumulative effects of multiple facilities over regional scales. Mercury is indeed a local pollutant—the importance of US EGUs as a source of domestic deposition may in fact be underestimated in the RIA (Zhou et al. 2017; Sunderland et al. 2016). However, due to its different chemical forms, mercury also has regional impacts. As a result, regions with a high density of facilities that may be below the Tier 1 screening threshold emission rate, may still experience substantial mercury inputs. Further, the complex dynamics of mercury in aquatic and terrestrial ecosystems mean that some landscapes are more sensitive to mercury inputs than others (Perlinger et al. 2018). These dynamics are not captured in the exposure and risk screening assessment used by the EPA.

Assessment methods that better capture the complex biogeochemistry of mercury indicate that highly- exposed populations (such as subsistence fishers and Tribal Nations where harvesting fish is an important dietary, socio-cultural, and rights-based practice) may be exposed to methylmercury at levels exceeding EPA’s reference dose, even with policies like MATS in place. In Perlinger et al. (2018), with colleagues, we model changes in atmospheric deposition and fish mercury concentrations in the Great Lakes resulting from policies at multiple jurisdictional scales (including MATS). We explore the environmental justice implications of these policy changes for a tribe in Michigan’s Upper Peninsula (UP) with a high fish consumption rate, the Keweenaw Bay Indian Community (KBIC). We find that even with regional, domestic, and global policy, KBIC members may be exposed to levels above the EPA reference dose at their desired level of fish consumption (454 g/day).1 The UP landscape is highly sensitive to Hg deposition, with nearly 80% of lakes estimated to be impaired. Sensitivity to mercury is caused primarily by the region's abundant wetlands. This finding suggests that landscape factors should be considered in the risk assessment process.

Assessment methodologies such as the one described in the Proposed Finding also do not take into account legacy emissions (i.e., recycling of previously deposited mercury), and therefore underestimate health risks (Angot et al. 2018). Because it is persistent in the environment, mercury once emitted can circulate for decades to centuries. In Angot et al. (2018), with colleagues, using an integrated modelling approach, we quantify the impact of growing legacy reservoirs of mercury and find that due to this legacy penalty, aggressive maximum feasible reductions in emissions are needed for fish mercury concentrations to approach target levels by 2050 for high fish-consuming communities such as the Aroostook Band of Micmacs.

1 Research suggests that there is no threshold level of methylmercury exposure below which neurodevelopmental effects do not occur—consequently, the EPA reference dose is a non-conservative benchmark for health risk. In closing, we urge that the EPA consider up-to-date information in both its appropriate and necessary finding, and its assessment of residual risk.

Respectfully,

Amanda Giang Assistant Professor of Environmental Modelling and Policy Institute for Resources, Environment and Sustainability Department of Mechanical Engineering University of British Columbia

Noelle Selin Associate Professor Institute for Data, Systems and Society and Department of Earth, Atmospheric and Planetary Sciences Massachusetts Institute of Technology

With signatories:

Hélène Angot Research Associate Institute of Arctic and Alpine Research University of Colorado, Boulder

Hugh Gorman Professor of Environmental History and Policy Department of Social Sciences Michigan Technological University

Noel Urban Professor of Environmental Engineering Department of Civil and Environmental Engineering Michigan Technological University

Valoree S. Gagnon Director, University-Indigenous Community Partnerships Great Lakes Research Center Michigan Technological University

Judith Perlinger Professor Civil & Environmental Engineering Department Michigan Technological University

[Attachments: Giang & Selin 2016; Perlinger et al. 2018; Angot et al. 2018]

Works Cited

Angot, Hélène, Nicholas Hoffman, Amanda Giang, Colin P. Thackray, Ashley N. Hendricks, Noel R. Urban, and Noelle E. Selin. 2018. “Global and Local Impacts of Delayed Mercury Mitigation Efforts.” Environmental Science & Technology 52 (22): 12968–77. https://doi.org/10.1021/acs.est.8b04542. Genchi, Giuseppe, Maria Stefania Sinicropi, Alessia Carocci, Graziantonio Lauria, and Alessia Catalano. 2017. “Mercury Exposure and Heart Diseases.” International Journal of Environmental Research and Public Health 14 (1): 1–13. https://doi.org/10.3390/ijerph14010074. Giang, Amanda, and Noelle E. Selin. 2016. “Benefits of Mercury Controls for the United States.” Proceedings of the National Academy of Sciences 116 (2): 286–91. https://doi.org/10.1073/pnas.1514395113. Grandjean, Philippe, and Martine Bellanger. 2017. “Calculation of the Disease Burden Associated with Environmental Chemical Exposures: Application of Toxicological Information in Health Economic Estimation.” Environmental Health: A Global Access Science Source 16 (1): 1–14. https://doi.org/10.1186/s12940-017-0340-3. Karagas, Margaret R, Anna L Choi, Emily Oken, Milena Horvat, Rita Schoeny, and Elizabeth Kamai. 2012. “Evidence on the Human Health Effects of Low-Level Methylmercury Exposure.” Environmental Health Perspectives 120 (6): 799–806. O’Neill, Catherine A. 2004. “Mercury, Risk, and Justice.” Enironmental Law Review 34: 11070–115. Perlinger, J A, N R Urban, A Giang, N E Selin, A N Hendricks, H Zhang, A Kumar, et al. 2018. “Responses of Deposition and Bioaccumulation in the Great Lakes Region to Policy and Other Large- Scale Drivers of Mercury Emissions.” Environmental Science: Processes & Impacts 20 (1): 195–209. https://doi.org/10.1039/C7EM00547D. Ranco, Darren J., Catherine a. O’Neill, Jamie Donatuto, and Barbara L. Harper. 2011. “Environmental Justice, American Indians and the Cultural Dilemma: Developing Environmental Management for Tribal Health and Well-Being.” Environmental Justice 4 (4): 221–30. https://doi.org/10.1089/env.2010.0036. Rice, Glenn E, James K Hammitt, and John S Evans. 2010. “A Probabilistic Characterization of the Health Benefits of Reducing Methyl Mercury Intake in the United States.” Environmental Science & Technology 44 (13): 5216–24. https://doi.org/10.1021/es903359u. Roman, Henry a, Tyra L Walsh, Brent a Coull, Éric Dewailly, Eliseo Guallar, Dale Hattis, Koenraad Mariën, et al. 2011. “Evaluation of the Cardiovascular Effects of Methylmercury Exposures: Current Evidence Supports Development of a Dose-Response Function for Regulatory Benefits Analysis.” Environmental Health Perspectives 119 (5): 607–14. https://doi.org/10.1289/ehp.1003012. Sunderland, Elsie M., Charles T. Driscoll, James K. Hammitt, Philippe Grandjean, John S. Evans, Joel D. Blum, Celia Y. Chen, et al. 2016. “Benefits of Regulating Hazardous Air Pollutants from Coal and Oil-Fired Utilities in the United States.” Environmental Science and Technology 50 (5): 2117–20. https://doi.org/10.1021/acs.est.6b00239. Sunderland, Elsie M., Miling Li, and Kurt Bullard. 2018. “Decadal Changes in the Edible Supply of Seafood and Methylmercury Exposure in the United States.” Environmental Health Perspectives 126 (2): 029003. https://doi.org/10.1289/EHP3460. Zhou, Hao, Chuanlong Zhou, Mary M. Lynam, J. Timothy Dvonch, James A. Barres, Philip K. Hopke, Mark Cohen, and Thomas M. Holsen. 2017. “Atmospheric Mercury Temporal Trends in the Northeastern United States from 1992 to 2014: Are Measured Concentrations Responding to Decreasing Regional Emissions?” Environmental Science and Technology Letters 4 (3): 91–97. https://doi.org/10.1021/acs.estlett.6b00452.

Benefits of mercury controls for the United States

Amanda Gianga,1 and Noelle E. Selina,b aInstitute for Data, Systems, and Society, Massachusetts Institute of Technology, Cambridge, MA 02139; and bDepartment of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139

Edited by Catherine L. Kling, Iowa State University, Ames, IA, and approved November 18, 2015 (received for review July 21, 2015)

Mercury pollution poses risks for both human and ecosystem substantial gaps exist in scientific understanding of the processes health. As a consequence, controlling mercury pollution has become that mercury undergoes through long-range transport. Thus, it a policy goal on both global and national scales. We developed an has historically been difficult to quantitatively estimate pro- assessment method linking global-scale atmospheric chemical trans- spective domestic benefits from global environmental treaty- port modeling to regional-scale economic modeling to consistently making in ways that can be compared with socioeconomic evaluate the potential benefits to the United States of global (UN analyses designed to support domestic environmental decision- Minamata Convention on Mercury) and domestic [Mercury and Air making. Here, we use an assessment approach that enables Toxics Standards (MATS)] policies, framed as economic gains from tracing this pathway, accounting for best-available scientific un- avoiding mercury-related adverse health endpoints. This method derstanding and addressing uncertainties and knowledge gaps attempts to trace the policies-to-impacts path while taking into with policy-appropriate assumptions. account uncertainties and knowledge gaps with policy-appropriate Mercury is a naturally occurring element, but human activities bounding assumptions. We project that cumulative lifetime benefits such as mining and coal combustion have mobilized additional from the Minamata Convention for individuals affected by 2050 are amounts, enhancing the amount of mercury circulating in the atmosphere and surface oceans by a factor of three or more $339 billion (2005 USD), with a range from $1.4 billion to $575 billion (3, 4). Mercury previously deposited to land and water can in our sensitivity scenarios. Cumulative economy-wide benefits to revolatilize over decades to centuries. Thus, human activities the United States, realized by 2050, are $104 billion, with a range have fundamentally altered the global biogeochemical cycle of from $6 million to $171 billion. Projected Minamata benefits are mercury (5). Deposited mercury in aquatic systems can be con- more than twice those projected from the domestic policy. This verted to more toxic methylmercury (MeHg), which bio- relative benefit is robust to several uncertainties and variabilities, accumulates. People are then exposed to MeHg by eating ≈ with the ratio of benefits (Minamata/MATS) ranging from 1.4 to 3. contaminated fish. Effects of MeHg exposure include IQ deficits However, we find that for those consuming locally caught freshwa- in prenatally exposed children (6–8) and may include cardiovas- ter fish from the United States, rather than marine and estuarine cular effects in adults (7, 9). Scientific uncertainty and variability fish from the global market, benefits are larger from US than global are substantial throughout this pathway, including but not limited action, suggesting domestic policies are important for protecting to atmospheric chemistry, deposition patterns, methylation pro- these populations. Per megagram of prevented emissions, our do- cesses, bioaccumulation and food web dynamics, dietary patterns of mestic policy scenario results in US benefits about an order of mag- exposure, and dose–response relationships. Despite these uncer- nitude higher than from our global scenario, further highlighting tainties, scientific analyses have been conducted to support decision- theimportanceofdomesticaction. making, and state-of-the-art models exist for many of these steps. Some studies have previously traced the pathway from mer- mercury | policy | impacts assessment | Minamata Convention | economic cury emissions to human impacts. These studies are limited in benefits how completely they have represented physical processes, and how they have accounted for knowledge gaps. First, many do not oxic contamination from human activities is a global prob- explicitly consider spatial transport through the environment on a lem. Although some countries have regulated toxic sub- global scale, and so do not explicitly link emissions to exposure T changes (10–14). Timescales associated with bioaccumulation stances such as heavy metals and persistent organic pollutants for through ecosystems also are often not taken into account, making several decades, chemical contamination has still been identified as a key planetary boundary at risk for exceedance in the context of global change (1). To address this challenge, existing global Significance environmental treaties try to manage the entire life cycle of chemical contaminants (2). The newest of these is a global treaty Mercury is a globally transported pollutant with potent neu- on mercury, the Minamata Convention. In November 2013, the rotoxic effects for both humans and wildlife. This study intro- United States became the first country to fulfill the requirements duces an assessment method to estimate the potential human necessary to become a party to the convention. health-related economic benefits of global and domestic mer- In the United States, analyses to support domestic environ- cury control policies. It finds that for the US population as a mental decision-making include socioeconomic valuations of whole, global mercury controls could lead to approximately impacts as part of the regulatory process. However, these eval- twice the benefits of domestic action by 2050. This result is uations can be both scientifically challenging and politically robust to several uncertainties and variabilities along the contentious, particularly given uncertainties and knowledge gaps emissions-to-impacts path, although we find that those con- (as noted in arguments in a recent case heard in the US Supreme suming locally caught freshwater fish in the United States Court, Michigan v. Environmental Protection Agency, 2015, could benefit more from domestic action. addressing analysis of the costs and benefits of mercury regula- tion). These challenges are especially difficult for contaminants Author contributions: A.G. and N.E.S. designed research; A.G. performed research; A.G. such as mercury, which cross temporal and spatial scales and and N.E.S. analyzed data; and A.G. and N.E.S. wrote the paper. have both domestic and global sources. The chain of analysis The authors declare no conflict of interest. from policies, through emissions, to impacts involves a complex This article is a PNAS Direct Submission. pathway, which for mercury includes industrial activities, atmo- Freely available online through the PNAS open access option. spheric chemistry, deposition processes, bioaccumulation, and 1To whom correspondence should be addressed. Email: [email protected]. human exposure. Existing approaches have not fully combined This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. information and knowledge from these disparate fields, and 1073/pnas.1514395113/-/DCSupplemental.

286–291 | PNAS | January 12, 2016 | vol. 113 | no. 2 www.pnas.org/cgi/doi/10.1073/pnas.1514395113 it difficult to evaluate how the timing of emissions changes affects the US Environmental Protection Agency projected trend from benefits (15). Few studies have explicitly included more uncertain, 2016 to 2020 (15, 21) linearly, resulting in 2050 US emissions of − but potentially important, health endpoints such as cardiovascular 46 Mg·y 1. Our NP case for the United States includes no further effects in their estimates (12, 16). For instance, the US Environ- improvements in emissions control technology or policy, and mental Protection Agency (15) focused on only IQ-related MeHg thus results in an approximate doubling of 2005 emissions by effects in their analysis of the Mercury and Air Toxics Standards 2050 (19). Benefits of the Minamata Convention to the United (MATS) in the United States. Finally, methods used for previous States are calculated as the difference between the global studies were not designed to highlight the relative importance of Minamata and NP scenarios, holding US emissions constant at the uncertainties throughout the policies-to-impacts path. MATS scenario. Benefits of MATS to the US are calculated as the We explicitly incorporate uncertainty and sensitivity analysis difference between the US NP and MATS scenarios, holding for key steps along the policies-to-impacts pathway to assess the emissions in the rest of the world constant at the NP scenario. relative importance of policy-relevant uncertainties. We combine Under our Minamata case, mercury deposition to the United best available models to trace projected global mercury policy sce- States and to the global oceans are 19% and 57% less than under narios to their US impacts. We use atmospheric modeling to project NP in 2050, respectively. Fig. 1 maps these deposition differences the amount of mercury depositing to the US and global seafood over the contiguous United States. We model the atmospheric source regions with and without global policy. We incorporate as- transport and deposition of mercury using the global, 3D land- sessment of timescales associated with bioaccumulation through ocean-atmosphere mercury model GEOS-Chem v.9-02, at 4° × ecosystems. We then link atmospheric mercury models to economic 5° resolution globally and 0.5° × 0.667° resolution over the valuation models, generating a representation of mercury impacts United States (22–26). We use net total deposition as a measure that takes into account environmental and human response time- of mercury ecosystem enrichment (27). SI Appendix, Chemical scales. We use this assessment approach to present what is, to our transport modeling gives additional details on the modeling ap- knowledge, a first assessment of potential US benefits, defined in proaches. For our MATS case, deposition to the United States is economic terms, from the Minamata Convention. We explicitly 20% less than under NP, and deposition to the global oceans is compare benefits of global and US policies, using consistent 6% less. Although the modeled avoided deposition over the methodology, and analyze the relative impacts of these policies on entire United States is similar under MATS and Minamata, the the US population. We first present results from a base case analysis distribution of these differences varies, as shown in Fig. 1. of mercury policy to 2050, using our integrated model. We then Avoided deposition under MATS is more highly concentrated in present our sensitivity analyses, assessing the influence of uncer- the Northeast, where there are significant coal-fired emission tainties on our base case results. sources. In contrast, US deposition benefits under the Minamata Convention follow precipitation patterns, as policy avoids in- Results and Discussion creases in the global background mercury concentration. Tracking the Policies-to-Impacts Pathway: Base Case. Globally, our Because mercury is persistent in the environment, anthropo- emissions projections under the Minamata Convention will genic emissions also enrich reservoirs of mercury in the subsur- − result in 2050 in emissions of 1,870 Mg·y 1, which is roughly face ocean and soils. Mercury from these pools can enhance − equivalent to the present-day level, but 2,270 Mg·y 1 less than reemissions, contributing further to deposition. Our GEOS-Chem our no policy (NP) scenario (17). The largest sources of an- simulations take into account the effect of anthropogenic emis- thropogenic mercury emission are stationary coal combustion, sions changes on concentrations of mercury in surface reservoirs artisanal and small-scale gold mining, and metals production only, and consequently underestimate the total deposition benefits (18). Under NP, emissions are projected to more than double, attributable to policy. To roughly estimate the extent of this un- largely as a result of growth in coal use in Asia (19); thus, the derestimation, we use a seven-box, biogeochemical model de- main differences in policy and NP projections depend on as- veloped by Amos et al. (28, 29), which captures the deep ocean sumptions about emission controls for coal (20). Air quality and soil reservoirs, but not the spatial distribution of impacts (SI abatement technologies such as flue gas desulfurization can Appendix, Chemical transport modeling). We find that globally, capture mercury as a cobenefit. For global emission projections deposition reductions under policy are ∼30% larger when taking under the Minamata Convention, which requires the application into account enrichment of these subsurface pools. of best available technologies, taking into account technical and Recent research suggests that fish concentrations in ocean (30–32) economic feasibility, we assume the application of flue gas de- and freshwater (33–36) fish will likely respond proportionally to sulfurization or similar technology outside of the United States changes in atmospheric inputs over years to decades, although the (17, 19). In the United States, our policy scenario is based on magnitude and timing of a full response may be variable, depending MATS (currently under legal challenge), which was designed to on the region (see refs. 32 and 37–39 for examples). For our base control Hg emissions from power generation, with full imple- case scenario, we assume that fish MeHg in both freshwater and mentation by 2016 (15). In the United States, emissions in 2005 marine ecosystems responds after 10 y to proportionally reflect − were ∼90 Mg·y 1 (15). Under our MATS projection, we extend changes in atmospheric inputs (we test the response to this assumption SCIENCE SUSTAINABILITY

Fig. 1. Projected net deposition benefits (Δμg/m2∙y) of MATS and the Minamata Convention over NP over the contiguous United States, at 0.5° × 0.667° resolution. Global results, at 4° × 5° resolution, are shown in SI Appendix, Fig. S4.

Giang and Selin PNAS | January 12, 2016 | vol. 113 | no. 2 | 287 (VSL) approach, estimates projected lifetime (LT) benefits of avoided exposure for those born by 2050 and is consistent with US regulatory practice; the second, a human capital approach, estimates economy-wide (EW) benefits realized by 2050 from avoided labor productivity and wage losses. Given differences in methodology, results from these two approaches are not directly comparable (see SI Appendix, Economic modeling of health impacts for more details). To estimate LT benefits of avoided health effects, we apply estimates of projected lost wages and medical costs for IQ deficits and nonfatal acute myocardial infarctions (heart attacks), and VSL for premature fatalities resulting from myocardial infarctions (see ref. 12 and examples listed in ref. 44 of studies that use this approach), for each year’s projected birth cohort (IQ) and affected adult population (heart attacks). The second method uses the US Regional Energy Policy model, a computable general equilibrium model of the US econ- omy (45). Consistent with previous work valuing economic effects Fig. 2. Trajectories of welfare benefits under global and domestic policy of air pollution through computable general equilibrium modeling until 2050, discounted at 3%. (Top) Modeled EW benefits realized in a given (46), we take into account the effects of IQ deficits and fatal and year. (Bottom) Projected LT benefits for that year’s affected population. Base nonfatal heart attacks on the labor force, and its cumulative effect cases are indicated with markers. Unmarked lines show the range of trajec- over time. Base case cumulative EW benefits of the Minamata tories from sensitivity cases. Convention to the United States by 2050 are $104 billion (2005 USD) (Fig. 2, Top, blue line), and cumulative LT benefits for those born by 2050 are $339 billion (Fig. 2, Bottom, green line). EW in our sensitivity analysis) (30, 37). We specify base year blood benefits from our MATS scenario (Fig. 2, Top, red line) are $43 MeHg, as a biomarker for MeHg exposure, by region, based on billion by 2050, and LT benefits are $147 billion (Fig. 2, Bottom, the National Health and Nutrition Examination Survey (40). We purple line). Both EW and LT benefits are dominated (>90% for then scale blood concentrations based on the change in intake of LT and >99% for EW) by avoided cardiovascular effects, consis- fish MeHg (change in deposition plus time lag), taking into ac- tent with previous studies, including these health endpoints (12, count consumption of domestic freshwater and imported fish 16). Relative to US domestic action, estimated cumulative benefits species from global fisheries, using data from US seafood market from the Minamata Convention are more than twice as large. studies (41) and data compiled by the US Environmental Pro- Considered per unit of avoided emissions, however, the pro- tection Agency on noncommercial anglers (15, 42). Because of jected benefits of MATS to the United States are larger than those data limitations, we consider noncommercial mercury intake from of the Minamata Convention: $324 million/Mg compared with local, freshwater fish only. We treat noncommercial marine an- $46 million/Mg for EW benefits by 2050, and $1.1 billion/Mg glers as average US consumers of marine and estuarine fish. This compared with $150 million/Mg for LT benefits for those born by may slightly underestimate the benefits of MATS in our work; 2050. Given its global scope, the Minamata Convention is likely to however, further data are necessary to quantify the MeHg intake prevent more emissions than MATS. However, as mercury pollu- of noncommercial anglers in different US coastal regions (see tion has effects on both local and global scales, avoided emissions SI Appendix, Changes in human exposure for detailed methods). within the United States, on a per unit basis, lead to larger benefits. Calculated average US mercury intake in 2050, assuming a 10-y time lag between deposition changes and fish response, as well Policies-to-Impacts Sensitivity Analysis. We assess uncertainty and as constant fish intake patterns, is 91% less under our Minamata variability along the policies-to-impacts pathway by identifying scenario than under NP (SI Appendix,Fig.S5). Our MATS scenario key drivers of uncertainty in our base case integrated model, and reduces intake by 32% compared with the NP case. Although the calculating how changes in assumptions affect our quantification deposition decreases over the United States are roughly equivalent of US benefits from the Minamata Convention, MATS, and between the MATS and Minamata scenarios, changes to modeled relative benefits. Key assumptions addressed here include the ef- mercury intake are larger under the latter. More than 90% of the fect of atmospheric chemistry, ecosystem time lags, dietary choices, US commercial fish market, and the majority of US mercury intake, dose–response parameters linking MeHg exposure and health ef- comes from marine and estuarine sources, particularly from Pacific fects, economic costs, and discount rates. We run the integrated and Atlantic Ocean basins (41, 43). These regions are heavily model for realistic and policy-relevant low and high bounds for influenced by emissions from non-US sources, including East and these assumptions. Fig. 2 shows the range of calculated benefits South Asia. In addition, even locally caught freshwater fish are from these sensitivity scenarios, described further here. The un- affected by the long-range transport of mercury emissions. Re- certain range spanned by these cases is illustrated by the lines in gional differences in the geographic source of dietary fish (SI Fig. 2; however, the bounds delineated by these lines for the Appendix, Changes in human exposure) and deposition lead to Minamata (blue/green) and MATS (red/purple) scenarios are not variations in intake change patterns across scenarios, as shown independent. Some sensitivity scenarios lead to the same di- in SI Appendix,Fig.S4. The majority of modeled MeHg intake rectional change in benefits over the base case for both the do- in the North Central region (SI Appendix,Fig.S5)isfromself- mestic and global scenario, such that the magnitude of cumulative caught, local freshwater fish, leading to a diminished intake benefits for the Minamata scenario remain larger than for MATS. benefit from the Minamata scenario relative to the MATS This result is illustrated in Fig. 3, which shows the range in ratio of scenario. The opposite pattern holds for New York. These benefits between Minamata and MATS, under different sensitivity differences in intake lead to corresponding differences in IQ scenarios. Details of the low and high cases addressed are pre- deficits and cardiovascularoutcomes(seeSI Appendix, IQ sented in SI Appendix,TableS7andSensitivity analysis. effects; Cardiovascular impacts; and Health impacts for health Our low and high cases for atmospheric chemistry bound un- impacts methods and results, respectively). certainty about the form of mercury emissions and atmospheric Annual US economic benefits to 2050 (applying a 3% discount redox reactions. Although policies address total mercury emis- rate) from avoided health impacts under domestic and global sions, emissions of mercury occur as different chemical species mercury policies under our base case assumptions, relative to with different atmospheric lifetimes. Mercury emitted in its NP, are presented in Fig. 2. We use two economic valuation elemental form, Hg(0), has an atmospheric lifetime of 6 mo to a approaches: the first, a cost-of-illness and value of statistical life year, enabling it to transport globally before its oxidation and

288 | www.pnas.org/cgi/doi/10.1073/pnas.1514395113 Giang and Selin benefits are therefore highly sensitive to the temporal scope of analysis. For instance, EW benefits from IQ effects are primarily accrued when those in birth cohorts with reduced exposure are of working age (see SI Appendix, Economic modeling of health impacts), and consequently are not fully captured by our 2050 time horizon. Population growth and discounting assumptions (we use a 3% discount rate; see SI Appendix, Economic valuation for others) also influence our cumulative benefit assessment. Timing effects are further discussed in SI Appendix, Economic valuation.Ourlower bound incorporates an instantaneous response, which is the as- sumption commonly used in regulatory analyses (15, 42), and that may be roughly consistent with the behavior of certain classes of freshwater bodies (37). Our upper bound is 50 y, consistent with the high range of estimated response times for surface open ocean waters (30), where MeHg production and biomagnification are hypothesized to occur (31), and midrange estimates for watershed- fed coastal ecosystems and some lake systems, which may be the slowest to respond to changes in atmospheric deposition (32, 36). Population dietary choice between local freshwater and global market fish alters our Minamata base case cumulative EW ben- efits from $17 billion (2005 USD) to $127 billion, and cumulative LT benefits from $56 billion to $418 billion. Our base case as- sumes that population dietary choices between local fish and global market fish remain constant over time. For low and high bounds, respectively, we assume that people’s diets are 100% from either local freshwater or global sources. Where US seafood consumers eat a larger fraction of market marine and estuarine fish, benefits from Minamata are higher. Under the 100% local freshwater diet assumption, benefits from MATS exceed those of Minamata (Minamata/MATS ratio of 0.4 in Fig. 3). With different assumptions about pharmacokinetics and dose– response functions between mercury intake and human health effects, our results for the Minamata scenario vary from $6 million to $160 billion (2005 USD) in EW benefits, and from Fig. 3. (Top) Range in cumulative benefits of the Minamata scenario to $1.4 billion to $498 billion in LT benefits. Although convincing 2050. Note the different scales for LT and EW benefits. (Bottom) Range in evidence is present to associate MeHg with adverse human ef- ratio of cumulative benefits to 2050 (Minamata Benefits/MATS Benefits). fects at low to medium doses, particularly for IQ deficits (7, 50), Blue and green lines show base case results for LT and EW benefits, re- there may be variability in the magnitude of this effect; for in- spectively. Bars indicate the sensitivity of cumulative benefits to high and stance, because of genetic variability (51). As a result, we use low case assumptions for uncertain parameters. 95% confidence interval bounds for high and low cases for bio- marker and dose–response parameters (SI Appendix, Table S3). Associations between mercury exposure and cardiovascular im- subsequent deposition. Mercury emitted in its oxidized form, Hg(II), pacts are less certain than IQ effects (9). Previous studies have in the gas phase, or Hg(P) in the particle phase, is more soluble and expressed this uncertainty, using an expected value approach can deposit closer to its source. In addition, the speciation of present- taking into account both the plausibility of a relationship be- day mercury emissions is uncertain. Reduction reactions may convert tween MeHg and cardiovascular impacts and uncertainties in the Hg(II) to Hg(0), lengthening its lifetime; this process may occur in parameters of the relationship (12). Our lower bound does not the atmosphere in the aqueous phase (47), or in power plant plumes include cardiovascular impacts, whereas our base case and upper (48, 49). However, the mechanism of potential reduction is unknown. bound do, with the 97.5 percentile estimate of the relationship To bound this uncertainty, we assume for our low case that 90% of between hair mercury and heart attack risk used in the high case global Hg reductions over NP occur as Hg(II) or Hg(P), and for the (SI Appendix, Sensitivity Analysis) (52). A more detailed review of high case, that 90% of reductions occur as Hg(0). This results in a the epidemiological evidence contributing to these parameteri- range of cumulative EW benefits for Minamata between $102 billion zations is given in SI Appendix, IQ effects and Cardiovascular (low) and $123 billion (high) in 2005 USD, and a range of LT impacts. Although using different exposure–response functions benefits of $338 billion to $405 billion. That the low case results in leads to the largest absolute range in cumulative benefits among only a small difference from the base case reflects the emphasis on the sensitivity cases considered (Fig. 3), the relative benefits control technologies that capture oxidized mercury in the base case between Minamata and MATS do not change as substantially. assumptions (19). The relative benefits of Minamata versus our High and low assumptions for the economic valuation of MATS case vary to a factor of 2.9 from the base case. If policy mercury-related health effects lead to a range of $58 billion to prevents primarily Hg(0) emissions, or there is a high rate of in-plume $121 billion (2005 USD) in EW benefits from the Minamata reduction, there is greater long-range benefit to the United States scenario by 2050, and a range of $87 billion to $518 billion in LT SCIENCE and global oceans from avoided emissions occurring elsewhere. benefits. Our sensitivity scenarios for EW benefits address only SUSTAINABILITY If fish MeHg responds rapidly and quantitatively to changes morbidity, and not mortality, effects: medical costs associated in deposition, cumulative EW and LT benefits to 2050 from with heart attacks, and the relationship between IQ deficits and Minamata are projected to be $171 billion and $575 billion (2005 lost earnings. We use the 95% confidence interval for the IQ to USD), whereas a slower response reduces projected EW and LT income relationship and the range of estimates for medical costs benefits to $18 billion and $60 billion. Although reductions in from the literature as bounding cases (SI Appendix, Table S7). mercury deposition, all else equal, will eventually result in For LT valuations, we use central and range estimates for VSL and decreased environmental and fish concentrations, benefits within LT lost income from regulatory literature (15, 53). The valuation a given time horizon, which in this case is 2050, will depend on uncertainties considered have the smallest effect on the ratio of how long ecosystems take to respond. Estimated economic benefits between global and domestic scenarios (Fig. 3).

Giang and Selin PNAS | January 12, 2016 | vol. 113 | no. 2 | 289 Implications for Policy Evaluation. We developed and applied an subsistence fishers, and recreational anglers. In addition, it assessment method to examine the complex pathways from highlights the policy need for analysis and data collection on the policies to environmental effects for global toxic pollution from evolving patterns in fisheries production and fish consumption mercury that accounts for uncertainties and knowledge gaps in a (43). It has been noted that dietary guidance on fish selection structured way. We showed, using this method, that by 2050, the and consumption frequency could be part of an adaptation Minamata Convention could have approximately twice the benefit strategy to minimize mercury exposure (57), and our results of our scenario simulating domestic actions ($104 billion compared point toward their potentially large effect as a policy lever. with $43 billion in cumulative EW benefits, and $339 billion com- However, dietary advice is highly complex. Fish consumption, pared with $147 billion in cumulative LT benefits). The relative and specific fish selection, can have substantial benefits, both benefit is robust to several uncertainties assessed along the policies- nutritional (58, 59) and sociocultural (60). Balancing the risks to-impacts pathway, including atmospheric chemical processes, and benefits of fish consumption therefore requires careful ecosystem time lags, and exposure–response relationships; how- consideration of contextual factors. Even with such adaptive ever, we find that domestic action has a larger benefit when dietary approaches, there is continued need to mitigate future emissions. fish is sourced from local freshwater bodies. Per megagram of Although uncertainties related to chemical speciation of avoided emissions, the benefits to the United States of domestic emissions reductions led to the smallest range in cumulative action are nearly an order of magnitude larger than global action, benefits for the Minamata scenario, interactions between these highlighting that although mercury is a global pollutant, local pol- uncertainties and variabilities in dietary fish source could affect icies contribute strongly to local benefits. As shown in SI Appendix, the relative benefits of global versus domestic action. At this Fig. S4, avoided emissions associated with the Minamata Conven- time, our ability to constrain these speciation uncertainties is tion outside of the United States may lead to large benefits in Asia partially limited by measurement challenges (61). Improved and Southern Europe. Abatement costs will also vary by region. measurement techniques could provide insight into distributional Although we have conducted what is, to our knowledge, the first aspects of control policies. global-scale attempt to link future emissions trajectories to domestic Differences in valuation methods for health endpoints could impacts, our ability to incorporate detailed models of the entire lead to substantial variation in benefits estimates. Our two val- pathway is limited by existing scientific knowledge. In addition uation approaches highlight some of these potential variations: to these knowledge gaps, there are also variabilities in mercury’s Our EW approach emphasizes compounding economy-wide behavior across ecosystems and regions, as well as in human gains over time, but considers only effects to the economy (not responses (physical and social). Our approach uses bounding individuals) realized within the 2050 time horizon; in contrast, assumptions along the policies-to-impacts pathway as a proxy to our LT approach more closely resembles regulatory studies, assess the relative influence of various uncertainties, from a taking into account projected lifetime and nonmarket effects to range of disciplines. In a number of previous analyses, range in individuals (e.g., pain and suffering). As highlighted previously, the benefits of mercury reduction has been specified by the range economic benefits estimates are very sensitive to choices of in exposure–response functions (12, 13). Although our analysis temporal scope of analysis and discounting. Estimates are also underlines the importance of these uncertainties, particularly sensitive to the endpoints considered: In addition to the health those related to cardiovascular effects, it also suggests that pre- effects considered here, there may be other human and wildlife vious approaches miss other potentially large contributors to health endpoints not included in this study that, although not uncertainty in economic effects (particularly within a given time well characterized at this time (7), may also have economic ef- horizon), such as marine and freshwater ecosystem dynamics and fects. No less important, there may be dimensions of individual dietary intake variabilities. and community health and well-being that are not quantifiable Although, all else being equal, mercury emissions reductions within this economic framework, which should be considered in a will ultimately result in exposure reductions, our analysis in- holistic assessment of policy benefits (62). dicates that uncertainties in ecosystem dynamics affecting the Our assessment of US benefits from global and domestic policy timescale of these reductions will strongly influence benefits is designed to be illustrative, drawing attention to uncertainties in within a given time horizon. Many of the processes affecting the estimating economic benefits and methods to take these uncer- conversion of inorganic mercury to MeHg and subsequent tainties into account. As a consequence, our estimates should not uptake in biota are poorly understood, particularly in marine be taken as a comprehensive projection of impacts. However, as ecosystems (54, 55). In addition, there is variability among eco- scientific knowledge evolves, many uncertainties can be addressed system types, both freshwater (37) and marine and estuarine using similar methodology. Policies-to-impacts analyses similar to (32), in how quickly these systems and biota within them respond the one presented here can be valuable for synthesizing available to changes in deposition. As described previously, our analysis information, identifying its limitations, and when combined with focuses on changes to mercury in surface reservoirs, and ac- sensitivity analysis, suggesting areas where scientific data collec- counting for these effects could increase benefits estimates by tion to narrow uncertainty would lead to uncertainty reduction of ∼30%. Future research should more fully address the timescales importance to policy-making. of reemissions from subsurface reservoirs, both land and ocean, and their effects on benefits estimates. Better understanding of Materials and Methods mercury cycling, methylation and bioaccumulation processes, Brief explanations of methods have been included throughout Results and their variability, and the potential effects of global changes to Discussion.IntheSI Appendix, Supplementary methods,weprovideadetailed climate, land use, and other environmental contaminants will be description of methodology and data sources for emissions projections, critical for improving policy evaluation (56), particularly for chemical transport modeling, translating changes in deposition to changes in better understanding the distribution of benefits between current human exposure, IQ and cardiovascular impacts modeling, economic modeling of health impacts, and sensitivity analysis. Institutional review and informed and future generations. consent were not necessary for this modeling study, as all human health and Our analysis also reveals the importance of social factors in ecosystem input data were drawn from published sources. estimating the absolute and relative benefits of different policies. Dietary choices, including fish selection and consumption rate, ACKNOWLEDGMENTS. This work was supported by the National Science can have a potentially larger influence on the ratio of benefits Foundation (Awards 1053648 and 131755), the J. H. and E. V. Wade Fund from global compared with domestic action than substantial (Massachusetts Institute of Technology), and fellowships from the Natural scientific uncertainties about mercury′s environmental behavior. Science and Engineering Research Council of Canada and the MIT Sociotechnical This sensitivity result suggests that domestic actions may be Systems Research Center Stokes Fund (to A.G.). This work used a modeling tool (US Regional Energy Policy) developed by the MIT Joint Program on the Science particularly important for reducing exposure for communities and Policy of Global Change, which is supported by a number of federal agencies that consume mostly fish sourced from the contiguous United and a consortium of 40 industrial and foundation sponsors. A complete list of States, such as certain Indigenous peoples and immigrant groups, sponsors is available at globalchange.mit.edu.

290 | www.pnas.org/cgi/doi/10.1073/pnas.1514395113 Giang and Selin 1. Rockström J, et al. (2009) A safe operating space for humanity. Nature 461(7263): 33. Harris RC, et al. (2007) Whole-ecosystem study shows rapid fish-mercury response to 472–475. changes in mercury deposition. Proc Natl Acad Sci USA 104(42):16586–16591. 2. Selin H (2014) Global Environmental Law and Treaty-Making on Hazardous Sub- 34. Evers DC, et al. (2011) Mercury in the Great Lakes region: Bioaccumulation, spatio- stances: The Minamata Convention and Mercury Abatement. Glob Environ Politics temporal patterns, ecological risks, and policy. Ecotoxicology 20(7):1487–1499. 14(1):1–19. 35. Drevnick PE, et al. (2012) Spatial and temporal patterns of mercury accumulation in 3. Lamborg CH, et al. (2014) A global ocean inventory of anthropogenic mercury based lacustrine sediments across the Laurentian Great Lakes region. Environ Pollut 161: on water column measurements. Nature 512(7512):65–68. 252–260. 4. Amos HM, et al. (2015) Observational and modeling constraints on global anthro- 36. Vijayaraghavan K, Levin L, Parker L, Yarwood G, Streets D (2014) Response of fish pogenic enrichment of mercury. Environ Sci Technol 49(7):4036–4047. tissue mercury in a freshwater lake to local, regional, and global changes in mercury 5. Driscoll CT, Mason RP, Chan HM, Jacob DJ, Pirrone N (2013) Mercury as a global emissions. Environ Toxicol Chem 33(6):1238–1247. – pollutant: Sources, pathways, and effects. Environ Sci Technol 47(10):4967 4983. 37. Knightes CD, Sunderland EM, Craig Barber M, Johnston JM, Ambrose RBJ (2009) 6. Grandjean P, Satoh H, Murata K, Eto K (2010) Adverse effects of methylmercury: Application of ecosystem-scale fate and bioaccumulation models to predict fish Environmental health research implications. Environ Health Perspect 118(8): mercury response times to changes in atmospheric deposition. Environ Toxicol Chem – 1137 1145. 28(4):881–893. 7. Karagas MR, et al. (2012) Evidence on the human health effects of low-level meth- 38. Selin NE, Sunderland EM, Knightes CD, Mason RP (2010) Sources of mercury exposure ylmercury exposure. Environ Health Perspect 120(6):799–806. for U.S. seafood consumers: Implications for policy. Environ Health Perspect 118(1): 8. Sheehan MC, et al. (2014) Global methylmercury exposure from seafood consumption 137–143. and risk of developmental neurotoxicity: A systematic review. Bull World Health 39. Drenner RW, et al. (2013) Effects of mercury deposition and coniferous forests on the Organ 92(4):254–269F. mercury contamination of fish in the South Central United States. Environ Sci Technol 9. Roman HA, et al. (2011) Evaluation of the cardiovascular effects of methylmercury 47(3):1274–1279. exposures: Current evidence supports development of a dose-response function for 40. Mahaffey KR, Clickner RP, Jeffries RA (2009) Adult women’s blood mercury concen- regulatory benefits analysis. Environ Health Perspect 119(5):607–614. trations vary regionally in the United States: Association with patterns of fish con- 10. Trasande L, Landrigan PJ, Schechter C (2005) Public health and economic conse- sumption (NHANES 1999-2004). Environ Health Perspect 117(1):47–53. quences of methyl mercury toxicity to the developing brain. Environ Health Perspect 41. Sunderland EM (2007) Mercury exposure from domestic and imported estuarine and 113(5):590–596. – 11. Spadaro JV, Rabl A (2008) Global health impacts and costs due to mercury emissions. marine fish in the U.S. seafood market. Environ Health Perspect 115(2):235 242. Risk Anal 28(3):603–613. 42. US Environmental Protection Agency(2005) Regulatory Impact Analysis of the Final 12. Rice GE, Hammitt JK, Evans JS (2010) A probabilistic characterization of the health Clean Air Mercury Rule, (US Environmental Protection Agency, Office of Air Quality benefits of reducing methyl mercury intake in the United States. Environ Sci Technol Planning and Standards, Research Triangle Park, NC). 44(13):5216–5224. 43. Karimi R, Fitzgerald TP, Fisher NS (2012) A quantitative synthesis of mercury in 13. Bellanger M, et al.; DEMO/COPHES (2013) Economic benefits of methylmercury exposure commercial seafood and implications for exposure in the United States. Environ control in Europe: Monetary value of neurotoxicity prevention. Environ Health 12:3. Health Perspect 120(11):1512–1519. 14. Griffiths C, McGartland A, Miller M (2007) A comparison of the monetized impact of 44. Bell ML, Morgenstern RD, Harrington W (2011) Quantifying the human health ben- IQ decrements from mercury emissions. Environ Health Perspect 115(6):841–847. efits of air pollution policies: Review of recent studies and new directions in ac- 15. US Environmental Protection Agency (2011) Regulatory Impact Analysis for the Final countability research. Environ Sci Policy 14(4):357–368. Mercury and Air Toxics Standards (US Environmental Protection Agency, Office of Air 45. Rausch S, Metcalf GE, Reilly JM (2010) Distributional Implications of Alternative U.S. Quality and Planning Standards, Research Triangle Park, NC). Greenhouse Gas Control Measures, (National Bureau of Economic Research Working 16. Rice GE, Hammitt JK (2005) Economic Valuation of Human Health Benefits of Con- Paper 16053, Cambridge, MA). trolling Mercury Emissions from U.S. Coal-Fired Power Plants. (Northeast States for 46. Saari RK, Selin NE, Rausch S, Thompson TM (2015) A self-consistent method to assess Coordinated Air Use Management, Boston, MA). air quality co-benefits from U.S. climate policies. J Air Waste Manag Assoc 65(1): 17. Selin NE (2014) Global change and mercury cycling: Challenges for implementing a 74–89. global mercury treaty. Environ Toxicol Chem 33(6):1202–1210. 47. Subir M, Ariya PA, Dastoor AP (2012) A review of the sources of uncertainties in at- 18. United Nations Environment Programme. (2013) Global Mercury Assessment 2013: mospheric mercury modeling II. Mercury surface and heterogeneous chemistry – A Sources, Emissions, Releases and Environmental Transport, (UNEP Chemicals Branch, missing link. Atmos Environ 46:1–10. Geneva, Switzerland). 48. Landis MS, Ryan JV, ter Schure AFH, Laudal D (2014) Behavior of mercury emissions 19. Streets DG, Zhang Q, Wu Y (2009) Projections of global mercury emissions in 2050. from a commercial coal-fired power plant: The relationship between stack speciation – Environ Sci Technol 43(8):2983 2988. and near-field plume measurements. Environ Sci Technol 48(22):13540–13548. 20. Giang A, Stokes LC, Streets DG, Corbitt ES, Selin NE (2015) Impacts of the minamata 49. Tong Y, et al. (2014) Comparison of heterogeneous photolytic reduction of Hg(II) in convention on mercury emissions and global deposition from coal-fired power gen- the coal fly ashes and synthetic aerosols. Atmos Res 138:324–329. – eration in Asia. Environ Sci Technol 49(9):5326 5335. 50. NRC (2000) Toxicological Effects of Methylmercury. (National Research Council, National 21. US Environmental Protection Agency. (2011) IPM Analysis of the Final Mercury and Air Academy Press, Washington, DC). Toxics Standards (US Environmental Protection Agency, Clean Air Markets Division, 51. Basu N, Goodrich JM, Head J (2014) Ecogenetics of mercury: From genetic polymor- Washington, DC). phisms and epigenetics to risk assessment and decision-making. Environ Toxicol Chem 22. Selin NE, et al. (2008) Global 3-D land-ocean-atmosphere model for mercury: Present- 33(6):1248–1258. day versus preindustrial cycles and anthropogenic enrichment factors for deposition. 52. Virtanen JK, et al. (2005) Mercury, fish oils, and risk of acute coronary events and Global Biogeochem Cycles 22(2):1–13. cardiovascular disease, coronary heart disease, and all-cause mortality in men in 23. Soerensen AL, et al. (2010) An improved global model for air-sea exchange of mer- eastern Finland. Arterioscler Thromb Vasc Biol 25(1):228–233. cury: High concentrations over the North Atlantic. Environ Sci Technol 44(22): 53. RTI International. (2015) Environmental Benefits Mapping and Analysis Program - 8574–8580. Community Edition: User’s Manual Appendices, (Research Triangle Park, NC). 24. Holmes CD, et al. (2010) Global atmospheric model for mercury including oxidation by 54. Lamborg C, et al. (2014) Mercury in the Anthropocene Ocean. Oceanography (Wash bromine atoms. Atmos Chem Phys 10(24):12037–12057. DC) 27(1):76–87. 25. Amos HM, et al. (2012) Gas-particle partitioning of atmospheric Hg(II) and its effect 55. Hsu-Kim H, Kucharzyk KH, Zhang T, Deshusses MA (2013) Mechanisms regulating on global mercury deposition. Atmos Chem Phys 12(1):591–603. mercury bioavailability for methylating microorganisms in the aquatic environment: 26. Zhang Y, et al. (2012) Nested-grid simulation of mercury over North America. Atmos – Chem Phys 12:6095–6111. A critical review. Environ Sci Technol 47(6):2441 2456. 27. Corbitt ES, Jacob DJ, Holmes CD, Streets DG, Sunderland EM (2011) Global source- 56. Sundseth K, Pacyna JM, Banel A, Pacyna EG, Rautio A (2015) impacts receptor relationships for mercury deposition under present-day and 2050 emissions on environmental and human exposure to mercury in the arctic. Int J Environ Res – scenarios. Environ Sci Technol 45(24):10477–10484. Public Health 12(4):3579 3599. 28. Amos HM, Jacob DJ, Streets DG, Sunderland EM (2013) Legacy impacts of all-time 57. Xue J, et al. (2015) Modeling tribal exposures to methyl mercury from fish con- – anthropogenic emissions on the global mercury cycle. Global Biogeochem Cycles sumption. Sci Total Environ 533:102 109. 27(2):410–421. 58. Oken E, et al. (2012) Which fish should I eat? Perspectives influencing fish con- 29. Amos HM, et al. (2014) Global biogeochemical implications of mercury discharges sumption choices. Environ Health Perspect 120(6):790–798. from rivers and sediment burial. Environ Sci Technol 48(16):9514–9522. 59. Mozaffarian D, Rimm EB (2006) Fish intake, contaminants, and human health: Eval- SCIENCE –

30. Sunderland EM, Mason RP (2007) Human impacts on open ocean mercury concen- uating the risks and the benefits. JAMA 296(15):1885 1899. SUSTAINABILITY trations. Global Biogeochem Cycles 21(4):GB4022. 60. O’Neill CA (2004) Mercury, Risk, and Justice. Environ Law Rev 34(12):11070–11115. 31. Mason RP, et al. (2012) Mercury biogeochemical cycling in the ocean and policy im- 61. Jaffe DA, et al. (2014) Progress on understanding atmospheric mercury hampered by plications. Environ Res 119:101–117. uncertain measurements. Environ Sci Technol 48(13):7204–7206. 32. Chen C, et al. (2012) Sources to Seafood: Mercury Pollution in the Marine 62. Ranco DJ, O’Neill CA, Donatuto J, Harper BL (2011) Environmental Justice, American Environment (Toxic Metals Superfund Research Program, Dartmouth College, Hano- Indians and the Cultural Dilemma: Developing Environmental Management for Tribal ver, NH). Health and Well-being. Environ Justice 4(4):221–230.

Giang and Selin PNAS | January 12, 2016 | vol. 113 | no. 2 | 291 Environmental Science Processes & Impacts

View Article Online PAPER View Journal | View Issue

Responses of deposition and bioaccumulation in the Great Lakes region to policy and other large- Cite this: Environ. Sci.: Processes Impacts,2018,20,195 scale drivers of mercury emissions†

J. A. Perlinger, *a N. R. Urban,a A. Giang, ‡b N. E. Selin, b A. N. Hendricks,a H. Zhang,§c A. Kumar,c S. Wu,c V. S. Gagnon, d H. S. Gormand and E. S. Norman e

Mercury (Hg) emissions pose a global problem that requires global cooperation for a solution. However, neither emissions nor regulations are uniform world-wide, and hence the impacts of regulations are also likely to vary regionally. We report here an approach to model the effectiveness of regulations at different scales (local, regional, global) in reducing Hg deposition and fish Hg concentrations in the Laurentian Great Lakes (GL) region. The potential effects of global change on deposition are also modeled. We focus on one of the most vulnerable communities within the region, an Indigenous tribe in Michigan's Upper Peninsula (UP) with a high fish consumption rate. For the GL region, elements of global change (climate, biomass burning, land use) are projected to have modest impacts (<5% change from the year 2000) on Hg deposition. For this region, our estimate of the effects of elimination of anthropogenic emissions is a 70% decrease in deposition, while our minimal regulation scenario increases emissions by 35%. Existing policies have the potential to reduce deposition by 20% with most of the reduction attributable to U.S. policies. Local policies within the Great Lakes region show little effect, and global policy as embedded in the Minamata Convention is projected to decrease deposition by approximately 2.8%. Even within the GL region, effects of policy are not uniform; areas close to emission sources (Illinois, Indiana, Ohio, Pennsylvania) experience larger decreases in deposition than other areas including Michigan's UP. The UP landscape is highly sensitive to Hg deposition, with nearly 80% of lakes estimated to be impaired. Sensitivity to mercury is caused primarily by the region's abundant wetlands. None of the modeled policy scenarios are projected to reduce fish Hg Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. Received 13th November 2017 concentrations to the target that would be safe for the local tribe. Regions like Michigan's UP that are Accepted 2nd January 2018 highly sensitive to mercury deposition and that will see little reduction in deposition due to regulations DOI: 10.1039/c7em00547d require more aggressive policies to reduce emissions to achieve recovery. We highlight scientific rsc.li/espi uncertainties that continue to limit our ability to accurately predict fish Hg changes over time.

Environmental signicance Here, we project the responses within the Great Lakes region of rates of atmospheric deposition and sh mercury concentrations to changes in policy at multiple jurisdictional scales and to global change. We demonstrate that even within this region, responses to policy and global change drivers vary spatially with the largest reductions in deposition occurring in closer proximity to major Hg emission sources. Projected responses to atmospheric deposition are greater for U.S. policies than for either local or global Hg control policies. We focus further on the environmental justice implications of policy changes for sh mercury contamination facing one of the most vulnerable communities within the region, a tribe in Michigan's Upper Peninsula with a high sh consumption rate. The Upper Peninsula is shown to be a highly sensitive landscape that readily converts atmospherically deposited Hg to bioaccumulated methylmercury, but one that will respond little to any of the policies evaluated. Thoughtful and more aggressive policy changes would be required to alleviate the existing environmental injustice.

aCivil and Environmental Engineering Department, Michigan Technological eNative Environmental Science Department, Northwest Indian College, Bellingham, University, Houghton, MI 49931, USA. E-mail: [email protected] WA 98226, USA bInstitute for Data, Systems, and Society, Department of Earth, Atmospheric and † Electronic supplementary information (ESI) available. See DOI: Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, 10.1039/c7em00547d USA ‡ Currently at the Institute for Resources, Environment and Sustainability and cGeological and Mining Engineering and Science Department, Michigan Department of Mechanical Engineering, University of British Columbia, Technological University, Houghton, MI 49931, USA Vancouver, BC V6T 1Z4 dSocial Sciences Department, Michigan Technological University, Houghton, MI § Currently at the Center for Global and Regional Environmental Research, 49931, USA University of Iowa, Iowa City, IA, 52242, USA.

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 195 View Article Online Environmental Science: Processes & Impacts Paper

1. Introduction Superior, with the fewest upwind local/regional sources, was the least impacted by these sources (27%). Recently, Zhou et al. Mercury (Hg) is a toxic pollutant that, when emitted to the envi- showed that decreased atmospheric Hg concentrations ronment, undergoes volatilization, long-range transport and measured during the past decade in the northeastern U.S. are deposition in its oxidized form, transformation to its more toxic consistent with decreased Hg emissions from regional point form, methylmercury (MeHg), and bioaccumulation in ecosys- sources, and that increasing global emissions have not over- 22 ff tems. It is at this point that Hg poses threats to human health whelmed those decreases. Potential e ects of local, regional through consumption of sh containing high levels of MeHg. In and global regulation of Hg emissions on deposition across the the Laurentian Great Lakes (GL) region, sh tissue concentrations GL region are investigated in this paper. are unsafe for GL residents as evidenced by statewide sh The concept of landscape sensitivity to Hg deposition is well 23–25 consumption advisories.1–3 Fishing communities are burdened established in the literature. It is clear that landscapes with with the majority of negative impacts,4–8 and Native American abundant forests and wetlands, as well as low nutrients and ff shing rights and cultures, in particular, have been severely alkalinity in runo are predisposed to having high MeHg –  24,26 impacted by Hg contamination;4,7,9 12 thus Hg contamination is an concentrations in lakes and sh. Such regions require lower environmental and social justice issue. Most stakeholders involved rates of atmospheric Hg deposition than other regions to avoid  in the effort to reduce Hg contamination recognize that action is high MeHg concentrations in sh. Studies also suggest that some neededattheglobalaswellastheregionallevel.Towardthisend, of these regions may respond more slowly to decreases in Hg nations have signed and ratied the Minamata Convention on deposition because the Hg accumulated in organic-rich soils may mercury, which sets expectations associated with controls on continue to be mobilized to lakes by the high and potentially 27–29 emissions of Hg.13 A challenge now is to estimate when it will be increasing concentrations of dissolved organic matter in ff possible to safely consume sh in places such as the GL region as runo typical of these regions. Multiple studies have shown that 30–32 nations implement the Minamata Convention. In partnership some lakes can recover quickly from Hg contamination or 33–37 with the Keweenaw Bay Indian Community (KBIC), an Indigenous reduced Hg or sulfate deposition. However, modeling 38–40 36 41 tribe in Michigan's Upper Peninsula (UP), we formulated14 and studies, large surveys and synthesis studies revealed conducted an integrated assessment to address the question, “Can considerable variability among individual lakes that is mediated we, by 2050, meet the goal of safe sh consumption?” by biogeochemical conditions in a lake and/or its watershed as Because of the atmospheric residence times of different Hg well as by the mechanism for Hg input. In northern Wisconsin, species, the answer to this question is location dependent. Much a seepage lake with little catchment showed a rapid decline in  of the Hg deposition to forests is via dry deposition of Hg0 which water and sh Hg over the same time that a drainage lake with because of its long atmospheric residence time (2.7–12 considerable wetland in the catchment showed a decline in total 42 months)15,16 can be due to either regional or global sources. In Hg but little change in MeHg. Within Michigan, concentrations contrast, deposition (wet and dry) of the much shorter lived (0.5– of Hg in multiple lakes and rivers increased from 1998 to 2009 43 2days)15 Hg(II) species is affected primarily by local and regional while atmospheric deposition of Hg and sulfate decreased. In Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. emissions. Understanding the importance of Hg deposition from England, the leaching of Hg from catchment soils continues to 44 local and regional sources in the U.S. (the largest fraction of prolong lake recovery. The combination of variability among which comprise coal-red power plants), in particular within the lakes and our level of understanding of key watershed features GL region, continues to evolve. In the 1980s some thought that results in a wide range (a few years to centuries) in model- 45–47 atmospheric residence times of Hg were long (12–18 months),17 predicted times for lake recovery. In this paper, we high- ff and that atmospheric Hg deposition was due to global-scale light spatial variability in the landscape that results in di erent processes. In 1992 it was shown that local and regional anthro- responses of lakes to Hg deposition. pogenic sources of Hg in the GL region could determine Hg The objectives of this paper are to present an approach to ff concentrations in the litter and mineral soil horizons;18 those estimate e ects of policy and global changes on Hg concen-  concentrations paralleled changes in wet sulfate deposition and trations in sh, to apply this approach to the GL region, to ff human activity. In 2004, using a combination of a global chem- examine spatial di erences in response to global and policy ical transport model (CTM) and a nested regional CTM, Seigneur change, and to evaluate implications of the predicted lake et al.19 showed that Hg deposition in the contiguous U.S. was, on responses particularly for the inhabitants of Michigan's Upper average, comprised of 30% emissions from North America.19 By Peninsula. We restrict our regional scale comparisons to constraining the estimates of wet deposition by a global transport changes in atmospheric deposition and landscape sensitivity, model (GEOS-Chem) with measurements, Selin and Jacob20 two of the primary determinants of bioaccumulation in lakes. estimated that North American anthropogenic emissions We evaluate lake responses and their implications on a local contributed 50% of total deposition in the Midwest and Eastern scale, focusing on Michigan's UP. U.S. Cohen et al.21 used trajectory analysis (HYSPLIT) to discern differences in emission sources of Hg to the ve GL. They found 2. Experimental that Lake Erie, which is downwind of signicant local/regional 2.1 Description of study area emission sources, was the most impacted by local and regional emissions (58% of the base case total deposition), while Lake This study contrasted forcings (global change, policies) and responses on scales ranging from global to regional (hundreds

196 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

  to a few thousand kilometers) to local (tens to a few hundred mg Fish MeHg Conc: kilometers). We present here only the regional (Great Lakes kg wet wt: fish   region) and local (UP, Adirondack area) scale responses mg  ¼ RfD body wt:ðkgÞ (atmospheric deposition of Hg, sh Hg concentrations). kg d  We de ned the Great Lakes region as bounded by 40 and 1 3 50 N latitude and by 73 to 95 W longitude. The UP (44 10 Àkg wet wt: fishÁ km2)isdened spatially by the land border between fish consumption rate   d Wisconsin and Michigan to the west, Lake Superior to the mg 1 – ¼ 0:1 70 kg north, and Lake Michigan Huron to the south and east. The kg d kg wet wt: fish 3 2  0:454 11.3 10 km Adirondack area de ned for this project was d the rectangle bounded south to north by 43.7 to 44.3 Nand mg ¼ 15 ¼ 15 ppb ¼ 0:015 ppm east to west by 73.9 to 75.3 W. Both local areas are largely kg wet wt: fish covered with mixed deciduous-coniferous forests (UP – 77%, where, for illustration purposes, we use a body weight of 70 kg Adirondacks – 74%; National Land Cover Database101), and to be consistent with the sh consumption rate, and, as both have a high density of lakes (UP – 3.8% of area; Adir- – a benchmark, the U.S. Environmental Protection Agency (EPA) ondacks 5.1% of area). The UP has more wetlands (29% vs. mg reference dose48 RfD of 0:1 . We acknowledge that recent 11%; National Wetland Inventory), while the Adirondacks kg d have higher elevation, steeper topography, and higher rates of studies have identied mercury impacts at levels that are lower sulfate deposition. than the studies used to set the EPA's guideline,49 and that advocacy groups have called for lowering this limit.50 The target 2.2 Modeling approach to estimate atmospheric deposition concentration provided a benchmark against which to judge the ff and sh Hg responses to global change and emissions policies e ects of policy scenarios, but clearly the target varies with the target population, is not uniform across the region and will In general terms, our approach to estimating lake responses to evolve as knowledge of health impacts improves. forcings was to construct multiple scenarios embodying 2.2.b Formulation of global change scenarios. We used ff di erent regulatory approaches or global change conditions, to a scenario approach to consider the potential impact of four estimate Hg emissions for these scenarios, to use GEOS-Chem global change drivers on future atmospheric deposition and to predict the spatial distribution of Hg deposition rates sh tissue concentrations of Hg in the GL region: policy, climate within the Great Lakes region, and then to utilize the predicted change, land-use and land-cover change, and biomass burning. Hg deposition rates in a model of watershed and lake Hg Although there are interactions between these four drivers, we cycling. Scenarios used the approximate year 2000 as the evaluated the impact of each separately, holding the other “ ” present day situation with which to compare alternate emis- factors constant, to better isolate inuences on Hg biogeo-  ff sion scenarios. To obtain a signi cant di erence from the chemical cycling. These scenarios draw from previous publica- Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. present but avoid projecting too far into an uncertain future, we tions, so they are described only briey and summarized in selected the year 2050 as the common future time at which Table 1. scenarios would be compared. Because Hg emissions and 2.2.c Policy scenarios. To better understand the impact of  distributions are in uenced by other drivers of global change anthropogenic forcing on future atmospheric deposition and (climate change, land-use/cover, biomass burning), we created sh tissue concentrations, we considered present day emis- scenarios to estimate the changes in Hg deposition caused by sions, and three future emissions scenarios capturing a range of these forcings. For all scenarios, we evaluated the predicted policy ambition: aspirational, policy-in-action, and minimal- spatial changes in Hg deposition within the Great Lakes region, regulation. We implemented the anthropogenic emission and for selected policy scenarios we contrasted atmospheric scenarios dened below in spatially resolved emissions inven- deposition in two local areas to highlight spatial variability tories that are described fully by Corbitt et al.58 and Giang and within the region. Selin.51 Our present day, turn-of-the-21st-century emissions are  2.2.a Target MeHg concentration in sh. To contextualize based on year 2005 for the U.S. and Canada,51,59 and 2006  the current and predicted mercury concentrations in sh, emissions elsewhere in the world.58,60 project social scientists engaged our community partner, the Aspirational. Our aspirational scenario assumed cessation of 14  KBIC, as described in detail by Gagnon et al. The high sh anthropogenic emissions of Hg in 2050. Preventing these consumption rate of this tribal community renders it one of emissions would require transitions to Hg-free alternatives (e.g., the most vulnerable subpopulations within the Great Lakes renewable energy sources such as wind and solar, LED light  ‘ ’ region. The community partners de ned desired walleye bulbs), and advanced capture and control technologies, some of consumption as two 8-oz. (227 gram) meals per day, or a total which may not currently exist. In this scenario and all others, we  of 0.454 kg sh per day. This represents a rate typical of the assume that legacy emissions remain at their year 2000 value, to  spring walleye harvest, when regional shing activities are facilitate comparison of the direct anthropogenic component. 14  highest. This sh consumption rate was entered into the Policy-in-action. This scenario assumed that by 2050, recent  equation to compute a target MeHg concentration in sh as policy proposals targeting Hg emissions are fully implemented. follows.

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 197 View Article Online Environmental Science: Processes & Impacts Paper

Table 1 Summary of global change scenarios

Driver Scenario name Description

Policy51 2000s Anthropogenic emissions for 2005 over the U.S. and Canada and for 2006 elsewhere in the world are used as representative present day emissions at the turn of the 21st century. Future: aspirational Assumed cessation of anthropogenic emissions of Hg in 2050. Future: policy-in-action Recent policy proposals targeting Hg emissions (in the GL, in the U.S., and globally) are assumed to be fully implemented by 2050. Future: minimal regulation Limited global progress on Hg and other sustainability goals are assumed by 2050. Climate change52,53 2000s Meteorology simulated with the NASA GISS GCM. An average of simulated 1998– 2002 meteorology is used to represent present, turn of the 21st century climate. Future NASA GISS GCM used with greenhouse gas trends following the IPCC A1B scenario. An average of simulated 2048–2052 meteorology, is used to represent mid-century climate. Land-use and land-cover 2000s Land use simulated by the IMAGE model. Land cover (1998–2002) simulated using change54,55 the LPJ dynamic global vegetation model driven with the GISS GCM meteorology and CO2 concentrations. Future Future (2050s) represents the period 2048–2052. Changes in land use follow the IPCC A1B scenario. Land cover distributions generated from the LPJ model with GISS GCM meteorology and CO2 concentrations following the IPCC A1B scenario. Biomass burning56,57 2000s Biomass burning emissions estimated from a re emissions model. Average emissions for the period 1998–2002 representing present day are used. Future Refers to the 2050s. Fire emissions model driven with 2050s re frequencies, land use-land cover distributions, meteorology and burned area following the IPCC A1B scenario. Average emissions for the period 2048–2052 are used.

This scenario was developed at three nested geographic 2.2.d GEOS-Chem simulation of atmospheric deposition of domains: local (Lake Superior Basin), regional (U.S.), and total Hg. The GEOS-Chem coupled atmosphere-land-surface global. The global and regional scenarios are described fully in ocean Hg model, version v9-02, was used to simulate the Giang and Selin.51 Globally, we assumed the recently adopted impact of changes in emissions (anthropogenic, biomass United Nations Minamata Convention on mercury is fully in burning), land use/land cover change, and climate on atmo- force in the global community.13 Moreover, we assumed some spheric Hg deposition to the GL.65–68 Additional details on global cooperation on broad sustainability issues like climate model and simulation design are given by Zhang et al.,55 Giang change, following the IPCC SRES B1 scenario, as described by and Selin,51 and Kumar et al.57 Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. Streets et al.60 Current projections based on assessments of For anthropogenic emissions scenarios under all policy technology and energy trends suggest that with mitigation options, our GEOS-Chem simulations are driven with assimi- efforts consistent with the Minamata Convention text and lated meteorology from the NASA Goddard Earth Observing transitions away from coal combustion, emissions in 2050 System (GEOS-5). Using this meteorology allowed us to simulate could be stabilized roughly at current-day levels.60–62 Additional Hg chemistry and transport at a global resolution of 4 5, benets from more ambitious climate action could result in and at a ner 1/2 2/3 resolution over North America through further reductions, though we do not consider those here.63 a one-way nested grid model.59 For each of the four anthropo- Regionally, within the U.S., we assumed that regulations tar- genic emissions scenarios for 2050 dened above, we simulated geting emissions under the Clean Air Act are in place and meteorological years 2007–2011, using the rst two years as enforced through 2050, including the Mercury and Air Toxics initialization. By holding meteorology constant, we isolated the Standards (MATS).51 Locally, in the Lake Superior Basin, we effect of emissions changes alone. Results presented are inter- assumed that zero discharge goals have been achieved by annual averages for meteorological years 2009–2011. 2050.64 To examine the impacts from climate change and land use/ Minimal-regulation. In our minimal-regulation scenario, we land cover change in the coming decades, we carried out model modeled limited global progress on Hg and other sustainability simulations under various scenarios for climate, land use and goals by 2050. We assumed limited advances in the efficacy of land cover. For example, we compared simulation results with air pollution control technologies, limited adoption of existing the present (year 2000) and future (mid-century: 2050) climate control technologies globally, and continued global growth in to derive the potential impact of climate change on Hg depo- coal combustion consistent with the IPCC SRES A1B scenario, sition to the GL region. More details about the scenarios can be as described by Streets et al.60 Under these minimal-regulation found in Zhang et al.55 All GEOS-Chem simulations were driven conditions, Hg emissions in 2050 are close to double those of by meteorological elds from the NASA Goddard Institute for present day levels.60 Additional details for this scenario are Space Studies (GISS) General Circulation Model (GCM 3).69 The provided by Giang and Selin.51 coupling between GISS GCM3 and GEOS-Chem model is

198 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

described by Wu et al.52 We simulated meteorological years 1998–2002 for the present day and 2048–2052 for the future scenario at 4 5 horizontal resolution globally. We analyze the results based on ve-year averages. We also accounted for the potential impacts of global change (including changes in population, climate, land use and land cover) on Hg emissions from wildres and consequently on Hg deposition. Changes in land use/land cover can affect vegeta- tion type/availability while climate is an important determinant of conditions favorable for re occurrence and natural ignitions (lightning). In addition, human interactions with natural vegetation (e.g. use of re for vegetation clearing) are a source of anthropogenic ignitions for wildres while re suppression is greater in areas of high population density. We took advantage of a recently developed re emission model for Hg.57 We calculated the Hg emissions from wildres for the present day (2000s) and future (2050s), respectively, and implemented these emissions in the GEOS-Chem model to examine the perturba- tions to atmospheric Hg burden and deposition. The GEOS-Chem model as used here represents impacts Fig. 1 Schematic of mercury processes and pathways incorporated in from near-term changes in emissions, but does not capture the lake and watershed mercury model. The watershed is semi- alterations in longer-term legacy biogeochemical cycling distributed spatially to account for categories of land use (wetland, forest, urban). Results presented here assume runoff coefficients for changes in response to emissions changes, that were held each land use are fractions of annual deposition. In addition to the constant in all of our scenarios as noted above. The potential cycling of mercury, the model incorporates seasonal variations in implications of these changes that can impact legacy emissions temperature, lake mixing, DOC inputs, and phytoplankton abundance. are discussed further below. 2.2.e Lake mass balance modeling. The mass balance model for Hg of Hendricks et al.70 was used to predict aqueous loadings change.74,75 In the results presented in this paper, concentrations based on the characteristics of a lake (depth, uncertainties reect only the distribution of bioconcentration water retention time, trophic state) and its watershed (area, land factors, rather than the total error calculated via Bayesian cover and wetland extent), and the local rates of Hg atmospheric techniques by Hendricks.70 This model was validated for a lake deposition (Fig. 1). Hg species included elemental, divalent, and in Michigan's UP for which Hg concentrations had been MeHg in the lake compartments epilimnion, hypolimnion, and measured in the water, sediments and sh. Characteristics of

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. sediments; this resulted in a system of nine ordinary differential this lake are summarized in Table 2. equations (ODE) that was solved using a numerical ordinary The model requires inputs of atmospheric concentrations of differential equation solver package for R.71 The model was Hg(0), particulate bound mercury (PBM), MeHg, and reactive based loosely on EPA's spreadsheet model SERAFM,72 but it was gaseous Hg (RGM) to calculate deposition to the lake as well as adjusted for non-steady state conditions to enable prediction of wet and dry deposition uxes of all species to the watershed. In lake responses to changes in loadings. Chemical and physical contrast to many previous studies,76–78 we did not assume that processes of Hg in the lake that were modeled included redox dry deposition would be the same to lakes as to their water- reactions of divalent and elemental Hg, methylation, deme- sheds. The theoretical basis for our assumption is well thylation, photodemethylation, thermocline dispersion, parti- tioning (phases included DOC, abiotic solids, biotic solids, and sediments), diffusion in sediments, settling, burial, and resus- Table 2 Summary of characteristics of the modeled lake. Data from mixed sources as summarized by Hendricks86 pension. Seasonal changes in lake temperature, stratication, and ice cover were simulated. Redox reactions, methylation, Parameter Value or range and demethylation were corrected for these changes in temperature of the lake. Photodemethylation was corrected for Surface area 9.7 106 m2 8 3 the amount of light attenuation received in the water column. Volume 1.4 10 m Mean depth 15 m The rate of thermocline dispersion changed seasonally to allow Watershed area 1.9 108 m2  for mixing and strati cation in this dimictic lake. Atmospheric Wetland area 2.7 107 m2 deposition during ice cover “accumulated” on top of the ice pH 7.55 until spring melt when it entered the lake. The model does not DOC 7.4 mg L 1 – include sh population dynamics; rather, it uses bio- Lake temperature 1.2 18.1 C Water residence time 1 year accumulation factors for mixed feeders and piscivorous sh Sediment Hg 200–400 mgg 1 73 – taken from Knightes. Hence, the model ignores the 3 7 years Lake Hg 0.6–0.8 ng L 1 required for sh populations to come to steady state when

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 199 View Article Online Environmental Science: Processes & Impacts Paper

established. The roughness length for lakes is smaller than for Responses of sh Hg concentrations to scenarios of policy tree canopies resulting in a higher friction velocity.79,80 Alter- change were evaluated, but not to scenarios of change of natively, the difference may be understood in terms of the global climate, land use, and biomass burning. Modeling of resistance components to deposition; for tree canopies these atmospheric deposition was performed only for the present typically include surface (stomatal, cuticular), aerodynamic, and the year 2050. To determine the maximum lake change and quasi-laminar boundary resistances81 while water has likely under the chosen scenarios, we ran the lake model for different surface components. Additionally, in small lakes air– the 2050 conditions. Because the Hg residence time in the water gas exchange is facilitated by convective mixing such that modeled lake (excluding sh) is short (2 months),70 the lake transfer velocities are faster than those estimated based on responds quickly to changes in inputs. While the sediments boundary layer or surface renewal models.82 To estimate dry take longer to reach steady state (20–60 years), low organic deposition to lakes and watersheds, we used the Hg species- and matter content of the study lake results in low in-lake meth- landform-specic deposition velocities in GEOS-Chem.67 Gas ylation and a fast response in this lake to changes in inputs. As exchange with the lake was estimated using deposition veloci- mentioned, we do not include the time for Hg to stabilize in ties for RGM and MeHg, and with a boundary layer-based the food web, and to estimate the maximum possible lake transfer velocity for Hg0.70 Concentrations of inorganic Hg response we assumed a rapid response of the watershed to species were taken from the output of GEOS-Chem; MeHg alterations in deposition. concentrations in wet deposition were taken as a ratio to 2.2.f Virtual experiments: effect of landscape sensitivity. concentrations of total Hg as reported in the literature.83 Our Fish accumulation of Hg is a response to the interplay of rates of predicted wet deposition rates compared favorably with nearby atmospheric deposition with watershed (i.e., landscape sensi- measurements by the Mercury Deposition Network, and our tivity) and lake characteristics. To highlight the effect of land- estimates of dry deposition to the watershed were consistent scape sensitivity on lake response to deposition, we performed with nearby measured Hg uxes in litterfall.84,85 two virtual experiments. First, we compared predicted sh Hg The model also requires inputs from the watershed to the concentrations for a UP lake under two regimes of atmospheric lake of Hg species and dissolved organic carbon (DOC). Inputs deposition (policy-in-action deposition for the UP and Adir- of DOC are modeled as a sine function to reect seasonal ondacks). Second, we also changed the watershed conditions to variability; the mean and amplitude are specictothelake those typical of the Adirondacks area. To clarify the reasons for being modeled. It has been reported in the literature that rates different lake responses in the two regions, we present of total Hg runoff from catchments to lakes range from 1–35% a summary of lake and watershed characteristics and sh Hg of the rate of atmospheric Hg deposition.76,87,88 Multiple concentrations in the two locations. The primary data source models of watershed Hg runoff of varying complexity are for sh Hg in the UP was Michigan's Fish Contaminant Moni- available,89,90 but inter-model comparison has shown large toring Program (http://www.michigan.gov/deq/0,4561,7-135- discrepancies in predictions.91 It is also clear that there may be 3313_3681_3686_3728-32393–,00.html); only concentrations in considerable delays between deposition to a catchment and walleye are summarized here. To eliminate sh size as a cause

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. runoff to streams and lakes.41,92 TheapproachofsettingtheHg of variability among lakes, regressions of walleye Hg vs. walleye runoff from catchments as proportional to atmospheric length were performed for each lake, and concentrations at deposition72 assumes that only recently deposited Hg is a common length of 42 cm (the median size of all walleye susceptible to leaching; this approach yields a fast lake sampled) were calculated and are reported here. Data sources response to changes in Hg deposition. We use a semi- for water quality included Michigan Surface Water Information distributed spatial model that multiplies runoff coefficients Management System (http://www.mcgi.state.mi.us/miswims/), for areas of land use categories by annual Hg deposition rates EPA National Lake Assessment (https://www.epa.gov/national- to estimate the maximum possible lake response to changes in aquatic-resource-surveys/data-national-aquatic-resource-surveys), atmospheric deposition. This approach allows us to say if it is and a variety of reports and published papers.97–100 Although not possible to reach safe sh concentrations by 2050. many data sources are available for the Adirondacks, we Specically,fordivalentHgweusearunoff coefficient of 10% used only data presented by Yu et al.24 Data on the abun- for “sensitive upland landscapes”,andavalueof5%forless dance of lakes and wetland areas in lake catchments for both sensitive upland regions. Coefficients of 20% are used for regions were gathered from the National Wetland Inventory runoff of divalent Hg from wetlands in all areas, and a coeffi- (https://www.fws.gov/wetlands/Data/State-Downloads), and cient of 4.9 is used for MeHg runoff from wetlands.73 Urban forest cover was taken from the National Land Cover and agricultural areas were negligible in the modeled water- Database.101 sheds. Multiple factors (hydrology, owpaths, soil character- istics) regulate runoff of Hg from upland soils;89,92–94 because it 3. Results is widely observed that Hg runoff is proportional to runoff or concentrations of OC (both particulate and dissolved),87,89 and 3.1 Atmospheric deposition of Hg to the GL region, because runoff of DOC from forest soils is reported to be high Michigan's UP, and New York's Adirondacks region in the vs. in the UP95 the runoff coefficient for Hg from upland soils was present 2050 setat10%ofatmosphericdeposition rates based on the ratios Changes in total (wet + dry) Hg deposition estimated from reported by Babiarz et al.96 GEOS-Chem simulations from the present to the future (2050)

200 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

are summarized in Table 3, along with % contributions to the policy-in-action scenario from the three policy scales (global, U.S., or Lake Superior). Total deposition is estimated to decrease by 70% due to the aspirational policy change scenario in the GL region (Fig. S1†), while it is estimated to decrease by 65% in Michigan's Upper Peninsula (UP) and 78% in the New York Adirondack region. These large estimated decreases are licy-in-action scenario reasonable given that for the aspirational scenario, all anthro- pogenic emissions in the 2050 simulation were turned off. % contribution to policy-in-action scenario Because legacy emissions are held constant, this scenario does )

not represent pre-industrial emissions. The combined policy-in-action scenario leads to 20%, 15%, and 29% decreases in atmospheric deposition in 2050 in the GL, Michigan's UP, and New York Adirondack regions, respec- tively (Table 3). Fig. 2 presents the effects of individual scales of policies (Lake Superior, U.S., and global) on the combined policy-in-action deposition benet for the GL region. Approxi- 78 mately 85% of the total decrease is due to U.S. Clean Air Act 29 Increase (+) or decrease ( in total Hg deposition (%) regulations including MATS, while the local (Lake Superior Basin) and global (Minamata Convention) policies each account for 1% and 14% of the total decrease, respectively. Fig. 2B demonstrates the effect of reducing U.S. emissions on GL regional deposition. The two times greater reduction in depo- sition in the Adirondacks compared to the UP for this scenario is therefore largely related to differences in the inuence of U.S. policies. % contribution to policy-in-action scenario Compared to the GL region policy-in-action scenario contributions of 14% from global policies, 85% from U.S., and 1% from Lake Superior, the UP receives 25% of total deposition decrease from global sources, 70% from the U.S., and 5% from Lake Superior (Table 3). In the Adirondacks region, approxi- mately 97% of the total decrease in deposition for this scenario is due to Clean Air Act regulations in the U.S., while local (Lake Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. ) in total Hg 65 Superior Basin) and global (Minamata Convention) each 15 Increase (+) or decrease ( deposition (%) account for 2.9% and 0.1% of the total decrease, respectively. These values clearly show the large differences in contributions that proximity to up-wind sources make. The global contribu- tion (as a percentage) of emissions reductions to decreased atmospheric deposition is over two orders of magnitude greater for the UP than for the Adirondacks (Table 3). Sources within the U.S. (mostly coal-red power plants) provide a greater frac- tion of the deposition to the Adirondacks as compared to the % contribution to policy-in-action scenario UP. The minimal-regulation scenario leads to increases in total )

deposition of 35% and 34% in the GL (Fig. S2†) and Michigan's UP region, respectively (Table 3). The similarity in these esti- mates is likely coincidental. Based on the policy-in-action 70 20 0.2 +0.7 Great LakesIncrease (+) or decrease ( Michigan Upper Peninsula New York Adirondacks in total Hg deposition (%) scenario, differences in deposition benet are largely related to U.S.-scale changes in emissions, and the largely U.S. emis- sions that comprise the minimal-regulation total deposition estimates happen to lead to similar increases in deposition. The climate change, land use/land cover change, and biomass burning scenarios lead to smaller total Hg deposition changes in 2050 as compared to the policy scenarios (Table 3). Change in total Hg atmospheric deposition from the present to 2050 for the given region and scenario, and % contribution of a given scale region to the po In the GL region, these estimates are 3.8% (Fig. S3†), 0% (Fig. S4†), and 1.9% (Fig. S5†), respectively, whereas for Michi- Region Scenario Table 3 2050 aspirational 2050 policy-in-action gan's UP, the changes are 5.2%, 0%, and 2.3%, respectively. WorldU.S.Lake Superior2050 minimal-regulation2050 climate change2050 land +35 use change 2050 biomass burning +3.8 +1.9 1 14 85 +34 +5.2 +2.3 5 25 70 0.1 2.9 97

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 201 View Article Online Environmental Science: Processes & Impacts Paper Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM.

Fig. 2 Lake Superior (A), U.S. (B), global (C), and combined (D) policy total Hg deposition benefits from present day. The panels on the left represent the policy benefits in Dmg per m2 per year, while those on the right represent them as % change from the present.

The climate change increase is primarily due to an increase in deposition from biomass burning is a result of higher biomass Hg(0) dry deposition and total wet deposition uxes. The burning emissions in the boreal parts of North America and negligibly small land use/land cover changes are a result of both Western U.S. These increases are, in turn, a result of greater re increases and decreases in land cover in 2050 that inuence the activity caused by more vegetation availability and a warmer surfaces to which Hg deposits (Fig. S4†). The increase in climate in 2050.

202 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

3.2 Modeled responses of sh MeHg concentrations to changes in deposition and watershed characteristics The lake Hg cycling model was validated with measured mercury concentrations in Torch Lake.70 Measured Hg concentrations in Torch Lake water and sediments were used in calibrating the model, and the more numerous measure- ments of sh Hg concentrations were used as validation. Hg concentrations in brown bullhead, walleye, northern pike and smallmouth bass have been measured approximately every six years since 1988.102 Model predictions for both mixed feeders and piscivorous sh overlapped with the range of measured concentrations in these sh groups, although predictions underestimated mean sh concentrations. The model does Fig. 4 Spatial distribution of walleye Hg concentrations in Michigan's  not explicitly account for sh size; rather, it predicts concen- UP. Depicted are all 74 lakes for which data were located. Concen- trations for sh with a distribution of sizes about an invariant trations in each lake were calculated for walleye of 42 cm length. mean. In the UP lake, modeled sh Hg content decreases in 3.3 Spatial distribution of Hg in Michigan's UP lakes and proportion to changes in atmospheric deposition. The aspira- walleye tional scenario reduced sh Hg content by 65%, and the policy- in-action scenario led to an 11% decrease (Fig. 3). These If the 74 lakes sampled for walleye Hg are representative of the 103 responses are lake-specic and result from the combination of 15 000 UP lakes, the survey results (Fig. 4) suggest that 77% atmospheric deposition, in-lake processes, and catchment of UP lakes are impaired with respect to EPA's sh Hg criterion characteristics. If policy-in-action rates of atmospheric deposi- (0.3 ppm wet wt). The percentage of impaired lakes in the UP is tion from the Adirondacks area are applied, the model predicts considerably above the national average of 49% (ref. 104) and sh concentrations 34% higher than those with UP deposition above the 23% reported as impaired in the Adirondacks.24 rates for the same scenario. If, in addition to Adirondacks Possibly as a result of historical mining activities,105 concen- deposition rates, watershed characteristics are changed to trations of Hg in walleye are higher in western than in eastern values representative of a landscape less sensitive than the UP UP lakes (t-test, p ¼ 0.037), although fewer lakes in the east were (i.e., reduced wetland area by 38%, reduced runoff from upland sampled than in the west. Multiple linear regression and prin- soils from 10% of deposition to 5%), then sh Hg concentra- cipal component analyses indicated that sh Hg concentrations tions are predicted to be only 70% of the original prediction for are best predicted by watershed area and trophic state for large 2 UP conditions. In other words, the effects of varying landscape (>3 km ) lakes, and by pH, wetland area in catchment, lake area, characteristics within the GL region are larger than the effects of maximum depth and total phosphorus for small lakes.106 Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. the variable rates of atmospheric deposition found in this region. 4. Discussion 4.1 Regional nature of deposition responses to policy changes Mapping of the projected changes in deposition due to envi- ronmental changes (climate, land use/land cover, biomass burning) and policy clearly shows that even within the GL region there is spatial variability (Fig. 2 and Table 3). Areas in close proximity to upwind sources show a larger response to controls of those sources than do areas far from sources. In all locations, the majority of the simulated decrease in deposition is due to decreased Hg(II) deposition rather than Hg(0) deposi- tion; for the GL region as a whole, 75% of the decrease is due to Hg(II) and 25% due to Hg(0). However, in the Adirondacks, 80%

Fig. 3 Predicted Hg concentrations in piscivorous fish in the UP lake of the decrease is due to Hg(II) while in the UP 70% is due to under three policy scenarios and two virtual experiments. The aspi- Hg(II). This projected outcome results from the short atmo- rational scenario decreases fish Hg by 65% while the policy-in-action spheric residence time of Hg(II) emissions (0.5–2 days15) and scenario leads to an 11% decrease. The first virtual experiment does not reect regional differences in availability of oxidants increases deposition over the policy-in-action scenario by 32% and for Hg(0).16 The projected large decrease in Hg(II) deposition is fish Hg increases by 35%. The second virtual experiment maintains the high deposition rate but considers a less sensitive watershed (fewer at odds with the absence of a trend in wet deposition of Hg but 37 wetlands, less runoff from forests); fish Hg is only 70% of the policy-in- an observed decrease in litterfall Hg uxes in the Adirondacks action scenario with UP conditions. that are postulated to result from recent declines in Hg(0)

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 203 View Article Online Environmental Science: Processes & Impacts Paper

deposition. According to our scenario modeling, those areas median value for UP walleye. This comparison supports the currently receiving the highest rates of atmospheric deposition assertion that Michigan's UP is a highly sensitive landscape show the largest percentage decreases in deposition due to with respect to promoting Hg bioaccumulation in sh. enactment of existing policies in the U.S. In contrast, the UP is The virtual experiments, performed to quantify the magni- not downwind of major Hg emission sources. U.S. policy-in- tude of the landscape effect, suggest that landscape sensitivity action changes in emissions still represent the majority of differences within the GL region may cause larger differences in changes in deposition in the UP (70%), but the rest of the world sh Hg than do the variations in rates of atmospheric Hg contributes 1.8 times more to UP deposition than to the GL deposition. The virtual experiments kept conditions within the region, and 250 times more than to the Adirondacks. lake (Hg process rates, pH, alkalinity, sulfate, etc.) the same, but One implication of these regional differences in trajectories added sequentially an increase in atmospheric deposition of Hg deposition is that it will be easiest to detect system (32%) followed by a decrease in landscape sensitivity (reduced responses to policy in those regions projected to show the wetland area, reduced Hg runoff from forests). As expected, an largest responses. This study suggests that it will be difficult to increase in atmospheric deposition increased sh Hg concen- detect policy-related changes in lake monitoring data in Mich- trations (35%); the similarity in increases in deposition rates igan's UP over the next 50 years, unless additional policy steps and sh Hg concentrations reects the lake and watershed are taken beyond those included in the scenarios of this project. conditions used. Importantly, when the landscape sensitivity is Similarly, because the predicted changes in deposition due to changed, the sh Hg drops more than enough to counter the climate- and land-use changes are small relative to those asso- increased atmospheric deposition. The magnitude of the drop ciated with policy changes, the likelihood is small that those again reects the particular conditions that were modeled changes can be detected in the GL region unless anthropogenic (watershed : lake area ratio, wetland areas, runoff coefficients emissions stabilize. Within the GL region, changes will be used), but the conditions were selected to be representative of easiest to detect in northern Illinois, Indiana, Ohio and Penn- areas within the GL region. sylvania (Fig. 2). The factor rendering UP lakes vulnerable to high bio- accumulation of MeHg is primarily the abundance of wetlands. In the UP, 29% of the land area is occupied by wetlands; among 4.2 Local nature of lake sensitivity 85 lakes, the median percentage of the catchment occupied by Surveys23,24,107,108 have shown that lakes in certain geographic wetlands was 18%. Wetlands are important sites of Hg meth- areas are more susceptible to having high MeHg concentrations ylation; the MeHg may then be transported to lakes.114,115 in sh. This sensitivity is manifested in the geographic distri- Wetlands also export DOC to lakes; the DOC can bind with and bution of state-wide sh consumption advisories for Hg in the transport inorganic Hg to lakes.114 UP lakes are characterized by U.S.;109 each U.S. state bordering the GL or Canada has a state- relatively high DOC; among 61 UP lakes, the median DOC was wide advisory although many of these states have rates of 7.7 mg L 1. High DOC in lakes results in reduced light pene- atmospheric deposition lower than states to the south.20 Factors tration that can reduce algal production, and microbial degra-

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. responsible for this sensitivity have been shown to include dation of DOC can consume oxygen in lake hypolimnia thereby prevalence of forests and wetlands in the catchment, low alka- rendering them suitable for Hg methylation. It is instructive to linity, low pH, high sulfate loading, low nutrients and long compare characteristics promoting Hg bioaccumulation water residence times.23,107,110 Importantly, the susceptibility of between the UP and the Adirondacks (Table 4); data for 44 lakes to having high sh MeHg content is determined more by Adirondack lakes were taken from Yu et al.24 The sampled lakes factors leading to production of MeHg than by the total supply in both areas are similar in having largely forested catchments, of Hg to lakes,111 although factors affecting accumulation of and being of comparable size. The Adirondack lakes have low MeHg in biota (e.g., sh growth rate, length of food web, OC pH as expected in low alkalinity waters at high elevations that content of sediments) are also important.110,112 receive signicant acid deposition. The UP lakes have high This study suggests that Michigan's UP is a sensitive land- DOC, consistent with the high percentage of wetlands in the scape that readily converts atmospherically deposited Hg into catchments, and higher chlorophyll. Yu et al. concluded that MeHg that is bioaccumulated in sh. Among 74 lakes for which low pH and low alkalinity were the most prominent factors walleye Hg concentrations are available, 77% had concentra- contributing to Hg bioaccumulation in the Adirondack lakes. tions above the U.S. EPA criterion value of 0.3 ppm, and the Based on comparison with the Adirondack lakes, these factors median concentration was 0.43 ppm.109 Comparisons among do not appear to be the controlling factors in the sampled UP regions are difficult because of different sh species present, lakes; rather, the UP lakes are distinguished by having higher different sampling strategies, and different analytical protocols. DOC and higher wetland areas. Yu et al.24 measured Hg concentrations in multiple sh species The UP lakes will not “recover” to the point that Indigenous (not including walleye but including large- and small-mouth peoples can safely consume sh in the quantities desired by bass, sh of comparable trophic level) from 44 Adirondack 2050. Even low rates of atmospheric Hg deposition will lead to lakes, a region repeatedly called a biological hotspot for high bioaccumulation in sh. The modeled lake does not Hg.23,24,113 They found that in only 10 of 44 lakes were sh found approach the target concentration of <0.015 ppm under any with concentrations above 0.3 ppm, and in less than 10% of scenario. The aspirational scenario reduced Hg deposition lakes were sh found with concentrations above 0.42 ppm, the approximately three-fold, nearly the magnitude of the increase

204 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

Table 4 Comparison of features affecting Hg bioaccumulation in UP particularly salient for shing communities.122 Fish consumption a and Adirondack lakes advisories for Hg are widely disseminated in the GL region, but advisories are minimally effective for populations that are most Parameter UP lakes Adirondack lakesb dependent on sh.4–8 The socio-cultural reliance and dependence Non-wetland forest area (%)101 77% 74% on sh oen leads to a lack of adherence to advisory recom- Wetland in catchment (%) 18% (12%, 85) 11% mendations where Indigenous shing communities are put in 1 DOC (mg L ) 7.7 (4.9, 61) 3.9 (1.4, 44) averydifficult position. To forgo harvesting and sh consumption PH 7.5 (0.7, 99) 6.6 (0.6, 44) practices suppresses important cultural norms and prevents Alkalinity (meq L 1) 854 (820, 146) 86 (82, 44) Chlorophyll a (mg m 3) 3.7 (4.8, 105) 2.0 (3.0, 44) important knowledge transmission to future generations, yet THg (ng L 1) 0.8 (2.5, 61) 1.5 (0.9, 44) consuming sh places individuals at increased risk of physical Lake area (ha) 116 (795, 117) 132 (481, 44) harm. As a result, many Anishinaabe people feel that they cannot, – a Reported values are the median followed by the standard deviation or will not, adhere to advisory recommendations it is in essence and number of lakes in parentheses. Forest area is region-wide. b Data a ‘false choice’. 24 from Yu et al. Consumption advisories have been used to address the problem of sh contamination for almost ve decades in the U.S., however, it is important to know that these advisories were only meant to be short-term means to manage sh contamina- commonly observed in lake sediments since pre-industrial tion.119 Initially, within the GL there was an assumption that Hg times.116,117 However, our modeling approach used for the sources were regional and that regional policy could reduce and aspirational scenario would not have brought Hg deposition prevent Hg contamination to a point that advisories would no down to pre-industrial levels, because of the continuing inu- longer be necessary. We now know that atmospheric sources of ence of legacy emissions. The insensitivity of the UP to the Hg to the UP contain a signicant fraction (25%) of global source, policy-in-action scenario (15% decrease in deposition) indicates and, based on the modeling results presented here, policy that this region will require more aggressive policies to achieve oriented toward the Lake Superior Basin will only be minimally signicant reductions in deposition. The combination of the effective in achieving safe sh for vulnerable populations in the sensitive landscape and the insensitivity of Michigan's UP to UP. Although it is important to consider implications for current policies are likely to considerably prolong the time vulnerable populations in landscape sensitivity assessments, we required to see substantial decreases in sh Hg. must also emphasize the socio-cultural and political factors that signicantly increase the magnitude of Hg contamination 4.3 Implications for Indigenous communities consequences and Hg reduction policy required at differing These research results are particularly important for Indigenous scales to address them. The ‘xity to place’ (i.e.,inabilityto communities that rely on harvesting and consuming sh as relocate) that many Indigenous peoples have to their homelands, a means of maintaining cultural identity and socio-cultural both because of long-term connections and reserved treaty rights,

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. well-being; the sh harvest has been sustaining the original requires a Hg policy approach that considers the spatial distri- GL people for nearly two millennia.118 The sensitivity of the UP bution of vulnerable populations worldwide.123,124 landscape to increased Hg deposition, methylation, and MeHg Multiple limitations remain in our ability to model lake bioaccumulation is further confounded by the sensitivity of the responses to changes in Hg emissions. The scenarios used here population. The presence of other airborne bioaccumulative do not include potential longer-term changes in legacy emis- toxics in sh further compounds the problem. Native American sions of Hg; hence our predictions overestimate the real-world tribes have some of the highest documented sh consumption rate of response to changes in anthropogenic emissions, and rates in the U.S., with GL Indigenous populations currently omit the impact of legacy emissions occurring between 2000 consuming MeHg at rates that are well above human health and 2050.125,126 Legacy emissions of Hg to surface waters also criteria.118,119 For the KBIC in particular, more than seventy-ve- were not included in our modeling effort; work elsewhere percent of tribal members presently report sh as a primary indicates that such emissions prolong the time to recovery by source of subsistence.118 Subsistence harvesting is a political decades.45–47 Application of more complicated watershed right. As the oldest and largest federally-recognized Indian tribe models91 will be required to predict the timeframe of watershed in Michigan, the KBIC retains treaty-protected homelands and responses to changes in deposition. The lake Hg model can only harvesting rights across ten-million acres of the Lake Superior crudely predict responses to changes in lake pH; such changes watershed, including the majority of the western UP118,120,121 (the are likely in response to changes in atmospheric deposition of KBIC is one of thirty-six Anishinaabe tribes that retain harvest acids and sulfate and to changes in DOC inputs to lakes.127,128 treaty rights throughout the GL region). The model does not yet include sulfate deposition as a driver of Harvesting sh is not only considered a treaty right, but also an Hg methylation in wetlands or in lakes.35,42,129 All of these inherent right. That is, harvesting from the region's waters is current shortcomings, however, do not negate our prediction integral to maintaining Anishinaabe identity rooted in a tradi- that recovery will not occur in response to the policy scenarios tional way-of-life. Contamination of the sh threatens this way-of- considered here; they do prevent us from predicting the time- life and also compromises food sovereignty – an issue that is frame for recovery. common for Indigenous communities throughout the world, but

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 205 View Article Online Environmental Science: Processes & Impacts Paper

5. Conclusions References

The approach for assessing the effects of policy on safety of 1 U. S. EPA, Great Lakes Fish Monitoring and Surveillance sh consumption developed in this project provided a means Program: Background, Trends, and General Conclusions, to use state-of-the-art science tools (rigorous modeling using accessed September 9, 2017, https://www.epa.gov/great- carefully craed scenario analysis) to answer a community- lakes-monitoring/great-lakes-sh-monitoring-and- relevant question. While uncertainties in atmospheric, catch- surveillance. ment, and lake models preclude an explicit prediction of 2 U. S. EPA, National Listing of Fish Advisories, Fish and recovery times for lakes, the conclusion that recovery will not Shellsh Advisories and Safe Eating Guidelines, accessed follow quickly aer implementation of the policies modeled in September 9, 2017, https://www.epa.gov/choose-sh-and- this study is robust. Our application of this approach to the GL shellsh-wisely/sh-and-shellsh-advisories-and-safe- region points out that recovery from Hg contamination is eating-guidelines. spatially variable. Regions such as Michigan's UP that are both 3 GLIFWC, Guidance for Safe Consumption of Walleye from highly sensitive to inputs from atmospheric deposition and Inland Lakes within the Ceded Territories of Wisconsin, not close to the emission sources of that deposition will Michigan, and Minnesota, http://www.glifwc.org/Mercury/ require careful policy development to recover from current mercury.html, accessed September 9, 2017. levels of contamination. 4 J. Cassady, Med. Anthropol. Q., 2010, 24, 451–471. 5 J. Donatuto and B. Harper, Risk Anal., 2008, 28, 1497–1506. 6 J. Donatuto, T. Sattereld and R. Gregory, Health Risk Soc., Author contribution 2011, 13, 103–127. 7 C. A. O'Neill, J. Environ. Law Litigation, 2007, 22, 131–151. J. Perlinger contributed to the writing of the manuscript and its 8 D. J. Ranco and D. Suagee, Antipode, 2007, 39, 691–707. review and revisions. N. Urban conducted lake modeling, 9 V. S. Gagnon, Water History, 2016, 8, 365–384. contributed to the writing of the manuscript, and its review and 10 E. Hoover, Ecol. Process, 2013, 2,4. revision. A. Giang conducted GEOS-Chem modeling of policy 11 E. S. Norman, Area, 2013, 45, 179–187. scenarios and contributed to the Experimental section. N. Selin 12 D. J. Ranco, C. A. O'Neill, J. Donatuto and B. Harper, contributed to GEOS-Chem modeling of policy scenarios, and Environ. Justice, 2011, 4, 221–230. reviewed and revised the manuscript. A. Hendricks conducted 13 UNEP, Minamata Convention on Mercury, http:// lake modeling. H. Zhang conducted GEOS-Chem modeling of www.mercuryconvention.org/Portals/11/documents/ climate and land use/land cover scenarios and contributed to Booklets/Minamata%20Convention%20on%20Mercury_ the Experimental section. A. Kumar conducted GEOS-Chem booklet_English.pdf, accessed November 13, 2017. modeling of the biomass burning scenario and contributed to 14 V. S. Gagnon, H. S. Gorman and E. S. Norman, Gateways: the Experimental section. S. Wu contributed to GEOS-Chem International Journal of Community Research and Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. modeling related to the climate, land use, land cover, and Engagement, 2017, 10, 164–184. biomass burning scenarios and reviewed the Experimental 15 C. T. Driscoll, Y. J. Han, C. Y. Chen, D. C. Evers, section. V. Gagnon conducted university-community research K. F. Lambert, T. M. Holsen, N. C. Kamman and and engagement, and contributed to writing the manuscript, R. K. Munson, Bioscience, 2007, 57,17–28. and its review and revision. H. Gorman conducted social 16 H. M. Horowitz, D. J. Jacob, Y. X. Zhang, T. S. Dibble, science research and contributed to writing the manuscript. F. Slemr, H. M. Amos, J. A. Schmidt, E. S. Corbitt, E. Norman conducted social science research and contributed E. A. Marais and E. M. Sunderland, Atmos. Chem. Phys., to writing the manuscript. 2017, 17, 6353–6371. 17 W. Seiler, C. Eberling and F. Slemr, Pure Appl. Geophys., fl 1980, 118, 964–974. Con icts of interest 18 E. A. Nater and D. F. Grigal, Nature, 1992, 358, 139–141. There are no conicts to declare. 19 C. Seigneur, K. Vijayaraghavan, K. Lohman, P. Karamchandani and C. Scott, Environ. Sci. Technol., 2004, 38, 555–569. Acknowledgements 20 N. E. Selin and D. J. Jacob, Atmos. Environ., 2008, 42, 5193– 5204. The authors acknowledge the valuable contributions of the 21 M. D. Cohen, R. R. Draxler, R. S. Artz, P. Blanchard, anonymous reviewers to this paper. The authors also acknowl- M.S.Gustin,Y.J.Han,T.M.Holsen,D.Jaffe, P. Kelley, edge the Keweenaw Bay Indian Community, with particular H.Lei,C.P.Loughner,W.T.Luke,S.N.Lyman,D.Niemi, gratitude to the Tribal Council, the Natural Resources Depart- J.M.Pacyna,M.Pilote,L.Poissant,D.Ratte,X.Ren, ment, and the Ojibwa Community College for sharing knowl- F. Steenhuisen, A. Steffen, R. Tordon and S. J. Wilson, edge and expertise with our research team throughout the Elementa, 2016, DOI: 10.12952/journal.elementa.000118. project. This research was funded by the U.S. National Science Foundation through Grant #ICER-1313755.

206 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

22 H. Zhou, C. L. Zhou, M. M. Lynam, J. T. Dvonch, J. A. Barres, 41 J. Munthe, R. A. Bodaly, B. A. Branreun, C. T. Driscoll, P. K. Hopke, M. Cohen and T. M. Holsen, Environ. Sci. C. C. Gilmour, R. Harris, M. Horvat, M. Lucotte and Technol. Lett., 2017, 4,91–97. O. Malm, Ambio, 2007, 36,33–44. 23 D. C. Evers, Y. J. Han, C. T. Driscoll, N. C. Kamman, 42 C. J. Watras and K. A. Morrison, Can. J. Fish. Aquat. Sci., M. W. Goodale, K. F. Lambert, T. M. Holsen, C. Y. Chen, 2008, 65, 100–116. T. A. Clair and T. Butler, BioScience, 2007, 57,29–43. 43 MDEQ, Michigan's Water Chemistry Monitoring Program: A 24 X. Yu, C. T. Driscoll, M. R. Montesdeoca, D. Evers, Report of Statewide Spatial Patterns 2005-2009 and Fixed M. Duron, K. Williams, N. Schoch and N. Kamman, Station Status and Trends 1998-2008, W. R. Division Report Ecotoxicology, 2011, 20, 1543–1554. MI/DEQ/WRD-13/005, Michigan Department of 25 D. C. Depew, N. M. Burgess and L. M. Campbell, Environ. Environmental Quality, Lansing, MI, 2013. Pollut., 2013, 176, 234–243. 44 H. D. Yang, S. Turner and N. L. Rose, Environ. Pollut., 2016, 26 K. R. Rolus, J. G. Wiener, R. J. Haro, M. B. Sandheinrich, 219, 1092–1101. S. W. Bailey and B. R. Seitz, Sci. Total Environ., 2015, 514, 45 M. Meili, K. Bishop, L. Bringmark, K. Johansson, J. Munthe, 192–201. H. Sverdrup and W. de Vries, Sci. Total Environ., 2003, 304, 27 C. D. Evans, P. J. Chapman, J. M. Clark, D. T. Monteith and 83–106. M. S. Cresser, Global Change Biol., 2006, 12, 2044–2053. 46 L. Hakanson, Ecol. Model., 1996, 93, 251–262. 28 M. Erlandsson, N. Cory, S. Kohler and K. Bishop, J. Geophys. 47 K. Vijayaraghavan, L. Levin, L. Parker, G. Yarwood and Res.: Biogeosci., 2010, 115, G03004. D. Streets, Environ. Toxicol. Chem., 2014, 33, 1238–1247. 29 N. R. Urban, E. S. Verry, S. J. Eisenreich, D. F. Grigal and 48 National listing of Fish Advisories, Fish and Shellsh S. Sebestyen, Element cycling in upland/peatland Advisories and safe eating guidelines, viewed November watersheds, Peatland Biogeochemistry and Watershed 2016, https://www.epa.gov/choose-sh-and-shellsh-wisely/ Hydrology at the Marcell Experimental Forest, ed. R. K. sh-and-shellsh-advisories-and-safe-eating-guidelines, Kolka, S. Sebestyen, E. S. Verry and K. Brooks, CRC Press, accessed November 13 2017. Boca Raton, FL, 2011, pp. 213–241. 49 E. Oken, R. O. Wright, K. P. Kleinman, D. Bellinger, 30 L. Lindestrom, Ambio, 2001, 30, 538–544. C. J. Amarasiriwardena, H. Hu, J. W. Rich-Edwards and 31 M. J. Parsons, D. T. Long and S. S. Yohn, Appl. Geochem., M. W. Gillman, Environ. Health Perspect., 2005, 113, 1376– 2010, 25, 1676–1687. 1380. 32 R. C. Harris, J. W. M. Rudd, M. Amyot, C. L. Babiarz, 50 E. Groth, An Overview of Epidemiological Evidence on the K. G. Beaty, P. J. Blancheld, R. A. Bodaly, B. A. Branreun, Effects of Methylmercury on Brain Development, and A C. C. Gilmour, J. A. Graydon, A. Heyes, H. Hintelmann, Rationale for a Lower Denition of Tolerable Exposure Zero J. P. Hurley, C. A. Kelly, D. P. Krabbenho,S.E.Lindberg, Mercury Working Group, Brussels, Belgium, 2012. R. P. Mason, M. J. Paterson, C. L. Podemski, A. Robinson, 51 A. Giang and N. E. Selin, Proc. Natl. Acad. Sci. U. S. A., 2016, K. A. Sandilands, G. R. Southworth, V. L. S. Louis and 113, 286–291.

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. M. T. Tate, Proc. Natl. Acad. Sci. U. S. A.,2007,104, 16586– 52 S. Wu, L. J. Mickley, D. J. Jacob, J. A. Logan, R. M. Yantosca 16591. and D. Rind, J. Geophys. Res., 2007, 112, D05302. 33 C. J. Watras, K. A. Morrison, R. J. M. Hudson, T. M. Frost 53 X. M. Zhang, K. R. Rygwelski, M. D. Rowe, R. Rossmann and and T. K. Kratz, Environ. Sci. Technol., 2000, 34, 4051–4057. R. G. Kreis, J. Great Lakes Res., 2016, 42,62–69. 34 T. R. Hrabik and C. J. Watras, Sci. Total Environ., 2002, 297, 54 S. Wu, L. J. Mickley, J. O. Kaplan and D. J. Jacob, Atmos. 229–237. Chem. Phys., 2012, 12, 1597–1609. 35 P. E. Drevnick, D. E. Caneld, P. R. Gorski, 55 H. Zhang, C. D. Holmes and S. Wu, Atmos. Environ., 2016, A. L. C. Shinneman, D. R. Engstrom, D. C. G. Muir, 141, 230–244. G. R. Smith, P. J. Garrison, L. B. Cleckner, J. P. Hurley, 56 Y. X. Huang, S. L. Wu and J. O. Kaplan, Atmos. Environ., R. B. Noble, R. R. Otter and J. T. Oris, Environ. Sci. 2015, 121,86–92. Technol., 2007, 41, 7266–7272. 57 A. Kumar, S. Wu, Y. Huang, H. Liao and J. Kaplan, Atmos. 36 N. Gandhi, R. W. K. Tang, S. P. Bhavsar and Environ., 2017, 173,6–15. G. B. Arhonditsis, Environ. Sci. Technol., 2014, 48, 5404– 58 E. S. Corbitt, D. J. Jacob, C. D. Holmes, D. G. Streets and 5414. E. M. Sunderland, Environ. Sci. Technol., 2011, 45, 10477– 37 J. R. Gerson and C. T. Driscoll, Environ. Sci. Technol., 2016, 10484. 50, 10943–10950. 59 Y. Zhang, L. Jaegle, A. van Donkelaar, R. V. Martin, 38 C. D. Knightes, E. M. Sunderland, M. Craig Barber, C. D. Holmes, H. M. Amos, Q. Wang, R. Talbot, R. Artz, J. M. Johnston and R. B. Ambrose, Environ. Toxicol. S. Brooks, W. Luke, T. M. Holsen, D. Felton, E. K. Miller, Chem., 2009, 28, 881–893. K. D. Perry, D. Schmeltz, A. Steffen, R. Tordon, P. Weiss- 39 C. D. Knightes and R. B. Ambrose, Environ. Toxicol. Chem., Penzias and R. Zsolway, Atmos. Chem. Phys., 2012, 12, 2007, 26, 807–815. 6095–6111. 40 N. E. Selin, E. M. Sunderland, C. D. Knightes and 60 D. G. Streets, Q. Zhang and Y. Wu, Environ. Sci. Technol., R. A. Mason, Environ. Health Perspect., 2010, 118, 137–143. 2009, 43, 2983–2988.

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 207 View Article Online Environmental Science: Processes & Impacts Paper

61 P. Rafaj, I. Bertok, J. Cofala and W. Schopp, Atmos. Environ., 81 T. R. Khan and J. A. Perlinger, Geosci. Model Dev., 2017, 10, 2013, 79, 472–479. 3861–3888. 62 N. E. Selin, Environ. Toxicol. Chem., 2014, 33, 1202–1210. 82 J. S. Read, D. P. Hamilton, A. R. Desai, K. C. Rose, 63 J. M. Pacyna, O. Travnikov, F. De Simone, I. M. Hedgecock, S. MacIntyre, J. D. Lenters, R. L. Smyth, P. C. Hanson, K. Sundseth, E. G. Pacyna, F. Steenhuisen, N. Pirrone, J. J. Cole, P. A. Staehr, J. A. Rusak, D. C. Pierson, J. Munthe and K. Kindbom, Atmos. Chem. Phys., 2016, 16, J. D. Brookes, A. Laas and C. H. Wu, Geophys. Res. Lett., 12495–12511. 2012, 39, DOI: 10.1029/2012GL051886. 64 Lake Superior Binational Program, Lake Superior Zero 83 B. D. Hall, H. Manolopoulos, J. P. Hurley, J. J. Schauer, Discharge Demonstration Program and Critical Chemical V. L. St Louis, D. Kenski, J. Graydon, C. L. Babiarz, Reduction Milestones, 2017, https://www.epa.gov/sites/ L. B. Cleckner and G. J. Keeler, Atmos. Environ., 2005, 39, production/les/2015-11/documents/lake-superior-zero- 7557–7569. discharge-demonstration-program-2012-8pp.pdf, accessed 84 M. R. Risch, J. F. Dewild, D. A. Gay, L. Zhang, E. W. Boyer November 13, 2017. and D. P. Krabbenho, Environ. Pollut., 2017, 228,8–18. 65 H. M. Amos, D. J. Jacob, C. D. Holmes, J. A. Fisher, Q. Wang, 85 M. R. Risch, D. A. Gay, K. K. Fowler, G. J. Keeler, R. M. Yantosca, E. S. Corbitt, E. Galarneau, A. P. Rutter, S. M. Backus, P. Blanchard, J. A. Barres and J. T. Dvonch, M. S. Gustin, A. Steffen, J. J. Schauer, J. A. Graydon, Environ. Pollut., 2012, 161, 261–271. V. L. St Louis, R. W. Talbot, E. S. Edgerton, Y. Zhang and 86 A. Hendricks, A Model to Predict Concentrations and E. M. Sunderland, Atmos. Chem. Phys., 2012, 12, 591–603. Uncertainty for Mercury Species in Lakes, M.S. thesis, 66 C. D. Holmes, D. J. Jacob, E. S. Corbitt, J. Mao, X. Yang, Michigan Technological University, 2018. R. Talbot and F. Slemr, Atmos. Chem. Phys., 2010, 10, 87 D. F. Grigal, Environ. Rev., 2002, 10,1–39. 12037–12057. 88 H. Hsu-Kim, C. S. Eckley, D. Acha, X. Feng, C. Gilmour, 67 N. E. Selin, D. J. Jacob, R. M. Yantosca, S. Strode, L. Jaegle S. Jonsson and C. P. J. Mitchell, Ambio, in review. and E. M. Sunderland, Global Biogeochem. Cycles, 2008, 89 M. N. Futter, A. E. Poste, D. Buttereld, P. J. Dillon, 22, GB 2011. P. G. Whitehead, A. P. Dastoor and D. R. S. Lean, Sci. 68 A. L. Soerensen, E. M. Sunderland, C. D. Holmes, Total Environ., 2012, 424, 219–231. D. J. Jacob, R. M. Yantosca, H. Skov, J. H. Christensen, 90 H. E. Golden, C. D. Knightes, P. A. Conrads, G. M. Davis, S. A. Strode and R. P. Mason, Environ. Sci. Technol., 2010, T. D. Feaster, C. A. Journey, S. T. Benedict, M. E. Brigham 44, 8574–8580. and P. M. Bradley, J. Geophys. Res.: Biogeosci., 2012, 117, 69 D. Rind, J. Lerner, J. Jonas and C. McLinden, J. Geophys. G01006. Res.: Atmos., 2007, 112, D09315. 91 H. E. Golden, C. D. Knightes, P. A. Conrads, T. D. Feaster, 70 A. N. Hendricks and N. R. Urban, Ecol. Model., in prep. G. M. Davis, S. T. Benedict and P. M. Bradley, Environ. 71 K. Soetaert, T. Petzoldt and R. W. Setzer, J. Stat. Soware, Toxicol. Chem., 2013, 32, 2165–2174. 2010, 33,1–25. 92 R. B. Mills, A. M. Paterson, J. M. Blais, D. R. S. Lean,

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. 72 C. D. Knightes and R. B. Ambrose, Development of an J. P. Smol and G. Mierle, Can. J. Fish. Aquat. Sci., 2009, 66, Ecological Risk Assessment Methodology for Assessing 187–200. Wildlife Exposure Risk Associated with Mercury- 93 J. T. Bushey, C. T. Driscoll, M. J. Mitchell, P. Selvendiran Contaminated Sediments in Lake and River Systems, Report and M. R. Montesdeoca, Hydrol. Processes, 2008, 22, 4813– EPA/600/R-06/073, U.S. Environmental Protection Agency 4826. Office of Research and Development, Washington, D.C., 94 R. K. Kolka, D. F. Grigal, E. A. Nater and E. S. Verry, Soil Sci. 2006. Soc. Am. J., 2001, 65, 897–905. 73 C. D. Knightes, Environ. Model. Soware, 2008, 23, 495–510. 95 P. C. Frost, J. H. Larson, G. A. Lamberti, C. A. Johnston, 74 L. Hakanson, Water, Air, Soil Pollut., 2000, 124, 301–317. K. C. Young, P. A. Maurice and S. D. Bridgham, Aquat. 75 M. C. Barber, Environ. Toxicol. Chem., 2008, 27, 755–777. Sci., 2006, 68,40–51. 76 E. B. Swain, D. R. Engstrom, M. E. Brigham, T. A. Henning 96 C. Babiarz, S. Hoffmann, A. Wieben, J. Hurley, A. Andren, and P. L. Brezonik, Science, 1992, 257, 784–787. M. Shafer and D. Armstrong, Environ. Pollut., 2012, 161, 77 P. Lorey and C. T. Driscoll, Environ. Sci. Technol., 1999, 33, 299–310. 718–722. 97 P. A. Soranno, K. S. Cheruvelil, T. Wagner, K. E. Webster 78 P. E. Drevnick, D. R. Engstrom, C. T. Driscoll, E. B. Swain, and M. T. Bremigan, PLoS One, 2015, 10(8), e0135454. S. J. Balogh, N. C. Kamman, D. T. Long, D. G. C. Muir, 98 K. S. Cheruvelil, P. A. Soranno, K. E. Webster and M. J. Parsons, K. R. Rolus and R. Rossmann, Environ. M. T. Bremigan, Ecol. Appl., 2013, 23, 1603–1618. Pollut., 2012, 161, 252–260. 99 J. E. Breck, Compilation of Databases on Michigan Lakes 79 J. Kwan and P. A. Taylor, Bound.-Layer Meteorol., 1994, 68, MDNR Fisheries Division Lansing, MI, 2004. 339–356. 100 D. Knauer, G. Pelkola, S. Casey, D. Krabbenho and 80 D. E. Anderson, R. G. Striegl, D. I. Stannard, J. F. Dewild, The distribution of total and methyl mercury in C. M. Michmerhuizen, T. A. McConnaughey and Michigan's Upper Peninsula Lakes, Michigan Dept. J. W. LaBaugh, Limnol. Oceanogr., 1999, 44, 988–1001. Environmental Quality, Lansing, MI, 2011.

208 | Environ. Sci.: Processes Impacts,2018,20,195–209 This journal is © The Royal Society of Chemistry 2018 View Article Online Paper Environmental Science: Processes & Impacts

101 National Land Cover Database 2011, https://www.mrlc.gov/ 114 P. Selvendiran, C. T. Driscoll, J. T. Bushey and nlcd2011.php, accessed November 13, 2017. M. R. Montesdeoca, Environ. Pollut., 2008, 154,46–55. 102 MDEQ, Status of Fish Contaminant Levels in the Torch Lake 115 P. Selvendiran, C. T. Driscoll, M. R. Montesdeoca and Area of Concern 2013, W. R. D. Michigan Department of J. T. Bushey, J. Geophys. Res.: Biogeosci., 2008, 113, G00C01. Environmental Quality Report MI/DEQ/WRD-16/003, 116 D. R. Engstrom, E. B. Swain, T. A. Henning, M. E. Brigham Michigan Dept. Environmental Quality, Lansing, and P. L. Brezonik, Adv. Chem. Ser., 1993, 237,33–66. Michigan, 2016. 117 D. R. Engstrom, W. F. Fitzgerald, C. A. Cooke, 103 C. P. McDonald, J. A. Rover, E. G. Stets and R. G. Striegl, C. H. Lamborg, P. E. Drevnick, E. B. Swain, S. J. Balogh Limnol. Oceanogr., 2012, 57, 597–606. and P. H. Balcom, Environ. Sci. Technol., 2014, 48, 6533– 104 L. L. Stahl, B. D. Snyder, A. R. Olsen and J. L. Pitt, Environ. 6543. Monit. Assess., 2009, 150,3–19. 118 V. S. Gagnon, P. Nankervis and E. Johnston, Keweenaw Bay 105 W. C. Kerfoot, N. R. Urban, C. P. McDonald, H. Zhang, Indian Community Wildlife and Natural Resources Report, R. Rossmann, J. A. Perlinger, T. Khan, A. Hendricks, KBIC Natural Resources Department, Baraga, MI, M. Priyadarshini and M. Bolstad, Environ. Sci.: Processes 2013, http://nrd.kbic-nsn.gov/sites/default/les/KBIC% Impacts, 2018, in press. 20WILDLIFE%20and%20NR%20REPORT%2011_2013% 106 M. Priyadarshini, Factors Contributing to Elevated 20%28hyperlinked%20TABLE%20of%20CONTENTS%29_0. Concentrations of Mercury and PCBs in Fish in the pdf, accessed November 13, 2017. Inland Lakes of Michigan's Upper Peninsula and Lake 119 C. A. O'Neill, Mercury, risk & justice, Environmental Law Superior, M.S. thesis, Michigan Technological University, Institute, Environmental Law Reporter News and Analysis, M.S., 2017. 2004. 107 C. A. Eagles-Smith, J. T. Ackerman, J. J. Willacker, 120 A. McCammon-Soltis and J. S. Kekek, Fullling Ojibwe M. T. Tate, M. A. Lutz, J. A. Fleck, A. R. Stewart, treaty promises—an overview and compendium of J. G. Wiener, D. C. Evers, J. M. Lepak, J. A. Davis and relevant cases, statutes and agreement, Great Lakes Indian C. F. Pritz, Sci. Total Environ., 2016, 568, 1171–1184. Fish & Wildlife Commission Division of Intergovernmental 108 D. C. Evers, J. G. Wiener, C. T. Driscoll, D. A. Gay, N. Basu, Affairs, 2009, http://www.glifwc.org/minwaajimo/Papers/ B. A. Monson, K. F. Lambert, H. A. Morrison, J. T. Morgan, Legal%20Paper%20-%20DIA.pdf, accessed November 13, K. A. Williams and A. G. Soehl, Great Lakes Mercury 2017. Connections: The Extent and Effects of Mercury Pollution in 121 C. F. Wilkinson, Blood Struggle: The Rise of Modern Indian the Great Lakes Region, Biodiversity Research Institute, Nations, NY, 2005. Gorham, Maine, 2011. 122 E. S. Norman, Soc. Nat. Resour., 2017, 30, 537–553. 109 National Listing of Fish Advisories, https://www.epa.gov/ 123 E. S. Norman, Water Alternatives, 2012, 5, 138–160. sites/production/les/2015-06/documents/maps-and- 124 E. S. Norman, Governing Transboundary Waters: Canada, the graphics-2011.pdf, accessed November 13, 2017. United States, and Indigenous Communities, Routledge,

Published on 23 January 2018. Downloaded 4/17/2019 6:00:29 PM. 110 C. Y. Chen, C. L. Folt, R. S. Stemberger, N. C. Kamman and Florence, KY, 2014. B. M. Mayes, Ecotoxicology, 2005, 14, 135–147. 125 H. M. Amos, D. J. Jacob, D. G. Streets and E. M. Sunderland, 111 C. A. Eagles-Smith, J. G. Wiener, C. S. Eckley, J. J. Willacker, Global Biogeochem. Cycles, 2013, 27, 410–421. D. C. Evers, M. Marvin-DiPasquale, D. Obrist, J. A. Fleck, 126 M. Jiskra, J. G. Wiederhold, U. Skyllberg, R. M. Kronberg, G. R. Aiken, J. M. Lepak, A. K. Jackson, J. P. Webster, I. Hajdas and R. Kretzschmar, Environ. Sci. Technol., 2015, A. R. Stewart, J. A. Davis, C. N. Alpers and J. T. Ackerman, 49, 7188–7196. Sci. Total Environ., 2016, 568, 1213–1226. 127 M. Erlandsson, I. Buffam, J. Folster, H. Laudon, 112 C. Y. Chen, M. Dionne, B. M. Mayes, D. M. Ward, S. Sturup J. Temnerud, G. A. Weyhenmeyer and K. Bishop, Global and B. P. Jackson, Environ. Sci. Technol., 2009, 43, 1804– Change Biol., 2008, 14, 1191–1198. 1810. 128 B. L. Skjelkvale, H. Borg, A. Hindar and A. Wilander, Appl. 113 D. C. Evers, R. P. Mason, N. C. Kamman, C. Y. Chen, Geochem., 2007, 22, 1174–1180. A. L. Bogomolni, D. L. Taylor, C. R. Hammerschmidt, 129 J. D. Jeremiason, D. R. Engstrom, J. E. Almendinger, S. H. Jones, N. M. Burgess, K. Munney and K. C. Parsons, E. B. Swain, B. A. Monson, E. A. Nater, B. M. Johnson and EcoHealth, 2008, 5, 426–441. R. K. Kolka, Environ. Sci. Technol., 2006, 40, 3800–3806.

This journal is © The Royal Society of Chemistry 2018 Environ. Sci.: Processes Impacts,2018,20,195–209 | 209 Policy Analysis

Cite This: Environ. Sci. Technol. 2018, 52, 12968−12977 pubs.acs.org/est

Global and Local Impacts of Delayed Mercury Mitigation Efforts † ‡ † § ∥ ⊥ Helénè Angot,*, Nicholas Hoffman, Amanda Giang, , Colin P. Thackray, Ashley N. Hendricks, ⊥ † ‡ Noel R. Urban, and Noelle E. Selin , † Institute for Data, Systems, and Society, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States ‡ Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States § Institute for Resources, Environment and Sustainability, University of British Columbia, Vancouver, British Columbia Canada V6T 1Z4 ∥ Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, United States ⊥ Civil and Environmental Engineering Department, Michigan Technological University, Houghton, Michigan 49931, United States

*S Supporting Information

ABSTRACT: Mercury (Hg) is emitted to air by natural and anthropogenic sources, transports and deposits globally, and bioaccumulates to toxic levels in food webs. It is addressed under the global 2017 Minamata Convention, for which periodic effectiveness evaluation is required. Previous analyses have estimated the impact of different regulatory strategies for future mercury deposition. However, analyses using atmospheric models tradition- ally hold legacy emissions (recycling of previously deposited Hg) constant, and do not account for their possible future growth. Here, using an integrated modeling approach, we investigate how delays in implementing emissions reductions and the associated growing legacy reservoir affect deposition fluxes to ecosystems in different global regions. Assuming nearly constant yearly emissions relative to 2010, each 5-year delay in peak emissions defers by additional extra ca. 4 years the return to year 2010 global deposition. On a global average, each 5-year delay leads to a 14% decrease in policy impacts on local-scale Hg deposition. We also investigate the response of fish contamination in remote lakes to delayed action. We quantify the consequences of delay for limiting the Hg burden of future generations and show that traditional analyses of policy impacts provide best-case estimates.

1. INTRODUCTION impact on the global Hg cycle.10 Legacy emissions (i.e., Mercury (Hg) is an environmental toxicant dangerous to recycling of previously deposited Hg) from soil and oceanic Downloaded via UNIV OF BRITISH COLUMBIA on April 17, 2019 at 17:00:52 (UTC). reservoirs account for about three-fifths of Hg annually emitted human health and the environment. Because of its long lifetime 11 − 1,2 to the atmosphere. Even if anthropogenic emissions stay

See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles. in the atmosphere (0.3 1 year), Hg travels regionally and constant, Hg deposition will continue to increase due to legacy globally in its gaseous elemental form (Hg(0)). It deposits to 11,12 ecosystems by wet and dry processes as Hg(0) and gaseous/ emissions. particulate divalent Hg (Hg(II)), and converts to highly toxic Parameterizing legacy emissions in atmospheric models is methylmercury (MeHg) which bioaccumulates in aquatic challenging due to a paucity of models which can capture both systems.1,3 Fish consumption is thus a main source of exposure three-dimensional atmospheric transport and oceanic/terres- − 13,14 to Hg for the general population.4 6 trial cycling simultaneously. The majority of policy Regulatory actions to reduce human exposure to Hg aim to analyses conducted using atmospheric models therefore only reduce anthropogenic inputs to the environment. In that reflect changes in direct anthropogenic emissions, and do not context, the Minamata Convention on Mercury7 entered into consider the effect of changing legacy emissions. For example, force in August 2017 and has 101 parties as of November Pacyna et al.9 estimated that year 2035 global anthropogenic 2018. Under Article 8, parties “shall take measures to control, emissions would be reduced by 940 tons compared to 2010 and where feasible, reduce emissions of Hg to the atmosphere”, and anthropogenic Hg emissions to the atmosphere are Received: August 13, 2018 projected to begin to decrease based on current and enhanced Revised: October 17, 2018 policy efforts.8,9 However, emitted Hg circulates for decades to Accepted: October 30, 2018 centuries, and anthropogenic emissions have a long-lasting Published: October 30, 2018

© 2018 American Chemical Society 12968 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

Figure 1. Conceptual framework. Due to a challenging parametrization, chemical transport models (CTMs, GEOS-Chem in this study) traditionally hold legacy emissions constant at present-day (2010) levels when making future projections. The latter thus only reflect changes in direct anthropogenic emissions. Using a fully coupled seven-reservoir global biogeochemical cycle model11,16 we account for future legacy emissions as a result of both past and future emissions by adjusting GEOS-Chem outputs. The effective policy impacts in terms of local Hg deposition are then dependent on policy delay. The adjusted deposition flux is then used as input to a lake Hg model to evaluate the response of fish contamination to delayed action. Current Policy (CP), New Policy (NP) and Maximum Feasible Reduction (MFR) refer to future global emissions scenarios developed by Pacyna et al.9

Table 1. Simulations Performed with the Chemical Transport Model (CTM) GEOS-Chem and the Seven-Reservoir Global a Biogeochemical Cycle (GBC) Model11,16

simulation CTM meteorological year simulated: 2010b GBC model years simulated: 2000 BCE-2100 CE BASE 2010 AMAP/UNEP inventory17 and emissions controls18c Street et al.19 from 2000 BCE to 2008 CE. CP from 2009 onward

PRE-2010 Primary anthropogenic emissions zeroed out Primary anthropogenic emissions completely eliminated as of 2010 LEGACY

FUTURE Future (CP, NP or MFR) emissions inventories9 NP or MFR implemented in YYYY = 2010, 2020, 2025, 2030, 2035, or 2050

PRE-YYYY Same as PRE-2010 LEGACY since 2010-YYYY emissions are Primary anthropogenic emissions completely eliminated as of: YYYY = 2020, LEGACY not taken into account 2025, 2030, 2035, or 2050 aCurrent Policy (CP), New Policy (NP), and Maximum Feasible Reduction (MFR) refer to future global emissions scenarios developed by Pacyna et al.9 A number of simulations listed here were performed to check the robustness of the method (see SI Section 1.1). bThe model was run for meteorological years 2007−2010, with the first three years used as the initialization period. We used consistent 2007−2010 meteorology for present and future runs to isolate the effect of emissions. cIn order to evaluate present-day model outputs against observations and account for interannual variability, this simulation was also performed for meteorological years 2009−2015 following a three-year spin-up. and Hg deposition to ecosystems by 20−30% (except in India) We further examine consequences of delayed global mitigation if policy commitments and plans are fully implemented. efforts in different regions of the world to illustrate However, in their analysis, while anthropogenic and natural implications relevant to environmental justice concerns. For sources keep emitting Hg during the period 2010−2035 and one of these regions, a Native American community in the the Hg legacy reservoir grows, legacy emissions were simulated U.S., we examine the impact on fish contamination in remote at 2010 level.9 Another recent analysis15 of future policy, lakes where the main source of contamination is atmospheric investigating projected Hg deposition in Asia by 2050, also did deposition from the global Hg pool. not account for legacy Hg. However, using a biogeochemical cycle model, the authors estimated that legacy Hg changes 2. MATERIALS AND METHODS could alter the magnitude of calculated policy impacts by 30%, 2.1. Conceptual Framework. The conceptual framework ff but they did not resolve this e ect spatially. of this study is presented in Figure 1. Present-day (2010) and Here, using an integrated approach that combines future simulations were performed using the chemical biogeochemical cycle modeling with global-scale chemical transport model (CTM) GEOS-Chem (see Section 2.3). As transport modeling, we investigate the impacts of delayed in traditional future projections with other atmospheric global action on global- and local-scale Hg deposition and models,9 legacy emissions were initially assumed constant to evaluate associated changes in policy impacts, including their 2010 level to calculate the effects of changes in direct spatial resolution. We test the hypothesis that a longer delay in anthropogenic emissions (see Section 2.2). A global bio- near-term peak emissions will lead to a larger Hg legacy pool geochemical cycle (GBC) model was used to calculate a global and thus a measurable influence on expected policy impacts. legacy penalty as a function of the amount of emissions since

12969 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

2010 (see Section 2.4). This global legacy penalty was then we did not include additional 1850−2008 atmospheric Hg spatially distributed and added to future deposition fluxes from emissions from commercial products (105 Gg) proposed by the CTM (see Section 2.4). Using this approach, adjusted Hg Horowitz et al.22 For 2009-onward primary anthropogenic deposition from the CTM and the effective policy impacts emissions, we used future emissions scenarios developed by differ as a function of the total amount of mercury emitted, Pacyna et al.9 as described above. whichisgreaterforlongerpolicydelay.AdjustedHg 2.3. Chemical Transport Modeling. The global CTM deposition fluxes were then used as inputs to a lake Hg GEOS-Chem (www.geos-chem.org) was used to project model (see Section 2.5) to investigate the influence of delayed present-day and future total (wet+dry) gross Hg deposition global action on lacustrine fish contamination. fluxes to ecosystems. The model is driven by assimilated Throughout this paper, we compare policy impacts assuming meteorological data from the NASA GMAO Goddard Earth an immediate (i.e., traditional method−no delay/legacy Observing System.23 MERRA-2 data were used for the penalty) or delayed global action to reduce emissions. For simulations (https://gmao.gsfc.nasa.gov/products/). GEOS- each grid box of the CTM, policy impacts (PI, in Δμg/m2/yr) Chem is a global-scale model that couples a 3D atmos- are calculated as the difference in total deposition between phere,2,24,25 a 2D surface-slab ocean,26 and a 2D terrestrial future and present-day simulations. The percent change (PC) reservoir27 at a 2°×2.5° horizontal resolution. A two-step in policy impacts due to a global action delayed to year YYYY oxidation mechanism of Hg(0) initiated by Br was used. The (YYYY = 2020, 2025, 2030, 2035, or 2050) is calculated second-stage HgBr oxidation is mainly by the NO2 and HO2 according to eq 1: radicals using the new mechanism for atmospheric redox 2 fi PI− PI chemistry developed by Horowitz et al. Oxidant elds from action delayed to YYYY traditional method Schmidt et al.28 have 4°× 5° horizontal resolution. Photo- PCYYYY = × 100 PItraditional method reduction of aqueous-phase Hg(II)−organic complexes is (1) dependent on the local concentration of organic aerosols, the ffi NO2 photolysis frequency, and an adjusted coe cient The mean percent change in policy impacts, given by eq 2,is − − − (K_RED_JNO2) set to 9.828 × 10 2 m3 μg 1 here for the average percent change due to a near-term (2020 2035) fi 5-year delay: simulations with MERRA-2 meteorological elds and use of the slab ocean.29 For further details, a comprehensive (PC−+−+− PC ) (PC PC ) (PC PC ) 30 PC = 2035 2030 2030 2025 2025 2020 description of the model is available elsewhere. 3 2.4. Legacy Penalty. We used a previously published fully (2) coupled seven-reservoir GBC model11,16 (available at https:// 2.2. Present-Day and Future Emissions Scenarios. github.com/SunderlandLab/gbc-boxmodel) to scale legacy Simulations performed with the CTM and the GBC model are emissions. The model allows full coupling of the atmosphere, listed in Table 1. The GEOS-Chem present-day (2010) BASE ocean (surface, subsurface, and deep ocean), and terrestrial simulation was performed using the AMAP/UNEP inven- ecosystems (fast terrestrial, slow and armored soils). Hg cycles tory,17 applying emission controls to U.S., Canadian, Euro- between reservoirs and is ultimately removed by burial in deep pean, and Chinese emissions from coal fired power plants.18 marine sediments. In order to evaluate the impact of delayed The CTM PRE-2010 LEGACY simulation was performed to action on global Hg deposition, incremental 5-year delays in quantify deposition from current (2010) legacy emissions and implementing a NP or MFR scenario were tested (see Table 1, FUTURE simulations). We assumed a CP scenario (i.e., a 3.02 evaluate its spatial pattern. FUTURE simulations were − performed using gridded emissions inventories developed by Mg yr 1 increase) until implementation of a NP or MFR Pacyna et al.9 Briefly, the Current Policy (CP) scenario scenario. A total of six simulations were performed with the projects that annual Hg emissions will slightly increase in 2035 GBC model (PRE-2010 and PRE-YYYY LEGACY in Table 1) (ca. +75 Mg compared to 2010, i.e., +3.02 Mg yr−1). in order to quantify the contribution to global Hg deposition Increasing energy demand contributing to increased emissions of emissions during the policy delay Δt. The Δt-dependent globally will be offset by the implementation of additional global legacy penalty was defined as the difference in global control measures. The more stringent New Policy (NP) deposition between PRE-YYYY and PRE-2010 LEGACY scenario indicates that annual emissions will significantly simulations (where YYYY = 2020, 2025, 2030, 2035, or decrease by 2035 (ca. −820 Mg compared to 2010, i.e., −32.7 2050). The global legacy penalty was then spatially distributed Mg yr−1). This scenario assumes that policy commitments and (see below) and added to future deposition fluxes from the plans announced by countries worldwide to reduce greenhouse CTM in each grid-box (see Figure 1). Rather than assuming a gas emissions and phase out fossil fuel subsidies are fully globally homogeneous distribution of legacy emissions (i.e., implemented. Additionally, this scenario assumes that the use global legacy penalty evenly divided among all 2°×2.5° grid of Hg in products will be reduced by 70% by implementing boxes of the CTM), we used the spatial distribution of the Article 4 of the Minamata Convention on Hg-added products. legacy emissions contribution to Hg deposition (PRE-2010 Finally, the Maximum Feasible Reduction (MFR) scenario LEGACY simulation with the CTM, see Table 1). Spatial leads to a dramatic decrease of annual Hg anthropogenic differences in legacy impact relate to atmospheric transport, emissions (ca. −1500 Mg compared to 2010, that is, −59.9 Mg geographic factors (e.g., land vs ocean) and reemissions. In yr−1). In this scenario, all countries reach the highest feasible GEOS-Chem, reemission of Hg previously deposited to land reduction efficiency in each emission sector. follows the deposition patterns of current sources.31 The GBC model was driven by 2000 BCE to 2008 CE Local consequences of delayed global action for Hg primary anthropogenic emissions from Streets et al.,19 Hg deposition fluxes to ecosystems are discussed (see Section discharges from rivers (held constant at present-day levels),16 3.2) at four selected sites located at varying distance from and global geogenic emissions (90 Mg yr−1).20,21 For anthropogenic sources: (A) tribal areas of eastern Maine, consistency with the emissions inventories used in the CTM, representative of remote regions and used to illustrate

12970 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

Figure 2. (a) Global primary anthropogenic emissions of Hg to the atmosphere (in Mg). A New Policy (NP) is implemented in 2010 (black), 2020 (blue), 2025 (red), 2030 (green), 2035 (yellow) or 2050 (purple). The NP annual emissions target9 is reached at a rate of −32.7 Mg yr−1even in case of delayed global action. (b) Global atmospheric Hg deposition to ecosystems (in Mg). Each 5 year delay in implementing NP delays by additional extra ca. 4 years the return of Hg deposition to its year 2010 level (chosen for illustrative purposes) due to legacy emissions. implications relevant to environmental justice concerns (see Information (SI) Section 1.2.a). This approach assumes that Section 3.3); (B) Ahmedabad, the largest city of the Indian only recently deposited Hg is susceptible to leaching leading to state of Gujarat and the location of two coal-fired power plants a fast response to changes in Hg deposition.48 More of more than 1000 MW electricity generation;32 (C) Shanghai, information regarding the parametrization used here and the China’s biggest city and one of the main industrial centers, model performance can be found in SI Section 1.2 and Figure where elevated atmospheric Hg concentrations have been S1. − reported;33 35 and (D) an area of the Southern Pacific known Although some lake Hg contamination can be attributed to 51,52 for albacore tuna fisheries.36 Sunderland et al.37 recently direct inputs from local sources, we focus here on remote estimated that seafood harvested from the Equatorial and lakes where the main source of contamination is atmospheric South Pacific Ocean accounts for 25% of the U.S. population- deposition from the global Hg pool (see Section 3.2). More wide MeHg intake. specifically, we concentrate on remote tribal regions of eastern 2.5. Fish Contamination. To investigate the influence of Maine (see SI Figure S2) since Native Americans are delayed global action on fish contamination, we used a recently particularly vulnerable to Hg contamination due to traditional fi 53 developed implementation of the mechanistic model SERAFM subsistence shing. In order to investigate the response to fi (Spreadsheet-based Ecological Risk Assessment for the Fate of changes in atmospheric deposition, the model was rst run for − Mercury) of Hg in aquatic environments developed by the U.S. 10 years (2000 2010) to reach steady-state, and then Environmental Protection Agency (EPA).38,39 This model has transiently using the adjusted deposition values from the − been widely used and evaluated since its development.40 46 CTM described in Section 2.1. Here, we evaluated the fi The implementation of SERAFM used here was developed by response of sh contamination to delayed global action Hendricks47 in the programming language R and modified for assuming everything else (e.g., food web structure, nutrient loading) constant. While Hg biogeochemical cycling will be nonsteady state conditions to enable prediction of lake ff 14,54−56 responses to changes in loadings. A description of this mass a ected by climate and land-use change, this is not balance model is provided by Perlinger et al.48 Briefly, the taken into account here for consistency and ease of comparison model is a three-reservoir box model (epilimnion, hypolimn- with other traditional policy impact studies. ion, and sediments). It enables prediction of aqueous Hg concentrations based on the characteristics of the lake of 3. RESULTS AND DISCUSSION interest (e.g., depth, retention time), its watershed (e.g., 3.1. Impact of Delayed Action on Global Hg surface area), and the local Hg atmospheric deposition flux. Deposition. The impact of delayed action on global Hg The model does not include fish population dynamics and deposition was quantified using the fully coupled seven- solves for an annually averaged MeHg concentration in the reservoir GBC model11,16 (FUTURE simulations, see Table water column that is multiplied by bioaccumulation factors 1). Figure 2a shows global anthropogenic emissions to the (BAFs) for mixed feeders or piscivorous fish38 to give a atmosphere from 1950 onward. Emissions rise steadily after distribution of MeHg concentrations in fish. We therefore 1950 due to increased coal use and artisanal gold mining.19 neglect the time required (3−7 years) for fish populations to Over recent years, decreasing emissions in Europe and North − reach steady state following a change in Hg loadings.48 50 America due to domestic regulation have been offset by an Additionally, Hg runoff from catchments is set to be increase in East Asia, leading to an overall increase.9,19 Global proportional to atmospheric deposition (see Supporting Hg deposition is depicted in Figure 2b. The atmosphere

12971 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

Table 2. Policy Impact on Local Hg Deposition (Δμg/m2/yr, Percent in Parentheses) Assuming an Immediate a Implementation and No Legacy Penalty (Traditional Method) and Percent Change in Policy Impact Depending on Year of b Implementation, i.e., Length of the Delay

policy impact in Δμg/m2/yr (% change vs present) percent change in policy impact (%) year of implementation traditional method 2020 2025 2030 2035 2050 mean Global Average NP −1.4 (−13.6%) −33.6 −48.5 −62.7 −76.2 −114.1 −14.2 MFR −2.5 (−23.8%) −19.1 −27.5 −35.5 −43.2 −64.6 −8.0 global legacy penalty (Mg) 0.0 272 392 506 615 921 114.5 Maine (A) NP −2.9 (−15.3%) −29.5 −42.6 −55.0 −66.9 −100.0 −12.5 MFR −4.9 (−25.8%) −17.5 −25.3 −32.7 −39.7 −59.5 −7.4 Ahmedabad, India (B) NP +4.2 (+25.9%) −12.7 −18.4 −23.7 −28.9 −43.2 −5.4 MFR −6.3 (−38.4%) −8.6 −12.4 −16.0 −19.5 −29.2 −3.6 Shanghai, China (C) NP −6.3 (−55.1%) −2.7 −3.9 −5.0 −6.1 −9.1 −1.1 MFR −7.8 (−68.3%) −2.2 −3.1 −4.1 −4.9 −7.4 −0.9 South Pacific (D) NP −1.7 (−12.9%) −38.3 −55.3 −71.4 −86.9 −130.1 −16.2 MFR −3.0 (−22.1%) −22.8 −32.8 −42.4 −51.6 −77.2 −9.6 aThe policy impact is calculated as the difference in deposition between FUTURE and BASE simulations. bThe percent change in policy impact is calculated according to eq 1. The mean percent change in policy impact is the average percent change due to a near-term (2020−2035) 5 year delay, calculated according to eq 2. New Policy (NP) and Maximum Feasible Reduction (MFR) refer to future global emissions scenarios developed by Pacyna et al.9. responds relatively quickly (though not proportionally) to 3.2. Local Consequences and Percent Change in decreasing emissions. If a NP scenario is implemented in 2020, Policy Impacts. Using the integrated modeling approach deposition begins to decrease by 2021. In this scenario, global described in Section 2.1, we calculated the local consequences primary emissions decrease by ca. 980 Mg from 2020 to 2050 of delayed global action for Hg deposition fluxes to ecosystems while deposition decreases by ca. 635 Mg. These results reflect at the four selected sites (see Section 2.4). Following a the balance and cycling of Hg between the various reservoirs. traditional atmospheric modeling method (i.e., no delay/legacy Return to year 2010 deposition (arbitrary threshold) is penalty−see Figure 1), the NP scenario leads to a 15.3%, achieved in 2038, that is, 18 years after NP implementation. 55.1%, and 12.9% decrease in Hg deposition (vs present-day fi On the other hand, return to year 2010 deposition is achieved levels) in Maine, Shanghai, and the South Paci c, respectively in 2027 if a more stringent MFR scenario is implemented in (see Table 2). In contrast, a 25.9% increase is observed in Ahmedabad, India due to projected growth of regional 2020 (see SI Figure S3). 9 To evaluate the impact of delayed global action, a NP anthropogenic emissions. The MFR scenario leads to scenario was implemented for various years between 2020 and consistent global-scale Hg deposition reduction, with a 25.8%, 38.4%, 68.3%, and 22.1% decrease in Maine, 2050. Return to year 2010 deposition level is reached in 2038, fi 2047, 2056, or 2064 if a NP scenario is implemented in 2020, Ahmedabad, Shanghai, and the South Paci c, respectively. 2025, 2030, or 2035, respectively. On average, each near-term These results are consistent with those reported by Pacyna et al.9 despite the use of the a different CTM with a formulation, 5 year delay in implementing a NP scenario in turns delays by spatial resolution, and physical and chemical process para- additional extra ca. 4 years a return to its year 2010 level (this metrizations considerably different from GLEMOS and level is not the goal of policy action but used here for 30 ECHMERIT. illustrative purposes). Each near-term 5 year delay leads to a As expected, policy impacts are lower in remote regions, far ca. 2.2% increase of the atmospheric reservoir mass, mainly due from emissions sources.9,48,57 While North America contrib- to the feedback from legacy emissions. utes a significant fraction of global anthropogenic emis- Based on these results, we also compare the emissions sions,17,58,59 Hg emissions are low in Maine60 (∼50 kg yr−1) reduction rate needed in order to reach the same given and in the neighboring New England states (see SI Figures S4 deposition target at the same time but assuming delayed global and S5) due to a lack of major emitting sources as well as the action. According to Figure 2b, return to year 2010 deposition adoption in 1998 of a regional Hg action plan with aggressive level is reached in 2038 if emissions reduction is initiated in 61 − emission reduction goals. Based on Figure 1 in Giang and − 1 57 2020 at a 32.7 Mg yr reduction rate (NP). To reach the Selin, little impact is expected from domestic U.S. regulations same deposition target (year 2010 level) at the same time − in eastern Maine in terms of avoided deposition. As inferred by (2038), reduction rates of −48.0, −83.0, and −230.0 Mg yr 1 2007−2016 hourly air back-trajectories computed with the are needed if emissions reduction is initiated in 2025, 2030, or HYSPLIT model62 (see SI Figure S6), Maine tribal areas are 2035, respectively. In other words, emissions reduction must mainly influenced by air masses originating from Canada and be ca. 1.5, 2.5, or 7.0 times more stringent if initiated in 2025, the Arctic (Hudson Bay), that is, the Northern Hemisphere 2030, or 2035, respectively, instead of 2020 due to legacy atmospheric background, rather than U.S. emissions. Sunder- emissions and increasingly shortened recovery periods. land et al.63 also showed that, in the early 2000s,

12972 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis anthropogenic emissions in the U.S. and Canada resulted in According to a study performed in Maine tribal areas,69 total ∼30% of Hg deposition to the Gulf of Maine, with the rest Hg concentration in predatory fish exceeds the 0.3 mg kg−1 (∼70%) from global anthropogenic and natural sources. In U.S. EPA threshold70 in 16 out of 20 lakes (see SI Figure S7) that context, a decrease of Hg deposition in Maine tribal areas despite their remoteness from emissions sources (see Section can only be achieved through the reduction of the global 3.2). As discussed by Perlinger et al.,48 the susceptibility of background Hg concentration, that is, through global action. lakes to being contaminated depends on the total supply of Hg Potential additional effects of global change (climate, biomass to lakes but more importantly on factors leading to production burning, land use) on Hg deposition have recently been and accumulation of MeHg (e.g., prevalence of forests and investigated in Michigan’s Upper Peninsula and projected to wetlands in the catchment, low alkalinity, pH or nutrients, and have modest impacts compared to changes in direct long water residence time). In order to navigate the gap anthropogenic emissions.48 between safe and desired fish consumption levels for Figure 3 depicts the mean percent change in policy impacts populations with significant exposure to Hg, it is necessary due to a near-term (2020−2035) 5 year delayed implementa- to model changes in fish contamination over time71 and to investigate the response to delayed global action. The median response (over 20 lakes, see SI Section 1.2)of lacustrine predatory fish contamination to changes in atmospheric deposition can be seen in Figure 4. All the lakes

Figure 3. Mean percent change in policy impacts due to a near-term (2020−2035) 5-year delayed implementation of a New Policy (NP) scenario. Results are discussed at selected sites with varying impact from emissions sources, focusing on (A) tribal areas of eastern Maine, (B) Ahmedabad, India, (C) Shanghai, China, and (D) an area of the Southern Pacific known for albacore tuna fisheries. This Figure was made using the R package autoimage.64 tion of a NP scenario. On a global average, each 5 year delay leads to a ca. 14% decrease in NP impacts. The consequences Figure 4. Median response of eastern Maine lacustrine predatory fish of delayed global action depend on the stringency of the policy contamination to delayed implementation of a New Policy (NP) or as each 5 year delay leads to a ca. 8% decrease in MFR impacts Maximum Feasible Reduction (MFR) scenario. Black dashed line: year 2010 MeHg concentration. Red dashed line: U.S. EPA reference (global average, see Table 2). Remote regions are proportion- −1 ally more impacted by delayed global action than regions close dose for MeHg (0.3 mg kg ). to emission sources and a clear gradient between the Northern and Southern Hemispheres can be observed (see Figure 3). within the study area respond with a rapid decrease in MeHg While a 5 year delay leads to a −12.5 and −16.2% change in concentration over a decade, followed by a slower decline NP impacts in Maine and South Pacific, respectively, it induces toward steady state. Using the SERAFM model, Knightes et a −5.4 and −1.1% change in Ahmedabad and Shanghai, al.39 modeled the response to a hypothetical 50% decline in respectively (see Table 2 and Figure 3). This can be explained deposition across a range of lake types and also found a similar by the relatively lower policy impact in remote regions (see response. However, the time response reported here is at the above) and therefore proportionally higher influence of the upper range of those reported in Knightes et al.39 Assuming an legacy penalty. Consequences in terms of human exposure immediate implementation of a NP scenario (i.e., no delay/ through fish consumption are further discussed in the next legacy penalty−traditional method), the median MeHg section, with a specific focus on remote inland waters of concentration in predatory fish rapidly declines by ca. 11%. eastern Maine, USA. This suggests that even in the case of an immediate NP 3.3. Local Impacts on Fish Contamination: A Tribal implementation and of a rather fast response time to changes Case Study. In the U.S., rates of fish consumption and type of in deposition, the median MeHg concentration is still above fish consumed vary widely. Whereas fish forms a small the U.S. EPA threshold. These results are in line with those component of the diet of many Americans, some groups recently reported in Michigan’s Upper Peninsula.48 Substantial such as Native American tribes eat fish as frequently as and rapid response of fish contamination to reduced emissions − − daily.65 67 Additionally, fishing is an important component of have been observed near emissions sources.72 74 However, the cultural and religious practice for many Native Americans.68 relatively lower policy impacts in remote areas (see Section Therefore, fish contamination poses special risks for tribal 3.2) are likely to considerably prolong the time required to see members and is an issue relevant to environmental justice.68 substantial decreases in fish Hg concentrations. Under an

12973 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis immediate implementation of a MFR scenario, the median uncertainty on the percent change in policy impact. We − MeHg concentration drops to about 0.3 mg kg 1, that is, the calculated a mean legacy penalty for each near-term 5-year U.S. EPA threshold. A desired subsistence fish consumption of delay of 107 Mg, 115 Mg (see Table 2), or 123 Mg assuming 300−500 g per day requires a safe level target of ∼0.018 mg constant emissions (+ 0 Mg yr−1), a CP scenario (+3.02 Mg − kg 1.48,66,71 The predatory fish Hg concentrations under the yr−1), or a 2 × CP (+6.04 Mg yr−1) rate. We find that, on a MFR scenario therefore indicate that, even under the strictest global average, each 5-year delay leads to a 13%, 14%, or 15% global Hg regulations, a traditional-subsistence diet high in decrease in NP policy impacts, respectively. Perturbations to predatory fishes (e.g., brook trout, brown trout, burbot, the legacy penalty therefore have little impact on our results. landlocked salmon, smallmouth bass) will lead to unsafe 3.5. Implications. Our results show that traditional MeHg exposure in Maine tribal areas. While flawed from the spatially resolved analyses of prospective policy impacts from standpoint of environmental justice, a diet that shifts toward mercury reductions that do not consider future changes in mixed feeders (e.g., white sucker, also known as mullet or bay legacy emissions can overestimate changes driven by policy fi sh) would reduce MeHg exposure (see SI Figure S1). This implementation by up to 110% by 2050 and should be fi 75 suggestion also holds true for riverine sh in the study area. considered as best-case estimates. Though legacy impacts have Although not a true substitute for a pristine environment, previously been evaluated at global scale using global another alternative is the increasing consumption of lower-Hg biogeochemical cycling models, these effects are not widely fi 55 containing sh from aquaculture. The Aroostook Band of appreciated by policy-makers. Selin12 recently proposed a Micmacs located in Presque Isle (see SI Figure S2) has global metric to help policy-makers better understand the recently made this choice and created a recirculating implications of policy options by taking into account near-term aquaculture brook trout fish hatchery.76 fi changes in legacy Hg. Using the integrated modeling approach The impact of delayed global action on MeHg sh described here, the effect of legacy Hg changes on policy concentrations was then evaluated. Return to year 2010 fi impact can be resolved spatially. Future work could build upon MeHg level (an arbitrarily de ned threshold used here for this study by (re)examining the impact of other or upcoming illustrative purposes) is achieved in 2021, 2027, 2033, or 2043 future emission scenarios beyond those developed by Pacyna if a NP scenario is implemented in 2020, 2025, 2030, or 2035, et al.9 evaluated here. Future policy analyses should account for respectively. In other words, the longer the delay, the longer it future legacy emissions and associated deposition to better takes to reach the same MeHg concentration target. While the inform policy decision-making; the approach outlined here U.S. EPA threshold is reached in a decade if a MFR scenario is provides a straightforward methodology to estimate this effect implemented in 2010 (see Figure 4), this target is never − reached in case of delayed MFR implementation (see SI Figure without relying on advanced coupled atmosphere ocean S8). models. Alternately, an even simpler approach could scale legacy emissions and resulting deposition globally using a 3.4. Uncertainties in Atmospheric Hg Modeling. 12 Uncertainties in atmospheric Hg modeling for policy global-scale estimate of biogeochemical model output if evaluation, in particular for linking sources to receptors, have running a global biogeochemical cycle is infeasible. 13 Our results highlight the benefits of near-term aggressive Hg recently been thoroughly discussed by Kwon and Selin. ff Major uncertainties arise from biogeochemical cycling, mitigation e orts. Return to year 2010 global deposition is atmospheric chemistry, and anthropogenic emissions. There achieved 2.6 times faster under a MFR vs NP scenario (see is for example ongoing controversy in the literature and a Section 3.1). Contrary to the NP scenario, the MFR scenario rapidly evolving understanding of Hg pool sizes and fluxes in leads to consistent global-scale Hg deposition reduction (see − the global Hg cycle.13,22,77 81 The fully coupled seven- Section 3.2). We also show that each near-term delay in taking reservoir GBC model11,16 used in this study to scale legacy global action to reduce emissions has a non-negligible emissions is based on Streets et al.’s19 all-time emission influence on expected policy impacts due to legacy emissions. inventory, which assumes a major atmospheric Hg impact from Global emissions reduction must be ca. 1.5, 2.5, or 7.0 times late 19th century Gold Rush mining in North America. Recent more stringent if initiated in 2025, 2030, or 2035, respectively, studies, including historical documents on Hg use, ore instead of 2020 (see Section 3.1). On a global average, each 5- geochemistry and a large array of ice and lake sediment year delay leads to a ca. 14% decrease in NP impacts (see records, have challenged this account, as documented in a Section 3.2). Finally, while the median MeHg concentration in critical review by Outridge et al.77 This synthesis argues for a predatory fish in Maine lakes is still ca. 25% too high for safe “low-mining emissions” scenario which translates into smaller fish consumption in case of an immediate NP implementation, legacy pools in the oceans and soils than considered until now. the U.S. EPA threshold is achieved under immediate In order to investigate the consequences of a smaller legacy implementation of a MFR scenario (see Section 3.3). pool on the calculated legacy penalty, we cut historical (1850− However, this level is never reached if policy is delayed. It 1920 CE) mining emissions in the Streets et al.19 inventory by should also be emphasized that under a business-as-usual 50%, as proposed by Engstrom et al.78 This “low-mining scenario (CP), deposition fluxes to ecosystems will gradually emissions” scenario did not lead to any significant change in increase. Even if moderately delayed, NP and MFR scenarios the influence of delayed action on NP policy impacts. As lead to reductions in Hg deposition and MeHg concentration, suggested by Amos et al.,79 atmospheric deposition is most highlighting the positive impact of concerted global action. sensitive to the profile of anthropogenic emissions in recent decades, and results presented here are robust to the ■ ASSOCIATED CONTENT uncertainty in historical emissions. Our results also depend on the emissions scenario used from year 2010 until *S Supporting Information implementation of the NP scenario. We conducted a The Supporting Information is available free of charge on the perturbation analysis to investigate the effect of this ACS Publications website at DOI: 10.1021/acs.est.8b04542.

12974 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

Results of tests performed to check the robustness of our (6) Sunderland, E. M. Mercury Exposure from Domestic and Imported Estuarine and Marine Fish in the U.S. Seafood Market. method and a detailed description of the SERAFM − model parametrization are available. Additional Tables Environ. Health Perspect. 2007, 115 (2), 235 242. (7) UNEP. Text of the Minamata Convention on Mercury, Available and Figures are also available: Summary of some at: Http://Mercuryconvention.Org/Portals/11/Documents/ characteristics of the modeled lakes (Table S1), Lake ’ Booklets/COP1%20version/Minamata-Convention-Booklet-Eng- geometry values from a lake in Mighigan s Upper Full.Pdf (accessed 18 April 2018. 2017). Peninsula (Table S2), Ratios of lake characteristics from (8) Ancora, M. P.; Zhang, L.; Wang, S.; Schreifels, J. J.; Hao, J. a lake in Mighigan’s Upper Peninsula (Table S3), Lake Meeting Minamata: Cost-Effective Compliance Options for Atmos- Hg model evaluation (Figure S1), Location of Maine pheric Mercury Control in Chinese Coal-Fired Power Plants. Energy and Tribal communities (Figure S2), Global impact of Policy 2016, 88 (C), 485−494. NP vs MFR implementation (Figure S3), Hg emissions (9) Pacyna, J. M.; Travnikov, O.; De Simone, F.; Hedgecock, I. M.; in Maine (Figure S4), Year 2011 U.S. state-level Hg Sundseth, K.; Pacyna, E. G.; Steenhuisen, F.; Pirrone, N.; Munthe, J.; emissions (Figure S5), Origin of air masses influencing Kindbom, K. Current and Future Levels of Mercury Atmospheric Pollution on a Global Scale. Atmos. Chem. Phys. 2016, 16 (19), Maine tribal areas (Figure S6), Total Hg concentration − fi fi 12495 12511. in predatory sh llets collected in Maine tribal areas (10) Selin Noelle, E. Global Change and Mercury Cycling: (Figure S7), and Median response of Maine lacustrine fi Challenges for Implementing a Global Mercury Treaty. Environ. predatory sh contamination to delayed implementation Toxicol. Chem. 2014, 33 (6), 1202−1210. of a MFR scenario (Figure S8) (PDF) (11) Amos, H. M.; Jacob, D. J.; Streets, D. G.; Sunderland, E. M. Legacy Impacts of All-Time Anthropogenic Emissions on the Global Mercury Cycle. Global Biogeochem. Cycles 2013, 27 (2), 410−421. ■ AUTHOR INFORMATION (12) Selin, N. E. A Proposed Global Metric to Aid Mercury Corresponding Author Pollution Policy. Science 2018, 360 (6389), 607−609. *E-mail: [email protected]. (13) Kwon, S. Y.; Selin, N. E. Uncertainties in Atmospheric Mercury Modeling for Policy Evaluation. Curr. Pollution Rep 2016, 2 (2), 103− ORCID 114. Helénè Angot: 0000-0003-4673-8249 (14) Obrist, D.; Kirk, J. L.; Zhang, L.; Sunderland, E. M.; Jiskra, M.; Notes Selin, N. E. A Review of Global Environmental Mercury Processes in fi Response to Human and Natural Perturbations: Changes of The authors declare no competing nancial interest. Emissions, Climate, and Land Use. Ambio 2018, 47 (2), 116−140. (15) Giang, A.; Stokes, L. C.; Streets, D. G.; Corbitt, E. S.; Selin, N. ■ ACKNOWLEDGMENTS E. Impacts of the Minamata Convention on Mercury Emissions and Global Deposition from Coal-Fired Power Generation in Asia. This work was supported by the MIT Leading Technology & − Policy Program, a core center grant P30-ES002109 from the Environ. Sci. Technol. 2015, 49 (9), 5326 5335. − (16) Amos, H. M.; Jacob, D. J.; Kocman, D.; Horowitz, H. M.; National Institute of Environmental Health Sciences Zhang, Y.; Dutkiewicz, S.; Horvat, M.; Corbitt, E. S.; Krabbenhoft, D. National Institute of Health, the MIT EAPS Undergraduate P.; Sunderland, E. M. Global Biogeochemical Implications of Mercury Research Opportunities Program, the U.S. National Science Discharges from Rivers and Sediment Burial. Environ. Sci. Technol. Foundation through grant #ICER-1313755, and the National 2014, 48 (16), 9514−9522. Institute of Environmental Health Sciences Superfund Basic (17) AMAP/UNEP. Technical Background Report for the Global Research Program − National Institute of Health (P42 Mercury Assessment 2013, Arctic Monitoring and Assessment ES027707). We thank J. Pacyna and S. Cinnirella for providing Porgramme, Oslo, Norway; UNEP Chemicals Branch, Geneva, future emissions inventories. H.A. acknowledges F. Corey, D. Switzerland, 2013; p 263. Macek, S. Venno, A. Ajmani, B. Longfellow, C. Johnson, N. (18) Zhang, Y.; Jacob, D. J.; Horowitz, H. M.; Chen, L.; Amos, H. Dalrymple, and K. M. Vandiver for rewarding discussions M.; Krabbenhoft, D. P.; Slemr, F.; Louis, V. L. S.; Sunderland, E. M. during field trips to Maine. Observed Decrease in Atmospheric Mercury Explained by Global Decline in Anthropogenic Emissions. Proc. Natl. Acad. Sci. U. S. A. 2016, 113 (3), 526−531. ■ REFERENCES (19) Streets, D. G.; Devane, M. K.; Lu, Z.; Bond, T. C.; Sunderland, (1) Selin, N. E. Global Biogeochemical Cycling of Mercury: A E. M.; Jacob, D. J. All-Time Releases of Mercury to the Atmosphere Review. Annual Review of Environment and Resources 2009, 34 (1), from Human Activities. Environ. Sci. Technol. 2011, 45 (24), 10485− 43−63. 10491. (2) Horowitz, H. M.; Jacob, D. J.; Zhang, Y.; Dibble, T. S.; Slemr, F.; (20) Pirrone, N.; Cinnirella, S.; Feng, X.; Finkelman, R.; Friedli, H.; Amos, H. M.; Schmidt, J. A.; Corbitt, E. S.; Marais, E. A.; Sunderland, Leaner, J.; Mason, R.; Mukherjee, A.; Stracher, G.; Streets, D.; E. M. A New Mechanism for Atmospheric Mercury Redox Chemistry: Telmer, K. Global Mercury Emissions to the Atmosphere from Implications for the Global Mercury Budget. Atmos. Chem. Phys. Anthropogenic and Natural Sources. Atmos. Chem. Phys. 2010, 10, 2017, 17 (10), 6353−6371. 5951−5964. (3) Gustin, M. S.; Lindberg, S. E.; Weisberg, P. J. An Update on the (21) Bagnato, E.; Aiuppa, A.; Parello, F.; Allard, P.; Shinohara, H.; Natural Sources and Sinks of Atmospheric Mercury. Appl. Geochem. Liuzzo, M.; Giudice, G. New Clues on the Contribution of Earth’s 2008, 23, 482−493. Volcanism to the Global Mercury Cycle. Bull. Volcanol. 2011, 73 (5), (4) Mergler, D.; Anderson, H. A.; Chan, L. H. M.; Mahaffey, K. R.; 497−510. Murray, M.; Sakamoto, M.; Stern, A. H. Methylmercury Exposure and (22) Horowitz, H. M.; Jacob, D. J.; Amos, H. M.; Streets, D. G.; Health Effects in Humans: A Worldwide Concern. Ambio 2007, 36 Sunderland, E. M. Historical Mercury Releases from Commercial (1), 3−11. Products: Global Environmental Implications. Environ. Sci. Technol. (5) Du, B.; Feng, X.; Li, P.; Yin, R.; Yu, B.; Sonke, J. E.; Guinot, B.; 2014, 48 (17), 10242−10250. Anderson, C. W. N.; Maurice, L. Use of Mercury Isotopes to Quantify (23) Bey, I.; Jacob, D. J.; Yantosca, R. M.; Logan, J. A.; Field, B. D.; Mercury Exposure Sources in Inland Populations, China. Environ. Sci. Fiore, A. M.; li, Q.; Liu, H. Y.; Mickley, L. J.; Schultz, M. G. Global Technol. 2018, 52 (9), 5407−5416. Modeling of Tropospheric Chemistry with Assimilated Meteorology:

12975 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

Model Description and Evaluation. Journal of geophysical research Sediments (SERAFM). Environmental Modelling & Software 2008, 2001, 106 (D19), 23,073−23,095. 23 (4), 495−510. (24) Holmes, C. D.; Jacob, D. J.; Corbitt, E. S.; Mao, J.; Yang, X.; (39) Knightes, C. D.; Sunderland, E. M.; Barber, M. C.; Johnston, J. Talbot, R.; Slemr, F. Global Atmospheric Model for Mercury M.; Ambrose, R. B. Application of Ecosystem-Scale Fate and Including Oxidation by Bromine Atoms. Atmos. Chem. Phys. 2010, Bioaccumulation Models to Predict Fish Mercury Response Times 10, 12037−12057. to Changes in Atmospheric Deposition. Environ. Toxicol. Chem. 2009, (25) Amos, H. M.; Jacob, D. J.; Holmes, C. D.; Fisher, J. A.; Wang, 28 (4), 881−893. Q.; Yantosca, R. M.; Corbitt, E. S.; Galarneau, E.; Rutter, A. P.; (40) Brown, S.; Saito, L.; Knightes, C.; Gustin, M. Calibration and Gustin, M. S.; Steffen, A.; Schauer, J. J.; Graydon, J. A.; Louis, V. L.; Evaluation of a Mercury ModelforaWesternStreamand Talbot, R. W.; Edgerton, E. S.; Zhang, Y.; Sunderland, E. M. Gas- Constructed Wetland. Water, Air, Soil Pollut. 2007, 182 (1−4), Particle Partitioning of Atmospheric Hg(II) and Its Effect on Global 275−290. Mercury Deposition. Atmos. Chem. Phys. 2012, 12 (1), 591−603. (41) Selin, N. E.; Sunderland, E. M.; Knightes, C. D.; Mason, R. P. (26) Soerensen, A. L.; Sunderland, E. M.; Holmes, C. D.; Jacob, D. Sources of Mercury Exposure for US Seafood Consumers: J.; Yantosca, R. M.; Skov, H.; Christensen, J. H.; Strode, S. A.; Mason, Implications for Policy. Environ. Health Perspect. 2010, 118 (1), R. P. An Improved Global Model for Air-Sea Exchange of Mercury: 137−143. High Concentrations over the North Atlantic. Environ. Sci. Technol. (42) Hong, Y.-S.; Rifkin, E.; Bouwer, E. J. Modeling Mercury 2010, 44 (22), 8574−8580. Distribution in the Sarasota Bay Ecosystem Using SERAFM and (27) Selin, N. E.; Jacob, D. J.; Yantosca, R. M.; Strode, S.; Jaegle,́ L.; Stable Isotope Ratios of Nitrogen (Δ15N) in Biota. Environ. Eng. Sci. Sunderland, E. M. Global 3-D Land-Ocean-Atmosphere Model for 2014, 31 (3), 135−147. Mercury: Present-Day versus Preindustrial Cycles and Anthropogenic (43) Hope Bruce, K.; Louch, J. Pre-anthropocene Mercury Residues Enrichment Factors for Deposition. Global Biogeochem. Cycles 2008, in North American Freshwater Fish. Integr. Environ. Assess. Manage. 22 (2), GB2011. 2013, 10 (2), 299−308. (28) Schmidt, J. A.; Jacob, D. J.; Horowitz, H. M.; Hu, L.; Sherwen, (44) Hope, B. K. Efficacy of Translators for Establishing Mercury T.; Evans, M. J.; Liang, Q.; Suleiman, R. M.; Oram, D. E.; Le Breton, Total Maximum Daily Loads. Hum. Ecol. Risk Assess. 2011, 17 (6), M.; Percival, C. J.; Wang, S.; Dix, B.; Volkamer, R. Modeling the 1263−1278. Observed Tropospheric BrO Background: Importance of Multiphase (45) Gidley, P. T.; Kreitinger, J. P.; Zakikhani, M.; Suedel, B. C. Chemistry and Implications for Ozone, OH, and Mercury. J. Geophys. Methylmercury Screening Models for Surface Water Habitat Restoration: Res. Atmos. 2016, 121 (19), 2015JD024229. A Case Study in Duluth-Superior Harbor, ERDC/EL TR-17−19; (29) Hg and POPs working group. GEOS-Chem mercury ERDC-EL Vicksburg United States; ERDC-EL: Vicksburg United simulations http://wiki.seas.harvard.edu/geos-chem/index.php/ States, 2017. ̌ Mercury#K_RED_JNO2 (accessed May 15, 2018). (46) Zhu, S.; Zhang, Z.; Zagar, D. Mercury Transport and Fate (30) Travnikov, O.; Angot, H.; Artaxo, P.; Bencardino, M.; Bieser, J.; Models in Aquatic Systems: A Review and Synthesis. Sci. Total D’Amore, F.; Dastoor, A.; De Simone, F.; Dieguez,́ M. D. C.; Environ. 2018, 639, 538−549. Dommergue, A.; Ebinghaus, R.; Feng, X. B.; Gencarelli, C. N.; (47) Hendricks, A. N. A Model to Predict Concentrations and Hedgecock, I. M.; Magand, O.; Martin, L.; Matthias, V.; Mashyanov, Uncertainty for Mercury Species in Lakes. Open Access Master’s Thesis, N.; Pirrone, N.; Ramachandran, R.; Read, K. A.; Ryjkov, A.; Selin, N. Michigan Technological University: http://digitalcommons.mtu.edu/ E.; Sena, F.; Song, S.; Sprovieri, F.; Wip, D.; Wangberg,̈ I.; Yang, X. etdr/585, 2018. Multi-Model Study of Mercury Dispersion in the Atmosphere: (48) Perlinger, J. A.; R. Urban, N.; Giang, A.; E. Selin, N.; N. Atmospheric Processes and Model Evaluation. Atmos. Chem. Phys. Hendricks, A.; Zhang, H.; Kumar, A.; Wu, S.; S. Gagnon, V.; S. 2017, 17 (8), 5271−5295. Gorman, H.; Norman, E. Responses of Deposition and Bioaccumu- (31) Selin, N. E.; Jacob, D. J.; Park, R. J.; Yantosca, R. M.; Strode, S.; lation in the Great Lakes Region to Policy and Other Large-Scale Jaegle,́ L.; Jaffe, D. A. Chemical Cycling and Deposition of Drivers of Mercury Emissions. Environmental Science: Processes & Atmospheric Mercury: Global Constraints from Observations. J. Impacts 2018, 20 (1), 195−209. Geophys. Res. 2007, 112, D02308. (49) Hakanson,̊ L. The Derivation and Use of a Dynamic Model for (32) Guttikunda, S. K.; Jawahar, P. Atmospheric Emissions and Mercury in Lake Fish Based on a Static (Regression) Model. Water, Pollution from the Coal-Fired Thermal Power Plants in India. Atmos. Air, Soil Pollut. 2000, 124 (3−4), 301−317. Environ. 2014, 92, 449−460. (50) Barber, M. C. Dietary Uptake Models Used for Modeling the (33) Duan, L.; Wang, X.; Wang, D.; Duan, Y.; Cheng, N.; Xiu, G. Bioaccumulation of Organic Contaminants in Fish. Environ. Toxicol. Atmospheric Mercury Speciation in Shanghai, China. Sci. Total Chem. 2008, 27 (4), 755−777. Environ. 2017, 578, 460−468. (51) Thevenon, F.; Graham, N. D.; Chiaradia, M.; Arpagaus, P.; (34) Han, D.; Zhang, J.; Hu, Z.; Ma, Y.; Duan, Y.; Han, Y.; Chen, X.; Wildi, W.; Pote,́ J. Local to Regional Scale Industrial Heavy Metal Zhou, Y.; Cheng, J.; Wang, W. Particulate Mercury in Ambient Air in Pollution Recorded in Sediments of Large Freshwater Lakes in Shanghai, China: Size-Specific Distribution, Gas−Particle Partition- Central Europe (Lakes Geneva and Lucerne) over the Last Centuries. ing, and Association with Carbonaceous Composition. Environ. Pollut. Sci. Total Environ. 2011, Complete (412−413), 239−247. 2018, 238, 543−553. (52) Glass, G. E.; Sorensen, J. A.; Schmidt, K. W.; Rapp, G. R. New (35) Chen, X.; Balasubramanian, R.; Zhu, Q.; Behera, S. N.; Bo, D.; Source Identification of Mercury Contamination in the Great Lakes. Huang, X.; Xie, H.; Cheng, J. Characteristics of Atmospheric Environ. Sci. Technol. 1990, 24 (7), 1059−1069. Particulate Mercury in Size-Fractionated Particles during Haze Days (53) Kuhnlein, H. V.; Chan, H. M. Environment and Contaminants in Shanghai. Atmos. Environ. 2016, 131, 400−408. in Traditional Food Systems of Northern Indigenous Peoples. Annu. (36) Lehodey, P.; Senina, I.; Nicol, S.; Hampton, J. Modelling the Rev. Nutr. 2000, 20 (1), 595−626. Impact of Climate Change on South Pacific Albacore Tuna. Deep Sea (54) Eagles-Smith, C. A.; Silbergeld, E. K.; Basu, N.; Bustamante, P.; Res., Part II 2015, 113, 246−259. Diaz-Barriga, F.; Hopkins, W. A.; Kidd, K. A.; Nyland, J. F. (37) Sunderland, E. M.; Li, M.; Bullard, K. Decadal Changes in the Modulators of Mercury Risk to Wildlife and Humans in the Context Edible Supply of Seafood and Methylmercury Exposure in the United of Rapid Global Change. Ambio 2018, 47 (2), 170−197. States. Environ. Health Perspect. 2018, 126 (1), 017006. (55) Krabbenhoft, D. P.; Sunderland, E. M. Global Change and (38) Knightes, C. D. Development and Test Application of a Mercury. Science 2013, 341 (6153), 1457−1458. Screening-Level Mercury Fate Model and Tool for Evaluating Wildlife (56) Ahonen, S. A.; Hayden, B.; Leppanen,̈ J. J.; Kahilainen, K. K. Exposure Risk for Surface Waters with Mercury-Contaminated Climate and Productivity Affect Total Mercury Concentration and

12976 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977 Environmental Science & Technology Policy Analysis

Bioaccumulation Rate of Fish along a Spatial Gradient of Subarctic (74) Zhou, H.; Zhou, C.; Lynam, M. M.; Dvonch, J. T.; Barres, J. A.; Lakes. Sci. Total Environ. 2018, 637−638, 1586−1596. Hopke, P. K.; Cohen, M.; Holsen, T. M. Atmospheric Mercury (57) Giang, A.; Selin, N. E. Benefits of Mercury Controls for the Temporal Trends in the Northeastern United States from 1992 to United States. Proc. Natl. Acad. Sci. U. S. A. 2016, 113 (2), 286−291. 2014: Are Measured Concentrations Responding to Decreasing (58) Muntean, M.; Janssens-Maenhout, G.; Song, S.; Selin, N. E.; Regional Emissions? Environ. Sci. Technol. Lett. 2017, 4 (3), 91−97. Olivier, J. G. J.; Guizzardi, D.; Maas, R.; Dentener, F. Trend Analysis (75) U.S. Fish and Wildlife service. Contaminant Assessment of White from 1970 to 2008 and Model Evaluation of EDGARv4 Global Suckers from Eight Rivers in the Gulf of Maine Distinct Population Gridded Anthropogenic Mercury Emissions. Sci. Total Environ. 2014, Segment for Atlantic Salmon; FY07-MEFO-1-EC; Maine Field Office, 494−495, 337−350. 2007; p 44. (59) Muntean, M.; Janssens-Maenhout, G.; Song, S.; Giang, A.; (76) US Fish and Wildlife service. First Micmac Fish Harvest fi Selin, N. E.; Zhong, H.; Zhao, Y.; Olivier, J. G. J.; Guizzardi, D.; https://usfwsnortheast.wordpress.com/2017/06/28/ rst-micmac- fi Crippa, M.; Schaaf, E.; Dentener, F. Evaluating EDGARv4.Tox2 sh-harvest/ (accessed May 8, 2018). Speciated Mercury Emissions Ex-Post Scenarios and Their Impacts on (77) Outridge, P. M.; Mason, R. P.; Wang, F.; Guerrero, S.; ̈ Modelled Global and Regional Wet Deposition Patterns. Atmos. Heimburger-Boavida, L. E. Updated Global and Oceanic Mercury − Budgets for the United Nations Global Mercury Assessment 2018. Environ. 2018, 184,56 68. − (60) U.S. EPA. 2011 National Emissions Inventory (NEI) Data Environ. Sci. Technol. 2018, 52 (20), 11466 11477. https://www.epa.gov/air-emissions-inventories/2011-national- (78) Engstrom, D. R.; Fitzgerald, W. F.; Cooke, C. A.; Lamborg, C. emissions-inventory-nei-data (accessed May 8, 2018). H.; Drevnick, P. E.; Swain, E. B.; Balogh, S. J.; Balcom, P. H. (61) Smith, C. M.; Trip, L. J. Mercury Policy and Science in Atmospheric Hg Emissions from Preindustrial Gold and Silver Northeastern North America: The Mercury Action Plan of the New Extraction in the Americas: A Reevaluation from Lake-Sediment Archives. Environ. Sci. Technol. 2014, 48 (12), 6533−6543. England Governors and Eastern Canadian Premiers. Ecotoxicology (79) Amos, H. M.; Sonke, J. E.; Obrist, D.; Robins, N.; Hagan, N.; 2005, 14 (1−2), 19−35. Horowitz, H. M.; Mason, R. P.; Witt, M.; Hedgecock, I. M.; Corbitt, (62) Stein, A. F.; Draxler, R. R.; Rolph, G. D.; Stunder, B. J. B.; E. S.; Sunderland, E. M. Observational and Modeling Constraints on Cohen, M. D.; Ngan, F. NOAA’s HYSPLIT Atmospheric Transport Global Anthropogenic Enrichment of Mercury. Environ. Sci. Technol. and Dispersion Modeling System. Bull. Am. Meteorol. Soc. 2015, 96 − − 2015, 49 (7), 4036 4047. (12), 2059 2077. (80) Beal, S.; Osterberg, E. C.; Zdanowicz, C.; Fisher, D. An Ice (63) Sunderland, E. M.; Cohen, M. D.; Selin, N. E.; Chmura, G. L. Core Perspective on Mercury Pollution during the Past 600 Years. Reconciling Models and Measurements to Assess Trends in − − Environ. Sci. Technol. 2015, 49 (13), 7641 7647. Atmospheric Mercury Deposition. Environ. Pollut. 2008, 156, 526 (81) Zhang, Y.; Jaegle,́ L.; Thompson, L.; Streets, D. G. Six 535. Centuries of Changing Oceanic Mercury. Global Biogeochemical Cycles (64) French, J. P. Autoimage: Multiple Heat Maps for Projected 2014, 28 (11), 1251−1261. Coordinates. RJ.2017, 9 (1), 284−297. (65) O’Neill, C. Mercury, Risk, and Justice, SSRN Scholarly Paper ID 1004769; Social Science Research Network: Rochester, NY, 2007. (66) U.S. EPA. Wabanaki Traditional Cultural Lifeways Exposure Scenario https://www.epa.gov/tribal/wabanaki-traditional-cultural- lifeways-exposure-scenario (accessed May 8, 2018). (67) Xue, J.; Zartarian, V.; Mintz, B.; Weber, M.; Bailey, K.; Geller, A. Modeling Tribal Exposures to Methyl Mercury from Fish Consumption. Sci. Total Environ. 2015, 533, 102−109. (68) Ranco, D. J.; O’Neill, C. A.; Donatuto, J.; Harper, B. L. Environmental Justice, American Indians and the Cultural Dilemma: Developing Environmental Management for Tribal Health and Well- Being. Environmental Justice 2011, 4 (4), 221−230. (69) U.S. EPA. Region 1 Fish Tissue Contamination QA Plan | Regional Environmental Monitoring and Assessment Program | U.S. EPA https://archive.epa.gov/emap/archive-emap/web/html/reg1_ rpt.html (accessed May 8, 2018). (70) US EPA. Water Quality Criteria: Notice of Availability of Water Quality Criterion for the Protection of Human Health: Methyl- mercury https://www.federalregister.gov/documents/2001/01/08/ 01-217/water-quality-criteria-notice-of-availability-of-water-quality- criterion-for-the-protection-of-human (accessed May 8, 2018). (71) Gagnon, V.; Gorman, H.; Norman, E. Power and Politics in Research Design and Practice: Opening up Space for Social Equity in Interdisciplinary, Multi-Jurisdictional and Community-Based Re- search. Gateways: International Journal of Community Research and Engagement 2017, 10 (0), 164−184. (72) Hutcheson, M. S.; Smith, C. M.; Rose, J.; Batdorf, C.; Pancorbo, O.; West, C. R.; Strube, J.; Francis, C. Temporal and Spatial Trends in Freshwater Fish Tissue Mercury Concentrations Associated with Mercury Emissions Reductions. Environ. Sci. Technol. 2014, 48 (4), 2193−2202. (73) Brigham, M. E.; Sandheinrich, M. B.; Gay, D. A.; Maki, R. P.; Krabbenhoft, D. P.; Wiener, J. G. Lacustrine Responses to Decreasing Wet Mercury Deposition RatesResults from a Case Study in Northern Minnesota. Environ. Sci. Technol. 2014, 48 (11), 6115− 6123.

12977 DOI: 10.1021/acs.est.8b04542 Environ. Sci. Technol. 2018, 52, 12968−12977