Analysis of potential mercury policies: The impact of stream basin characteristics on susceptible populations


  • Caroline Chan,

    Corresponding author
    1. Environmental and Occupational Health Sciences, School of Public Health and Information Sciences, University of Louisville, Louisville, Kentucky, USA
    2. Kentucky Division of Water, Watershed Management Branch, 200 Fair Oaks Ln, Frankfort, Kentucky, 40601, USA
    • Environmental and Occupational Health Sciences, School of Public Health and Information Sciences, University of Louisville, Louisville, Kentucky, USA
    Search for more papers by this author
  • Robert R Jacobs

    1. Environmental and Occupational Health Sciences, School of Public Health and Information Sciences, University of Louisville, Louisville, Kentucky, USA
    Search for more papers by this author


Human exposure to Hg through fish consumption from local waterways is an ongoing concern to regulatory decision makers. Previously described population exposure and bioaccumulation models were combined to analyze the impact of potential policies on susceptible populations. The combined model simulated the problem of Hg exposure by examining the system from the point of Hg in the water column to its concentration in population biomarkers. Evaluated policy scenarios included the protectiveness of fish consumption advisories, total maximum daily load changes, and watershed management strategies. Simulations indicated that the characteristics of a basin combined with the unique pattern of intake rates of susceptible populations determine the risk associated with fish consumption from a given waterway. Each population had a unique pattern of biomarker response to changes in fish tissue Hg. Management strategies that lowered bioaccumulation rates also reduced ecosystem services. Reducing fish tissue contamination through reductions in Hg loading to watersheds is a long-term solution. For the short-term, fish consumption advisories should be used to protect populations from adverse exposures. The combination of characteristics of the basin and the populations that fish from the waterway should be the determinant for setting advisories. Integr Environ Assess Manag 2013; 9: 124–133. © 2012 SETAC


Mercury (Hg) is a potent neurotoxin, with the developing fetus at greatest risk. Exposure occurs primarily through consumption of fish, a source of lean protein that, for many populations, is culturally important and can be an affordable, nutritious component of the diet. The problem of contamination of fish with Hg is complex. The degree of contamination cannot be detected by consumers, and information on which fish are safe to consume can be confusing. Unlike other types of food, fish consumption advisories recommend limits on the amount and species of fish to eat over a period of time. These advisories are specific to each waterway as fish of the same species but from different waterways may have dissimilar levels of contamination, and one species from a particular waterway may be safe to consume whereas another may not. The advisories are established to protect susceptible individuals from the adverse effects of consuming contaminated fish whereas regulators work to reduce the contamination as specified by the Clean Water Act (CWA).

In seeking to protect susceptible populations, decision makers must decide on a level of the contaminant that can reasonably be expected to protect human health. To establish this level of protection, the question of which population should be used to establish these levels becomes crucial. Susceptibility to a pollutant can originate from 2 fronts, increased sensitivity or increased exposure (UNEP WHO 2008). Certain populations are more sensitive to the effects of a contaminant because of life stage, altered immune status, or genetic susceptibility. Because of its developing nervous system, the fetus has been found to be the most sensitive life stage to adverse effects from methylmercury (MeHg) exposure. In addition to enhanced sensitivity, some populations are more susceptible because exposure is greater. Studies of fish consumption by ethnic and other subgroups have consistently found that certain populations consume substantially greater amounts of fish than the general population (Burger et al. 1999; Shilling et al. 2010; Lincoln et al. 2011). Executive Order 12898 (1994) mandates that federal agencies make environmental justice a part of their mission and address human health and environmental effects on minority and low income populations. As it applies to MeHg exposure from fish consumption, agencies must consider populations that are disproportionately impacted from Hg exposure when determining a level of safety. Because all populations include the most sensitive life stage, consideration of the consumption rates of populations with the highest intake rates is paramount.

Although downward trends in fish tissue Hg were realized during the 1970s and 1980s, this trend has not continued (Chalmers et al. 2011). With no apparent trends, regulators cannot make assumptions about changes in Hg contamination over time, requiring ongoing monitoring efforts. For most aquatic systems, the source of contamination is from the atmospheric deposition of inorganic mercury (IHg). Inorganic mercury washes into waterways where it can be converted to MeHg, considered the most toxic form of Hg. Methylmercury accumulates in the food web, and in some ecosystems, higher trophic level fish can accumulate levels of MeHg that may be unacceptable for consumption. Although some Hg is naturally occurring, much of the Hg found in the atmosphere comes from anthropogenic sources. For example, coal-fired utilities along with other industrial sources such as chlor-alkali and cement manufacturing plants, release Hg in stack emissions. Two-thirds of current atmospheric Hg is thought to be anthropogenic in origin (USEPA 2010).

In this study, the complex problem of human exposure to Hg through fish consumption was examined through a systems approach. Models exist that look at specific components of the problem of Hg in the environment. For instance, the Water Quality Analysis Simulation Program (WASP) model looks at Hg fate in stream and lake systems but does not include bioaccumulation in the food web (Wool et al. 2001). The Bioaccumulation and Aquatic System Simulator (BASS) predicts fish tissue Hg based on food web and fish community structure (Barber 2008). A unified model that examines how different parts of the problem interact, such as the model described here, may shed additional insight into the problem of Hg exposure and potential strategies for regulators in addressing this problem.

Two parts of the problem, population exposure and bioaccumulation in fish, were examined separately and described previously. The dynamics of population exposure was simulated with a model that projected biomarkers of exposure for the distribution of intake rates of specified populations given a fish tissue Hg concentration (Chan et al. 2011). The model used a streamlined, physiologically based pharmacokinetic approach. The compartments the model included were blood MeHg, hair MeHg, fetal blood MeHg (once every 5 years), tissue MeHg, tissue IHg, and hair IHg. Mercury entered the system from fish consumption and exited through several routes: defecation, excretion in hair, and excretion in urine. The movement of Hg through the system was based on studies that examined the relationship between intake and blood MeHg (Sherlock et al. 1984), blood MeHg and hair total Hg (THg) (Clarkson et al. 2007), and blood MeHg and fetal blood MeHg (Stern and Smith 2003). Studies examining the rates of excretion of Hg and the ratio between blood and tissue MeHg were also used to increase the accuracy of the simulation (Smith et al. 1994; Cernichiari et al. 2007; Clarkson et al. 2007).

The second model projected fish tissue Hg levels for defined stream basins given water quality and watershed variables from available regulatory data (Chan et al. 2012). The model was developed using stream basins in Kentucky for calibration and basins outside of Kentucky but within the Ohio River Valley for evaluation. Inputs include land cover, physical, and chemical attributes of the stream and its contributing basin. Analysis found that the primary drivers of bioaccumulation were forested coverage, wetlands coverage, and phosphorus (P) that was used as a surrogate for nutrient loads. Nutrient loads had an inverse relationship with bioaccumulation. Increased primary production has been found to lower Hg bioaccumulation through both a dilution and growth effect. Increased primary production distributes the mass of Hg through a larger biomass of algae leading to a lower concentration within the trophic level (Pickhardt et al. 2002; Chen and Folt 2005). Increased nutrient loads also result in a higher quality food source resulting in greater biomass production at the consumer level per unit ingestion (Karimi et al. 2007). These processes lead to an inverse relationship between nutrient loads and bioaccumulation. Forested coverage works to increase bioaccumulation. Although other processes may also be at work, the model simulates light limitation. For small to medium streams, canopy cover limits primary productivity driving the process of bloom dilution in the opposite direction: limited primary production results in less biomass and more concentrated Hg at the base of the food web (Hill and Larsen 2005; Allan and Castillo 2007). Wetlands result in increased bioaccumulation by enhancing redox conditions that favor methylation of IHg (Balogh et al. 2004; Brigham et al. 2009).

As a tool for regulators, model usage is dependent on the availability of input data. Regulatory agencies monitor Hg in water more frequently than wet deposition or ambient air data, making the water column a good starting point for modeling the system.

In this study, the population exposure and bioaccumulation models were combined. A schematic diagram showing the combined model can be found in the Supplemental Data. The model was run using stream basins from Kentucky and characteristics of susceptible populations. Output was analyzed from the perspective of a regulatory decision maker charged with protecting populations from the adverse impacts of Hg exposure. Potential policy scenarios were projected with the combined model to examine what could be learned from a systems approach to the problem of Hg exposure as well as to determine what better questions should be asked in the search for solutions to this problem.

Three types of policy scenarios were considered in this evaluation: 1) a Hg reduction scenario that simulates the implementation of a total maximum daily load (TMDL) for Hg in water, 2) watershed management strategies, and 3) the protectiveness of fish consumption advisories for literature-defined populations. Implementation of a TMDL would result in lower Hg contamination levels in the watershed and subsequently lowered exposure for populations that consume those fish. However, the timeframe from the implementation of the TMDL to a sufficient reduction in fish tissue levels of Hg to protect populations may take decades (Knightes et al. 2009). The long timeframe for remediation of contamination requires that regulators look at other interim options to protect populations.

Watershed characteristics can influence the efficiency of bioaccumulation in a basin. Although regulating land cover or land use would not be a likely policy strategy for mitigating Hg exposure, forecasting the outcome of proposed changes in a basin may be a factor in approving or disapproving land use activities. Because mitigation of Hg contamination has been difficult to achieve, the combination of fish consumption advisories and educational outreach is the current policy used to protect the public from possible adverse effects from Hg exposure.

For many areas, the primary source of Hg contamination is from atmospheric deposition (USEPA 2012a). Mercury emission regulations have been a source of debate by lawmakers, regulators, and industry for decades. Reducing atmospheric Hg loads is an obvious long-term solution to reducing fish contamination. However, approximately 67% of atmospheric Hg deposition in North America comes from emission sources from other continents (Travnikov 2005). The global source of much of this contamination makes it difficult to support the unilateral reduction of local emissions of Hg. However, this value varies substantially by region and when the variation is identified, reducing atmospheric loads may offer a strategy for intervention. For example, studies examining Hg deposition near emission sources found that local sources have local impacts on Hg deposition. Selin and Jacob (2008) found that although the mean wet deposition of Hg from North American sources was 27% for the entire United States, up to 60% of wet deposition in the northeast was from local or regional sources. Keeler et al. (2006) found that 70% of Hg wet deposition in eastern Ohio was from local sources.

A TMDL approach was used to examine the impact of reductions in Hg loading to streams. When waterways are out of compliance for a designated use, the CWA requires the establishment of a TMDL, the daily loading of contaminant into a stream that will bring the waterway into compliance for its specified use (USEPA 2011a). An implemented TMDL scenario is simulated by designating the unfiltered total mercury (THg) concentration in water and projecting how that change in THg water concentration will impact susceptible populations. The watershed management scenario looked at altering sensitive variables in basins to determine the impact on susceptible populations. Although not typically viewed as a strategy for reducing Hg exposure, changes within a basin can have a substantial impact on the efficiency of bioaccumulation and therefore were simulated to project impacts on populations.

The final policy analysis looked at the decision process of setting fish consumption advisories. Fish consumption advisories are the default policy that agencies use until progress is made in lowering contamination levels in the environment. This study focused on the process of setting a level of safety for the advisory.

Important aspects of any potential policy include defining the population to be protected, determining what portion of the population should be protected, environmental impacts, and benefits and costs. An economic analysis is beyond the scope of this study, but population considerations and environmental impacts are discussed.



A description of the development and documentation of the population exposure and bioaccumulation models are given in detail in Chan et al. (2011, 2012). The models were constructed with STELLA version 9.0.3 (isee Systems, Lebanon, NH), an icon based program. The models were combined by eliminating the user input of the fish tissue Hg concentration in the population biomarker model and, instead, linking the fish tissue concentration from the bioaccumulation model. A linked Excel file is required that contains land cover, water quality, and geophysical parameters as described in Chan et al. (2012). As with the individual models, the combined model requires the user to choose the population to be simulated and the mean trophic level of the fish that is consumed by that population.

The model was developed using stream basins within the state of Kentucky. The Environmental Quality Commission for the state formed a Mercury Task Force to address the problem of Hg contamination in fish (EPPC et al. 2006). The recommendations put forth by the Task Force included strengthened testing and analysis for Hg and working to correlate sampling in air, precipitation, and fish tissue. The availability of data in the various media and cooperation and interest from Kentucky Department of Environmental Protection employees made Kentucky an ideal state to focus on for model development.

Three stream basins representing a broad spectrum of land cover and water quality attributes were used in this analysis. These basins were Tygarts Creek, a heavily forested basin (69%) with low nutrient loads; Mayfield Creek, with 23% forested coverage and high nutrient loads; and the Mud River, which falls in the middle with 50% forested coverage and average nutrient loads. Six populations, representing 2 angler, 2 subsistence, and women and children consumers were used to represent populations that may be disproportionately impacted from local fish consumption. Although these populations are not specific to Kentucky, it is assumed that populations within the state have similarities to the examined populations. Reported intake rates represent the consumption of freshwater fish only. The intake rate of the mean, 90th, 95th, and 99th percentile for each population is input into the model. Brief descriptions of the basins and populations can be found in Tables 1 and 2, respectively, with more detailed information on the populations in Chan et al. (2011) and basins in Chan et al. (2012).

Table 1. Basin characteristics
BasinArea (km2)Fraction forestedFraction wetlandsMean annual P (mg/L)
Mayfield Creek7590.2340.0420.1972
Mud River6890.4970.0070.0607
Tygarts Creek7200.6880.0000.0193
Table 2. Population descriptions
Anglers 1Wisconsin sport anglers

USEPA 1997

Anglers 2Lake Ontario sport fishers

Connelly et al. 1996

Subsistence Fishers 1Rural, southern subsistence fishers on the Savannah River

Burger et al. 1998

Subsistence Fishers 24 Native American tribes in Washington State

USEPA 1997

Women ConsumersWomen of childbearing age who consume fish

USEPA 1997

Children ConsumersChildren ages 0–14 who consume fish

USEPA 1997

Options to perturb certain parameters were added to project the effects of possible regulatory or watershed management strategies. Sensitivity analysis of the bioaccumulation submodel found that the most sensitive input parameters were nutrient loads and the fractions of the watershed that were forested or in wetlands. These 3 variables were modified within their plausible range to examine potential watershed management strategies. The plausible range for all variables was determined by assessing the range found within the 14 basins used to calibrate and evaluate the model within the Ohio River Valley and comparing those values with ranges found in the literature. Forested coverage can vary substantially by basin; the plausible range was set at 5% to 95% (Hurley et al. 1995; Warner et al. 2005; Bell and Scudder 2007). Wetlands coverage was varied from 1% to 35%. Although the basins in the tested region ranged from 0% to 9% wetlands coverage, the higher 35% range was chosen based on higher wetlands coverage found in literature studies from other regions (Hurley et al. 1995; Warner et al. 2005; Bell and Scudder 2007). The range of P was set to 0.01 mg/L to 0.35 mg/L. This was the range found within the study basins and is supported by the range found in the assessment of nutrients in streams by Mueller and Spahr (2006). THg input was perturbed by reducing input to simulate potential regulatory strategies. When perturbing a variable that is input into the model as a monthly mean (THg input and nutrients), the value that is set is input for each month, eliminating any seasonal variations that may naturally occur.

Policy scenarios and analyses

Fish intake and population choice

Changing behavior by choosing a lower trophic level fish for consumption can modify exposure. This strategy examined how changes in intake might alter exposure by examining the response in population biomarkers to fish from the same basin but different trophic levels. Fish consumption advisories are issued as a means to encourage the public to modify the intake of fish from contaminated waterways. Although fish consumption advisories are typically set based on general criterion, to protect public health, decision makers must answer 2 questions: 1) which population has the greatest exposure? and 2) what portion of the population will be protected? An analysis was undertaken to demonstrate the effect of blood MeHg biomarker response to changes in trophic level of fish consumed. In addition, the response of the distribution of the blood MeHg biomarker of susceptible populations to changes in Hg concentration in fish tissue, and whether biomarkers for the examined populations fell below the US Environmental Protection Agency (USEPA) level of safety was assessed. Fetal blood MeHg and total hair Hg biomarkers were also examined. The patterns of change in biomarkers for populations were compared.

TMDL changes

Unfiltered THg input into the model was modified, simulating the implementation of a water column TMDL. This strategy allows the user to project the concentration of unfiltered THg in the stream that would bring fish tissue into compliance. Although a TMDL designates the daily loading into a stream that would bring the waterway into compliance, the model simulation begins with the THg concentration in the water. Calculation of the daily loading into the stream is beyond the scope of the model as developed, but the model does simulate the goal of the TMDL: the reduction in Hg concentration that will bring the waterway into compliance. The TMDL scenario was analyzed by projecting the outcomes in 3 streams with trophic level 4 fish for population intake.

Watershed management strategies

Previously, the most sensitive input parameters were found to be nutrient loads and the fraction of the watershed that was forested or in wetlands (Chan et al. 2012). The effects of perturbing these 3 variables on susceptible populations were examined. The impact on the 95th percentile intake rate of 6 populations was projected by modifying these basin variables from their measured values to plausible low and high values. Nutrient loads were tested at 0.0100 mg/L and 0.3500 mg/L. The fraction of the basin that was forested was modified to 0.05 and 0.95. The fraction wetlands coverage was set to 0.01 and 0.35. Trophic level 4 fish were chosen for the intake concentration for all analyses.


Fish intake and population choice

Table 3 shows the blood MeHg biomarker response to changes in the mean trophic level of the fish consumed for the Subsistence Fisher 1 population. As expected, the blood MeHg level decreased with decreasing trophic level.

Table 3. Response of Blood MeHg biomarker for the Subsistence Fisher 1 population to changes in trophic level of fish
Trophic levelStream system
Mayfield CreekMud RiverTygarts Creek

Figure 1 illustrates the blood MeHg biomarker for the intake rates of the mean, 90th, 95th, and 99th percentile of each population in response to a range of fish tissue Hg levels. Comparing the biomarker across populations, it is apparent that the angler populations had the lowest exposure levels, and the Children and Women Consumers had the highest exposures, illustrating the importance of population selection in adequately protecting the public. In addition to the general level of exposure, each population had a unique pattern of blood MeHg concentrations. For the Anglers 1 and 2 and Children Consumers populations, the pattern shows the 99th percentile intake rate had a much higher blood MeHg concentration than the mean, 90th and 95th percentiles. Conversely, for the Women Consumers and Subsistence Fishers 2 populations, the blood MeHg levels were fairly evenly spaced between each distribution rate. For the Subsistence Fishers 1 population, the 90th, 95th, and 99th percentile intake rates had very similar blood MeHg levels, well above the mean. Although the response pattern of each population was unique in the distribution of blood MeHg biomarker, for a given population, this pattern was the same for total hair Hg and fetal blood MeHg biomarkers (data not shown).

Figure 1.

Response of the blood MeHg biomarker to changes in fish tissue concentration for susceptible populations.

TMDL changes

Figure 2 shows the response of the blood MeHg biomarker for the Subsistence Fishers 1 population to reductions in unfiltered THg for 3 stream systems. For all 3 stream systems, the portion of the population that consumes fish at or below the mean intake rate for that population has a blood MeHg concentration below the level established as protective of the developing fetus by the USEPA for all THg concentrations. However, the percentiles with higher intake rates do not always fall below the USEPA level of safety. As the TMDL is decreased, the portion of the population that is protected is dependent on the stream system being simulated. Ninety-nine percent of the population is protected with a TMDL of 0.5 ng/L for the Mud River, but that level of reduction only affords protection for 90% of the population fishing from Mayfield Creek and does less than that for Tygarts Creek. For Mayfield and Tygarts Creeks, a TMDL of 0.3 ng/L would be needed to protect 99% of the Subsistence Fishers 1 population. Even though these 2 systems offer the same degree of protection at this level, the reduction in blood MeHg is much greater in Tygarts Creek than Mayfield Creek for a reduction from 1 ng/L to 0.3 ng/L of unfiltered THg in the water column.

Figure 2.

Reductions in THg. Arrows reflect changes in THg in water concentration from the measured mean monthly THg to 1.0, 0.5, 0.4, 0.3, 0.2, 0.1 ng/L.

Watershed management strategies

Figure 3 shows the response in population biomarkers to modification of sensitive watershed parameters. The degree of change in fish tissue Hg and subsequent population exposure resulting from changes in watershed variables depend on the modeled basin and the variable being modified. Whether this change in a watershed parameter has substantial impacts on population biomarkers also depends on the population being simulated.

Figure 3.

Impact of changes in sensitive basin parameters on populations. Depicted basin variables are (A) nutrients (low value 0.01 mg/L and high value 0.35 mg/L), (B) forested coverage (5% and 95% coverage), and (C) wetlands coverage (1% and 35% coverage). The x-axis depicts the 95th percentile intake rate for populations: 1 = anglers 1; 2 = anglers 2; 3 = subsistence Fishers 1; 4 = subsistence Fishers 2; 5 = children consumers; 6 = women consumers.

The previous analysis of the bioaccumulation model found that 3 watershed variables were sensitive to perturbations: nutrient loads and the fraction of the watershed that was forested or in wetlands (Chan et al. 2012). When nutrient levels were set at a minimal level (Figure 3A), population Hg biomarkers increased substantially for 4 of the 6 populations in the Mayfield Creek and Mud River basins whereas all blood MeHg levels remained relatively unchanged for Tygarts Creek. Nutrient levels were highest in Mayfield Creek (mean annual P of 0.1972 mg/L), average, relative to the other modeled basins, in the Mud River (mean annual P of 0.0607 mg/L), and low in Tygarts Creek (mean annual P of 0.0193 mg/L). As expected, the deviation in nutrient levels imposed by the user caused no change because Tygarts Creek already had low nutrient levels. Surprisingly, the magnitude of response for both Mayfield Creek and the Mud River were similar even though the default levels of P differed by a factor of 3. Setting a high level of P reduced population biomarkers for 4 of the 6 populations in Tygarts Creek but elicited little change in any populations in Mayfield Creek or the Mud River over the default level. These responses suggest a threshold range of P over which higher levels will not decrease fish tissue further and lower levels of the nutrient will not result in further increases in fish tissue. Within this range, nutrient levels impact the efficiency of bioaccumulation. When assessing a stream system, regulators can use this information to determine if changes in fertilizer use in the basin will modify bioaccumulation efficiency.

No response in population biomarkers was found for Mayfield Creek when the fraction of the basin that was forested was set to a low fraction (Figure 3B). This reflects the existing low forested coverage in the basin. When the fraction that was forested was increased, 4 populations had a greater than 200% increase in blood MeHg. The Anglers 1 and 2 populations increased to a lesser degree than the other populations. The Mud River showed a minimal response to a reduction in forested coverage for 4 populations, but a substantial increase in blood MeHg for those same populations when the fraction of forested coverage was increased. Tygarts Creek showed a greater response to a low forested coverage level for 4 populations, whereas no response was found for a higher fraction of forested coverage. Forested coverage increases bioaccumulation efficiency, substantially increasing exposure for populations with high consumption rates.

Manipulating wetlands coverage from low to high values showed a similar pattern (Figure 3C). Mayfield Creek, which had the highest wetlands coverage of the 3 streams at 4%, showed little change in blood MeHg when lowering coverage to 1%, but a strong response of increased blood MeHg when wetlands coverage was raised to 35% for 4 of the populations. Population blood MeHg responded similarly for the Mud River, although a slightly larger response was seen in the Anglers 1 and 2 populations for high wetlands coverage. Biomarker response was greatest for Tygarts Creek which has no wetlands coverage. Setting wetlands coverage to the low level, which is actually an increase in wetlands coverage for this basin, resulted in a slight increase in blood MeHg for 4 populations, and the high wetlands coverage setting showed substantial increases in all population biomarker levels. The presence of wetlands in a basin increases exposure to susceptible populations.


Fish intake and consumption advisories

The current strategy for protecting populations from the adverse impacts of Hg exposure from fish consumption is to limit exposure by issuing fish consumption advisories and educating the public about these advisories. However, fish consumption also provides substantial health benefits. It reduces coronary heart disease mortality and stroke and increases cognitive development in the developing fetus. The message these advisories attempt to convey is that women of childbearing age and young children should reduce consumption of fish with high levels of Hg contamination and substitute fish with lower concentrations, thus preserving health benefits while reducing risks. Table 3 demonstrates that reductions in the trophic level of the fish consumed resulted in lower exposures to susceptible populations without reducing the total amount of fish consumed. If the situation warrants, issuing a fish consumption advisory that encourages consumption of lower trophic level species while limiting consumption of higher trophic level fish could lower exposure without decreasing intake. The benefits of fish consumption could be realized while lowering risk. Some populations may have substantially higher exposures from consumption of locally caught fish than the general population. Some Native American tribes, Inuit, and certain Asian populations have been found to have significantly higher fish consumption rates than the general US population (Hightower et al. 2006; Shilling et al. 2010). The model assumes that 100% of fish consumption comes from the stream being modeled. Although this assumption is unrealistic for most populations because fish may come from multiple waterways or commercial sources, the assumption most closely reflects intake patterns for certain subsistence populations. Exposure may be underestimated or overestimated based on how closely this assumption is met and the degree of contamination of other sources.

Understanding the degree to which these populations are exposed and directing risk communication to those most affected is essential to protecting those most at risk (Burger et al. 1999; Shilling et al. 2010). This study used populations described in the literature that were not specific to the basins being simulated, although it is assumed that some of the fishers in the modeled basins would have characteristics similar to the literature derived populations. For decision makers, it is important to consider the unique populations that consume fish from the waterway being examined. Although the USEPA criterion for fish protected the 95th percentile for all of the populations examined in this study, there may be other subsistence populations that have higher intake rates. These populations may not be protected. For some of the simulated populations, the 99th percentile intake rate had substantially greater exposures, showing biomarkers well above the USEPA level of safety.

In addition to identifying the population most at risk from a particular water body, agencies must also choose a level of action. Commonly, the central tendency, 90th or 95th percentile of intake rates is selected (Shilling et al. 2010). Figure 1 includes both the USEPA level of safety for MeHg in blood (5.8 µg/L) (USEPA 2011b) and the USEPA criterion for MeHg in fish tissue (0.300 mg/kg) (USEPA 2011c). The figures illustrate that 95% of all the simulated populations are protected when fish tissue concentrations of MeHg are below the USEPA criterion of 0.300 mg/kg, confirming that model projections and USEPA guidelines on fish intake are similar. However, as evidenced by the number of fish consumption advisories, many stream systems exceed this level. In this model simulation, the 95th percentile of the subsistence and consumer populations exceeded the USEPA level of safety for MeHg in blood at levels just above the USEPA fish tissue criterion. Thus, although the criterion is protective for these populations, the margin of error is small. For the Subsistence Fishers 1 population, the 90th, 95th, and 99th percentile intake rates are very similar. Although only the 99th percentile exceeds the level of safety for blood MeHg at intake concentrations at the USEPA criterion, the 90th and 95th percentile exceed the level of safety at fish tissue concentrations just above the USEPA criterion. Clearly, the choice of population and the portion of the population to be protected are both critical decisions in determining the action level for issuing fish consumption advisories.

The unique distributions of intake rates for each population demonstrate that both the population examined and the percentile chosen for action contribute to the impact of a chosen policy. For decision makers, these population profiles evoke the better questions: what population using a particular waterway is most at risk? What portion of that population is protected by the USEPA fish tissue criterion? What percentile of intake rate adequately protects that population? Answers to these questions can lead to the development of more effective policies to protect susceptible populations.

TMDL changes

The CWA mandates that states and tribes set TMDLs for pollutants that are out of compliance for the designated use for a given waterway. As required by law, many states have established TMDLs for Hg in waterways as a result of noncompliance of fish tissue levels for Hg. The TMDL approach, however, is limited by the reality that setting a maximum loading rate for a nonpoint source pollutant does not actually mean that controls exist to limit loading. The CWA specifies watershed management strategies to reduce nonpoint source loading into waterways. Atmospheric deposition to aquatic systems is one form of nonpoint source pollution. Because atmospheric deposition is ubiquitous, watershed management programs cannot effectively reduce Hg loading in waterways. To reduce loading from the atmosphere, atmospheric sources must be reduced. Emission reduction regulations are the obvious means to reducing Hg contamination in fish, but these regulations have been difficult to promulgate. The Mercury and Air Toxics Standards (MATS) was issued in December of 2011 more than 20 years after the Clean Air Act Amendments of 1990 designated Hg as a hazardous air pollutant. This standard regulates Hg emissions from the largest industrial source—coal and oil fired electric utilities (USEPA 2012b). This rule, however, does not address global sources. The relationship between local, regional, national, and global sources and their impact on loading into waterways is complex and needs further study.

When a waterway exceeds an established criterion, the regulating agency determines the daily amount of contaminant the waterway can receive to come back into compliance. Because each stream system is unique, the TMDL will vary depending on the unique characteristics of the basin. Although regulating agencies set TMDLs to bring fish tissue Hg into compliance, the impact on a susceptible population will depend on several variables. The reduction in unfiltered THg inputs shown in Figure 2 can represent the results of possible TMDLs put forth by a regulating agency. Because of its greater forested coverage and lower nutrient levels, Tygarts Creek exhibited greater efficiency in bioaccumulation of MeHg than Mayfield Creek; therefore changes in Hg loads had a greater response in fish tissue and a greater subsequent impact on populations.

Watershed management strategies

Although the extent of the response in population biomarkers was different depending on the basin variable that was perturbed, a similar pattern of response was seen for the 6 populations examined. The magnitude of biomarker change for the Anglers 1 and 2 populations was less than the Subsistence Fishers and Consumers populations. This pattern was independent of the stream that was simulated or the variable that was perturbed and points to the importance of population choice for decision makers when setting guidelines for setting a level of safety. Having a clear understanding of how population intake rates were assessed is also necessary when evaluating a system. Intake rates for the angler populations and Subsistence 2 population were assessed over a year (CRITFC 1994; Connelly et al. 1996; USEPA 1997). The Subsistence 1 population was assessed over 8 months (Burger et al. 1998). Many anglers fish seasonally, and deriving a monthly rate from annual intake may underestimate intake for certain seasons. Conversely, the consumer population intake rates were extrapolated from a 3-day diary. Anyone who had consumed fish in this time was considered a consumer, and the monthly rate was then determined. This may have overestimated the monthly intake for these populations (USEPA 1997). Although many populations may have little exposure from local waterways, certain populations may have heavy exposures. Knowledge of the populations that consume fish from a stream system under consideration and their consumption patterns is essential to the decision making process.

Although changes in basin variables can significantly change bioaccumulation and subsequent population exposures, the environmental changes that decrease fish contamination would be viewed by many as detrimental because these same variables provide essential ecosystem services. An inverse relationship exists between nutrient loads and bioaccumulation efficiency. High nutrient loads, however, lead to eutrophication. Eutrophic stream systems provide a lower level of ecosystem services, with algal blooms, murky water, and, at high enough levels of nutrients, the loss of the ability to support aquatic life. Bioaccumulation is lower in basins with less forested coverage. Ecosystem services provided by forests include recreational opportunities, production of O2, C sequestration, soil production and retention, promotion of the recharging of groundwater, and regulation of flow regimes (UNEP 2009). Wetlands coverage is also correlated with bioaccumulation efficiency. Wetlands ecosystem services include flood control, groundwater recharge, sediment and nutrient removal and storage, and water purification (Ramsar 2011). For all 3 basin variables examined, changes that reduced contamination in fish also reduced the ecosystem services provided by that system. Because systems with greater bioaccumulation efficiency also have a greater response to changes in loading, targeting these systems for reductions in loading would provide the benefit of reduced contamination and greater services.


The dynamic model simulated the problem of Hg exposure by examining the system from the point of Hg in the water column to its concentration in population biomarkers. Combining the environmental and population exposure components permitted the assessment of the interplay of watershed and population characteristics on human exposure. Possible policy scenarios were evaluated while viewing the problem as one system, suggesting important questions to be considered when assessing a contaminated system. These questions include:

  • What population has the greatest exposure from this stream?

  • What does the pattern of intake of fish for this population look like?

  • What portion of this population is protected by the USEPA level of safety?

  • How efficient is bioaccumulation in this stream system?

Ultimately, the CWA requires the reduction of Hg levels in fish tissue to meet water quality criterion. One approach would be to alter watershed characteristics. Model results indicated that reducing forested or wetlands coverage or increasing nutrient loads would reduce Hg contamination in fish. However, these changes also reduce valued ecosystem services. The alternative approach to meeting water quality criterion would be to reduce the loading of Hg into waterways and this can only be achieved by reducing Hg emissions. Bringing these waterways into compliance through Hg emission regulations is expected to take decades to centuries, and therefore shorter-term policies are required to protect populations from adverse exposures (Selin 2011). For the short-term, fish consumption advisories must be used to protect populations from adverse exposures. Model results showed that each population had a unique fish intake profile. USEPA guidelines were protective of the examined populations, although some of the populations were close to exceeding the level of safety. Other populations with high consumption patterns may not be protected by USEPA guidelines and may not be protected. Ultimately, the unique characteristics of a basin combined with the unique pattern of intake rates of susceptible populations determine the risk associated with fish consumption from a given waterway, and this combination should be the determinant for setting advisories.


Schematic diagram of combined model.


Financial support for this project was provided by the STAR Fellowship Assistance Agreement no. FP-91711701-0 awarded by the USEPA. It has not been formally reviewed by USEPA. The views expressed in this publication are solely those of the authors, and USEPA does not endorse any products or commercial services mentioned in this publication.