Environ. Sci. Technol. 2002, 36, 238-246
Risk-Based Management of Contaminated Sediments: Consideration of Spatial and Temporal Patterns in Exposure Modeling I G O R L I N K O V , * ,†,‡ D M I T R I Y BURMISTROV,† JEROME CURA,† AND TODD S. BRIDGES§ Menzie-Cura & Associates, Inc., One Courthouse Lane, Suite Two, Chelmsford, Massachusetts 01824, Arthur D. Little, Inc., 20 Acorn Park, Cambridge, Massachusetts 02140 and United States Engineering Research and Development Center, Vicksburg, Mississippi 39180
This paper addresses interactions among foraging behavior, habitat preferences, site characteristics, and spatial distribution of contaminants in developing PCB exposure estimates for winter flounder at a hypothetical open water dredged material disposal site in the coastal waters of New York and New Jersey (NY-NJ). The implications of these interactions for human health risk estimates for local recreational anglers who fish for and eat flounder are described. The models implemented in this study include a spatial submodel to account for spatial and temporal characteristics of fish exposures and a probabilistic adaptation of the Gobas bioaccumulation model that accounts for temporal variation in concentrations of hydrophobic contaminants in sediment and water. We estimated the geographic distribution of a winter flounder subpopulation offshore of NY-NJ based on species biology and its vulnerability to local recreational fishing, the foraging area of individual fish, and their migration patterns. We incorporated these parameters and an estimate of differential attraction to a management site into a spatially explicit model to assess the range of exposures within the population. The output of this modeling effort, flounder PCB tissue concentrations, provided exposure point concentrations for an estimate of human health risk through ingestion of locally caught flounder. The risks obtained for the spatially nonexplicit case are as much as 1 order of magnitude higher than those obtained with explicit consideration of spatial and temporal characteristics of winter flounder foraging and seasonal migration. This practice of “defaulting” to extremely conservative estimates for exposure parameters in the face of uncertainty ill serves the decision-making process for management of contaminated sediments in general and specifically for disposal of dredged materials. Consideration of realistic spatial and temporal scales in food chain models can help support * Corresponding author telephone: (617)498-5317; fax: (617)4987021; e-mail: [email protected]
† Menzie-Cura & Associates, Inc. ‡ Present address: Arthur D. Little, Inc., 20 Acorn Park, Cambridge, MA 02140. § United States Army Corps of Engineers. 238
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002
sediment management decisions by providing a quantitative expression of the confidence in risk estimates.
Introduction Exposure estimates for wildlife in areas containing spatially localized contaminants are a function of spatial factors such as foraging area, size of the habitat, and distribution of contamination. Species exhibiting different foraging strategies may experience significantly different chemical exposures from the same site, even if their foraging areas overlap. Currently, exposure estimates and subsequent human health and ecological risk projections usually assume a static and continuous exposure of an ecological receptor to a contaminant concentration represented by some descriptive statistic, such as the mean or maximum. These assumptions are thus overly conservative and ignore some of the major advantages offered by risk assessmentsthe ability to account for site-specific conditions and to conduct iterative analyses. The importance of consideration of the spatial extent of site contamination in the terrestrial environment has recently attracted the attention of individual researchers (1-6) as well as regulatory agencies (7-9). These studies call for explicit incorporation of habitat sizes and foraging strategy for terrestrial receptors. Nevertheless, spatial scales in the risk assessment for aquatic ecosystems have not been widely considered. This paper proposes a framework for spatially explicit risk assessment associated with contaminated sediments. Many sediments are contaminated as a result of industrial development, and management decisions have to be made on their safe use as well as on their disposal (10). For example, the United States Army Corps of Engineers (U.S. ACE) or their permit recipients dredge about 400 million cubic yards of sediment annually to maintain 25 000 mi of navigational channels. About 60 million cubic yards of dredged material, including sediments that receive urban or agricultural runoff, are placed in more than 150 open water marine sites designated by the U.S. Environmental Protection Agency (U.S. EPA). The fact that open water management facilities are geographically restricted invites an ecological exposure analysis that accounts for the spatial and temporal aspects of a receptor’s biology. The proposed framework can be used to support risk-based decision making in the regulation and management of contaminated sediment. For some sediments and sites, bioaccumulation and biomagnification of hydrophobic organic contaminants, such as polychlorinated biphenyls (PCBs), may represent the primary source of environmental risks to aquatic organisms and their higher order predators, including humans. This paper addresses the interactions of various aspects of foraging behavior, habitat characteristics, and spatial distribution of contaminants in developing PCB exposure estimates for winter flounder at a hypothetical open water dredged material disposal site in the coastal waters of New York and New Jersey (NY-NJ). It then considers the implications of these interactions for human health risk estimates for local recreational anglers who fish for and eat those flounder. We also address the advantages of such spatially explicit modeling in environmental decision-making where sediment contamination poses risk to wildlife.
Modeling Approach and Parameters Conceptual Model. We developed a conceptual model to represent a predominantly sediment-driven food web that 10.1021/es010721d CCC: $22.00
2002 American Chemical Society Published on Web 12/01/2001
FIGURE 1. Modeling approach. is common for sites with contaminated sediments. The conceptual model is a simple food chain in which the contaminant of concern is total PCBs. We selected PCBs, which are highly lipophilic and hydrophobic, for this analysis because they (i) are often found in contaminated sediments, (ii) are known to biomagnify through food chains, and (iii) pose risk to both humans and ecological receptors. Although the current analysis addresses only PCBs, the general methodology and conclusions are applicable to a wide range of organic contaminants. The exposure media are surface water and sediments. A common polychaete, Nereis virens (sandworm), represents the prey base for Pseudopleuronectes americanus (winter flounder). The human receptors are recreational anglers who catch and eat the flounder. The analysis employs a spatially explicit foraging submodel (Figure 1) that provides a time series of sediment and water concentrations that a fish may encounter within its habitat. The model inputs are based on information on seasonal abundance of fish, habitat size for a species, size and location of the management area within the species habitat, size of the species foraging area, and concentrations of PCBs in sediment and water over the management site and in the surrounding areas. The model also uses a site-specific attraction factor that accounts for differential attraction to the management area. The outputs of the spatial submodel are combinations of sediment and surface water concentrations that the fish population may encounter while foraging in this habitat over time. The analysis then applies a bioaccumulation submodel for transfer of PCBs from sediments and surface water through a fish food chain. This analysis uses sandworms (N. virens) as the base of the food chain. The use of the sandworm represents a conservative estimate of invertebrate exposure because it is a deposit feeder that burrows and lives in sediment and moves only partially out of its burrow to feed (11). Winter flounder, which represent the next trophic level above sandworms, feed primarily on invertebrates in the sediment. Their feeding preferences vary with the age and size of the individual. The adult fish mostly consume annelids, molluscs, and cnidaria. Several investigators (12-14) noted that they are omnivorous, opportunistic feeders and prey upon various sediment-dwelling organisms such as sandworms, amphipods and isopods (crustaceans), pelecypods, and plant material. Steimle and Terranova (15) found that winter flounder from both contaminated and cleaner control areas fed primarily on the tentacular crowns of tube-dwelling anemones and sandworms. Within this conceptual model, we assumed that the flounder feed solely on sandworms. This assumption maximizes exposures because, among the flounder’s dietary components, the sandworm with its benthic habitat is likely to experience a relatively high exposure to sediment contaminants. Winter flounder is a reasonable representative fish species because it (i) is an important recreational and commercial
species; (ii) occurs abundantly in the New York/New Jersey coastal area; (iii) represents a higher order, bottom-feeding predator; and (iv) is a resident species with a relatively small foraging area. It will more frequently encounter localized contaminated sediments than other recreationally obtained species such as bluefish or striped bass, pelagic species that feed in the water column and forage over larger ranges. The output of the bioaccumulation submodel is a time series of fish tissue PCB residues. The human risk submodel (Figure 1) averages these time series and then calculates human risk resulting from ingestion of these fish, based on accepted human exposure parameters. Spatial Submodel. The approach used to design the spatial submodel is an extension and modification of prior methods (3, 4). The model, modified to perform a random walk analysis, can provide probabilistic outputs. The habitat is divided into a grid of 1 m × 1 m cells across the management area and the surrounding habitat (specified as a given distance from the center of the management area). In the current study, all cells within the limits of the disposal site are assigned a probability distribution for PCB concentrations in sediment (see Table 1 and discussion below). While the present analysis assumes that all cells outside the management area are free of PCBs, other model parametrizations (for example, inclusion of background concentrations outside the facility) are possible. The spatial submodel uses the habitat grid to calculate exposure point concentrations for fish via sediment and water. At specified time periods, each individual fish in the simulation is modeled foraging in randomly selected areas within the habitat. The exposure point concentration for each time step is the average concentration across the cells that a fish encounters within its foraging area for a specified time period. The current simulation uses a monthly time step because the data for fish abundance is available only in monthly increments. However, a time interval shorter than 1 month can be implemented. The exposure point concentrations depend on the sizes of the foraging area, facility, and habitat. Individuals with a small foraging area may feed exclusively within the facility boundaries and thus have a high accumulation of contaminants, while individuals feeding outside the facility may not be affected at all. On the other hand, a higher percentage of individuals in species foraging widely over the habitat may be exposed at some point to contaminated areas. These exposures would generally be of shorter duration. The outputs of the spatial submodel are the exposure point concentrations that individual fish encounter over time. The bioaccumulation submodel uses these time series to calculate PCB body burdens in fish over time. The input parameters for the spatial submodel include sediment and water PCB concentrations, size and location of the management site within the habitat, site attraction factor, seasonal abundance of fish, fish foraging area, and habitat size. VOL. 36, NO. 2, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
TABLE 1. Input Parameters parameter
Facility facility size (km2) attraction factor
point 3 cases
lipid content (%)
Lutza as derived in text
NA NA NA
115 0.33 NA
631 2.09 NA
NA NA NA
34 34 17
Sandworm Winter Flounder body weight (g) lipid content (%) seasonal abundance (no./ha)
triangular traingular point
foraging range (ha) habitat size (km2)
NA 3 cases
263 1.04 Jan-185, Feb-59, Mar-74, Apr-101, Mar-77, Jun-143, Jul-84, Sep-47, Oct-49, Nov-131, Dec-116 250 48
log Kow sediment concentration (total PCBs, ng/g dry wt) water concentration (total PCBs, ng/L) TOC (%)
body wt (kg) fish ingestion (g/day)
log normal log normal
36 50th and 95th percentilesc assumed
Sediment and Water
exposure duration (days) a
U.S. Army Corps of Engineers, unpublished data. c From ref 36.
Sediment and Water Concentrations.The key input parameter for sediment is the total PCB concentration in micrograms per kilograms dry weight. The sediment data in the analysis are from several sites in NY-NJ Harbor (collectively referred to as Dredged Material Management Plan, or DMMP, data). The relatively high concentrations of PCBs in the DMMP sediments represent a reasonably high conservative estimate for sediments proposed for dredging in NY-NJ Harbor. Greges (personal communication) and Wisemiller (U.S. ACE, unpublished data) provided these site-specific sediment concentrations. We assumed that sediment PCB concentrations outside the management area were zero, although alternative approaches are feasible. The key input parameter for the water column is the freely dissolved concentration of PCBs in water. This analysis assumes that the contaminated sediments and overlying water achieve equilibrium. For model inputs, we estimated water concentrations of PCBs from sediment concentrations based on equilibrium partitioning. This is a very conservative approach, ignoring processes such as dispersion, advection, and tidal mixing, all of which would result in a much lower steady-state concentration. Table 1 shows the modeled probability distributions for sediment and water concentrations derived from individual measured concentrations. In deriving distributions, we considered that, over the long term, humans consuming aquatic organisms would be exposed to average tissue concentrations overall. The mean sediment and water concentrations were thus characterized by normal distributions for the mean, with uncertainties characterized by the standard error of the mean. The correlation coefficient between the water and the sediment concentrations was set to 1.0 because the water concentrations were calculated under an equilibrium assumption. Size of the Management Site and Attraction Factor. We assumed that the sediments from these areas are dredged 240
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002
and subsequently placed at a management site offshore of NY-NJ Harbor. The assumed size of this site is 3.75 km3, close to the size of a typical management area (Lutz, U.S. ACE, personal communication). The model uses a differential attraction factor to quantify the likelihood of increased flounder population density at the management site due to disturbance, presence of topographic features, or organic enrichment of sediments at the site. Exposure assessments frequently recognize such differential attraction as a source of uncertainty but rarely address it explicitly other than to assume that a receptor spends all its time within the boundaries of a site. This is a common approach that can produce large overestimates of risk with large uncertainty. We define the attraction factor as the ratio of fish abundance within the boundaries of the management site (number of fish per unit area or catch per unit effort) to the fish abundance outside the facility boundaries. In the absence of habitat-specific information that describes the potential differential attraction of flounder to a management site or other disturbed sites, we assumed that an examination of spatial variation in fish abundance among historical sampling stations would reflect the range for such an attraction factor. We reviewed the difference in winter flounder abundance among several sampling stations within their habitat as reported in the literature. The data on distribution and abundance of adult winter flounder collected during the NEFSC bottom trawl survey in 1963-1997 (16) shows that fish abundance can vary across sampling station within the habitat, but the ratio of winter flounder abundance among many stations did not exceed 10. This ratio is also confirmed by a study of winter flounder abundance in Narragansett Bay (17). Therefore we assumed that the difference among stations represents the magnitude of the potential attraction and used 10 as the estimate for the attraction factor, but we varied this within the spatial submodel, from 1 to 100, to assess the sensitivity of the model to this uncertain parameter.
Winter Flounder Seasonal Movements and Abundance. Studies of seasonal migration of winter flounder show that adults live in cooler offshore waters during the summer and then move to shallower inshore waters in winter and early spring. The extent of offshore-inshore movements varies geographically. Many tagging studies (18-22) show that flounders remain in bays and harbors year-round, moving into deeper waters during the warmest weather. The model incorporated fish abundance (number of individual fish per unit area) as well as seasonal changes in population abundance due to seasonal migration between in-shore and off-shore areas that are characteristic of many fish species. Reported abundance for winter flounder varies widely. Pearcy (12) reports that the average abundance for juvenile winter flounder in the Upper Mystic River Estuary ranged between 0.1 and 0.01 juvenile fish/m2. The abundance of adult fish is likely to be about 10 times less than this density, based on the survival curve presented in their study. Haedrich and Haedrich (23) report 0.004, 0.017, 0.027, and 0.013 fish/ m2 in June, August, November, and May for the Mystic River Estuary. Black and Miller (24) observed about 0.005 fish/m2 near Lower Argyle, Nova Scotia. We used data from the most detailed study of winter flounder abundance, collected in Narragansett Bay, RI (17). This study reports average monthly abundance for winter flounder at 10 stations within the bay. The abundance varies from 0.005 fish/m2 in August-October to about 0.02 fish/m2 in January. We assumed that this range and pattern of seasonal change describe the flounder population in the NY-NJ area. Foraging Area and Habitat Size. A habitat is the area traversed by and providing shelter to a population. Foraging area is the area within the habitat that provides the population with food. The size of a population’s habitat (i.e., the area occupied by the population) and foraging area have a strong influence on fish exposure to the site contamination. If the area over which one defines the local population is large (i.e., fish routinely migrate large distances over a short time period), spatially localized contamination would not result in significant fish exposure and, thus, significant risks. On the other hand, a contaminated site that is relatively large as compared to habitat size could result in significant exposure to most or all of a population. We assumed that the size of the foraging area is an undirected component of fish movement and is characterized by a dispersion coefficient (25). We use a dispersion coefficient based on tagging experiments in Rhode Island Sound (21) to represent a foraging area. These estimated dispersion coefficients ranged from 1.74 to 2.85 km2 day-1. This study described the average dispersion coefficient of 2.33 km2, or about 250 ha, as typical for a winter flounder foraging area. We varied the size of the foraging area from 25 to 2500 ha to study the sensitivity of the model to this uncertain parameter. The size of the habitat can be defined biologically (i.e., the migration area over which individual fish move over a specified time period) or operationally to reflect a combination of ecological considerations and risk management judgments. The size of a winter flounder habitat defined biologically can be very large. Flounder can travel large distances if constantly moving, at an average speed of about 1.44 km/h (26). A study of tagged winter flounder movement in New York Bight found that fish can travel over 40 km in one season. One tagged fish traveled 328 km from the tagging site (27). In another study (28), fish tagged in Barnegat Bay, NJ, were recaptured over an area of 5000 mi2 over 2 yr. This large habitat size for winter flounder has been observed in other geographic areas as well. The average distance traveled by winter flounder around Cape Cod ranged from 5.7 to 42.2 km across 15 locations where fish were tagged (18).
The above data indicate that the habitat size for a winter flounder population can range from several hundred to several thousand square kilometers. If the spatial submodel used such a large habitat, large areas of relatively clean habitat surrounding a small management site would dilute the effect of localized contamination from a management facility, and risks would always be minimal. However, when considering the risks to human populations consuming local winter flounder, a definition of local habitat that incorporates the portion of the flounder habitat over which the human population is likely to obtain fish may be more appropriate. To provide a conservative (i.e., protective of human health) estimate of risk, we adopted an operational definition of the flounder subpopulation to which local consumers might be exposed. Even though flounder in this subpopulation may migrate far away, a conservative (i.e., much smaller) estimate for habitat size for this subpopulation can be derived based on the consumption pattern of local recreational anglers. The average annual catch reported for New Jersey is about 500 000 adult fish (29). The smallest spatial area required to support the production of this number of fish can be estimated by dividing the total catch by average fish abundance. Using an average abundance of 0.01 individuals/ m2 (12, 17, 23, 24) results in a total habitat size of about 50 km2. We used 25, 50, and 100 km2 to test the model’s sensitivity to the size of winter flounder habitat. Bioaccumulation Model. We developed a mechanistic, time-varying model based on the approach used by Gobas (43). This section presents model summary and parameters that were used in this study. (Please refer to refs 30 and 31 for more detailed model and parameter description.) The model predicts PCB accumulation in fish from direct gill uptake of PCBs from water as well as dietary consumption of contaminated prey. It relies on solutions to the following differential equations, which describe the time-varying uptake of PCBs using time series data for sediment and surface water PCB concentrations:
dCf ) k1Cwd + kdCdiet - (k2 + ke + km + kg)Cf dt
where k1 is the gill uptake rate constant (L kg-1 d-1), Cwd is the freely dissolved concentration of PCBs in water (ng/L), kd is the dietary uptake rate constant (d-1), Cdiet is the concentration of PCBs in the diet (µg/kg), k2 is the gill elimination rate constant (d-1), ke is the fecal egestion rate constant (d-1), km is the metabolic rate constant (d-1), kg is the growth rate (d-1), and Cf is the concentration of PCBs in fish (µg/kg). The model can be run deterministically (to predict point estimates of bioaccumulation in the food web) or probabilistically (by incorporating distributions for input parameters). These input parameters include the following: time series for sediment and water concentrations for PCBs, weight and lipid content of aquatic organisms, food ingestion rate and body weight of fish, total organic carbon in sediment, and Kow of the contaminant. The bioaccumulation model that predicts PCB concentrations in fish uses time series data for sediment and water concentrations, based on the output of the spatial exposure model, as explained above. Water concentrations were calculated from sediment concentrations using equilibrium partitioning. This is a conservative assumption since equilibrium is not likely in marine ecosystems. Data from the literature were used to develop distributions for species-specific input parameters, as described above. Model Constants. Several sources provided equations for the rate constants used in the model (30-32). TOC. We used DMMP site-specific measurements for TOC (Wisemiller, U.S. ACE, unpublished data) to derive a probability distribution (Table 1). VOL. 36, NO. 2, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
Body Weight and Lipid Concentrations. Lipid content and weight of fish for a particular age class (for example, adult fish) vary greatly and thus were treated using probability distributions derived from lipid concentrations and body weights available in the literature (33). Percent lipid distributions were specified as triangular for the winter flounder, using measurements in the NY-NJ area (34). Octanol-Water Partition Coefficient (Kow). PCBs were evaluated as “total” PCBs. Total PCBs represents a mixture of individual congeners, each of which has its own Kow. The Kow was treated as a model variable with an assigned triangular distribution, where the range is given by the minimum and maximum Kow for the individual PCB congeners analyzed, and the mode is estimated as the average of all the congeners in the mixture. The Kow data were obtained from Mackay et al. (35). Human Health Exposure and Risk Model. The cancer risk to adults is defined as
CSF × IRf × Cf ×ED BW × 106 × AT
where risk is the incremental individual lifetime cancer risk, CSF is the cancer slope factor (mg kg-1 day)-1, IRf is the annualized fish ingestion rate (g/day), Cf is the concentration of PCBs in fish (µg/kg), ED is the exposure duration (days), BW is the body weight (kg), AT is the averaging time (days), and 106 is the unit conversion factor. The noncancer risk was estimated using the hazard quotient approach, defined as
IRf × Cf × ED BW × RdD × AT × 106
where HQ is the toxicity hazard quotient, RdD is the reference dose (mg kg-1 day-1), BW is the body weight (kg), and 106 is the unit conversion factor. Exposure Duration. The outputs of the spatial and bioaccumulation submodels are predictions of PCB concentrations in the tissue of an individual fish over time. The averaging time and number of fish consumed are required to provide input for an estimate of human health risk from exposure through fish ingestion. An averaging time of 7300 days (i.e., 365 day/yr for 30 yr) was used to characterize lifetime exposure for cancer risk calculations. Annual averages (365 days) were used in characterizing noncancer risks. Fish Ingestion. The U.S. EPA Exposure Factors Handbook (36) provides a distribution for fish ingestion rates for adult recreational consumption of marine fish in the mid-Atlantic region. The fish ingestion rate for the average consumer was set at 6.5 g/day; the rate for a reasonably maximally exposed individual (RME) was set at 18.9 g/day (36). Body Weight. Body weight is set to 70 kg. This weight is commonly used in U.S. EPA risk assessments and is assumed in the derivation of CSFs (U.S. EPA Exposure Factors Handbook (36). Toxicity Factors. The cancer slope factor and reference dose are from the Integrated Risk Information System (37). These values are specified as point estimates following U.S. EPA guidance (33).
Results and Discussion Figure 2 presents the modeled spatial distribution of the fish foraging around the management site, displaying different degrees of attraction for the facility (AF; varies from 1 to 100). Each dot corresponds to the modeled spatial location during a specific month of one of the 1000 individual fish considered in this simulation. For a facility with no differential attraction (top graph), the fish are evenly distributed across the habitat. 242
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002
FIGURE 2. Foraging of winter flounder in the vicinity of the management site with different degrees of attraction. The management site is assumed to be 2500 m long and 1500 m wide and is centered at x ) 4000 m and y ) 3000 m. Each dot represents one of 1000 simulated fish in the population. The top plot represents a site that is equally attractive as compared to the neighboring areas. Foraging around more attractive sites (AF ) 10 and 100) are also shown. When AF is equal to 10 (middle graph), most of the predicted foraging occurs within the management site. When AF is equal to 100, the model predicts that only about 2% of the fish will forage away from the management site. The bars in Figure 3 represent the time-varying exposure point concentrations for three individual fish over the 3-yr time interval simulated by the spatial submodel. The high exposure point concentration corresponds to more frequent exposure to the contaminated site within the habitat. For example, fish 1 foraged within the site boundaries in August of year 1 and in August of year 2, while fish 2 happened to forage within the site boundaries in April of year 3. In each plot, the lower monthly averages for the exposure point concentrations correspond to less exposure to the site. These partial exposures occurred when fish foraged only in clean areas for some period, excluding contaminated sediments from the foraging habitat. The PCB tissue concentration in fish (simulated by the bioaccumulation submodel) is presented as solid lines in Figure 3. The figure clearly illustrates rapid PCB accumulation when fish forage in contaminated areas with much slower depuration after leaving the site. For example, when fish 3 foraged close to the facility in January-February of year 1, the PCB concentration in its tissue slowly decreased to the background level over the remainder of the year when the fish foraged only over clean sediments. Figure 4 compares cancer risk and hazard quotients for the highly conservative (but often employed) scenario in
FIGURE 3. PCB bioaccumulation in winter flounder. Bars in each graph represent exposure point concentrations (i.e., average sediment concentration the individual fish exposed in the corresponding month) calculated in the spatial submodel for three random individual fish. Solid lines present resulting temporal pattern for PCB bioaccumulation. which there are no spatial considerations for exposure to three scenarios that accommodate various spatially explicit assumptions about fish foraging and/or site characteristics. The x-axis presents exposure scenarios with various modeling assumptions for habitat size, attraction factors, and foraging areas. The y-axis shows cancer risk and hazard quotients resulting from fish consumption by recreational anglers under the varying assumptions. Under most regulatory programs, a hazard quotient exceeding 1 and a cancer risk between 10-4 and 10-6 indicate potential risk. Box and whisker plots represent the distribution of risks corresponding to each exposure scenario. The leftmost plot in Figure 4 represents the distribution of risks under the biologically unrealistic scenario that does not incorporate spatial or temporal aspects of fish exposure (i.e., assumes that fish are exposed all the time). The plot shows an expected 75% probability of exceeding a hazard quotient of about 27 and a 95% probability or less of experiencing a hazard quotient of about 31 or less. Risk estimates decrease under the various scenarios that incorporate data on the spatial and temporal dimensions of flounder biology and the physical characteristics of the management site. The box and whisker plots demonstrate that increasing the habitat size from 24 to 96 km2 decreases the median hazard quotient and cancer risk by a factor of 3. A site with a high attraction factor (100) could result in almost 7 times the median cancer risk and hazard quotient as compared to a facility with no differential attraction (AF ) 1). The 95th percentile shows a higher sensitivity to the size of the foraging area for winter flounder. A change from a 25 to a 2500 ha foraging area decreases the 95th percentile by almost a factor of 7 for the cancer risk and hazard quotient. The median risk value changes by a factor of 3.
Risk assessments provide risk managers with estimates to evaluate potential human health risks associated with exposure to contaminants in dredged sediments. Most often, the risk assessment defaults to conservative exposure assumptions. For example, food chain models often use the average concentration in contaminated media without considering the spatial and temporal behavior of the receptors. U.S. EPA guidance explicitly requires that risk assessments address uncertainty in the underlying assumptions (33). The pragmatic question facing dredged material managers is “How confident can risk managers be that these estimates realistically represent exposure and risk?”. The present analysis shows that spatial factors of fish behavior (size of foraging area and habitat) and the characteristics of the management site (size and differential attraction) are important components in evaluating realistic exposure and risk to human receptors. We have presented a model that (i) is useful in examining these factors, (ii) demonstrates the variation in risk estimates that they engender, and (iii) provides a model framework for incorporating realistic assumptions into risk estimates. Our analysis assumed that all fish consumption for recreational fishermen comes from a flounder population whose habitat was conservatively modeled, on the basis of local fish catch statistics. A biologically defined habitat for flounder would be much larger, resulting in much lower risk estimates. Even under our conservative assumption, the incorporation of rational (i.e., data-driven) parameters in the exposure models results in significantly lower median health risks as compared to a spatially nonexplicit model. To obtain median risks close to the prediction of the spatially nonexplicit case, all spatial parameters would have to be taken to conservative extremes simultaneously. It is important to note that the spatially explicit approach does not ignore the possibility that some individuals may ingest fish that have foraged mostly in the contaminated area. The advantage of the approach is that it assigns a probability to the occurrence of this scenario. For example, Figure 4 shows that when the model incorporates a small foraging area (25 ha), the 95th percentile for cancer risk and the hazard quotient are close to the 95th percentile observed in the scenario lacking spatial considerations. The reason is that if the foraging area is small, a fraction of the fish population will forage exclusively within the site boundaries and thus receive higher exposure. There is some probability, however small, that some individuals may eat only these fish, and this probability increases as the number of highly contaminated fish increases. We also tested the confidence of our model predictions by varying spatially explicit parameters over a wide range. Our results suggest that 95% estimates for cancer and noncancer risk are nearly always lower than the median risk estimate for the nonspatially explicit case. Scenarios with varying habitat sizes and attraction factors also result in 95th percentile values lower than in the spatially nonexplicit case. Even though the presented model incorporates simplified assumptions about the nature of the spatial behavior of ecological receptors, it is useful for capturing some of the major components of an exposure and risk analysis for contaminated sites. Since the current paper illustrates the general framework for a hypothetical case, a rigorous model validation cannot be presented here and will be the subject of a subsequent publication. Nevertheless, a qualitative evaluation of the available concentration measurements for the NY-NJ Harbor Estuary supports the validity of the model. Figure 5 presents PCB measurements in eel, winter flounder, and bluefish collected in six sampling areas in the NY-NJ Harbor Estuary in fall 1993 or early winter 1994 (38). VOL. 36, NO. 2, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
FIGURE 4. Hazard quotient and cancer risk for human consuming winter flounder. Scenarios with different assumptions for winter flounder spatial behavior are presented. Box and whisker plots are used to represent the uncertainty in risk estimates for each scenario. The six areas have varying sediment PCB contamination resulting from industrial discharges and disposal of waste materials. The three fish species represent different foraging strategies. Eels spend most of the time foraging in the same area, while bluefish are known to cover large distances within short time periods. As discussed in this paper, winter flounder is a residential fish; its foraging area is quite large as compared to that of eel but much smaller than that of bluefish. The figure shows that the average PCB concentration in eel varies over 3 orders of magnitude among sampling areas, while the range for bluefish contamination is less than 1 order of magnitude. The range of body burdens within the same sampling area shows a similar trend. Winter flounder caught within the same general area exhibit quite a wide range of PCB concentrations, while individual bluefish do not show as much variation in tissue PCB concentrations. 244
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002
Individual eels exhibit a smaller range of PCB concentrations as compared to winter flounders, which seems to contradict model predictions. However, individual eels were collected in exactly the same location, so the variations among these individuals are small and cannot be compared to variation among individual bluefish and winter flounder, which were collected at different locations within the same general area. These trends in concentration variation can be explained by the fact that fish with small foraging areas are likely to reflect local sediment contamination. Eel that happen to forage in a contamination hot spot are likely to be heavily contaminated, while other eel collected in a noncontaminated area are likely to be clean. Since bluefish forage over large spatial areas as well as consume fish that forage over extended spatial areas, they are affected by both clean and
FIGURE 5. PCB concentration in fish (wet weight and lipid normalized) collected within the New York-New Jersey Harbor Estuary by Skinner et al. (38). Sampling stations: 1, Upper Bay; 2, East River; 3, Kills; 4, Jamica Bay; 5, Lower Bay; 6, NY Bight. contaminated areas. No matter where bluefish are captured, they reflect the average contamination of a large habitat. Figure 5 shows that winter flounder falls between these two extremes. This paper illustrates the use of spatial modeling in risk analysis. If used in a realistic fashion, it can more fully inform the decision-making process for the management of contaminated sediments. The model could be also modified to incorporate additional complexities and numbers of sites within a habitat, different site shapes and contamination profiles, and preferential migration of ecological receptors. However, we note that the ability to use and interpret such models is often limited by the state of knowledge concerning the spatial behavior of ecological receptors. Nevertheless, probabilistic treatment of the model parameters (43), coupled with sensitivity analyses, should provide a rigorous basis for making sound environmental decisions.
Acknowledgments The authors Mr. Monte Greges and Mr. Bryce Wisemiller of the NY District Office of U.S. ACE for access to the DMMP data sets. Fruitful discussions and reviews by Ms. K. von Stackelberg; Drs. D. Vorhees, B. Hope, and S. Kane Driscoll; Ms. L. Williams; and two anonymous reviewers are also gratefully acknowledged. This study was supported by the U.S. Army Corps of Engineers, Dredging Operations Environmental Research Program (DOER). Permission was granted by the Chief of Engineers to publish this material.
Literature Cited (1) Suter, G. W. Ecological Risk Assessment; Lewis Publishers: Ann Arbor, MI, 1993. (2) Clifford, P. A.; Barchers, D. E.; Ludwig, D. F.; Sielken, R. L.; Klingensmith, J. S.; Graham, R. V.; Banton, M. I. Environ. Toxicol. Chem. 1995, 14, 895-906. (3) Freshman, J. S.; Menzie, C. A. Hum. Ecol. Risk Assess. 1996, 2 (3), 481-496. VOL. 36, NO. 2, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
(4) Schell, W. R.; Linkov, I. Priorities in Forest Radioecology: Model Development for Integrated Assessment of Toxic Element Transport. In Contaminated Forests: Recent Developments in Risk Identification and Future Perspectives, Linkov, I., Schell, W. R., Eds.; Kluwer: Amsterdam, 1999; pp 1-14. (5) Akc¸ akaya, H. R. Population Ecol. 2000, 42, 45-53. (6) Hope, B. K. Risk Anal. 2000, 20 (5), 573-89. (7) Suter, G. W.; Efroymson, R. A.; Sample, B. E.; Jones, D. S. Ecological Risk Assessment for Contaminated Sites; Lewis Publishers: Boca Raton, FL, 2000. (8) U.S. Environmental Protection Agency. Ecological Risk Assessment Guidance for Superfund: Process for Designing and Conducting Ecological Risk Assessments; EPA-540/R-97-006; U.S. Government Printing Office, Washington, DC, 1997. (9) U.S. Environmental Protection Agency. Guidelines for Ecological Risk Assessment. Fed. Regist. 1998, 63 (93), 26846-26924. (10) United States Environmental Protection Agency and United States Army Corps of Engineers. Evaluation of Dredged Material Proposed for Ocean Disposal: Testing Manual; EPA-503/8-91/ 001; U.S. Government Printing Office: Washington, DC, 1991. (11) Wilson, W. H., Jr.; Ruff, R. E. Species Profiles: Life Histories and Environmental Requirements of Coastal Fishes and Invertebrates (Mid-Atlantic)-Sandworm and Bloodworm; FWS/OB2-82/ 11.80); U.S. Fish and Wildlife Service: Washington, DC, 1988. (12) Pearcy, W. G. Bull. Bingham Oceanogr. Collect. 1962, 18, 1-78. (13) MacPhee, G. K. Feeding habits of the winter flounder, Pseudopleuronectes americanus (Walbaum), as shown by stomach content analysis. M.A. Thesis, Boston University, Boston, 1969, 66 pp. (14) Frame, D. W. Biology of young winter flounder Pseudopleuronectes americanus (Walbaum): Feeding habits, metabolism and food utilization. Ph.D. Thesis, University of Massachusetts, Amherst, MA, 1972, 109 pp. (15) Steimle, F. W.; Terranova, R. Trophodynamics of select demersal fishes in the New York Bight; NOAA Technical Memorandum NMFS-F/NEC-84; U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Northeast Region, Northeast Fisheries Science Center: Woods Hole, MA, July 1991. (16) Pereira, J. J.; Goldberg, R.; Ziskowski, J. J.; Berrien, P. L.; Morse, W. W.; Johnson, D. L. Essential Fish Habitat Source Document: Winter Flounder, Pseudopleuronectes americanus, Life History and Habitat Characteristics; NOAA Technical Memorandum NMFS-NE-138; U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Northeast Region, Northeast Fisheries Science Center: Woods Hole, MA, 1999. (17) Oviatt, C. A.; Nixon, S. W. Estuarine Coastal Mar. Sci. 1973, 1, 361-378. (18) Howe, A. B.; Coates, P. G. Trans. Am. Fish. Soc. 1975, 104, 1329. (19) Azarovitz, T. R. In Fish Distribution MESA New York Bight Monograph No. 15; Grosslein, M. D., Azarovitz, T., Eds.; New York Sea Grant Institute: Albany, NY, 1982. (20) Pierce, D. E.; Howe, A. B. Trans. Am. Fish. Soc. 1977, 106, 131139. (21) Saila, S. B. Limnol. Oceanogr. 1961, 6, 292-298. (22) Van Guelpen, L.; Davis, C. C. Trans. Am. Fish. Soc. 1979, 108 (1), 26-37. (23) Haedrich, R. L.; Haedrich, S. O. Estuarine Coastal Mar. Sci. 1974, 2, 59-73.
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 2, 2002
(24) (25) (26) (27) (28)
(30) (31) (32) (33)
(39) (40) (41) (42) (43)
Black, R.; Miller, R. J. Environ. Biol. Fishes 1991, 31, 109-121. Jones, R. J. Cons., Cons. Int. Explor. Mer. 1959, 25, 58-72. MacDonald, J. S. Can. J. Zool. 1983, 61 (3), 539-585. Phelan, B. A. Trans. Am. Fish. Soc. 1992, 121, 777-784. Scarlett, P. G. Life history investigations of marine fish: occurrence, movements, food habits and age structure of winter flounder from selected New Jersey estuaries; Technical Series; New Jersey Department of Environmental Protection, Division of Fish, Game, and Wildlife, Marine Fisheries Administration, Bureau of Marine Fisheries: 1988; pp 88-20. New Jersey Department of Environmental Protection. Fish consumption patterns by New Jersey consumers and anglers; New Jersey Marine Sciences Consortium, Sandy Hook, NJ, and New Jersey Department of Agriculture, Trenton, NJ, 1994. Gobas, F. A. P. C. Ecol. Modell. 1993, 69, 1-17. Gobas, F. A. P. C.; Z’Graggen, M. N.; Zhang, X. Environ. Sci. Technol. 1995, 29, 2038-2046. Burkhard, L. P. Environ. Toxicol. Chem. 1998, 17, 383-393. U.S. Environmental Protection Agency. Risk Assessment Guidance for Superfund, Volume 1-Human Health Evaluation Manual, Part A, Interim Final; EPA/540/1-89/0002; U.S. EPA, Office of Emergency and Remedial Response: Washington, DC, 1989. Northeast Fisheries Science Center. 25th Northeast Regional Stock Assessment Workshop (25th SAW): Stock Assessment Review Committee (SARC) Consensus Summary of Assessments; Northeast Fisheries Science Center Reference Document 97-14; 1997; 143 pp. Mackay, D.; Shiu, W. Y.; Ma, K. C. Illustrated Handbook of Physical-Chemical Properties and Environmental Fate for Organic Chemicals. Volume I. Monoaromatic Hydrocarbons, Chlorobenzenes, and PCBs; Lewis Publishers: Chelsea, MI, 1992. U.S. Environmental Protection Agency. Exposure Factors Handbook, Volume I: General Factors; EPA/600/P-95/002Fa; U.S. EPA, Office of Research and Development: Washington, DC, 1997. U.S. Environmental Protection Agency. Integrated Risk Information System Database (IRIS), http://www.epa.gov/iris, 2000. Skinner, L. C.; Jackling, S. J.; Kimber, G.; Waldman, J.; Shastay, J., Jr.; Newell, A. J. Chemicals in Fish, Shellfish and Crustaceans from the New York-New Jersey Harbor Estuary: PCB, Organochlorine Pesticides and Mercury; New York State Department of Environmental Conservation, Division of Fish, Wildlife and Marine Resources: November 1996. Schrock, M. E.; Barrows E. S.; Rosman, L. B. Chemosphere 1997, 34, 1333-1339. Lemieux, H.; Blier, P. U.; Dufresne, F.; Desrosiers, G. Mar. Ecol. Prog. Ser. 1997, No. 156, 151-156. Briggs, D. E. G.; Kear, A. J. Paleobiology 1993, 19, 107-135. Rosman, L. B. Personal communication with Igor Linkov regarding sandworms, 1999. Linkov, I.; von Stackelberg, K.; Burmistrov, D.; Bridges, T. S. Sci. Total Environ. 2001, 274, 255-269.
Received for review March 8, 2001. Revised manuscript received September 19, 2001. Accepted October 12, 2001. ES010721D