Assessment and Remediation of Contaminated Sediments (ARCS) Program
Table of Contents
- Chapter 1
- Chapter 2
- Chapter 3
- Chapter 4
- Chapter 5
- Chapter 6
- Chapter 7
- Chapter 8
- Chapter 9
- Chapter 10
- List of Tables
- List of Figures
Assessment Guidance Document
US Environmental Protection Agency. 1994. ARCS Assessment Guidance Document. EPA 905-B94-002. Chicago, Ill.: Great Lakes National Program Office.
Table Of ContentsDATA PRESENTATION AND INTERPRETATION
- SEDIMENT QUALITY DESCRIPTION AND MAPPING
SEDIMENT CLASSIFICATION METHODS
- Whole Sediment Toxicity Testing
- Spiked Sediment Toxicity Testing
- Interstitial Water Toxicity Identification Evaluation
- Equilibrium Partitioning
- Tissue Residues
- Benthic Macroinvertebrate Community Structure
- Sediment Quality Triad
- Apparent Effects Threshold
- National Status and Trends Program Effects-Based Approach
- Use of the Sediment Classification Approaches
- NUMERICAL RANKING OF HAZARDOUS SEDIMENTS TO PRIORITIZE SITES FOR REMEDIAL ACTION
- CONCLUSIONS AND RECOMMENDATIONS
The interpretation of sediment quality data is an iterative process that proceeds throughout the course of a lengthy sediment assessment program. Ideally, it begins with the compilation, review, and synthesis of available information on the AOC. The rationale for such interpretation of existing data was recently summarized by USEPA (1990):
Before full-scale, potentially costly sediment assessment programs are begun, the initial identification of areas containing probable contamination problems should be attempted. The contamination of sediments is a process influenced by a number of variables including contaminant source, contaminant type, sedimentary and hydrologic environment, sediment grain size distribution and composition, presence and type of aquatic life, and historical influences. The likelihood of there being a sediment contamination problem at a particular site needs to be appraised based on readily available information. Such information may be available from ongoing monitoring or regulatory programs, previous site characterizations, dredging records, discharge permits, area maps, fishing advisories, reports of spills, fish kills and beach closings, etc.
This early data interpretation effort should be designed to identify the chemicals of concern for a given AOC, and any data gaps that must be filled to provide a more complete characterization of any sediment quality problems. This should serve to focus further assessment efforts and make the best possible use of available resources.
Subsequent data collection efforts may be tiered, and each tier may be followed by a separate data interpretation effort. The purpose of each tier of field sampling and laboratory analyses should be to refine and fine-tune the understanding of sediment contamination problems within the AOC. The overall goals of a sediment assessment program can generally be summarized as answering three questions:
- What is the nature and spatial extent of chemical contaminants in sediments relative to appropriate reference conditions?
- What sediments have sufficiently high concentrations of chemical contaminants so as to present unacceptable risks to humans or aquatic biota, and therefore must be considered for remediation?
- How should priorities for remediation be assigned to various sites within an overall AOC?
The first question above is addressed by collecting representative sediment samples, analyzing them for the chemicals of concern, and then accurately recording the resultant data in an easily interpreted form. The first section below describes a number of considerations and recommendations for describing and mapping sediment quality data in easily readable forms that enable the data user to interpret the 3-dimensional distribution of important sediment quality parameters.
The second question above is addressed through the collection of data for an integrated sediment assessment, encompassing sediment chemistry, physical characteristics, and biological effects data. The second section below discusses various approaches that are potentially applicable to the interpretation of data resulting from an integrated sediment assessment. Human health and ecological risk assessments are not addressed in this guidance document, but are also considered to be of vital importance in answering the question of whether sediments present unacceptable risks. Additional information on these topics is discussed in the ARCS Risk Assessment and Modeling Overview Document (USEPA 1993a).
The last question is especially important, given the high cost of dredging and other forms of sediment remediation. Cost considerations may well limit future remediation activities to only the most hazardous sediments, however they may be defined. It is therefore vitally important that a detailed, accurate characterization of the area be performed to focus remedial efforts where they will most efficiently lower risks posed by the most hazardous sediments. The third section below summarizes a strategy that is potentially applicable to prioritizing sites within an overall AOC for remediation.
It is impossible to recommend a single set of data presentation and interpretation techniques that would be applicable in all cases. The approach to be taken will necessarily be a function of both the types of data collected and the specifics of the AOC under consideration. Nevertheless, the discussion that follows is intended to give the reader an overview of potentially applicable data presentation and interpretation techniques that may be useful for individual sediment assessment programs.
Sediment quality is often highly variable in all three dimensions; representing this variability through sampling and other means is a key component of the overall sediment assessment. Highly contaminated sediment deposits can be quite localized, as noted in a national overview of sediment quality data (Lyman et al. 1987):
The combined effect of varied source locations, and variable hydrology and sediment characteristics, has led to large variability in the concentrations of in-place [sediment] pollutants within a water course or water body. The more contaminated spots are often referred to as "hot spots."
The report continues:
An important qualification . . . is that in each location, the actual areas of high contamination may be extremely localized. These localized areas with high levels are often related to the location of the sources of contamination, e.g., at the end of a sewage or industrial outfall. In general, however, they are difficult to identify and pinpoint. Their locations appear to vary due to the movements of currents and other disturbances, e.g., ship traffic or dredging. The high mobility of sediments in some water bodies is a complicating issue. Pollutants discharged in the upper reaches of a watershed may travel tens or hundreds of miles before finding a relatively permanent "home" in an open harbor, lake or bay. Even here, however, permanent or episodic (e.g., storm generated) currents can result in significant sediment redistribution. In some areas, older contaminated sediments may become buried by cleaner material as part of the natural sedimentation processes.
Even so, much of the reported information about contaminated sediments, which is based on grab samples of surficial sediments, gives the impression that this is largely a two-dimensional problem. In fact, as ARCS and other coring studies have shown, the most highly contaminated sediments may be located well below the sediment surface (i.e., in older sediments). Consequently, it is essential to have some means of representing contaminant distributions in three dimensions.
In general, sediment quality data are more easily interpreted when presented in map form, because the goal is to understand how sediment contaminants and toxicity are distributed within a particular AOC. Quantitative mapping provides valuable insights on the extent and variability of contaminant zones. It also aids in providing some basis for prioritizing or ranking sites within an AOC.
The cartographic representation of sediment quality data begins with the selection of a suitable base map upon which to plot the data. The base map should be of such a scale as to balance the need for detail with the area to be covered and of an accuracy commensurate with the intended use of the final map product. For example, smaller scale maps are more appropriate for detailed data analysis (e.g., hot-spot mapping). Potential sources of base maps include USGS 1:24,000 and 1:100,000 topographic quadrangles, NOAA nautical charts, and, for rivers and harbors where navigational dredging takes place, Corps project maps. The latter, used by the Corps to plan navigational dredging, may provide the most useful base maps for sediment mapping (see Data Set Mapping below). In some cases, it may be useful to indicate historical information, such as former industrial sites or effluent sources, on the base maps. Base maps considered useful for the ARCS priority AOCs included USGS 7-1/2 minute quadrangles (for watershed information), NOAA nautical charts (for navigation and harbor surroundings), city maps (for local road access), and Corps project maps.
Maps generated using computer mapping software require a plane coordinate reference grid to plot information. The two most common plane coordinate systems are the State Plane Coordinate System and the Universal Transverse Mercator Grid System. Regardless of which grid system is chosen, all position data collected in the field using the spherical geographic coordinate system (i.e., latitude/longitude) need to be converted to the appropriate plane coordinate system before they can be plotted on a map. A variety of coordinate conversion programs are available, and many computer mapping programs have built-in conversion capabilities. The locations of sampling stations in the ARCS field surveys were determined accurately using DGPS (see Chapter 3) and plotted on digitized Corps project maps in a state plane coordinate system.
It is sometimes useful to include bathymetric data, normalized to a common datum (i.e., water depth relative to low water datum), on maps because such data establish the contours of the upper surface of the sediments. Bathymetric data may already be available from the Corps and NOAA charts, or measured with sounding instruments (fathometers) in the field.
The distribution of sediment quality parameters can be represented on maps in various ways. Single-value quantitative point symbol maps can be used to represent the values of a single parameter at various locations. A simple geometric symbol, such as a circle, can be placed at the desired location, and its size, color, color intensity, or texture can be varied to indicate the magnitude of the parameter. This type of mapping lends itself well to samples collected from a particular depth horizon, such as surface grab samples (Figure 9-1).
Sediment core data, which represent multiple values at a given point, can be graphically depicted using icons that are diagrammatic representations of data. The advantage of icons is that they can express large amounts of information concisely in a small space and, if carefully designed, are easily understood. The apparent simplicity of icons, however, belies the fact that their generation can be extremely labor intensive. Icons can be used to show multiple values of a single parameter (Figure 9-2), multiple values of many parameters (Figure 9-3), or both qualitative and quantitative data for multiple values for one or many parameters (Figure 9-4).
Contour mapping or surface modeling is a tool that can be used to predict continuous, 2-dimensional distributions of data from discrete point data. A number of contouring software programs exist with each package usually containing several different contouring algorithms. As Figure 9-5 shows, different contouring algorithms applied to the same data set can result in different data distribution patterns. A more thorough description of contour mapping and alternative algorithms is provided by Baudo (1990).
One option available with many contour mapping software packages is the ability to create pseudo 3-dimensional surface representations. These surface representations lend themselves extremely well to the visualization of sediment topography (Figure 9-6). In addition, some contouring packages allow contour maps of different parameters to be draped over these generated surfaces (Figure 9-7). With some contouring packages, it is also possible to estimate sediment volumes.
Contour mapping is a powerful tool for visualizing the spatial distribution of sediment quality data, but it is important to recognize that the resulting contour map is only a model of the actual surface distribution based on interpolation and extrapolation of values at selected sampling points. The more accurately the sampling points represent the concentration range and distribution of the parameter of interest, the more accurate will be the contour map. Consequently, an effort should be made to incorporate sampling points that anticipate the distribution and range of the target parameter into the sampling strategy. For example, selecting a single sampling point upstream or downstream of a known point source might bias the resulting contour map in the area of that point source, without representing the true spatial variability. By sampling both upstream and downstream of the point source, an abrupt change in parameter values will be constrained to the appropriate area.
The producer of contour maps must exercise some discretion in deciding what can be reasonably contoured and how it should be done. Contouring large areas with few data points may yield maps that are useful for planning purposes but are too inaccurate for other purposes (e.g., for making remedial decisions). Producing contour maps with data collected from a linear array of stations along the shore will reflect variability along the shore but will not give an accurate representation of variability outward from the shore, especially in areas where navigational dredging has occurred.
A GIS combines cartographic display, data management, and spatial analysis capabilities in one software package. In addition to producing maps using the techniques discussed above, a GIS allows the spatial analysis of existing maps using various analytical tools to produce new maps with new or enhanced information.
Options for true 3-dimensional mapping of sediment quality data include the geologic modeling program (GMP, Dynamic Graphics Corporation, California), which runs on a Personal Iris 4D/2D graphics workstation (Silicon Graphics Corporation, Mountain View, California). On the workstation, sediment data representing specified intervals of coring data can be interpolated onto a 3-dimensional grid, concentration ranges can be color-coded, and the resulting data model (contour map) of contaminant zones can be displayed on an outline map of the site. Once displayed, the model can be manipulated on the screen: it can be scaled up or down, stretched vertically, rotated on three axes, viewed transparently, "peeled" away zone by zone, and sectioned along different planes. GMP can also be queried for point concentration values and volume calculations from the various displays of contaminant zones. This is a useful way to simulate different dredging scenarios and estimate their costs.
The ultimate goal of the sediment assessment techniques discussed in this guidance document is to assess whether and to what extent sediments are "contaminated" or have the potential to adversely affect the environment. Chemical analyses of sediment samples can demonstrate whether chemical concentrations in a specific area of interest are elevated relative to a reference or background area. However, elevated chemical concentrations alone are insufficient to demonstrate adverse environmental effects. The focus of attempting to classify sediments as "contaminated" or "uncontaminated" may be on the protection of ecological receptors, human receptors, or, more typically, both. The sediment assessment techniques described in this document can be used together to help interpret integrated sediment assessment data (i.e., combining measurements of sediment chemistry, physical characteristics, and various indicators of biological effects). As indicated earlier, human health and ecological risk assessment procedures are not addressed in this guidance document, but are discussed in the ARCS Risk Assessment and Modeling Overview Document (USEPA 1993a).
A number of different approaches are potentially applicable to the assessment of the adverse effects of sediment contamination on ecological receptors. The following approaches, recently reviewed and evaluated in the USEPA's Sediment Classification Methods Compendium (USEPA 1992), are summarized here for potential application to sediment assessments in Great Lakes AOCs.
Whole sediment toxicity testing can be used to predict whether sediments can have adverse effects on benthic biota (USEPA 1992; see also Chapter 6). Test organisms are exposed in the laboratory under controlled conditions to field-collected sediments. To measure toxicity, a specific biological endpoint (e.g., mortality, reductions in growth or reproduction) is used to assess the response of the organisms to contaminants in the sediments. It is assumed that the toxicity of chemicals measured in the test sediments is similar to that in natural in situ sediments. One of the benefits of whole sediment toxicity testing is that it integrates the effects of all sediment contaminants. That is, the interactions (e.g., synergism, additivity, antagonism) of various chemicals can be taken into account without the need to measure their concentrations in the sediments, and without any a priori knowledge of specific pathways of interaction between sediments and test organisms. Although whole sediment toxicity testing can be used to demonstrate adverse effects, such biological testing cannot be used alone to identify the chemical contaminant(s) responsible for the observed effects or to generate sediment quality values (SQVs) for individual chemical contaminants.
Spiked sediment toxicity testing can be used to predict the concentrations of specific chemicals that would be expected to be harmful to resident biota under field conditions (USEPA 1992). Test organisms are exposed in the laboratory under controlled conditions to uncontaminated sediments that are spiked with known concentrations of specific chemical contaminants. The results are evaluated to establish cause-and-effect relationships between chemicals and specific adverse effects (e.g., mortality, reductions in growth or reproduction). The results can also be evaluated to establish dose-response relationships and to generate SQVs for individual chemicals. While it is theoretically possible to evaluate the interactions (e.g., synergism, additivity, antagonism) among various chemicals by combining those chemicals in the spiked sediment samples, it would rarely be possible to mimic the complex mixtures of chemicals typically found in natural sediments. Another difficulty with this approach is that the site-specific factors that affect the bioavailability of chemical contaminants are not always known and would be difficult to simulate in the laboratory, especially for a wide variety of field-collected sediments (e.g., varying grain sizes, TOC content). It is also difficult to determine whether the contaminants are at equilibrium with the sediments. Evaluation of a large number of chemical contaminants by this method would also be very expensive.
The interstitial water toxicity approach is a multiphase procedure for assessing sediment toxicity using interstitial (pore) water separated from field-collected sediment samples (USEPA 1992). Interstitial water is used because of the supposition that it more accurately represents the contaminant concentrations that an organism is exposed to in the environment. The toxicity of the pore water is first quantified in laboratory toxicity tests, and then TIE procedures are used to identify and quantify the chemical constituents of the interstitial water responsible for the sediment toxicity. The TIE procedures are implemented in three phases to characterize the nature of the interstitial water toxicant(s), identify the suspected toxicant(s), and confirm identification of the suspected toxicant(s). These procedures, developed primarily for the evaluation of municipal and industrial effluents, are not as readily applied to sediments because of the difficulty in collecting sufficient volumes of interstitial water for toxicity testing. Typically, the TIE approach has only been used to assess the acute toxicity of sediment samples (e.g., <4-day tests).
The equilibrium partitioning approach focuses on predicting the chemical interactions among sediments, interstitial water, and contaminants, and assumes that the chemical contaminant concentrations in interstitial water are acceptable predictors of adverse biological effects (USEPA 1992). Based on equilibrium partitioning theory, the chemical contaminant concentrations in interstitial water are predicted from the bulk sediment chemical contaminant concentrations. If the predicted chemical contaminant concentrations in interstitial water exceed applicable water quality criteria or any other effect concentration, the sediment is predicted to have adverse biological effects. Many other sediment contaminants may have toxic effects that cannot be predicted using this approach. The equilibrium partitioning approach can be used to generate SQVs for individual chemicals. The USEPA is currently developing specific regulatory uses of SQVs based on this approach; however, the widespread application of the approach will be dependent on the development of water quality criteria for many more potentially toxic chemicals and an appropriate determination of uncertainty for site-specific applications. The equilibrium partitioning approach is also not capable of evaluating the synergistic, additive, or antagonistic effects of mixtures of sediment contaminants, such as those found in most naturally occurring sediments.
In the tissue residue approach, sediment chemical concentrations are determined that would result in unacceptable residues in the tissues of organisms of concern (i.e., either ecological or human receptors) (USEPA 1992). The chemical concentrations that represent unacceptable tissue residues may be derived from toxicity tests performed during generation of chronic water quality criteria, from bioconcentration factors derived from the literature or generated by experimentation, or by comparison with human health risk criteria associated with consumption of aquatic organisms. The tissue residue approach can be used to generate SQVs, and is most applicable for nonionic organic and organometallic compounds. However, this approach can also be used to evaluate metals and polar organic compounds. The approach has recently been applied to the calculation of the sediment concentration of TCDD that would be necessary to attain acceptable concentrations of TCDD in fish in Lake Ontario (Cook et al. 1990). The acceptable TCDD concentration in sediment is being used as the criterion for determining the remedial action necessary to reduce incremental loading of TCDD to the lake from a Superfund site (Carey et al. 1989).
Documentation of the structure of benthic macroinvertebrate communities through the taxonomic identification and enumeration of field-collected organisms may be used to assess sediment quality (USEPA 1992; see also Chapter 7). Benthic macroinvertebrates are relatively sedentary organisms that inhabit or depend on the sedimentary environment for their various life functions. Therefore, they may be sensitive to both long-term and short-term changes in habitat, sediment, and water quality. Unlike laboratory toxicity tests, assessments of the structure of benthic macroinvertebrate communities provide direct evidence of the effects of sediment contaminants on naturally occurring communities. Deviations from expected community characteristics (such as may be demonstrated by statistical comparisons with reference area conditions) may be attributable to the presence of chemical contaminants. However, they may also be attributable to other factors (e.g., sediment grain size, organic content) unrelated to chemical contamination. Therefore, it is generally considered essential to make comparisons with the benthic macroinvertebrate communities in reference areas with similar sediment characteristics except for the presence of chemical contaminants. Evaluations of benthic macroinvertebrate community structure cannot be used alone to generate SQVs, but may be an important part of an integrated sediment assessment.
The Sediment Quality Triad approach is an effects-based approach to describing sediment quality that incorporates measures of sediment chemistry, sediment toxicity, and benthic macroinvertebrate community structure (Chapman 1986, 1989; Chapman et al. 1992; USEPA 1992). All three measures are evaluated for samples of field-collected sediments from the same location. The Sediment Quality Triad can provide strong, complementary evidence for the degree of contamination-induced degradation in aquatic communities. The Sediment Quality Triad also provides a direct assessment of sediment quality and can be applied to all chemicals of concern, although it does not prove a cause-and-effect relationship between the concentrations of individual chemicals and adverse biological effects. This approach is most commonly used to describe sediment characteristics qualitatively.
The results of the three measures can be arrayed in a matrix to facilitate interpretation of the results (Table 9-1). Sediment Quality Triad data can also be plotted on triaxial graphs (Figure 9-8) to provide a visual representation of the data (Chapman et al. 1991). The data for each individual measure are first scaled proportionally between 1 and 100 (with 100 being the greatest effect; i.e., highest concentration of chemical contaminants, highest toxicity, or most altered benthic macroinvertebrate community) to keep the relative magnitude of the differences consistent for the three measures. Relative sediment quality can be evaluated by the sizes and shapes of the triangles. Large triangles are indicative of more contaminated or more impacted sites. More equilateral triangles indicate that the data from the three measures agree.
The AET approach employs synoptically collected sediment samples that are analyzed for both sediment chemistry and biological effects (Barrick et al. 1988; USEPA 1992). The biological effects used to date in the generation of AET values have included both assessments of benthic community structure and several different whole sediment toxicity tests. The significance of adverse biological effects is assessed by statistical comparisons with suitable reference or control sediments. The biological effects data are then considered in conjunction with the paired sediment chemistry data. For a given data set, the AET value for a given chemical contaminant is the sediment concentration above which a particular adverse biological effect has always been found to be statistically significant relative to reference conditions.
The AET values can be used as predictors of adverse biological effects for sediment samples where only sediment chemistry data are available. If the concentration of any chemical in a given sediment sample exceeds its AET value for a particular biological indicator, an adverse biological effect is predicted for that indicator. If the concentrations of all chemicals in a given sediment sample are below their respective AET values for a particular biological indicator, then no adverse effect is predicted for that biological indicator. The AET approach does not prove a cause-and-effect relationship between the concentrations of individual chemicals and adverse biological effects, but it provides a valuable tool for screening out samples where there is only a low likelihood of such effects.
To ensure the reliability of the AET values generated using this approach, a relatively large database (generally more than 30, and preferably at least 50 stations) is recommended, spanning a wide range of chemical contaminant mixtures and concentrations (Barrick et al. 1988). The AET values generated using this approach should appropriately only be applied within the geographic region where the AET database was collected. To date, the AET approach has been used in the State of Washington for the generation of marine SQVs used in sediment regulatory programs, and has been initially examined for similar use by the State of California.
NOAA has employed this approach to develop "informal, effects-based guidelines" for the assessment of sediment quality (USEPA 1992). It involves the identification of ranges in sediment chemical concentrations associated with biological effects based on a weight of evidence from many studies. In this approach, the data for many individual chemicals are assembled from modeling, laboratory, and field studies to determine ranges in chemical concentrations that are rarely, sometimes, and usually associated with adverse biological effects (e.g., toxicity). The approach has been used to calculate, based on the statistical distribution of a large amount of effects-based data, a "no-effects range," a "possible effects range," and a "probable effects range" of sediment contaminant concentrations for individual chemicals. Two slightly different methods have been used to determine these ranges.
Long and Morgan (1990) initially assembled a large database that included both data demonstrating biological effects and data demonstrating no biological effects. Included were field and laboratory data for both freshwater and saltwater organisms. Long and Morgan (1990) defined an Effects Range-Low (ER-L) value as the lower 10th percentile concentration for those sediment chemical contaminant concentrations associated with biological effects. Sediment chemical contaminant concentrations below the ER-L value were considered to represent the "no effects range." An Effects Range-Median (ER-M) value was defined as the 50th percentile concentration for those sediment chemical contaminant concentrations associated with biological effects. Sediment chemical contaminant concentrations between the ER-L and the ER-M values were considered to represent the "possible effects range" (i.e., at concentrations above the ER-L value, adverse effects may begin or are predicted to occur among sensitive lifestages or species as determined by sublethal tests). Sediment chemical contaminant concentrations above the ER-M value were considered to represent the "probable effects range" (i.e., at concentrations above the ER-M value, adverse effects are frequently or always observed or predicted to occur among most species).
Long and Morgan (1990) indicated that the ER-L and ER-M values were intended "only for use by NOAA as general guidance in evaluating the NS&T [NOAA's National Status & Trends] Program data." They also cautioned that "there is no intent expressed or implied that these values represent official NOAA standards." Nevertheless, others have attempted to use the ER-L and ER-M values as SQVs in other applications and in ways not intended by Long and Morgan (1990). Such uses should be attempted with caution.
More recently, MacDonald (1992) and Long et al. (in press) have refined the application of the Long and Morgan (1990) approach. MacDonald (1992) segregated saltwater data from freshwater data, and identified the three effects ranges with a method that used both the concentrations associated with biological effects and those associated with no observed effects. Based on statistical manipulations of the chemical contaminant concentration data, MacDonald (1992) then defined a no-observed-effect-level (NOEL), a threshold effects level (TEL), and a probable effects level (PEL). The "effects" and "no effects" databases still contain a wide variety of biological tests.
Long et al. (in press) have applied similar refinements to the approach and have expanded the original Long and Morgan (1990) database. The revised Long et al. (in press) database is limited to saltwater and estuarine data, however, and is therefore not applicable to the Great Lakes region.
Although different in their approach, the ER-L values defined by Long and Morgan (1990) are roughly equivalent to the NOEL values defined by MacDonald (1992), while the ER-M values are roughly equivalent to the PEL values. Just as for the AET values, neither the ER-L and ER-M values developed by Long and Morgan (1990) nor the TEL and PEL values developed by MacDonald (1992) prove a cause-and-effect relationship between the concentrations of individual chemicals and adverse biological effects. Nevertheless, they may be useful for screening sediment samples to determine the likelihood of such effects. Neither the ER-L and ER-M values nor the TEL and PEL values should be used alone as SQVs for establishing whether a given sediment is "contaminated" or "uncontaminated."
For an extensive discussion of the sediment classification approaches described above, including their applicability, advantages, disadvantages, and level of acceptance, as well as for additional references to pertinent source documents on these approaches (to 1992), the reader is referred to USEPA (1992). Although not discussed by USEPA (1992), it should be noted that USEPA and the Corps are jointly developing guidelines for the evaluation of dredged material from inland waters (USEPA-USACOE 1993), but these guidelines are not yet final.
The various approaches described above for classifying sediments as "contaminated" or "uncontaminated" can generally be categorized as numeric, descriptive, or a combination of numeric and descriptive approaches (USEPA 1992). Numeric methods (e.g., spiked sediment toxicity testing, interstitial water TIE, equilibrium partitioning, tissue residues) can be used to derive chemical-specific SQVs. Descriptive methods (e.g., whole sediment toxicity testing, benthic community structure) cannot be used alone to generate numerical SQVs for individual chemicals but do provide important information on ecological effects. Although both numeric and descriptive approaches can be used in assessing sediment quality, none of these approaches alone is considered adequate for a comprehensive sediment assessment. An integration of several methods using a weight-of-evidence approach is needed to assess the effects of chemical contaminants associated with sediment. The approaches that integrate data from whole sediment toxicity testing, chemical analyses, and benthic community assessments (e.g., the Sediment Quality Triad or AET approaches) provide strong complementary evidence of the degree of contaminant-induced degradation in aquatic communities and are therefore recommended for future studies of Great Lakes AOCs.
Under the ARCS Program, the integrated sediment assessment approach developed by the Toxicity/Chemistry Work Group included chemical analyses of sediments (see Chapter 5), whole sediment toxicity testing (see Chapter 6), and analyses of benthic community structure (see Chapter 7). Some toxicity tests were conducted using interstitial water and elutriate samples collected from the sediment samples, but they were not conducted in a phased manner with TIE procedures to identify and quantify the chemical constituents responsible for observed adverse effects. Because SQVs and equilibrium partitioning are tools for manipulating and interpreting the results of chemical analyses rather than for generation of chemical data directly, they were applied during data interpretation, classification of sediments as "contaminated" or "uncontaminated," and for intra-site ranking. Methods for developing SQVs based on tissue residues were investigated by the ARCS Risk Assessment and Modeling Work Group, and are discussed in the ARCS Risk Assessment and Modeling Overview Document (USEPA 1993a).
Two other important types of information collected under the ARCS Program were data on fish tumors and abnormalities (see Chapter 8) and on bioaccumulation in fishes. While neither is strictly part of the integrated sediment assessment approach, both provide important complementary information on the health of ecological communities that may potentially be related to sediment contamination.
Efforts to interpret the ARCS Program's integrated sediment assessment data using the various sediment classification approaches are continuing and are not yet ready for publication.
One goal of the ARCS Program was to develop a ranking method by which the relative risks associated with contaminated sediment from different sites can be compared within AOCs, among AOCs, or both. A general method of ranking contaminated sediment sites based on whole sediment chemistry was proposed by Kreis (1989), which put all sediment chemical concentration variables on the same scale so that they could be compared and combined. The numerical ranking system developed by Kreis (1989) was intended for use by managers in regulatory and remediation decision-making for contaminated Great Lakes sediments. Kreis (1988) had previously shown that the ranking process can be an effective tool for determining which sites, of a range of contaminated sites, need the most immediate attention. Thus, the results of the ranking process can be used to prioritize sites for remediation, which is desirable because of the high cost of sediment remediation. As resources become available, the sediments needing remediation could each be "cleaned-up" in the order of their ranking. However, other important factors that are not included in this ranking scheme (e.g., human health, economic factors) must also be considered. The actual remediation technology or combination of remediation technologies chosen is site-specific and would depend on ecological, chemical, economic, and engineering considerations that are independent of the site ranking process.
The Kreis (1989) methodology was modified for use in the ARCS Program by incorporating estimates of contaminant bioavailability, toxicity, and potential for effects on benthic community structure for the sediment contaminants of concern (Wildhaber, in press). The basic elements of this ranking method are described in the following sections. It should be recognized that the method is still undergoing development and is not yet ready for routine application. Nevertheless, it introduces some of the concepts considered desirable in any contaminated sediment site-ranking method that may ultimately be selected.
In the ranking system proposed by Kreis (1989) for sediment chemistry, each chemical or group of chemicals (e.g., metals, dioxins) analyzed is ranked independently of each other. The measured concentrations of each chemical or group of chemicals for each site under consideration are scaled from 1 to 100, relative to each other; the lowest contaminant concentration for a given chemical or group of chemicals becomes 1 and the highest concentration for that chemical or group of chemicals becomes 100. The equation used to calculate the ranks for each chemical or group of chemicals is:
|Rank = 1 +||Site Value - Minimum Value||x 99|
|Maximum Value - Minimum Value|
The ranks calculated for each chemical or group of chemicals are then averaged (arithmetic mean) for each site. The result is an average rank for each site based on all measured chemicals for which ranks were assigned. One problem with this ranking process is that the chemicals analyzed are scaled relative to each other based only on the concentrations present; it does not scale those chemicals based on a true measure of concern, such as their toxicity and availability to aquatic organisms. Another problem is that the ranks for each chemical or group of chemicals at a given site are not necessarily independent of one another, especially if the chemicals have a common source or sources.
The alternative approach considered under the ARCS Program differs substantially from that proposed by Kreis (1989) in that it uses toxicological and ecological information as well as estimated contaminant bioavailability to scale the chemicals before their ranks are combined. In this approach, each chemical or group of chemicals analyzed is not independently ranked. Instead, all the chemicals are put on a common toxicity scale and totaled among chemicals for each site; this total toxicity is then ranked. The result is a relative ranking of the sites under investigation based on what is known about the toxicity and potential bioavailability of the compounds found in the sediments.
Before the data for different chemicals can be combined, they must be put on the same toxicity scale, which is achieved through the use of individual bioavailability and toxicity estimates for each chemical measured in the sediments. The toxicity of chemicals in sediments is believed to be at least in part a function of how tightly bound the chemicals are to the sediments, or, conversely, how readily the chemicals can dissolve in the pore water. Different sediments with the same total quantities of individual toxic chemicals may exhibit varying toxicities because other sediment properties may influence the extent to which the chemicals are bound to the sediments.
For nonionic organic chemicals, the organic carbon content of the sediments is believed to be a primary determinant of the distribution of the chemicals between the solid and aqueous phases. Hence, the pore water concentration of nonionic organic chemicals can be estimated based on the whole sediment concentration of the chemical, the TOC content of the sediment, the partition coefficient for sediment organic carbon, and the assumption of equilibrium partitioning (Di Toro et al. 1991). There are, of course, many situations where the sediment and pore water may not be in equilibrium, but for the purposes of this estimation, the assumption is necessary.
The sorption of metals to sediments is potentially more complex, and may be influenced by the presence of oxides of iron and manganese, organic carbon, and sulfides. Di Toro et al. (1990) have suggested that the solubility (and therefore bioavailability and toxicity) of divalent metals may be primarily determined by the AVS phase (i.e., the solid-phase sediment sulfides that are soluble in cold acid). If there is more metal present on a molar basis than sulfides on a molar basis, then the metal may exist in the aqueous phase and be available in pore water. If the reverse is true, all of the metal may be present as a solid metal sulfide.
Once estimates have been made of the pore water concentrations of nonionic organic compounds and divalent metals, it may then be possible to estimate the relative toxicities of different sediment samples. The relative toxicity of each analyte in a sediment sample can be defined as the ratio (expressed in "toxic units") of the estimated equilibrium pore water concentration of the analyte to the ambient water quality criterion (AWQC) for aquatic life (USEPA 1986a) for that analyte:
|Toxic Unit =||Pore Water Concentration|
Toxic units for those chemicals with AWQCs are easily estimated. For those chemicals without AWQCs, it is necessary to use relative comparisons of toxicity (e.g., toxic equivalency factors [TEFs]; Safe 1990) to those chemicals with AWQCs.
Once toxic units have been estimated for each analyte at each site, the toxic units for each site are totaled over all analytes. Kreis' (1989) ranking process, as described above, is then used to rank the sites based on their total toxic units. The result is a relative toxicity ranking for the group of sites under investigation based on total estimated potential toxicity at each site.
The data from laboratory sediment toxicity tests and multiple endpoints measured within some tests must be put on a common risk scale before they can be combined (arithmetic mean). To accomplish this, the different measured responses associated with each of the tests is divided by the response observed for the control or reference sediment. Adjusting each response for the control response not only puts each measure on the same scale (i.e., proportion of the control), but it also adjusts each measured response for the laboratory conditions at the time of the test. Adjusting for the conditions at the time of the test is necessary to account for variations in test methods resulting from tests being run at different times, in different locations, by different investigators, or combinations of these factors. Analysis of control sediments alone does not take into account differences in toxicity test responses that may be attributable to differences in physical (e.g., sediment grain size) or other factors (e.g., ammonia, TOC content) between sediments. Analysis of reference sediment samples appropriately matched with the test sediment samples is necessary to take such factors into account.
Calculation of the proportional laboratory toxicity response for each measured value is as follows:
|Proportional Laboratory Toxicity Response =||Endpoint Value for Test Sediment|
|Endpoint Value for Control or Reference Sediment|
The estimates of risk for each toxicity test measure are then averaged over all measured endpoints at each site to estimate the average (arithmetic mean) risk at a site based on laboratory toxicity. Again, it is this average estimated risk that is ranked among sites. For this approach to be effective, the endpoints measured at each site should be similar, if not identical.
As for laboratory toxicity tests, the different measures of benthic community structure should ideally be put on one scale for evaluation. For the ARCS Program, the different measurements of benthic community structure were the percentages of the benthic community (i.e., as percentages of the total number of organisms) represented by each invertebrate order observed among all the sites. The use of the full list of observed orders is appropriate as long as the potential list of orders for the set of sites under consideration is similar among sites.
Since all values of these benthic community structure variables are percentages, they are already on the same general scale. To put the observed percentages for each invertebrate order on a relative risk scale, it is desirable to adjust their abundances by their relative tolerance to contamination. There is an implicit assumption in this approach that differences in the abundances of the various invertebrate orders are attributable to differences in sediment contamination, and not to differences in physical or other factors between sites. Future refinements to this ranking method may need to take this fact into account. Adjusting each order's percentage of the benthic community by its tolerance to contamination ensures that the presence of less tolerant orders (i.e., that may therefore be present in relatively low abundances) still influences a site's ranking.
Several different indices of tolerance to contamination have been proposed. Hilsenhoff (1987) proposed a biotic index for aquatic invertebrates in Wisconsin streams that was related to their tolerance of organic enrichment. Lenat (1993) proposed a biotic index for aquatic invertebrates in North Carolina streams that was related to their tolerance of chemical contamination. Although Lenat (1993) cautioned against using his biotic index outside its intended geographic range (i.e., southeastern United States stream environments), this index is currently the only index of tolerance to chemical contamination available. Until an index such as Lenat's (1993) index is developed for the Great Lakes, it remains the best measure of chemical contamination tolerance for benthic organisms. The use of such a biotic index to adjust the abundances of the various invertebrate orders would be as follows:
|Tolerance-adjusted Benthic Community Response =||% Benthic Community for an Order|
|Contamination Tolerance of that Order|
The index of contamination tolerance is structured such that the more tolerant orders receive a higher value, thereby adjusting their abundances downward in relationship to those of less tolerant orders. The estimates of risk generated by the tolerance-adjusted benthic community response for each invertebrate order would then be averaged (arithmetic mean) over all orders for each site. This average contamination tolerance-adjusted benthic community response represents an estimate of the relative risk at each site based on the assumed toxicity of the sediments to the benthic community. Again, it is this average estimated risk that would then be ranked among all sites under consideration.
The rankings that result from the different types of information discussed (i.e., sediment chemistry, laboratory toxicity tests, and benthic community structure) can be combined to produce an overall ranking for each site. At this point, each type of information is on a scale from 1 to 100. The estimate of relative risk for the sites under investigation, based on all three types of information, is just the average (arithmetic mean) of the three ranks. A simple average of these three ranks implicitly assumes that a range in values of 1 to 100 has the same meaning for each variable. This may represent a gross simplification, but still should allow a comparison of overall risk among sites.
The only requirement necessary before the three different rankings (i.e., sediment chemistry, laboratory toxicity tests, and benthic community structure) can be combined is that all three ranks must order the sites in the same manner (i.e., 1 = least toxic and 100 = most toxic). The chemistry rank already ranks the sites in the appropriate manner, but the laboratory toxicity tests and benthic community structure ranks must be reversed. To reverse any of the ranks, the following equation is used:
Site Rank = 1 + (1 - Rank/100) x 99
The purpose of the described ranking process is to allow different types of data, measured on different scales, to be combined into one overall estimate of relative risk for the set of contaminated sites under investigation. The scaling done to each class of data (i.e., sediment chemistry, laboratory toxicity tests, benthic community) allows for the incorporation into the estimates of relative risk as much information as is currently available in the scientific literature. The result is the current best estimate of relative risk associated with sediment contamination for the sites under investigation. This approach enables the comparison and combination of sediment contamination information, measured on different scales, on one relative scale that has a foundation in environmental chemistry, toxicology, and ecology.
The ranking process is dynamic; as more information becomes available about sediment processes, chemical fates, toxicity, etc., new information can be incorporated into the ranking model. Thus, the estimates of relative risk become more robust as the base of knowledge increases.
One very important set of assumptions associated with this process is that each measure within each class of data is considered independent of all the other measures in its class (i.e., effects are strictly additive). This is not necessarily the case for all the measures (e.g., lead and zinc; Schmitt et al. 1993). As information becomes available that contradicts these assumptions, the interactions that are present can also be incorporated into the ranking method.
Finally, the process does not have to be limited to the types of data described above. Other, less scientifically based classes of data (e.g., aesthetics, recreational potential) could potentially be incorporated into the ranking method. The ranking method could potentially also be extended to other quantifiable risks, such as carcinogenicity.
This chapter provides an overview of potentially applicable data interpretation techniques that may be useful for individual sediment assessment programs. Included are examples of techniques for mapping sediment quality data, classifying sediments as "contaminated" or "uncontaminated," and ranking of sites for consideration for remediation. It is not possible to recommend specific data interpretation techniques for each and every sediment assessment program. The data interpretation techniques selected for a given sediment assessment program will be a function of the program under which the assessment is being conducted, as well as of the types of data collected and the specifics of the AOC under consideration.
It is important that the reader understand that efforts to interpret the sediment quality data collected under the ARCS Program are continuing, and that more is yet to be learned about the most appropriate ways of analyzing these data. It was not the intent of the ARCS Program to select specific sediment classification methods for application in the Great Lakes AOCs, but the ARCS Program has considered how these methods could be applied. Similarly, it was not the intent of the ARCS Program to select specific methods for the ranking of contaminated sediment sites, but the ARCS Program is continuing to explore application of such methods in an effort to show how they might be applied in other programs.