Marine Pollution Bulletin 51 (2005) 649657 www.elsevier.

com/locate/marpolbul

Assessing cause and eect of multiple stressors on marine systems

S. Marshall Adams
*
Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA

Abstract An operational framework is developed to serve as a guideline for investigating causal relationships between environmental stressors and eects on marine biota. Because of the complexity and variability of many marine systems, multiple lines of evidence are needed to understand relationships between stressors and eects on marine resources. Within this framework, a weight of evidence approach based on multiple lines of evidence are developed and applied in a sequential manner by (1) characterizing the study system which involves determining if target biota are impaired, assessment of food and habitat availability, and measuring contaminant levels in the environment, (2) assessing direct eects of contaminant exposure on target biota using biomarkers and assessing indirect eects of exposure using suites of bioindicators, and (3) applying standard causal criteria based on epidemiological principles and diagnostic health proling techniques to assess potential causes. Use of multiple lines of evidence should also reduce the risk of false positives (Type I error or falsely concluding that there is a causal relationship when there is none) and false negatives (Type II error or falsely concluding there is not a causal relationship when there actually is). Understanding causal relationships and the mechanistic processes between environmental stressors and eects on biota is important in the eective management and restoration of impaired marine ecosystems. 2004 Elsevier Ltd. All rights reserved.
Keywords: Causal relationships; Multiple stressors; Indirect eects; Bioindicators; Biomarkers; Environmental diagnosis

1. Introduction Marine ecosystems, and especially near-shore coastal areas such as estuaries, are typically subjected to a variety of stressors, both natural and anthropogenic, which can impair the health and tness of resident biota. Multiple stressors including pollutants, nutrients, hypoxia, turbidity, suspended sediments, and altered habitat and hydrologic regimes can impact resources through single, cumulative, or synergistic processes. Responses of biota to environmental stressors are the integrated result of both direct and indirect processes which can be ultimately manifested as changes in abundance, diversity, and tness of individuals, populations, and communities. Establishing causal relationships between

Tel.: +1 865 574 7316; fax: +1 865 576 8543. E-mail address: adamssm@ornl.gov

stressors and eects on marine resources is dicult because of the physicochemical and biological complexity of these systems, the variety of biotic and abiotic factors that can modify responses of biota to stressors (McCarty and Munkittrick, 1996; Wolfe, 1996), compensatory mechanisms that operate in populations (Power, 1997), the orders of magnitude involved in extrapolation over temporal and spatial scales (Holdway, 1996), and the many pathways by which stressors can disrupt the normal functioning of ecosystems. For example, in natural systems, not only can contaminants aect biological components directly, but indirect eects can result from other inuential factors such as habitat and food availability, predatorprey interactions, and competition (Adams et al., 1998; Preston, 2002; Fleeger et al., 2003; Chapman, 2004). In addition, time lags between the initial cause and expression of cumulative eects can be long (Vallentyne, 1999), and interdependence among disturbance events, ecosystem properties,

0025-326X/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.marpolbul.2004.11.040

650

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

and biological invasions often make causal relationships dicult to discern (Bart and Hartman, 2000). Establishing denitive causal relationships between stressors and observed eects in natural systems is also dicult not only due to the many intrinsic environmental factors which can complicate this process, but because there are no widely accepted and proven approaches for establishing such relationships. The ability to establish causality between factors in the environment and biological eects is particularly important in environmental management and regulation. Proof of causality could reduce the uncertainty of management decisions resulting in less costly environmental policies and streamlining of regulatory and compliance procedures. The complexity of marine systems, their inherent high variability, and the inuence of multiple environmental factors or stressors suggests that no single measure (or perhaps even a few measures) is adequate for assessing the eects of multiple stressors on biota and for establishing the mechanistic basis of these eects. An appropriate suite of endpoints that includes multiple lines of evidence within a weight of evidence approach is required for determining the biological signicance of stress and for understanding the underlying cause or mechanistic basis of observed eects (Adams, 2002, 2003; Attrill and Depledge, 1997; Hodson, 2002; Galloway et al., 2004). In many instances, simply documenting that a change has occurred in a system or measuring such a change with a few response parameters may not be adequate to assess causality. In such situations, it is also necessary to understand the mechanistic basis of a change or eect if more informed decisions are to be made regarding eective management and mitigation practices in impacted coastal systems. Over-reliance on any one, or even a few, indicators for assessing eects and their underlying cause(s) can result in environmental regulation that is less accurate and either under or over-protective of water resources (Yoder and Rankin, 1998). Some studies have utilized chemical criteria exclusively for assessing eects of stressors and in helping to identify the possible causes of observed eects on biota. Even though chemical and physical measures are important components of water quality assessment programs (Chapman, 1995a,b; Galloway et al., 2004), relying on chemical criteria alone for assessing the eects of stressors and understanding the mechanistic basis of these eects on surface water resources may provide an inadequate assessment of the biological and ecological condition of aquatic systems. For example, of the 645 stream and river segments analyzed in Ohio (USA) to determine water quality eects, impairment as revealed by use of biological indicators was evident in 50% of these sites whereas no impairments were observed based on using only chemical indicators (Yoder and Rankin, 1998). Use of chemical criteria alone to assess the eects of water quality on ecological systems can

be an incomplete foundation for legislation related to resource policy because they do not include broader ecological measures (Yoder and Rankin, 1998; Barbour et al., 1996). More importantly, however, chemical criteria alone fail to reect all the other factors in the environment that can impair aquatic ecosystems, such as sedimentation, alterations in habitat and natural ow regimes, varying temperature and oxygen regimes, and changes in ecological factors such as food availability and predatorprey interactions. Conversely, biological criteria (biocriteria) possess several attributes that are desirable for assessing the eects of environmental stressors and in helping to understand the mechanistic basis of these eects on ecosystems. Some types of biocriteria are not only reective of chemical exposure (i.e., biomarkers), but have the capacity to integrate many of the physical, chemical, and biological stressors that operate in ecosystems. In addition, many biocriteria are capable of integrating the eects of stressors on organisms both spatially and temporally (Barbour et al., 2000), and are thus more suited for measuring and interpreting the possible eects of multiple stressors on aquatic ecosystems. Biocriteria, therefore, can reect environmental problems that might otherwise be missed, or underestimated, by approaches that rely on chemical criteria alone simply because they integrate and account for natural ecological conditions and variability. Biocriteria can also be used to assess damage or injury to natural resources from environmental stressors. In marine systems, pollution has been dened as the environmental damage caused by wastes discharged into the sea (Clark et al., 2001). This denition inherently implies that environmental damage has to be demonstrated in order to prove that a site is polluted. Within this context, chemical criteria and biomarkers of chemical exposure should not be used alone to demonstrate environmental damage. For example, measuring levels of contaminants in the environment is basically documenting the level of contamination, while biomarker responses, even though they may provide some indication of damage at the cellular level and demonstrate exposure, do not provide assessments of environmental damage at ecologically-relevant levels of biological organization. Biologically relevant endpoints, which are a component of the risk characterization phase of the ecological risk assessment (ERA) process, are typically used as the basis of environmental regulatory and management decisions (EPA, 1998). Assessing biological damage or injury requires the use of methods that demonstrate measurable biological responses. Multiple exposure (biomarkers) and eects (bioindicators) measures applied together, however, can be useful for demonstrating that biota have been exposed to pollutants and for helping to identify the mechanistic basis of effects at higher levels of biological organization (Adams, 2002; Galloway et al., 2004).

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

651

Given this background, then, the objectives of this study are to develop and provide an operational framework for (1) assessing the eects of multiple stressors on marine systems, and (2) helping to identify the possible causes of these eects on marine resources using multiple lines of evidence. Understanding causal relationships and the mechanistic processes linking environmental stressors and eects on biota is important in the eective management and restoration of marine ecosystems.

2. Framework for assessing causality The operational framework shown in Fig. 1 is a suggested guideline for helping to establish causal relationships, if in fact they exist, between environmental stressors and eects on marine biota or resources. Even though less rigorous and denitive approaches using single lines of evidence could possibly be taken to investigate possible causal relationships, presented here is an approach that utilizes multiple lines of evidence to assess eects and to help identify the possible causes of these eects. The use of multiple lines of evidence for establishing causality is important in marine systems because of their complexity, their high spatial and temporal variability, the synergistic and cumulative processes operating in these systems, and the dierent pathways that multiple stressors can act on marine resources.

This assessment framework consists of three integrated components which are related and are sequentially dependent on each other (Fig. 1). The three components are (1) characterization of the study system which includes determining if the target biota or resources of interest are impaired or damaged, assessing the basic status of the food supply and habitat associated with the target biota/resource, and measuring contaminant levels in the sediment, water, and/or biota; (2) evaluation of the eects of stressors such as pollutants on biota utilizing measures of both direct indicators of exposure (biomarkers) and indirect indicators of eects (bioindicators), and (3) assessment of the potential cause of the observed eects by invoking the use of the epidemiological approach and associated standard causal criteria in conjunction with application of diagnostic exposure-eects proles for the target biota. These three components utilized together serve as multiple lines of evidence for establishing possible relationships between stressors and eects on marine systems. 2.1. System characterization A rst step in the process to establish causality is to conduct a baseline or screening study to determine if the resources of interest have been damaged or impaired due to environmental quality. Because it would be impractical to study all components of a marine system, selected or target species could be used to serve as sentinel indicators of the overall health or integrity of the system of interest (Depledge and Fossi, 1994; Sandstrom, 1994). The sentinel species or resources chosen for study should be those that occupy important trophic positions or niches, are keystone species, or contribute signicantly to the biomass or energy ow of a system. If the health or integrity of a resource appears to have been compromised, then further characterization of the affected system should be performed including evaluation of the status of the habitat and food supply for the key biota. This type of information is important in the assessment process to evaluate the pathways (direct and indirect) possibly responsible for observed eects. Identication of the pathways of eects is also critical in helping to identify possible causes of any observed effects. Another important component of the characterization process is to determine the levels of contaminants in the major compartments of the system including water, sediment, and biota. This information is also used to help evaluate the possible cause(s) of any observed eects. 2.2. Assessing eects of stressors Following characterization of the aected system, the next step in this process is to determine, using biomarkers of exposure and bioindicators of eects, if the biota

System Characterization
- health of target biota - status of food supply - status of habitat - contaminant levels

Evaluate effects

Direct

Indirect

Biomarkers of exposure

Bioindicators of effects

Assess potential cause

Apply causal criteria

Diagnostic health profiling

Investigate cause using weight-of-evidence

Fig. 1. Operational framework for establishing causal relationships between environmental stressors and eects on marine resources. Multiple lines of evidence are developed in a sequential manner to establish possible causal relationships between stressors and eects.

652

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

or resource of interest has been impaired by environmental stressors. Responses of organisms to stressors are the integrated result of both direct and indirect pathways. Direct pathways operate primarily through metabolic processes, such as toxic action of chemicals at receptor sites, that are initiated at the lower levels of biological organization and are propagated upward through increasing higher levels of organization (Fig. 2). Indirect pathways however, operate mainly through eects on the food chain, on habitat availability, or through behavioral modication of organisms (Fig. 2). The eects of stressors acting through direct mechanisms, such as toxic action of chemicals, occur initially at the molecular or subcellular level and can be expressed, for example, as changes in biomolecular, biochemical, and physiological components or processes such as protein and gene expression, DNA integrity, enzyme activity, and metabolism and respiration, respectively. Eects initiated at these lower levels can be propagated upward through increasing levels of biological complexity causing changes in metabolic processes such as lipid dynamics, immunocompetence, and hormonal regulation (Larsson et al., 1985). Ultimately, these eects may be manifested as changes at the organism, population, and community levels. Environmental stressors may also impact organisms indirectly through the food chain and through bioenergetic processes by inuencing the quality (energy and protein content) and quantity (biomass) of energy available to consumers. Indirect eects of stressors can also aect interspecic interactions through competition and predation which may cause trophic cascading eects in aquatic systems (Fleeger et al., 2003; Preston, 2002; Campbell et al.,

Environmental Stress
Direct Indirect

Biomolecular Biochemical

Food & habitat availability

Bioenergetic

Inter & intraspecific interactions

Behavior Competition Predation

Physiological Pathological

Organism
(growth, reproduction)
Biochemical Metabolic

Population

Community

Fig. 2. Relationships between environmental stressors and direct and indirect eects on biological systems. Direct pathways aect organisms primarily through biochemical and metabolic processes and indirect pathways inuence biota through eects on food and habitat availability and through intra- and interspecic interactions.

2003). In addition, stressors can indirectly impair the health of marine biota by aecting the quality and quantity of the habitat, resulting in altered behavior related to reproduction, feeding, or habitat selection (Reynolds and Casterlin, 1980; Little, 2002). The more ecologically-relevant parameters of aquatic systems, such as growth, reproduction, and population-level attributes, can therefore be inuenced by both direct and indirect pathways which involve the integrated eects of metabolic impairment, energy availability, behavioral, and interspecic responses to environmental stressors (Adams, 2002). Biomarkers of exposure and bioindicators of eects can be used together to assess the eects of environmental stressors on biota and to help evaluate the relative inuence of direct and indirect pathways on the health of marine systems. Biomarkers are generally dened as functional measures of exposure to stressors which are expressed at the suborganism level of biological organization (Huggett et al., 1992; Benson and DiGiulio, 1992). Bioindicators, however, are dened less precisely than biomarkers and can be viewed as either structural entities, such as sentinel species or communities (Van Gestel and Van Brummelen, 1996), or they can be considered, as they are in this study, as biological eects at higher levels of organization (Adams, 1990; Engel and Vaughan, 1996; Adams, 2002). One approach for helping to assess the relative inuence of direct and indirect pathways on biota is to measure a selected suite of biomarkers and bioindicators representing dierent types or categories of response variables (Table 1). Examples of the types and categories of indicators that could be measured are (1) direct indicators of exposure to stressors including biomarkers of exposure, (2) direct indicators of eects that include bioindicators of metabolic and bioenergetic impairment and dysfunction, and (3) indirect indicators of eects including nutrition and feeding indices, growth, reproduction, behavior, and lipid pool measurements (Table 1). Identication of the primary pathway responsible for any particular observed eect can therefore be assessed based on the relative proportion of responses that are measured (compared to an observed change from a reference or standard condition) in each of these three response categories. For example, if contaminants are aecting organisms primarily through indirect pathways such as the food chain, we would expect to see a preponderance of corresponding bioenergetic-related responses in the target biota such as alterations in nutrition and feeding indices, growth, and lipid pools. On the other hand, if eects are being manifested primarily through direct metabolic stress pathways, we might expect to see a higher proportion of these eects expressed as metabolic responses such as changes in molecular structure and function, organ dysfunction enzyme levels, protein and gene expression, osmogulatory ability, and

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

653

Table 1 Major categories of response indicators to environmental stressors representing direct indicators of exposure (biomarkers), direct indicators of eects (bioindicators), and indirect indicators of exposure/eects which can be used for helping to identify causes of eects due to environmental stressors in marine systems Direct indicators of exposure (biomarkers) Detoxication enzymes DNA damage Antioxidant enzymes Metallothioneins Stress proteins Protein and gene expression Direct indicators of eects (bioindicators) Lipid metabolism Organ dysfunction enzymes Immunocompetence Selected histopathologies Steroid hormones Metabolism/respiration Indirect indicators of exposure/eects Nutrition/feeding Lipid pools Growth Reproduction Behavior Bioenergetic

other biochemical responses.

and

physiologically-mediated

2.3. Assessment of potential cause(s) Assessment of potential causes for observed eects can also be addressed by applying standard causal criteria based on epidemiological principles in conjunction with application of health diagnostic proles of the target biota. Using these two methods together as multiple lines or weight of evidence strengthens the condence we have in claiming that causal relationships exist (or may not exist) between stressors and eects in aquatic ecosystems. Several studies have advocated the weight of evidence approach for helping to establish possible causal relationships between environmental stressors and effects including (1) a series of related studies using epidemiological causal criteria (Collier and Adams, 2003), (2) investigations which address causality in sediment quality assessments (Borgmann et al., 2001; Chapman et al., 2002; Forbes and Calow, 2004), and (3) assessments which describe application of the weight of evidence approach in ecological risk analysis (Suter et al., 2002; Forbes and Calow, 2002). 2.3.1. Application of causal criteria Because of various limitations typically associated with other methods, the epidemiological approach is a particularly attractive technique for establishing causality (Mac and Edsall, 1991). This relatively new forensic approach (forensic toxicology) has been used to integrate a wide variety of environmental information which has successfully led to denitive statements that describe the injury to populations of sh, wildlife, and humans in relation to specic stressors such as toxicants (Gilbertson, 1997). This approach provides a logical framework to draw together disparate threads of evidence and integrate them into a coherent whole so that defensible regulatory decisions can be made (Fox, 1991). Using the principle of ecoepidemiology, a causal relationship is assumed to exist whenever evidence indicates that environmental factors increase the probability of the occurrence of a disease and, conversely, a reduction in one or more

of these factors decreases the probability of occurrence of that disease (Fox, 1991). The ecoepidemiologial approach does not necessarily provide the kind of causal proof required by experimental designs or by rigorous hypothesis testing, but it does form a process upon which a logical and balanced judgment for causality can be built. By applying this approach, the available data either support or detract from a causal judgment which is usually subjective rather than one based on rigorous quantitative data and hypothesis testing. Epidemiological criteria have been used to infer causal relationships between exposure to environmental stressors and injury to biological systems. The ecoepidemiological approach has as its basis several assembly rules or criteria for providing and presenting causal arguments. These causal criteria are logical categories of evidence that are used to organize information concerning each hypothesized cause and which are consistently applied to support or refute a specic cause. Kochs postulates (Woodman and Cowling, 1987) or causal criteria have been used in the evaluation of causal relationships involving infectious agents in human populations, and these postulates have served as a model upon which modern concepts of causality are based (Fox, 1991). To date there have been less than a dozen assembly rules or criteria which have been used to establish causality based on the weight-of-evidence approach. The principle causal criteria which represent a complication of Kochs postulates and those proposed or modied by Hill (1965), Suter (1993), USEPA (1998, 2000) are (1) Strength of AssociationCause and eect coincide. A large portion of individuals are aected in the exposed (stressed area) compared to reference areas or the magnitude of an eect is high in a few individuals; (2) Consistency of Associationthe association between a particular stressor or stressors and an eect has been observed by other investigators in similar studies at other times and places; (3) Specicity of Associationthe eect is diagnostic of exposure. Could the eect be due to a dierent

654

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

(4)

(5)

(6)

(7)

cause? Could the proposed cause produce another type of eect? Time order or temporalitythe cause precedes the eect in time and also the eect decreases when the cause is decreased or removed; Biological Gradientthere is a doseresponse relationship either spatially or temporally within the system. The risk of an eect is a function of magnitude of exposure; Experimental Evidencevalid experimental studies (such as laboratory, microcosms, and mesocosms) support the proposed causeeect relationship; Biological Plausibilitythere is a credible or reasonable biological and/or toxicological basis for the proposed mechanism linking the proposed cause and eect.

2.3.2. Diagnostic health proling In addition to applying standard causal criteria based on epidemiological principles, potential causal relationships between stressors and eects can also be investigated using diagnostic health proling techniques (Adams, 2001). This technique, used together with the causal criteria and information from the baseline characterization study, should provide multiple lines of evidence for the establishment of cause and eect, if, in fact, such a relationship exist. Conceptually, construction of diagnostic health proles for wildlife species involves use of the health diagnostic principles used in the medical profession to help determine the cause of disease and pathologies in human patients. As a background for the conceptual development of environmental health proling, marine systems can be impacted by a variety of anthropogenic activities including point source discharges of domestic and industrial pollutants; atmospheric deposition; agricultural practices; land-use activities including urban development and runo; modication of water ow regimes; dredging practices, and physical habitat alteration. Many of these activities are associated with specic stressors such as contaminants, nutrients, or sediments which characterize the operation of that particular activity. These groups of stressors, which are unique or characteristic for each type of activity, can be used to help separate or identify the eects of these activities from each other. For example, point-source discharges from paper mills are typically characterized by chlorophenolic and resin acid compounds, dioxin-type contaminants, and high nutrient loading. In contrast, non-point source agricultural activities can contribute pesticides, nutrients, and sediment to receiving coastal systems which typically result in specic or predictable types of biological responses. Because certain types of stressors cause predictable types of responses in biological systems, environmental

proling or diagnosis methods can be used to help partition or separate out the eects of natural vs anthropogenic stressors on biota and thus help establish causal relationships. The use of multi-response bioindicators for environmental diagnosis or proling is conceptually similar to approaches used by the medical profession to diagnose the health of human patients. In human subjects, a battery of medical tests and procedures are performed such as chemical proling of blood and urine, and the results compared to standardized norms for diagnosis of pathology and disease. In humans, diagnosis of health is relatively straightforward because the individual is the ultimate endpoint of interest. In considering ecosystem health, however, the endpoints of interest are typically populations, communities, and ecosystems and diagnosing the causes of eects observed at these higher levels of biological organization is particularly problematic because of the interacting eects of biotic and abiotic factors, high temporal and spatial variability, and compensatory mechanisms that operate in ecological systems. To demonstrate use of the diagnostic proling approach in the identication of causal relationships in marine systems, exposure-eects response proles were constructed for dierent exposure-eect relationships corresponding to each major type or source of anthropogenic stress shown in Fig. 3. A combination of biomarkers of exposure and bioindicators of eects were used to construct exposure-eects proles for organisms exposed to these dierent sources of stress. The principle types of stressors associated with each activity shown in Fig. 3 were rst identied and then matched with the types of biomarker responses typically characteristic of that specic activity. Since certain stressors, and in particular various types of contaminants, are associated

Biomarkers of Exposure
MFO Enzymes Bile Metabolites Metallothionein DNA Damage Antioxidant Enzymes
Gene expression Stress proteins Glycolytic enzymes

Acetylcholinesterase

Exposure-Effects Profiles
Reproduction

xx

Bioindicators of Effects

Bioenergetics

increased

Heavy metals
x x

Domestic Sewage

Growth

x x

Agriculture
x x

l ica em ch o tr Pe

x s nt ie tr x-nu ia ox yp H

decreased

Histopath Gill

x x

x x

Liver

Fig. 3. Exposure-eects proles for diagnosing and identifying sources of stress to marine biota exposed to multiple stressors.

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

655

with specic responses at the biomolecular, biochemical, or physiological levels, this analysis matched each major type of stressor or activity to their corresponding eect. Once the exposure responses (biomarkers) associated with each anthropogenic activity were identied, a stress exposure-biological eects prole was generated for each of these activities by plotting exposure biomarkers on one axis (X-axis) and bioindicator responses or effects on the other axis (Y-axis) (Fig. 3). The principle higher-level responses (bioindicators) typically associated with each major type of anthropogenic activity were identied from literature reviews. Cross marks placed within the exposure-eects prole (ellipse) for each activity indicate those specic bioindicators of effects on the Y-axis that are associated with the various biomarkers of exposure on the X-axis. For example, the principle biomarkers of exposure for petrochemical activities are induction of P450 detoxication enzymes and production of aromatic biliary metabolites while biomarkers that appear to be associated with hypoxia stress are various types of glycolytic enzymes and specic gene and protein expressions such as the HSP 70 proteins and hemocyanin. Some of the major bioindicator responses related to petrochemical activities are reported to be various gill and liver pathologies, impaired reproductive function, and decreased growth, while the more typical bioindicator responses associated with hypoxia stress are impaired reproduction and bionergetic function along with decreased growth. Thus specic combinations of biomarkers and bioindicators, utilized together, can provide a relative unique response prole that is diagnostic of a particular activity or type of stressor providing, therefore, useful information relative to identifying the possible causes of eects in aquatic systems.

3. Establishing causality using multiple lines of evidence Because many marine systems, and particular nearshore coastal zones, are relatively complex, spatially and temporally variable, and subjected to multiple stressors, multiple lines of evidence could be useful for assessing eects and in helping to establish causal relationships between stressors and eects on target resources. These lines of evidence are based and developed on information provided from each of the three major components of the operational framework shown in Fig. 1. In the characterization component of this framework, knowledge of the health of the target biota, food and habitat availability, and levels of contaminants in the environment provide initial clues regarding the potential causes of observed eects. If, in fact, the target biota, whether individuals or populations are impaired, the relationship of food availability, habitat status, and contaminant exposure can suggest whether organisms

are being aected primarily through direct or indirect exposure pathways. For example, if the health of a resource is determined to be impaired to some degree, and both food and habitat availability are found to be good but contaminant levels are relatively high, this particular combination of situations would suggest that eects are being mediated primarily through direct effects of pollutants (Fig. 2). Conversely, if the immediate environment of contaminant-exposed organisms are characterized by good habitat and low food availability, then exposure eects mediated through indirect pathways such as the food supply may be implicated (Fig. 2). Furthermore, a combination of poor habitat and low food supply could suggest that the principle pathway of contaminant eects was not through direct mechanisms. The initial characterization studies should provide some clues or indications relative to (1) if biota are being aected by stressors, and (2) if contaminants are responsible for these eects, based on evidence of injury through direct or indirect pathways. If contaminants appear to be implicated then both biomarkers of contaminant exposure and bioindicators of eects can be measured in the target organisms. A biomarker response suggests that an organism has been exposed to a contaminant, and in some cases, if the proper biomarkers are being measured, such responses will also be diagnostic of specic classes or groups of contaminants such as chlorinated hydrocarbons, PAHs, pesticides, metals, lead, etc. Even though most bioindicators of eects are integrative in nature and are not specic or diagnostic of specic stressors, certain groups of bioindicators, taken together, can be used to evaluate whether stressor eects are acting through direct or indirect pathways (Table 1). Information provided from the characterization component of the assessment framework such as the relative importance of direct vs indirect pathway exposure can be compared to the studies which measure biomarkers of exposure and bioindicators of eects to determine if they are consistent in their results. For example, if the initial characterization studies indicate that indirect pathways are implicated because food availability is low and contaminant levels are high in the environment (Fig. 2), then many of the bioindicators (i.e., bioenergetic-related responses) measured in step 2 (Table 1) should also reect that organisms are being aected through indirect pathways (Fig. 2). Thus, in this example, two separate lines of evidence should demonstrate that eects of stressors are mediated primarily through indirect mechanisms. Both the standard epidemiological criteria and the diagnostic health proling technique also provide independent lines of evidence for helping to establish causality. If there is a relatively strong association or relationship between stressors and responses in biota,

656

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657

then both of these methods should provide convincing evidence of these associations. However if causal relationships are relatively weak, then causality based on both the more subjective epidemiological criteria and the exposure-eects proles based on empirical measurement of biomarkers and bioindicators should also provide evidence of a rather weak causal association. Combined with the information provided from components 1 and 2 (system characterization and assessment of eects, respectively) and using a combination of the epidemiological criteria and diagnostic proling, use of these multiple lines of evidence increases the probability of detecting such causal relationships if they exist. Multiple lines of evidence should also reduce the risk of false positives (Type I error or falsely concluding that there is a causal relationship when there is none) and false negatives (Type II error or falsely concluding there is not a causal relationship when there actually is). Use of a single line of evidence, however, may not be adequate to reduce the probability of these types of errors. For example, Beliae and Burgeot (2002) reported that using a suite of biomarkers instead of a single biomarker in their marine studies could avoid false negatives. As environmental scientists we certainly want to minimize the probability of false negatives in protecting and managing marine systems.

References
Adams, S.M., 1990. Status and use of bioindicators for evaluating eects of chronic stress on sh. American Fisheries Society Symposium 8, 18. Adams, S.M., 2001. Biomarker/bioindicator response proles of organisms can help dierentiate between sources of anthropogenic stressors in aquatic ecosystems. Biomarkers 6, 3344. Adams, S.M., 2002. Biological indicators of aquatic ecosystem stress: introduction and overview. In: Adams, S.M. (Ed.), Biological Indicators of Aquatic Ecosystem Stress. American Fisheries Society, Bethesda, MD, pp. 111. Adams, S.M., 2003. Establishing causality between environmental stressors and eects on aquatic ecosystems. Human and Ecological Risk Assessment 9, 1735. Adams, S.M., Ham, K.D., LeHew, R.F., 1998. A framework for evaluating organism responses to multiple stressors: mechanisms of eect and importance of modifying ecological factors. In: Cech, J.J., Wilson, B.W., Crosby, D.G. (Eds.), Multiple Stresses in Ecosystems. Lewis Publishers, Boca Raton, FL, pp. 1322. Attrill, M.J., Depledge, M.H., 1997. Community and population indicators of ecosystem health: targeting links between levels of biological organization. Aquatic Toxicology 38, 183197. Barbour, M.T., Gerritsen, J., Grith, G.E., Frydenborg, R., McCarron, E., White, J.S., Bastian, M.L., 1996. A framework for biological criteria for Florida streams using benthic macroinvertebrates. Journal North American Benthological Society 15, 185 211. Barbour, M.T., Swietlik, W.F., Jackson, S.K., Courtemanch, D.L., Davies, S.P., Yoder, C.O., 2000. Measuring the attainment of biological integrity in the USA; a critical element of ecological integrity. Hydrobiologia 422423, 453464. Bart, D., Hartman, J.M., 2000. Environmental determinants of Phragmites australis expansion in a New Jersey salt marsh: an experimental approach. Oikos 89, 5969. Beliae, B., Burgeot, T., 2002. Integrated biomarkers response: a useful tool for ecological risk assessment. Environmental Toxicology and Chemistry 21, 13161322. Benson, W.H., DiGiulio, R.T., 1992. Biomarkers in hazard assessment of contaminated sediments. In: Burton, G.A. (Ed.), Sediment Toxicity Assessment. Lewis Publishers, Boca Raton, FL. Borgmann, U., Norwood, W.P., Reynoldson, T.B., Rosa, F., 2001. Identifying cause in sediment assessments: bioavailability and the sediment quality triad. Canadian Journal Fisheries Aquatic Science 58, 950960. Campbell, P.F.C., Hontela, A., Rasmussen, J.B., Giguere, A., Gravel, A., 2003. Dierentiating between direct (physiological) and foodchain mediated (bionergetic) eects on sh in metal-impacted lakes. Human Ecological Risk Assessment 9, 847866. Chapman, J.C., 1995a. Bioassay testing for Australia as part of water quality assessment programmes. Australian Journal of Ecology 20, 719. Chapman, J.C., 1995b. The role of ecotoxicology testing in assessing water quality. Australian Journal of Ecology 20, 2027. Chapman, P.M., McDonald, B.G., Lawrence, G.S., 2002. Weight-ofevidence issues and frameworks for sediment quality (and other) assessments. Human Ecological Risk Assessment 8, 1489 1515. Chapman, P.M., 2004. Indirect eects of contaminants. Marine Pollution Bulletin 48, 411412. Clark, R.B., Frid, C., Attrill, M.J., 2001. Marine Pollution, fth ed. Oxford University Press, Oxford, p. 237. Collier, T.K., Adams, S.M., (Eds.), 2003. Debate and commentary: establishing causal relationships between environmental stressors and biological eects in eld studies. Human Ecological Risk Assessment, vol. 9, pp. 15266.

4. Synthesis The operational framework shown in Fig. 1 provides a suggested guideline for investigating possible causal relationships between environmental stressors and effects on marine resources. Because of the complexity and variability of many marine systems, multiple lines of evidence used within a weight of evidence framework are needed to understand relationships between stressors and eects on natural resources. Multiple lines of evidence, based on this framework, are developed and applied in a sequential manner by (1) characterization of the study system which includes determining if target biota are impaired and assessment of food and habitat availability along with contaminant levels in the environment, (2) assessment of exposure on target biota using biomarkers and evaluation of eects using bioindicators, (3) application of standard causal criteria based on epidemiological principles and diagnostic health proling techniques to assess potential causes, and (4) use of all the information together in components (1)(3) above to determine, based on the weight or preponderance of the evidence, possible causal relationships between stressors and observed eects. Understanding causal relationships and the mechanistic processes between environmental stressors and eects is important in the eective management and restoration of damaged marine ecosystems.

S.M. Adams / Marine Pollution Bulletin 51 (2005) 649657 Depledge, M.H., Fossi, M.C., 1994. The role of biomarkers in environmental assessment (2), Invertebrates. Ecotoxicology 3, 161 172. Engel, D.W., Vaughan, D.S., 1996. Biomarkers, natural variability, and risk assessment: can they coexist? Human Ecological Risk Assessment 2, 257262. Fleeger, J.W., Carman, K.R., Nisbet, R.M., 2003. Indirect eects of contaminants in aquatic ecosystems. Science of the Total Environment 317, 207233. Forbes, V.A., Calow, P., 2002. Applying weight-of-evidence in retrospective ecological risk assessment when quantitative data are limited. Human Ecological Risk Assessment 8, 16251639. Forbes, V.E., Calow, P., 2004. Systematic approach to weight of evidence in sediment quality assessments: challenges and opportunities. Journal Aquatic Health Management 7, 339350. Fox, G.A., 1991. Practical causal inference for ecoepidemiologists. Journal of Toxicology and Environmental Health 33, 359373. Galloway, T.S., Brown, R.J., Browne, M.A., Dissanayake, A., Lowe, D., Jones, M.B., Depledge, M.H., 2004. A multibiomarkers approach to environmental assessment. Environmental Science and Technology 38, 17231731. Gilbertson, M., 1997. Advances in forensic toxicology for establishing causality between Great Lakes epizootics and specic persistent toxic chemicals. Environmental Toxicology and Chemistry 16, 17711778. Hill, A.B., 1965. The environment and disease; association or causation. Proceedings Royal Society of Medicine 58, 295300. Hodson, P.V., 2002. Biomarkers and bioindicators in monitoring and assessment: the state of the art. In: Adams, S.M. (Ed.), Biological Indicators of Aquatic Ecosystem Stress. American Fisheries Society, Bethesda, MD, pp. 591619. Holdway, D.A., 1996. The role of biomarkers in risk assessment. Human and Ecological Risk Assessment 2, 263267. Huggett, R.J., Kimerle, R.A., Mehrle, P.M., Bergman, H.L. (Eds.), 1992. Biomarkers. Lewis Publishers, Boca Raton, FL, p. 347. Larsson, A., Haux, C., Sjobeck, M.L., 1985. Fish physiology and metal pollution: results and experiences from laboratory and eld studies. Ecotoxicology and Environmental Safety 9, 250 281. Little, E.E., 2002. Behavioral measures of environmental stressors in sh. In: Adams, S.M. (Ed.), Biological Indicators of Aquatic Ecosystem Stress. American Fisheries Society, Bethesda, MD, pp. 431472.

657

Mac, M.J., Edsall, C.C., 1991. Environmental contaminants and the reproductive success of lake trout in the Great Lakes: an epidemiological approach. Journal of Toxicology and Environmental Health 33, 375394. McCarty, L.S., Munkittrick, K.R., 1996. Environmental biomarkers in aquatic toxicology: ction, fantasy, or functional? Human and Ecological Risk Assessment 2, 268274. Power, M., 1997. Assessing the eects of environmental stressors on sh populations. Aquatic Toxicology 39, 151169. Preston, B.L., 2002. Indirect eects in aquatic ecotoxicology: implications for ecological risk assessment. Environmental Management 29, 311323. Reynolds, W.W., Casterlin, M.E., 1980. The role of behavior in biomonitoring of shes: laboratory studies. In: Hocutt, C.H., Stauer, J.R. (Eds.), Biological monitoring of sh. Lexington Books, Lexington, MA, pp. 5982. Sandstrom, O., 1994. Incomplete recovery in a coastal sh community exposed to euent from a modernized Swedish bleached kraft mill. Canadian Journal Fisheries Aquatic Sciences 51, 21952202. Suter, G.W., 1993. Ecological Risk Assessment. Lewis Publication, Boca Raton, FL. Suter, G.W., Norton, S.B., Cormier, S.M., 2002. A methodology for inferring the causes of observed impairments in aquatic ecosystems. Environmental Toxicology Chemistry 21, 11011111. US Environmental Protection Agency (USEPA). 1998. Guidelines for ecological risk assessment. Oce of Research and Development, Risk assessment forum, Washington, DC, EPA/630/R-95002F. US Environmental Protection Agency (USEPA). 2000. Stressor identication guidance document. Oce of Water, Washington, DC, EPA-822-B-00-025. Vallentyne, J.R., 1999. Extending causality in the Great Lakes basin ecosystem. Aquatic Ecosystem Health Management 2, 229237. Van Gestel, C.A.M., Van Brummelen, T.C., 1996. Incorporation of the biomarker concept in ecotoxicology calls for a redenition of terms. Ecotoxicology 5, 217225. Wolfe, D.A., 1996. Insights on the utility of biomarkers or environmental impact assessment and monitoring. Human and Ecological Risk Assessment 2, 245250. Woodman, J.N., Cowling, E.B., 1987. Airborne chemicals and forest health. Environmental Science and Technology 21, 120126. Yoder, C.O., Rankin, E.T., 1998. The role of biological indicators in a state water quality management process. Environmental Monitoring and Assessment 51, 6188.