1993-95 PRIORITIES AND PROGRESS UNDER THE GREAT LAKES WATER QUALITY AGREEMENT

Chapter Two:Great Lakes Science Advisory Board

2.0 SCIENCE ADVISORY BOARD ACTIVITIES

• Introduction

• 2.2.1 Introduction

  Ecosystem
  Human Health
  Ecosystem Health
  Scientific Models and Root Metaphors
  "Measuring" Ecosystem Health
  Conclusions

• 2.2.2 Human Impacts on the Ecosystem

  Conclusions

• 2.2.3 Environmental Burden of Illness

  Indicators of Environmental Burden of Illness

• 2.2.4 Indicator Selection Criteria
• 2.2.5 Summary
• 2.2.6 Recommendations

• 2.3.1 Introduction

  Specific Objectives

• 2.3.2 Three Methodologies

  Inference as to Causality
  Risk Assessment
  Scientific or Expert Panels

• 2.3.3 Risk Management

  Civic Science
  Precautionary Principle

• 2.3.4 Weight of Evidence
• 2.3.5 Synthesis and Findings
• 2.3.6 Recommendations
• Appendix I

• 2.4.1 Statement of the Problem
• 2.4.2 IJC-Sponsored Wingspread Symposium
• 2.4.3 Overview Statement from Wingspread Symposium, January 13-14, 1995: Hormonal Effects on Neurobehavioural Function: Priorities for Future Research
• 2.4.4 Conclusions and Recommendations

• 2.5.1 Introduction
• 2.5.2 Terms of Reference
• 2.5.3 Summary of Background Studies
• 2.5.4 General Findings
• 2.5.5 Recommendations

• 2.6.1 The Evolution of the Policy Framework for Persistent Toxic Substances
• 2.6.2 Implementing the VETF Strategy -- Toward Sustainable Industry
• 2.6.3 Workshop Sponsored by the Workgroup on Parties Implementation
• 2.6.4 Recommendations

• 2.7.1 International Aspects

  Introduction
  Assessment
  Research
  The Hydrological Cycle
  Conclusions

• 2.7.2 Regional Assessment of Freshwater Ecosystems and Climate Change in North America
• 2.7.3 Binational Approach for the Great Lakes

  Research Objectives and Framework
  Research Opportunities

• 2.7.4 Conclusions and Recommendations

• 2.8.1 Introduction
• 2.8.2 Survey Within the Great Lakes Region

  Conclusions

• 2.8.3 Global Survey of Priorities Beyond the Great Lakes Region

  Introduction
  Findings
  Conclusion

• 2.8.4 Recommendations

1. Environmental Risk Characterization: The Relationship Between Risk Assessment and Risk Management
2. Scientific and technical professionals commonly balance their work among a number of mindsets.

   Transdisciplinary initiatives may iterate and integrate among all types of disciplines and professions, as well as mindsets and methodologies

3. Integration Framework for Binational Project

1. Criteria for the Evaluation of Epidemiological Studies Linking Environmental Toxicant Exposures and Health Effects
2. Ecosystem Health Indicator Selection Criteria Developed by the Council of Great Lakes Research Managers
3. Comparison of Assessment and Decision Approaches
4. 48 Chemicals with Widespread Distribution in the Environment Reported to Have Reproductive and Endocrine-Disrupting Effects
5. Comparison of United States/Canada Wellhead Protection Programs
6. Timeline of Great Lakes-St.lawrence Basin Project Activities
7. Emerging Issues: Survey Results and Current Research Effort Based on 1991/1992 Council of Great Lakes Research Managers Research Inventory
8. Science Advisory Board's Workgroup on Emerging Issues: Summary of International Responses
9. General Characteristics of Major Water Quality Issues on a Global Scale
10. Focal Points of the Canadian Global Change Program (CGCP) in Relation to Key Global Change Issues

2.0 Great Lakes Science Advisory Board Activities

Introduction

The Great Lakes Water Quality Agreement provides the Terms of Reference for the Science Advisory Board (SAB) as a joint institution to be "... the scientific advisor to the Commission and the Water Quality Board." Through an integrative approach, including the natural, physical and social sciences, the principle role of the Board relates to three areas:

• assessment and advice on Great Lakes Basin Ecosystem health, including the scientific underpinning of public policy
• review and evaluation of science policy and programs related to the Parties implementation of the Agreement
• identification and evaluation of emerging issues and future priorities.

To meet these responsibilities, the Board comprises 18 members appointed by the Commission on the basis of their experience and expertise, to provide independent scientific advice under the Agreement. Science Advisory Board members are recruited from industry, academia, government and nongovernment organizations, and the Board is multi-disciplinary in its expertise.

To substantively address matters referred to it by the Commission under the 1993-95 Priorities, the Board continued to employ three workgroups composed of SAB members and non-Board members with pertinent expertise. The Board workgroups focus on Ecosystem Health, Parties Implementation, and Emerging Issues. Generally, each workgroup met quarterly and reported to the Board at its regular meetings. During the 1993-95 Biennium, the IJC priorities assigned to the Board and Board initiatives regarding emerging issues and technology assessment were delegated to the workgroups as follows:

• Workgroup on Ecosystem Health: Toxicological Mechanisms; Weight of Evidence; and Measuring Ecosystem Health
• Workgroup on Parties Implementation: Parties Toxics Reduction Programs; and Followup on the Virtual Elimination Task Force
• Workgroup on Emerging Issues: Climate Change; Emerging Issues; and the technology component of the Water Quality Board's Workshop on Pollution Prevention

Workgroup reports, conclusions and recommendations are reviewed, compiled and approved by the Board for submission to the Commission in its Biennial Report.

In addition to non-Board members serving on the workgroups, the SAB also benefitted from the involvement of Water Quality Board and Council of Great Lakes Research Managers members serving on the workgroups in a support and/or liaison role. This involvement provided consultation and coordination on advisory activities under the priorities.

2.2 Human Health in Ecosystem Health: Issues of Meaning and Measurement

The Science Advisory Board recommends that:

• the Commission, in its priority activities and its advice to the Parties, support further research to determine ambient levels of exposure to toxic chemicals in the Great Lakes basin and incorporate the following general principles for further development of environmental burden of illness indicators:
• continued monitoring of toxins in media, including trihalomethanes, nitrates, microbial contaminants in drinking water, PM-10, ozone and sulphates in air, and toxic bioaccumulative chemicals in general
• systematic synthesis of water sampling results for microbial contaminants that result in beach closings. Consider complementing these with information on symptoms among beach users
• inclusion of relevant ambient exposure factors (e.g. time outdoors, based on activity record) and consumption factors (e.g. freshwater fish and wildlife) in population-based health surveys. General population-based measures of body fluid levels of key contaminants (e.g. PCBs or DDE for the organochlorines in serum and breast milk, mercury and lead in whole blood for the metals) could be linked with these and other relevant social factors
• surveillance of established environmental health outcomes, such as asthma, such that these conditions may be considered as sentinels for pollution effects
• recognition that some human illness indicators are poorly suited to provide useful information on the impact of environmental matters on human health, e.g. most morbidity and mortality data that is routinely collected, including cancer rates
• development of longitudinal designs around exposures and conditions of interest to enable stronger inferences concerning relationship between exposure and health outcomes.
• the Commission support actions that would lower human exposure to persistent toxic substances such as PCBs and lower concentrations of these substances in human tissues.
• the Commission support the development of indicators and scales that measure the environmental component of illness and wellbeing and indices of environmental stress and environmental condition.
• the Commission continue to monitor state of the environment and sustainable development reporting in order to inform, in its recommendations to the Parties, regarding Great Lakes basin indicators. As these reports often take a broad-based approach to indicator selection, this monitoring is necessary to help ensure the integration of human exposure considerations into assessments of contamination in relevant fish and wildlife species.

2.3 Weight of Evidence Approaches to Decisionmaking in the Face of Uncertainty

The Science Advisory Board recommends that:

• scientific risk characterization formally include disclosure of: (1) choices embedded in the design of supporting research; (2) modifiers of risk factors used; and (3) all relevant uncertainties.
• risk characterizations prepared for environmental decisionmaking explicitly examine the potential indirect consequences resulting from the characteristics of the hazard, pathways and host response as outlined in Appendix I of this section.
• decisionmakers seek out or recommend relevant valuation assessments, legal and regulatory analysis, socio-economic assessments, equity analysis, ethical analysis and cumulative impact assessments as necessary inputs into risk management decisions.
• Commission weight-of-evidence decisions be clear as to evidence used, assumptions, values, uncertainties and consequences involved.
• the level of proof required (beyond a reasonable doubt, or more likely than not) be clearly stated.
• the risk of non-action be included in deliberations on risk management.
• Commission recommendations and decisions based on weight of evidence include parallel decisions on reasonable monitoring needed to serve as a measure of progress toward the desired goal, or conversely as an indicator of a wrong decision.
• Commission recommendations and decisions based on weight of evidence, because tentative, incorporate clear strategies for ongoing cooperation between scientists and managers.
• further development of an ethical basis for ordering and prioritizing goals of human health and/or environmental integrity, when there is a potential conflict between those goals, be undertaken.

2.4 Toxicological Mechanisms: Environmental Exposure to Chemicals Acting as Endocrine Modifiers

The Science Advisory Board recommends that:

• cooperative efforts occur between the governments, academia, the general public and industry to focus research:

  1. to identify which, if any, environmental exposures to chemicals are or have the potential to be endocrine modulators in humans. For those chemicals identified, what are the exposure and dose-response relationships that define the potential for adverse effects?
  2. to identify what effects and disease state in humans may be linked to endocrine modulation as a result of exposure to chemicals in the environment, and at what stage of development is the human most susceptible to these effects
  3. to identify the mechanisms of action of environmental exposures to chemicals relative to endocrine modulation, and how such knowledge can be factored into the risk assessment process
  4. to determine if structure/activity relationships can be developed to accurately predict which environmental exposures to chemicals have the potential to modulate the endocrine system
  5. to determine if sensitive biomarkers of endocrine modulation can be developed and validated for use in animals and humans exposed to chemicals in the environment
  6. to determine in animals if environmental exposures to chemicals that are endocrine modulators can be differentiated from other environmental stressors, such as loss of habitat, malnutrition, or changes in ecosystem dynamics that can similarly exert effects on the endocrine system
  7. to determine in humans if environmental exposures to chemicals that are endocrine modulators, can be differentiated from endocrine effects that are caused by endogenous, dietary or other lifestyle stressor factors (loss of jobs, etc.). How can their interactions be studied?
  8. to identify chemically-exposed cohorts that can be used to study the potential for environmental exposure to chemicals to alter endocrine function or endocrine responsive organ function
  9. to identify if technologies can be devised to control the release of endocrine modulators. Can more effective technologies be developed?

2.5 Federal and Provincial/State Toxic Reduction Programs and Related Activities in the Great Lakes Basin: A Preliminary Evaluation

The Science Advisory Board recommends that:

• the Commission consider toxics reduction programs as a priority for further action within the next biennial cycle. To further this priority item, the Commission should establish a special task force of the Science Advisory Board, in cooperation with the Water Quality Board and the Council of Great Lakes Research Managers, with a mandate to:

  (a) develop standardized binational mechanisms and criteria to assess toxic chemical management laws, programs and data collection activities
  (b) provide advice to the Commission on the design and implementation of such activities in order to assess toxics loadings to the Great Lakes basin.

• the Commission reiterate and re-emphasize to the Parties the recommendation from the Commission's Seventh Biennial Report on Great Lakes Water Quality, which stated:
  • Governments adopt a specific, coordinated binational strategy within two years with a common set of objectives and procedure for action to stop the input of persistent toxic substances into the Great Lakes environment, using the framework developed by the Virtual Elimination Task Force.

The Science Advisory Board recommends that:

• the Commission consider transition planning as a priority for further study and research within the next biennial cycle. Components of this research should include:
  • a study researching a number of case histories of where there were specific efforts to facilitate the transition of an industrial sector owing to some change in circumstance (such as the downsizing of the military establishment or the phaseout of certain substances such as CFCs) to determine the lessons learned from those experiences
  • an investigation to identify the major contributors of dioxin to the Great Lakes
  • the development of a transition plan, with the participation of the important stakeholders, for the virtual elimination of dioxin inputs from one of the major contributors. The development of the plan design would serve as a forum for a discussion on both the general framework and the key components of a transition plan. The terms of reference of the study would include:
    1. the definition of transition planning
    2. important principles that would be included in plan design (such as who should participate in the fashioning of a plan, when it is necessary, among many others)
    3. identification and feasibility of transition mechanisms (such as a transition fund) and the potential and obstacles for them to work in practice
    4. the need to establish, if at all, an institutional framework for the development, implementation and monitoring of the transition process, whether at a local, regional, national or international level.
• the Commission sponsor one or more roundtables to engage the dialogue of stakeholders in the topic, and to further elaborate on how the term should be interpreted and applied.
• the Commission actively seek avenues to participate in international dialogue both within North America and beyond, on transition planning.
• the Commission reiterate recommendation 20 in its Seventh Biennial Report to the government in particular, which stated that governments, industry and labour begin devising plans to cope with economic and social dislocation that may occur as a result of initiatives designed to promote virtual elimination.

Further to this recommendation, that the Commission recommend to the governments within the Great Lakes basin that:

• transition planning components, when deemed necessary in the sense that there is a risk of significant worker or community dislocation, be included into the specific commitments to those substances that are already candidates for phaseout, such as those identified under the Canada-Ontario Agreement, and the Lake Superior Binational Program
• governments report back to the Commission biennially on progress made in furthering these transition mechanisms.

The Science Advisory Board recommends that:

• the Parties be encouraged to support the completion of the binational implementation plan through to 2001 according to the scheduled timeline as indicated in Table 6.
• a quinquennial symposium on climate change in the Great Lakes basin be sponsored by the Parties and be sustained following the event planned for 1996, as an important scientific forum for discussion and to measure progress towards climate change assessment and adaptation.
• the recommendation from the 1993 Science Advisory Board report, that the Parties make a long-term commitment to climate change research under Annex 17 of the Great Lakes Water Quality Agreement, and report progress in a holistic and systematic fashion within the context of a State of the Great Lakes Basin Ecosystem report, receive further consideration and emphasis in the Commission recommendations to the Parties.

2.8 Identification and Assessment of Emerging Issues

The Science Advisory Board recommends that:

• the issues identified as potentially important for the Commission be considered as priorities for serious deliberations during the next biennium. They are complex issues, especially the issue of sustainability that reappeared in each survey, that would require Commission resources to address in terms of implications for progress under the Great Lakes Water Quality Agreement. The issues assessed as most salient include:

  1. sustainable development
  2. stability of water levels
  3. uv-B effects on biota
  4. various implications of North American Free Trade Agreement (NAFTA)
  5. lifestyle choices as a factor in ecosystem integrity
  6. incidence of endometriosis in women who eat fish from the Great Lakes

2.2.1 Introduction

The original task undertaken by the Subgroup on Measuring Ecosystem Health under the priorities for the 1993-1995 biennium was to prepare a discussion paper on methods for the diagnosis, prognosis, treatment and rehabilitation of ecosystems under stress. Most members attended the First International Symposium on Ecosystem Health and Medicine in Ottawa, on June 19-23, 1994. This symposium addressed the issues associated with this priority task in great detail. Based on the input from this symposium and discussion in the Subgroup, the scope of this task was focussed on clearly addressing the impact of ecosystems on human health and the role of human values in defining the "health" of an ecosystem. A contract was let with the Chair of Environmental Health, McMaster University to produce a monograph on this topic. This research chair within the Eco-Research Program under Environment Canada's Green Plan addresses environmental issues in an interdisciplinary way and includes exploration into the concept of ecosystem health among its research goals. The chairholder, John Eyles, a world renowned social geographer and his research associate, Donald Cole, an environmental epidemiologist, had the precise balance of skills to address the task. Their monograph, Human Health in Ecosystem Health: Issues of Meaning and Measurement has now been produced.

This section of the Science Advisory Board (SAB) Report draws heavily on the material contained in the monograph with the permission of the authors, as it was produced in parallel to the writing of the monograph. It is, however, not simply a condensation of the monograph. It represents the opinions of the Workgroup on Ecosystem Health of the SAB as accepted by the Board, and not necessarily in all aspects the opinions of Drs. Eyles and Cole.

The monograph discusses ecosystem health in relation to human environmental wellbeing in its broadest sense as an essential context for human health. It argues that a systematic review of quality of life indicators from a range of literature should be undertaken to develop appropriate health and wellbeing measures for Great Lakes basin populations that go beyond simple measures of environmental burden of illness. This section, because of its more limited scope, addresses human health primarily as defined as an absence of disease, i.e. to highlight for the SAB what is known about human disease that flows from exposure to agents within ecosystems, rather than being determined by human genetics, lifestyle behaviours, nutrition or social determinants such as class, poverty, education and self-esteem. The discussion of this environmental burden of illness and appropriate indicators for its measurement in the Great Lakes basin first requires a discussion of the concepts of ecosystem health and human health.

Ecosystem

Ecosystem can refer to an identifiable natural region; in this sense it is something real, an entity in itself rather than a human mental construct by which reality is understood. It is the geographical landscape and everything in it. As such it is something that we as humans can value and relate to, but it is not a model or analytical framework for scientific inquiry. The Great Lakes Basin Ecosystem, for example, has a geographic dimension that humans can understand and perceive. Ecosystem as a scientific model within which measurements can be made has been approached from two perspectives. One approach is the population-community approach, which focuses on the growth of populations, the structure and composition of communities of organisms and the interactions among individual organisms (O'Neill et al. 1986). This approach views ecosystems as networks of interacting living populations, so in effect the biota are the ecosystem while the non-living components are understood to be external influences or the backdrop in which biotic interactions occur. The other approach is the process-functional approach that emphasizes biophysical models of energy flows and nutrient cycling (e.g. Kay 1991).

Ecosystem can also be viewed as a completely abstracted management tool. Allen and Hoekstra (1992) argue that the observer uses a filter to engage the world. It involves not only definitions and identifying critical changes, but also the nature of measurement and the data collection process. The ecosystem is the system our measuring tools and information gathering techniques allow us to see. Put slightly differently, the human impact on ecosystems is dependent in part on how as well as what we observe (Bandurski 1994).

Human Health

A second "negative" definition of health is the absence of illness. Illness may or may not be associated with disease. The distinction often used is that disease is diagnosed by a physician or other health care professional, while illness is experienced. If an individual does not experience anxiety, pain or distress even if they are diseased, s/he is healthy. Conversely, even in the absence of disease, if an individual does experience anxiety, pain or distress, that individual is unhealthy. In the environmental arena the perception of risk related to exposure to chemicals or other agents in the environment is often a cause of anxiety and distress and thus a source of ill health. It is important to note, however, that the risk that is feared is that a toxic agent has already or will cause a disease. Our concept of what is good and bad in our modern science-oriented society is highly conditioned by the biomedical model. This definition of health does not easily translate to ecosystems. Our common anthropocentric belief is that humans are the most conscious organisms in the ecosystem and therefore we are more subject to pain and anxiety than other organisms.

There are four positive definitions of human health. First, health may be seen as that which enables people to achieve their maximum personal potential (Seedhouse 1986). Health requires basic necessities to be achieved but also provides the basis for higher human needs, such as caring and self-actualization. Dubos (1959) views health as the ability to adapt to new or changing circumstances. This capacity is seen as a fundamental human trait, part of which is humankind's ability and willingness to alter the environment or ecosystem for human purposes. The third positive definition is that of the World Health Organization (WHO; 1948) in which health is a "state of complete physical and social wellbeing and not merely the absence of disease or infirmity." Finally, Parsons' (1972) definition also emphasizes the ideal, seeing health as "the state of optimum capacity of an individual for the effective performance of the tasks and duties for which he/she has been socialized." All the positive definitions of health emphasize human capacity to function.

Ecosystem Health

"An ecological system is healthy and free from "distress syndrome" (the irreversible process of system breakdown leading to collapse) if it is stable and sustainable - that is, if it is active and maintains its organization and autonomy over time and is resilient to stress. Ecosystem health is thus closely linked to the idea of sustainability, which is seen to be a comprehensive, multiscale, dynamic measure of system resilience, organization and vigour. Accordingly, a diseased system is one that is not sustainable and will eventually cease to exist (p. 248)."

Parallel to the concept of ecosystem health is the concept of ecosystem (or ecological) integrity. The goal of the Great Lakes Water Quality Agreement (1978) is to "restore and maintain the chemical, physical, and biological integrity of the waters of the Great Lakes Basin Ecosystem." Ecosystem integrity refers to the ability of a natural system to function optimally. It is analogous to the positive definitions of human health. While ecosystem health implies the ability of a natural system to operate under normal environmental conditions, ecosystem integrity implies that the system can maintain an optimal operation point while stressed and can continue evolving and developing through a process of self-organization (Kay 1993).

Scientific Models and Root Metaphors

In the case of ecosystem health, "health" is a word normally applied to human individuals that is applied to ecosystems as entities encompassing interconnected populations of many species. In its broadest sense, such a metaphor is seeing something from the viewpoint of something else (Brown 1977), involving transferring one term from one system or level of meaning to another. It works when that term is consciously used in a different context. Thus metaphors must not only be significant but must also pretend not to be literally absurd. This is especially the case with root metaphors which put forward fundamental images and values about the world. Ecosystem health is such a metaphor, with fundamental, psychological importance linked to self (through health) and holism (through ecosystem).

Both models and metaphors describe human experience and encapsulate human observations, but they do so differently. Models capture those elements that can be measured, that are quantifiable. Metaphors capture those elements that enrich our understanding in one area by analogy with another area, but cannot be measured. Scientific models ultimately are mathematical relationships and give us the power of prediction of how the system will behave. Metaphors capture the similarities between things but they are not inherent properties of the systems being described. Models and metaphors are both derived a priori from our understanding of the world. Both represent strongly held beliefs about how the world operates.

Their difference lies in their testability: a scientific model is meant to be testable and falsifiable whereas a metaphor is part of a world view, challengeable only by revolutions in thought. Yet if we accept Allen and Hoekstra's (1992) view that observational techniques are filters, then it is important to understand the "humanness" of models. Models have meaning only in the context of the "boundaries of science" and their meaning is dependent not just on their findings but on the form of the model itself: its scientific code. Thus as Bateson (1972) argued, the structure of meaning is dependent on the code and how that is transformed into a message (scientific findings). If we share a code (a scientific model), we can understand missing parts - they are intelligible because we use the code to make sure all parts of the message fit.

"Measuring" Ecosystem Health

"[if] communities have fixed identities, [if] they are normative like organisms, we can easily apply the normative idea of health to them: if they are functionally and structurally similar to their abstract ideal, they are healthy; if they deviate significantly they are sick. If the idea that communities have a normative, equilibrium position, a balance point, were still widely accepted, then the idea of ecological health would pose few problems . . . but ecological concepts change . . . no longer are communities considered normative."

Kelly and Harwell (1990) lament that the analogy of ecological health to human health is strained, given that ecosystems are far more complex than human metabolism; exposure of an ecosystem to external disturbance of ten means differential exposure to only loosely connected parts of the system. Human tissues and organs, on the other hand, are strongly internally coordinated and highly interdependent.

Even with a characteristic set of normative ecosystem ideals, the health concept would still prove problematic. Just as the definitions of human health can vary between individuals, across cultures and over time, so can they vary for ecosystem health. The uncritical application of the concepts of ecosystem health and/or integrity can lead to the application of "medical diagnoses" to achieve an agreed upon state of "health." The "new ecology" (a term applied to describe a major theoretical shift in the field of biological ecology) which calls attention to the instability, disequilibria and chaotic fluctuations of environmental systems (Zimmerman 1994) may in fact make the ecosystem health concept problematic in scientific application. Although it may resonate with environmental action and policy debate and formulation, both Sagoff (1985) and Schrader-Frechette and McCoy (1993) have drawn attention to uncertainty in ecological science.

However, Fine and Sandstrom (1993) contend that people actually see and understand their world through simple slogans and metaphors like "ecosystem health," not through any complicated theories. Ecosystem health as a metaphor provides a commanding image of environmental concern in our ecological times (Worthington 1983) and the normative and personal nature of the health concept. Scientists respond to metaphor in much the same way as the general public (Gieryn 1983). They are guided by dominant cultural images in deciding suitable topics for research and in constructing limits around the "boundaries of science," which are of course also shaped by how observations can occur. The ecosystem health metaphor has indeed served as a point of departure, and as an important heuristic tool for scientific investigation into environmental diagnoses and prescriptions in general, and in the case of the North American Great Lakes in particular. For scientists and the lay public alike, the ecosystem health metaphor provides a method of common engagement, a "metaphorical resource" (Fine and Sandstrom, 1993, 26), packed with shared meaning and normative direction, that can be called upon to legitimate a cause or ignite an emotional response. Thus the ecosystem health metaphor encapsulates both the ecosystem approach to human health and as well, some notion that an ecosystem, like an organism, can react negatively to some external stressor and become diseased or "unhealthy."

Conclusions

• All indicators are goal-directed; they essentially monitor "system" change given desired outcomes. All indicators (as they are selected from an unknowable universe of all possible indicators) are normative.
• "Ecosystem" and/or "environment" is a core value of interest in the identity formation and concerns of populations in the Great Lake basin.
• The value sets that determine indicator selection for ecosystem health and human health should be clearly defined for any developed set of indicators.

The 20th Century has brought an increasing role for the physical and chemical sciences. Elucidation of temperature gradients and basic chemical parameters in water bodies was among the first descriptive work. For toxic substances in environmental media, methods have developed to quantify levels of gases, particulates and organic compounds in air (e.g. Ministry of Environment and Energy 1994) and a wide range of traditional inorganic (e.g. mercury) and organic compounds (e.g. combustion products) in soil and sediment. In water, sampling methods permit collection at distinct points within water columns of dissolved substances (e.g. phosphates), chemicals adsorbed to suspended particles (e.g. PAHs) and functional properties (e.g. biochemical oxygen demand).

Chemical analyses with increasing sensitivity have also enabled measurement of contaminants in many biological tissues of species that make up the food web (Environment Canada, Fisheries and Oceans, and Health Canada, 1991). Monitoring organochlorine pesticides and their metabolites in the fat of fish and bird species along with human foods, fat samples and breastmilk was initiated during the 1960s in response to local use and aerial transport of DDT. Neurotoxic metals also became important: mercury, because of the discovery of the role free-living bacteria play in transforming it to methyl mercury, increasing its bioavailability and subsequent concentration up the food chain; and lead because of its widespread dissemination as a gasoline additive.

Together these data on media and species have permitted sophisticated modelling of contaminant sources and movements within the ecosystem (e.g. review by McKay and Patterson, 1992). For biological species within a toxicological framework they provide the raw material to determine exposure to toxic substances, including calculations of dose based on the various routes of entry. Yet, after some of the more dramatic cases of contamination were mitigated (e.g. phosphate loading), the task of ascription of causal relationships between ecosystem observations and past or present human activities has become increasingly challenging, because of the complexity of ecosystem relationships and the political and economic implications involved.

While the increasing impact of humankind cannot be doubted (see Goudie 1994), nor should the power of human invention and innovation. In studying the effect of human activity on ecosystems, we must, therefore, not only examine the ecosystem but human adaptability as well. A focus of ecological anthropology (e.g. Geertz 1963; Vayda and Rappaport, 1976) is based on Steward's (1955, 1978) ideas on the causal connections between social structure and way of life. The nature and rate of environmental change (often degradation) cannot be divorced from this way of life, including needs, wants, technology and values.

Why does human activity in an environment take the form it does? This is a vital question for advocating particular changes in activity for ecosystem protection. Further, the form of activity is predicated on how a people perceive resources and their relationship to the environment. There are several ways to perceive that relationship; Kluckhohn (1953) suggests three:

• people as subjugated to nature, living at the mercy of a powerful and dominant environment
• people as over nature, dominating, exploiting and controlling the environment
• people as an inherent part of nature, trying to live in harmony with nature.

In the Great Lakes area, tension exists between the second and third, although it may be easier to understand the present status of the debate over ecosystem by asserting that the tension is exacerbated by the fear of the first, especially with respect to human health and wellbeing if control over our affairs is apparently reduced to the demands of ecosystem health.

These concerns are often considered when credible scenarios of potential outcomes are expressed using a range of tools. Ecological risk assessment and the more legally bound, environmental impact assessment, are increasingly being carried out on a wide range of human development projects and interventions. These tools permit explicit examination of trade-offs between human-oriented outcomes and environmental impacts. Although often cast in traditional cost-benefit terms with the cost of mitigation procedures weighed against the benefits of the particular development, other approaches to incorporating human interests and values in ecosystems are increasingly advocated (Public Health Coalition 1992). Ecological economics is one emerging field that questions the usual micro-economics approaches to valuations in development (Costanza et al. 1991). Among its practitioners, Daly (1991) has argued for the need to estimate and set limits on the maximum scale of human development activities possible within particular ecosystems up to the global scale.

If values are important in understanding how human activity affects the environment, it is perhaps also necessary to examine environment as a value in relation to other values and life-domains. Environment tends not to be valued highly in relation to other domains, such as family income and standard of living that are most highly valued (Eyles 1985; 1990). In one investigation in which people were asked the defining characteristics of where they lived, environment trailed such dimensions as social relationships, economic wellbeing, memories, roots, and even no opinion and nothing (Eyles 1985). This ranking reflects a lack of understanding that all of the valued dimensions depend ultimately on the environment.

Environment or ecosystem does not then necessarily engage significant life-domains or core values. The issue can, however, be considered differently. When does environment engage us? And what values are expressed? Our answers can only be suggestive. First, we are engaged when we are threatened. Edelstein (1988) in his work on contaminated communities makes the useful distinction between lifestyle and lifescape, the former referring to people s way of living, the latter to our fundamental understandings about what to expect from the world around us our social paradigm. When lifescape is threatened, core values are threatened. These ideas have not been fully developed, although some research suggests they include those things that indicate threats to our children's health, property values, fear of unknown, latent health effects (Eyles et al. 1993).

Second, the values expressed in environmental concern are again not well-articulated in empirical research. There has been some use of altruism to explain intentions to ameliorate environmental problems (Black et al. 1985). As Stern et al. (1993) explain, Altruism suggests that pro-environmental behaviour becomes more probable when an individual is aware of harmful consequences to others from a state of the environment and when that person ascribes responsibility to her/himself for changing the offending environmental condition. This is but one value orientation. Others include the land ethic, which emphasizes the welfare of non-human species (Heberlein 1972) or of the biosphere itself, as in deep ecology (Devall and Sessions, 1985). Still others implicate economic and socio-biological orientations (Hardin 1968; Olson 1965). Altruism seems the most likely value basis for environmental concern. Through it, concerns for the ecosystem are linked to concerns for other humans. Implicated in it are other fundamental human values such as community, equity and justice. Thus ecosystem health is indirectly pursued through human actions directed at humankind. But, this emphasis on ecosystem health through altruism is but one value orientation, and it is a fragile commitment. Human activity is geared toward human betterment, health and wellbeing. However, those who perceive the dependence of these on the environment tend to have strong environmental concerns and values.

Human choices are not free of the limits imposed by being part of the ecosystem. We cannot choose whatever kind of world we want. We can and do have models of the impact of human activities on ecosystems and the predictable consequences for humans if the ecosystem shifts from one state to another, e.g. arable land to desert or forest to eroded hillside. It is a human value choice whether we attempt to extend the lifespan of the human species, as much as possible, or view the human good as the maximum potentiation of the present -- getting the most out of our environment as it is now. The moral issue is not the extinction of the human species; the species will become extinct sooner or later. Rather, it is whether the extinction is at human hands or by natural forces, not the number of premature human deaths involved in the extinction process. The ethical issue is the lifespan of the species, the number of generations that enjoy this planet earth before the extinction occurs.

Human health, especially the positive definitions of human health, focus on the individual. The maximum potentiation of humans alive today may result in the rapid extinction of the species. A fundamental flaw with the concept of ecosystem health as the value for environmentalists to champion lies in the concept of human health itself. And yet human beings have devised social systems which incorporate societal as well as individual values. Environmentalists tend to extend these societal concerns to include concerns for other species. In the long run, human welfare depends on these species.

Conclusions

• Separate indicators of ecosystem health and human health are required since their goals and targets are different, in the former case ecosystem stability, persistence or resilience, in the latter the disease or illness state of individuals.
• There is a link, however, between indicators of the health of human populations (public health) and indicators of ecosystem health.

As stated earlier, human health can be defined positively and negatively. In a negative sense it can be considered an absence of disease (defined against objective/medical criteria of pathological processes), or an absence of illness (defined by the experience of the individual). In this section we will address the evidence for effects on health as an absence of disease related to exposure to environmental agents.

The evidence for an effect on health comes from environmental epidemiological studies when available. Such studies are limited by the difficulties in assessing the exposures to toxic agents at environmental exposure levels (i.e. accurately classifying who is relatively highly exposed and who is not). All epidemiological studies examine the difference in health outcomes between those who are highly exposed and those who have low exposure to the agent of concern. If a gradient of exposure cannot be found, epidemiological methods are useless, even though the consequences of the exposure may be very real and very severe. Consider the difficulty in knowing whether smoking was related to lung cancer if everyone smoked 20 cigarettes a day. Even if there is a gradient of exposure, we have to be able to correctly classify those who are highly exposed and those with low exposure to get some reasonable measure of the exposures. Otherwise the misclassification of exposure will lead to false negative results in studies. It is quite possible that some widely dispersed pollutants in the environment are having effects we cannot detect epidemiologically for precisely these reasons.

Epidemiological studies also require that the outcome - the health effect - be measured accurately. Much of the concern regarding environmental exposures relates to subtle effects - influences on neurobehavioural development, IQ, psychosexual development and fertility - that may be significant if they occur broadly throughout the whole population, although the impact or deficit for an individual is of little consequence. Other outcomes are of high significance for the individual - cancers, birth defects - but are at low risk at environmental levels of exposure. Because these outcomes can be caused by many factors, it is often difficult to determine if an environmental factor is adding to the burden of illness. Overlapping exposures, all of which in themselves seem to increase the risk of a particular symptom, would seem together to account for more than 100 percent of increases in symptoms. Appropriate statistical techniques must be used to adjust for the lack of independence between exposures, and interactions between exposures and personal characteristics (see Walters 1983). The criteria to assess environmental epidemiological studies are found in Table 1 (Frank et al. 1988).

Environmental health risks can also be estimated by risk assessment protocols using animal data on cancer and birth defect risks. In some situations health effects that have manifested themselves in occupational settings can reasonably be extrapolated back to environmental exposures. More importantly occupational epidemiology often confirms that health outcomes seen in animals will occur in humans if exposure is high enough. For example, Friberg (1984) discusses the evidence for the effect of cadmium on the kidney-linking animal and occupational health data.

The environmental burden of illness refers to the proportion of illnesses, of particular health outcomes that can be attributed to particular environmental exposures. If the relative risk of an outcome occurring in exposed individuals is known and the prevalence of exposure is known, the risk attribution to the exposure in the population, the population attributable risk, can be calculated. We will limit this discussion to the impacts that are or may be occurring in human populations living in the Great Lakes basin as a result of exposures in the ambient environment (exposure to outdoor air, drinking water, recreational water use, exposures to soil) or mediated by the ambient environment (exposure through food). We have included those toxic substances in this section for which there is good evidence for the health effects outlined and for which significant exposure and/or community concern exists in the Great Lakes basin. This section is not an exhaustive review of the evidence on any of the health effects listed. It is meant to cover those areas for which further research and prudent action is recommended. Health effects related to occupational exposures, indoor air quality (except radon), or major environmental disasters are not considered. These exposures can, however, be instructive with respect to risks that may be present from lower-level exposures in the ambient environment. Unfortunately, little precise information exists on exposures to toxic chemicals through the ambient environment in the Great Lakes basin.

Exposure to ozone and particulate (PM-10, i.e. particulate matter of 10 micrometers or less), especially sulphate particulate is widespread in the Great Lakes basin. Attributable risk estimates for the role of air pollution in hospital admissions and deaths for cardio-respiratory illnesses have advanced considerably over the past decade. A series of studies, including one in Detroit, have failed to detect a threshold for increases in deaths associated with small increases in particulates that can be inhaled fully into the lungs (particulate matter of 10 micrometers or less, PM-10) (Schwartz 1991). Similarly, subjecting environmental data on air pollution and hospital admission data to advance-time series analyses, Burnett et al. (1994) showed the increases above baseline admission rates attributable to ambient air pollution, ozone and sulphates in particular. Sulphates in air are widely monitored in Ontario, but sulphate may be an indicator of acid aerosol or PM-10 exposure, rather than sulphate itself causing the effect. This effect was present only for the warm months of May through August. Infants up to one year of age were the most affected, with 14.8% of all admissions for respiratory illnesses to hospitals in Ontario attributable to ozone or sulphate exposure. This study has generated the best attributable risk estimates for an ambient environmental exposure of any study in the Great Lakes basin. Given its major role in environmental burden of illness, extrapolation of these figures to particular Areas of Concern should be possible based on local air pollution data collected by provincial or state authorities.

The second most general ambient exposure of concern is exposure to certain organic compounds and metals in the air of the major industrial cities of the basin. In this case, risk assessment methodology generates estimates of cancer risks related to lifetime exposure. Comparison to total population risks for particular cancers (up to 10 percent for some cancers) reveals that the proportion attributable to individual air toxics is very small, not exceeding 1/10 of one percent. Given the large population exposed to the risks associated with a number of these compounds, their presence in our air is a public health concern. Cancer risks related to these air pollutants are well covered in the Windsor Air Quality Study (MOEE 1994) and the review of the outdoor air quality in the City of Toronto (Campbell 1993). The agents in the Windsor study with the major portion of the cancer risk range greater than one in 100,000 for lifetime exposure as an outdoor air pollutant are benzene, 1,3-butadiene (from car exhaust) and chromium VI. Cancer risks for diesel fumes are well established (Carey 1987) but the risk at ambient levels of exposure is unknown. Radon gas comes from the natural environment into homes and buildings and concentrates in indoor air. There is a very low risk of lung cancer from this indoor exposure, which has been difficult to demonstrate in epidemiological studies (Lubin 1994). Radon could be a problem in the portion of the Great Lakes basin that is on the Canadian shield, but it is also a community concern in the Port Hope area.

A considerable body of toxicological and epidemiological data has developed because of the stakes involved for the producers of chemicals and those exposed to chemicals, particularly in occupational settings. Higginson (1992) reviewed studies attributing portions of the cancer burden to different factors, but pointed out gaps on exposure information that required considerable assumptions to produce estimates, particularly with respect to physical environment and non-occupational exposures.

Expert groups, such as that brought together by the International Agency on Research in Cancer, have estimated the theoretical preventability of cancers (Tomatis 1990). Miller (1992) carried out a similar process for Canada, examining a series of actions that might reduce cancer incidence and comparing the reductions to those that are potentially preventable, based on intercountry comparisons of incidence. Melanoma from ultra-violet radiation stands out (40% reduction), although uv-B exposure is only one factor related to melanoma risk. The thinning of the ozone layer over the Great Lakes basin may be associated with increases in skin cancer and cataracts over time, but these effects have not yet been documented. We do not know the trend in personal exposure to sunlight in the Great Lakes basin, but the role of ultraviolet exposure from sunlight in skin cancer is well established (Ontario Task Force on the Primary Prevention of Cancer 1995).

Trihalomethanes are known to be carcinogenic in animals and are generated in the chlorination process for drinking water. The strongest evidence with respect to drinking water is increased risk of bladder and rectal cancer (Morris et al. 1992), but the carcinogenicity of chlorinated drinking water for humans cannot be considered proven. The major public health benefits of treating water with chlorination are well recognized (see Bellar et al. 1974; Morris et al. 1992); the same authors establish the carcinogenicity of trihalomethanes, using a meta-analytic approach based on case-control studies. The proportion attributable to drinking water would be very low, but most of the Great Lakes basin population drinks chlorinated water. An association between cancer incidence and water supply trihalomethane concentrations has yet to be demonstrated in a Great Lake state or Ontario (Gilman et al. 1992), partly due to the variable sources of drinking water among Great Lakes populations. Nevertheless, further exploration of the risks and benefits to human health of chlorination and its alternatives is clearly warranted.

Tritium is a hazardous substance in areas adjacent to nuclear power plants in Canada because of the use of heavy water in CANDU reactors (ACES 1994). The Advisory Committee on Environmental Standards (ACES) in Ontario recommended that the objective for tritium in water be immediately reduced to 100 becquerels/litre (in response to the recommendation by the Ontario Ministry of Environment and Energy to reduce the current objective of 40,000 Bq/L to 7,000 Bq/L) and be further reduced to 20 Bq/L within five years. Tritium concentrations in some Ontario drinking water supplies currently exceed the 20 Bq/L standard from time to time. This recommendation was made on the basis that tritium is a human carcinogen and that the same level of acceptable risk should be applied to it as to other chemicals that are human carcinogens. Exposure occurs through drinking water but also occurs through air and the food chain.

Diseases involving stomach and intestinal infection due to foods and water contaminated by micro-organisms are another major category for which attribution to environmental exposures is routinely made by public health authorities (Todd 1991). Outbreaks from contamination of municipal water supply systems by recently recognized protozoa (e.g. Moorehead et al. 1990) have constituted the largest clearly identifiable human burden of acute illness based on use of water from the Great Lakes or waters flowing into them. Both Milwaukee (MacKenzie et al. 1994), drawing from Lake Michigan, and Waterloo, drawing from the Grand River which flows into Lake Erie, have experienced difficult-to-control outbreaks of contamination by cryptosporidium species. These outbreaks are linked to contamination sources within watersheds that cannot be managed efficiently and effectively at the point of water treatment plants, but are better dealt with by watershed management schemes. Small outbreaks of giardia (another protozoan) and viral diseases such as hepatitis-A do occur, usually transmitted through food (Todd 1991). Giardia is consistently present in some wellwater supplies. Viral diseases transmitted through food in the Great Lakes basin are almost always imported, i.e. acquired by the initial case outside the basin. Exposures to sewage-contaminated waters during bathing (Fleisher et al. 1993) also result in illness, although Great Lakes basin cases are poorly documented.

Emerging literature such as that linking persistent organochlorine pesticide exposure and breast cancer (Wolff et al. 1993) have not been fully incorporated into standard cancer risk estimates, partly due to the ongoing controversy as to the significance of these findings (Ritter 1994; Kreiger et al. 1994). Risk assessment techniques have been used to estimate the cancer impact of eating Great Lakes fish contaminated with persistent organochlorines (Foran et al. 1989). Based on DDT and dieldrin levels in the fish and consumption rates, increases in cancer numbers for various concentrations are calculated. Yet these are difficult to relate to particular areas unless distributions of fish consumption are known; such data are often of variable quality and representativeness (Ebert et al. 1994).

The established effect of dioxins in animal models and the probable effect of DDT, PCBs and other persistent organochlorines on the immune system are likely to be an endocrine modulation effect (see Chapter 2.4). Exposure to dioxin is primarily through the food pathway because of distribution through the atmosphere (Davies 1988). Reliable risk estimates associated with this exposure are not available.

There is significant public concern regarding exposure to currently used pesticides. Organophosphate pesticides are used in institutions such as to control pests like cockroaches. Although case reports for health effects related to exposure do exist, these effects in the majority of the concerned population likely fall in the category of environmental hypersensitivity (see below). There is evidence that aldicarb, a carbamate pesticide, may impair immune function (Fiore et al. 1986). This exposure has occurred through well-water in Wisconsin. The International Agency for Research on Cancer (IARC) has classified several herbicides as possible human carcinogens and the recent report of the Ontario Task Force on the Primary Prevention of Cancer (1995) has recommended reasonable and measurable timetables to sunset these herbicides. Some fungicides have been shown to be carcinogenic in animals and significant exposure can occur through food, such as the consumption of pick-your-own strawberries (Mitchell et al. 1987). Use of these fungicides is now restricted in Canada and the United States.

Recent concern has focussed on neurobehavioural deficits resulting from in-utero exposure to persistent toxic substances. The effect of low levels of lead exposure are now well established (Needleman and Bellinger, 1991). Mercury is known from environmental disasters to produce neurobehavioural deficits in children, and modelling of fish consumption and methylmercury intake is feasible (Richardson and Currie, 1993). Epidemiological methods have not established effects in the Great Lakes basin. The role that aluminum exposure, primarily through drinking water, may have in the development of Alzheimer s Disease has been extensively reviewed (Nieboer et al. 1993), and although there are major weaknesses in the epidemiological evidence, a possible role cannot be ruled out by other scientific evidence. Infants of PCB-contaminated, fish-consuming mothers were smaller than controls and had behavioural deficits and impaired visual recognition (Fein et al. 1984; Jacobson et al. 1984; Jacobson and Jacobson, 1988), but the significance of these findings is still debated. Several research projects in progress in the basin are attempting to resolve this issue (ATSDR 1994). Limited evidence exists for direct neurotoxic effects related to exposure to organic solvents from waste dumps (e.g. Hertzman et al. 1987).

Determining the burden of reproductive problems expected at the levels of exposure thought to exist among human populations in the Great Lakes basin is fraught with uncertainties that have been highlighted in the Commission' s Seventh Biennial Report (IJC 1994). Reproductive outcomes refer to birth defects and to the impact on fertility. Cadmium, lead, mercury and chlorinated solvents are toxic to human reproduction, but at levels considerably above those found through environmental exposure in the basin. Controversy has surrounded the attribution of reported reductions of sperm counts in industrialized countries to increasing exposure to exogenous (from outside the human body) estrogens such as nonophenols, phthalates and persistent organochlorines (Carlsen et al. 1992; Bromwich et al. 1994). Studies are underway to examine contaminant levels in a range of angler, minority and other populations in the basin (ATSDR 1994) and new sensitive outcomes are being examined in relation to these levels (e.g. time to pregnancy). Some potential health effects such as changing the frequency of behaviours more common in boys or girls (dimorphic behaviours) due to environmental estrogens still remain unexamined.

It is beyond this scope of this discussion to outline the burden of illness related to environmental hypersensitivity. This illness has been increasingly attributed to physical environments (Ashford and Miller, 1991) but is likely associated with specific social environments as well. A set of psycho-social impacts (stress, anxiety, worry) may not be recognized as "disease" but may be significant in experiencing an environmental exposure (Edelstein 1988; Taylor et al. 1993). Other interpretive models than traditional epidemiological ones are required to understand the linkages between such experienced "illnesses" and ecosystem parameters. Other investigative methods, based on qualitative traditions, are also required (Eyles et al. 1993).

Indicators of Environmental Burden of Illness

It would be useful to determine the magnitude and trends in the impact of environmental factors on human health outcomes. A wide variety of morbidity and mortality statistics are kept, which are useful in health care service planning. These data do not, however, reveal the cause of the health effect. All health outcomes have a multitude of causes or risk factors; environmental agents are but one contributing factor. Several approaches have been tried to isolate the attributable risk associated with the environment (Walters 1983).

Cancer and birth defect data have been mapped in atlases (Gilman et al. 1992; Johnson et al. 1992; Mills and Semenciw, 1992). Geographic patterns do emerge in these atlases, but it is not only environmental risk factors that vary geographically. Rather than the atlas giving answers that the high rate of cancers or birth defects in a particular area is likely due to particular environmental conditions, we usually can explain the variation on the basis of what we already know about the risk factors for the specific health effects. Atlases, however, can at times be useful to generate a hypotheses to test by other means.

Cancer outcomes are poor indicators of environmental effect because the latency between exposure and outcome is usually several decades. The exception to this general rule is the use of childhood leukemia as an endpoint for exposures to radionuclides. The latency for some leukemias is two to ten years. Studies of the association between proximity to nuclear power plants and leukemia have shown a slight (but not statistically significant) trend toward increased leukemias (Clarke et al. 1991).

Birth defects (minor and major) occur in three to four percent of all pregnancies in Ontario (Mills and Semenciw 1992). Some of this effect is related to background exposure to teratogens, including radiation in the environment and naturally occurring chemicals in food. Although birth defects are relatively common, they are not good indicators of effect related to toxic agents added to the environment by human activities. There are a wide variety of birth defects and the effect of any specific agent will be only at critical stages in the development of the embryo. Individual types of birth defects have a low incidence in any geographic area. An environmental exposure over a relatively small geographic area, even if it were a strong teratogen, would produce few specific birth defects over a short period of several years. The increase in outcome will be less than the chance variation in outcome and thus not be detectable by epidemiological methods.

In some situations the environmental effect of a pollutant is the major source of variation in the health outcome. Thus a clear association between hospital admissions for asthma and respiratory problems has been demonstrated related to particulate, ozone and sulphate pollutants in the air (Burnett et al. 1994; Dockery and Pope, 1994). Indeed this effect varies directly with air pollution levels without evidence of a threshold. Hospital admissions for children under one year show the effect most clearly, and thus these hospital admissions would make a good indicator of the respiratory burden of illness. The lag time for the effect is up to three days. Compare this to several decades' latency for lung cancer and it is clear why trends in lung cancer do not give a good indication of effects related to those air pollutants that clearly do add to the burden of lung cancer.

Although the increased exposure to uv-B in the Great Lakes basin because of thinning of the ozone layer should result in increased skin cancers(squamous and basal cell carcinomas) and cataracts, this effect may also be moderated by measures individuals take to reduce their exposure. Skin cancers and cataracts have long latency periods, so that today's trends represent the effect of environmental and behaviour changes several decades ago. Trends in these outcomes deserve study, but they do not serve well as indicators of how our environment is doing today.

The practical reality is that the association of low-level environmental exposures to health effects can rarely be established by epidemiological methods. Risk assessment methods can give reasonable estimates of risk at levels of exposure in the ambient environment. The limitations of science, however, do not mean that health effects related to ambient environmental exposure are not occurring or are not of concern (see Chapter 2.4). For some toxic substances like lead, very good evidence from well-controlled epidemiological studies and/or risk assessments indicate that health effects are or very likely are occurring within the range of ambient exposure. In these situations, the most practical way of determining the trend in the environmental burden of illness then becomes measurement of the change in exposure, not measurement of the health outcome. Blood lead surveys have routinely been done for this purpose. The neurobehavioural impact of low levels of blood lead has been established in cohort studies (Needleman and Bellinger, 1991). Population surveys of blood lead levels in children can then establish the likely environmental health impact from lead. Population surveys of children's IQ and measurement of environmental concentrations of lead (air lead, soil lead) would never be able to establish the environmental burden of illness. Blood lead is a much better indicator of actual received dose than measures outside the body. Children's IQ is influenced by a wide variety of factors that, in general population monitoring would override any clear indication of a lead effect. Monitoring of other persistent toxic substances (such as PCBs in human tissues) would similarly be useful.

Blood lead is one example of a bioindicator, a measure of exposure. Bioindicators can also measure physiological changes as a result of exposure. Environmental agents are, however, not specific in producing these changes, so that the actual application of bioindicators of effect in Great Lakes basin studies has not proved to be useful (Kearney and Cole, personal communication).

What can summarize the best measures to monitor the potential of environmental exposures to produce human health effects are the actual monitoring of the agents known to produce the direct effects. Further studies of toxics in food similar to Davies (1988) are warranted. Blood lead surveys are appropriate, but with the elimination of lead in gasoline, lead exposure has become a less serious issue in the Great Lakes basin. A case can be made for creating a database to monitor persistent organochlorines in the population, but this recommendation may be influenced by the outcomes of current ATSDR-funded studies on PCBs and neurobehavioural effects (whether or not the results of the Michigan fisheater cohort studies are confirmed). New research on endocrine modulators is needed (see Chapter 2.4). Until there is a much clearer understanding of the effects of these chemicals, better characterization of exposure (serum total PCBs likely being the most cost effective and representative measure) is warranted.

In terms of actual effects of toxic agents in the environment on the health of humans in the Great Lakes basin, hospital admissions for children under one year of age for asthma/respiratory disease is the only precisely measurable indicator at this time.

Various attempts have been made to establish lists of criteria for indicators, recognizing that no single indicator is likely to meet all the criteria. The Council of Great Lakes Research Managers developed 16 criteria for indicators of ecosystem health (IJC 1991; see Table 2)Developed by the Council of Great Lakes Research Managers. In the report, Bioindicators as a Measure of Success for Virtual Elimination of Persistent Toxic Substances (1994) submitted to the IJC's Virtual Elimination Task Force (VETF), four criteria are suggested: specificity to the substances; placement in appropriate scales; ease and cost of measurement; and social relevance/public perception. This is a sensible short list. Eyles and Cole (1995) in their monograph use a simplified but more generic approach to indicator criteria applicable to ecosystems and human health. They propose two sets of indicator criteria: science based and use based, with the caveat that all indicators are goal directed and that good indicator selection is dependent on specifying the problem to be measured and managed. The science-based criteria are:

• Data availability and suitability. It is likely because of cost constraints that existing datasets must be used where possible, but it must be remembered that those data may have been collected for different purposes than now required.
• Validity and reliability. To be valid, an indicator must measure the phenomenon or concepts it is intended to measure. There are four types of validity:

  -Face validity (after evaluating the rationale behind indicator selection, is it a reasonable measure?)

  - Construct validity (does the measure behave as expected in relation to other variables in the scientific model in which it is being used?)

  - Predictive validity (does the measure correctly predict a situation which would be caused by the phenomenon being measured?)

  - Convergent validity (do several measures collected or structured in different ways all move similarly over time?)

  Reliability depends on the amount of error variance in an indicator measurement, and is determined by carrying out repeat measures of the same indicator.

• Indicator representativeness. Questions of data representativeness are quite easy to recognize, based as they are on sampling procedures, and size and population characteristics. More troublesome is the issue of indicator representativeness. Is it possible to select one or several indicators that cover the important dimensions of concern? Indicator representativeness may be enhanced by developing an index, combining indicators. However, even if the problems of combining indicators can be overcome, if the index rises or falls, it remains unstated which of its constituent indicators are rising or falling.
• Indicator comparability. Not only must data be available for several time periods, they must also mean roughly the same thing at those times. The sensitivity of measurement procedures or the nature of the population being studied may change.
• Disaggregating indicators. To be informative, indicators must be related to other variables such as age, sex, locale and various characteristics of the involved individuals or communities. If an indicator can be broken down by several variables, it tells us a great deal more, so long as the numbers do not become too small.

The use-based criteria for indicator selection are:

• Goal oriented. There should be as much clarity as possible in the definition of the relationship between the indicator and the goal (purpose, use, state) that it is meant to monitor.
• Feasibility. Are the data already collected? If they are, are they available for the right time periods and at the desired geographical scale? If they are not, how feasible is it to create surrogate or indirect indicators of the phenomenon of interest? If this is carried out, what happens to scientific validity? If the data are not collected, how expensive would it be to alter the information-gathering system?
• Desirability. Do the indicators inform on the state of the ecosystem or of health in ways that are perceived as important by those affected? Do the indicators enable residents of a particular region or the members of a particular population group to assess their needs and risks? Do the indicators enable them to make meaningful comparisons with similar groups of residents or populations members? A feature of desirability is in fact credibility (a user-version of validity).
• Gameability. If there is to be a link between public perceptions and indicators, then we must ensure that indicators are not gameable, i.e. that they cannot be "gamed" or altered by those with something to gain (while others lose) from the indicator being pushed in a certain direction at a particular pace. For example, if resources for improvements in water quality are dependent on a particular level of microorganisms, it may pay a municipality to defer reporting improvements until budgetary allocations are made.
• Manageability. The ability of human beings to process information is limited. Therefore, the number of indicators to be used should be as small as possible.
• Balance. There should be a rough balance among all of the phenomena of interest.
• Catalyst for action. We may choose to distinguish indicators that more or less act as catalysts for action, whether on the part of industry, government, communities or individuals. This criterion is also important in that it relates indicators firmly to the goals of monitoring.

These criteria act as criteria for the suitability of indicators in themselves and as criteria for specific indicator selection. They enable those concerned with monitoring ecosystems and human health in the Great Lakes basin to consider matters of proof (primarily, but not exclusively the scientific list) and of prudence (primarily, but not exclusively the use list) together.

It is necessary to ask continuously: how is human health relevant to the specific ecosystem issues under consideration? What "evidence" (scientific or philosophic) underpins the connection of human health and ecosystem health? How might we judge the significance of any identified connection? In answering such questions through identifying plausible indicators, we must always be aware of the normative basis and power of science.

We must recognize in our efforts to "measure" ecosystem health and human health as an integral part of it that "ecosystem," health and similar terms are abstracted notions with implications not only for what but also how we measure things. The notions that become powerful, that have resonance, take on metaphorical significance, hence the need for value clarification. We must also recognize that adoption of a prudent or precautionary stance towards the evidence of health effects must be open to scientific evidence. In our use of the metaphor ecosystem health, we must exercise caution concerning the connectionist view of the world contained in the metaphor. The utility of the connectionist, network approach to human health in relation to ecosystem is a frame-work -- an overarching recognition that warns of possible trade-offs, side effects, possible unintended consequences and unanticipated events. It should not be so overarching that it limits our capacity to act in subsystems or among subpopulations. For this, we must battle the power of metaphor.

Although it is difficult to attribute a specific proportion of overall burden of illness to the environment or ecosystem degradation, human health is a vital consideration in the ecosystem health paradigm. Ecosystem health internalizes human wellbeing as part of the environment, while a human health focus internalizes environment for individual and community wellbeing. The strength of the metaphor or paradigm is clear. Ecosystem health sees humans as integral parts of nature. The metaphors resonate strongly with core values about ourselves, our identify and our place in the world.

• The Science Advisory Board recommends that the Commission, in its priority activities and its advice to the Parties, support further research to determine ambient levels of exposure to toxic chemicals in the Great Lakes basin and incorporate the following general principles for further development of environmental burden of illness indicators:
  • continued monitoring of toxins in media, including trihalomethanes, nitrates, microbial contaminants in drinking water, PM-10, ozone and sulphates in air, and toxic bioaccumulative chemicals in general

  • systematic synthesis of water sampling results for microbial contaminants that result in beach closings. Consider complementing these with information on symptoms among beach users

  • inclusion of relevant ambient exposure factors (e.g. time outdoors, based on activity record) and consumption factors (e.g. freshwater fish and wildlife) in population-based health surveys. General population-based measures of body fluid levels of key contaminants (e.g. PCBs or DDE for the organochlorines in serum and breast milk, mercury and lead in whole blood for the metals) could be linked with these and other relevant social factors
  • surveillance of established environmental health outcomes, such as asthma, such that these conditions may be considered as sentinels for pollution effects
  • recognition that some human illness indicators are poorly suited to provide useful information on the impact of environmental matters on human health, e.g. most morbidity and mortality data that are routinely collected, including cancer rates

  • development of longitudinal designs around exposures and conditions of interest to enable stronger inferences concerning relationship between exposure and health outcomes.

• support actions that would lower human exposure to persistent toxic substances such as PCBs and lower concentrations of these substances in human tissues.

• support the development of indicators and scales that measure the environmental component of illness and wellbeing and indices of environmental stress and environmental condition

• continue to monitor state of the environment and sustainable development reporting in order to inform its recommendations to the Parties, regarding Great Lakes basin indicators. As these reports often take a broad-based approach to indicator selection, this monitoring is necessary to help ensure the integration of human exposure considerations into assessments of contamination in relevant fish and wildlife species.

2.3.1 Introduction

As more and more scientists venture into the arena of public policy, they are proving a valuable point: scientists, no matter how expert at their craft, are no wiser than anyone else when it comes to public policy.

David Sarokin, Washington, D.C. (From: "Letters," Science, Vol. 261, September 10, 1993, commenting on the editorial, "Pathological growth of regulations")

The Sixth Biennial Report on Great Lakes Water Quality of the International Joint Commission (IJC), in 1992, proposed as its first recommendation the application of a "weight of evidence" approach to identify and virtually eliminate persistent toxic substances. In that context, weight of evidence referred to considering together the many studies that indicate (or refute) injury or the likelihood of injury, to determine if the evidence is sufficient on which to base conclusions and policy decisions. The Commission elaborated on this proposal in its Seventh Biennial Report in 1994, suggesting a "pragmatic" definition drawn from both science and law. This definition noted that the cumulative weight of the many studies that address the question of injury or the likelihood of injury to living organisms should be considered. It was noted that this approach draws on formal science, logic and common sense. Many methodological and definitional questions remained, however. During a workshop at the 1993 Biennial Meeting, Commission members committed to making a priority for the 1993-95 cycle an analysis of how to proceed to clarify these questions. This report is a response to that priority.

Twenty years of experience with very diverse persistent, toxic and bioaccumulative hazards in the Great Lakes basin suggests the need for a systematized approach to evaluating the range of health and ecological effects putatively linked to these environmental exposures. Decisions as to possible measures to prevent or mitigate require explicit methods for dealing with the considerable uncertainty that often exists. Sometimes there is clear evidence of causation, in the form of "mature" epidemiological studies of human health outcomes. These studies can be analyzed by existing or new epidemiological criteria for assessing the quality of evidence on causation, such as those that Sir Austin Bradford Hill first proposed some 30 years ago. Typically, such studies estimate the strength of association between measured exposure to the contaminant and a specific, well-measured human health or natural resources outcome. One example is the evidence in the Great Lakes basin that the organochlorine family of PCBs has induced effects on wildlife. In this case harm to wildlife was observed first, and efforts to identify environmental agents associated with such harm and appropriate abatement measures followed.

However, experience in the Great Lakes also has required consideration of threats to biological diversity from factors other than conventional toxic substances. For these a more systematic and predictive approach is needed. All of the hazards have worrisome features, despite the fact that we do not yet fully understand them. Many toxic substances can lead to cascading complex effects in the Great Lakes Basin Ecosystem, with unpredictable but potentially severe outcomes after a long latent period. Effects can occur at the molecular or cellular level rather quickly, as in the case of estrogen-imitating compounds and we cannot yet evaluate the societal significance of such effects. More often in the past we have waited many years until reproductive failure is widespread among prominent species in the food chain, and at that point mitigation is problematical. Early detection and action are preferable, but that is also the period of greatest uncertainty. In these situations, we have only non-epidemiological evidence (i.e. laboratory studies or modelling) to consider potential harm to humans, animals and/or plant populations. Considerable potential exists for serious problems if longer-term ecological or human health effects are underestimated, and no action is taken to control the substances involved.

To develop criteria that could address these issues, a subgroup of the Workgroup on Ecosystem Health was formed in 1993, to undertake the weight of evidence priority set by the Commission. This term describes a synthetic integration to consider collectively all of the scientific evidence used in decisions to limit the risks from toxic substances. However, in the process of meetings held during 1994, the study group has come to realize that the expression "weight of evidence" is confusing in its similarity to legal scholars' approach to the adjudication of evidence in general, a rather different subject. As a result, the subgroup has, with support of the Workgroup on Ecosystem Health, adopted the subtitle shown above, "Approaches to Decisionmaking in the Face of Uncertainty."

Specific Objectives

Given the above background, this report has the following objectives:

• To review critically the current methodologies for assessing evidence of potential harm to humans or wildlife from putative toxic substances or other environmental interventions as the first step toward decisionmaking
• To identify those characteristics of environmental/human harm from putative toxic or other hazards that warrant adoption of broad, protective decisionmaking approaches
• To recommend, on the basis of this review, appropriate methodologies for assessing current and future Great Lakes environmental hazards as the first steps toward their control or remediation.

Three major methodologies have evolved, each of which in its own way systematizes the collection and use of scientific information and assists decisionmakers in the appropriate use and interpretation of that information. These are:

- inference as to causality
- risk assessment
- reports by interdisciplinary expert panels or study commissions.

We examine, comparatively, the strengths and weaknesses of each of these approaches as applied to water quality protection issues in the Great Lakes, particularly in how well a finding can be determined and the uncertainty quantified. The following sections discuss three contexts through which the information is reviewed and incorporated into decisions. This process is referred to in general as risk management.

Inference as to Causality

The steps required to infer causality have been understood since Koch's postulates were developed a century ago for assessing the microbial causation of infectious disease. More recently, Hill's criteria (1965) have been developed for a broad class of health studies. These criteria can be used by scientists not only to identify causal linkage, but often to describe the dose/exposure - response relationship in a mathematical way.

The basic principles of Hill's criteria of causality were developed for human epidemiology. They require that an association between a hazard and effect have most or all of the following characteristics:

Strength of Association. The rate of disease or other health effect in the exposed group of organisms must be higher, in a statistically significant sense(see Footnote 1), than in a control unexposed group -- preferably matched for age, sex, calendar year, etc. The actual strength of the association between hazard exposure and health outcome is measured by relative risk (RR): the proportionate increase in the risk of the outcome in question, in the exposed compared to an unexposed group. "Strong" associations have RRs of four or five or more, and are almost always causal in some sense; "moderate" associations (with RRs of two to four) and especially "weak" associations (relative risks of one to two) are often either due to imprecise identification of the exposed group or a non-causal association due to other confounding factors.

Consistency of Association. The same exposure/disease association is found in studies of other populations of the same species separated by geography, time and circumstances. In other words, the observed association is reproducible.

Specificity of Association. The uniqueness of the exposure's health effect is such that it strengthens one's confidence in causality. For example, Minimata disease occurs specifically following mercury poisoning. However, with environmental and other ubiquitous agents, specificity is often difficult or impossible to prove, since so few adverse health effects are caused by only one hazardous exposure. Thus, specificity is helpful when present, but is to be expected rarely.

Temporal Association. Exposure to the hazard must precede the disease or health effect. Latency periods for carcinogenesis (e.g. from smoking) and some other exposure-effect sequences, however, may extend over decades.

Dose-(Risk)-Response Relationship. Exposure increases should lead to stepwise increases in disease incidence or risk. In environmental epidemiological studies, this is often a crucial basis for inferring causation. This criterion can be modified in certain circumstances by competing causes of death, "all-or-nothing" biological responses that cannot be repeated, and resonance-related phenomena such as electromagnetic radiation, in which the dose-response relationship can change depending on frequency and field strength. In environmental studies it is often possible only to achieve an ordinal scale for dose, cf. "more" or "less," rather than a more accurate, continuous scale. Note here that the use of the term response relationship is an analog of, but not the same as, its use in toxicology/pharmacology where the response is not a population-based risk measurement, but rather some biological phenomena inside an organism.

Biological Plausibility of Association. The disease mechanism should be observable in animal models using the methods of laboratory disciplines (e.g. pathology, microbiology). Unfortunately, laboratory models do not exist for some human (and animal/plant) diseases. There should, however, be a plausible biological mechanism by which the exposure-effect sequence could conceivably occur.

Coherence of the Relationship. This criterion broadly implies that the hazard-exposure effect should be compatible with the known distribution of both the hazard and the health outcome over space and time. Thus childhood acute lymphocytic leukemia could not be primarily caused by electromagnetic fields if that condition is most common in rural developing countries where few strong electromagnetic fields are found. Similar studies should confirm the relationship, including studies in other branches of science and bioscience, such as those showing similar exposure-effect phenomena in animals or plants in the wild. The latter are often added as an analogy of additional criterion for inferring causation.

Experimental Confirmation. Laboratory studies on animals can test the cause of relationships between hazards and human disease. This obviously requires, however, that appropriate animal models are available to test the relationship. For those hazard-exposure effect sequences without such models, it is generally unethical and infeasible to conduct experiments in which humans are randomly exposed to a possible hazard, to see if the putative effect ensues. Thus environmental epidemiology is virtually always bereft of this most potent of Hill's criteria for causation. Some observers think of the demonstration of "reversibility" of health effects, after hazard abatement, as equally good "experimental" evidence of causation. However, many health effects are simply not easily reversed, or have long latencies after hazard exposure and before full expression, so one may not be able to so readily demonstrate "reversibility" in many situations.

In practice the inference of causality provides the framework on which to base decisions on elimination of harmful biophysical effects in contained situations such as a workplace. It is widely used in occupational health and is necessary to investigate less well-defined situations in an environment or ecosystem.

Scientific information generated through the study of humans in the workplace has provided one avenue for understanding the broader impact of such agents in the environment (air, water, soil or food chain) or the biota (living plants and animals), which together form the ecosystem. However, human health studies alone have proven insufficient to protect fixed ecosystems for one or more of the following reasons:

• Some routes of nonhuman exposure do not exist for humans
• Some routes of human exposure do not exist for nonhumans. Humans are exposed to all sorts of commercial products that are biologically active
• Chemicals are likely to be more toxic to some nonhuman organisms simply due to interspecies differences in sensitivity, physiology, biochemistry and/or lifestyle
• Mechanisms of action at the ecosystem level may not have human analogues
• Any environmental pollution is likely to result in much higher exposure to some nonhuman species than to humans, since humans have access to air conditioning, and food and water imports that protect them from local environmental hazards
• Most birds and mammals have a higher metabolic rate than humans so they may receive a larger exposure or dose per unit body mass of some hazards
• Some chemicals released into the environment are designed to kill nonhuman organisms (pests) and result in nontarget toxicity
• Nonhuman organisms are highly coupled to their environments and therefore may experience severe secondary effects such as loss of food or physical habitat.

In conclusion, the Inference of Causality is a methodology to synthesize epidemiological research on human, animal or plant populations and biochemical/physiological/microbiological knowledge of mechanisms. Thus the great value of the Inference of Causality lies in its integration of two disparate types of empirical evidence: the synthesis of multiple sorts of "basic science" evidence, and evidence from well-designed epidemiological studies. It represents careful scientific building of knowledge and leaves little uncertainty, but only where there has been a large enough population of organisms with measurable exposure to a well-defined hazard to generate sufficient cases of ill-health effects to allow a statistically significant relative risk compared to control unexposed populations. Unfortunately, all of these conditions fail to apply in many circumstances of health effects due to environmental hazards. And even when these conditions do apply, and the application of Hill's criteria is useful, the fact that one has waited until there is an actual human "body count" of affected individuals means that exposure to environmental hazards has obviously been allowed to go too far.

Limitations to this Approach: The Inference of Causality is used most appropriately with a single, well-defined and measured hazardous exposure, such as lead or methyl mercury, and a single biological effect, such as the neurotoxicity or Minimata disease. It begins with a null hypothesis, that the hazard does not cause the biological effect, and tests that hypothesis against observations using recognized statistical methodology, within the framework of particular epidemiological study designs (e.g. case-control or cohort).

While the research goals and experimental design are subject to choice, the Inference of Causality is generally considered value-free. It is also considered limited in its applicability. It usually does not, for example, include the study of mitigative activities, or the socio-economic consequences of such activities. However, when a full set of epidemiological studies and basic science information exists of relevance to a putative environmental hazard exposure and observed health effect, the application of Hill's criteria is invaluable for assessing this information impartially. Unfortunately, such full evidence is often missing in environmental hazard assessment, as has been noted above.

More frequently than not, epidemiological studies attempting to infer causality are inconclusive or have negative results, i.e. no adverse health effect is observed. This is often due to test design, the requirement of large samples, testing only one biological endpoint such as cancer death, or the effects of multiple toxic interactions. The human or ecosystem integrates all of the negative exposures received, and the resulting ill health may not be directly attributable to any one of the exposures taken in isolation. It is not possible for the science of inference, which deals best with single effects, to meet fully the challenges of multiple hazards, e.g. toxic chemicals such as occur in the Great Lakes basin.

Finally, as an approach to decisionmaking, the use of Hill's criteria is data-intensive and requires scientific products of a long period (often decades) of study. Thus, it is usually available for retrospective analysis, and less appropriate for new problems or other prospective decision situations.

Risk Assessment

The science of risk assessment addresses and quantifies, where possible and appropriate, hazard identification, dose-response (or exposure-response) relationships and exposure determination, which lead to risk characterization. Risk characterization is the primary scientific input into risk management, which will be discussed in the next section. These relationships are shown in Figure 1, taken from the International Joint Commission's (IJC) Workshop on Risk Assessment, Communication and Management, held February 1-2, 1993. Ecological risk assessment is defined as a process that evaluates the likelihood that adverse ecological effects may occur or are occurring as a result of exposure to one or more stressors.

Environmental Risk Characterization. The Relationship Between Risk Assessment and Risk Management (After Farland 1993)

The paradigm now used by the U.S. Environmental Protection Agency (U.S. EPA) for most human health risk assessment was developed by the U.S. National Academy of Sciences and published in 1983. Hazard identification is a part of the risk assessment process. Hazards to health or to ecological systems are those interactions with human products, activities or interventions at sufficient intensity to alter the functioning of human or ecological systems at some level of organization (including the cellular and molecular). Hazard identification depends on the collection of all relevant information derived from laboratory experimentation, epidemiology, toxicology and cytotoxicology, embryology, physiology, anatomy, biology and other relevant disciplines. From this hazard identification process, gaps in research can be identified and studies undertaken that would permit more confident statements to be made about the significance of the hazard.

Hazards other than persistent bioaccumulating toxic chemicals are recognized as sources of harm to the Great Lakes Basin Ecosystem. These hazards include recreational uses of the lakes, construction and dredging operations, invasion by exotic species, and other stressors. In 1993, the U.S. National Academy of Sciences updated its 1983 report on risk assessment to include a broader definition of hazard to include those to the ecosystem as well as hazards to human health. They cited the Georges Bank Fishery Assessment as "most complete." This assessment contained a determination of the qualitative effects of fishing on fish populations. While the assessment developed in this case was incorrect, the community dynamics described are clearly analogous to the determination of contaminant effects and can legitimately be called "hazard identification." Estimates of fishing effort and models of population response to exploitation are comparable to exposure or dose-response assessments of chemicals. The expression of outcomes in terms of future population sizes and yields carries risk characterization several steps further than was done in any of the contaminant studies in the 1983 report.

The most recent developments in hazard identification are in more broadly defined "adverse effects." Ecological risk assessments do not have an equivalent to the lifetime cancer risk estimate used in most health risk assessments. The ecological risks of concern differ qualitatively between different stresses, ecosystem types and locations. The value of avoiding these risks is not nearly as obvious to the general public as is the value of avoiding exposure to established carcinogens. Because few risk managers are trained as ecologists, effective communication between risk managers, technical staff and the public is essential in sound risk management decisions. Stressors of human health are also being identified in categories other than fatal cancers, for example endocrine disruptors or chemicals causing neurotoxicity. Often it is the public, e.g. native people, hunters and mothers, who perceive hazards first and call attention to them.

The second component of the risk assessment process is the evaluation of the dose response relationships. Valid data sets should be presented with the plausible models for extrapolation from high to low dose, and from tests in laboratory species to evaluate hazards and risks in humans. In ecosystems a more generalized relationship between exposure and response must be established. Dose response has medical-human connotations, and also is more correctly applied to chemical pollution. Exposure in an ecological sense may be episodic, such as invasion by exotic species, cyclical, for example, summer recreational boating, or continuous, like the chemical runoff from agricultural land. Exposure usually has a time dimension and a magnitude or concentration dimension. These will modify the assessment of ecosystem harm. The range of hazard potency should be included in risk assessment, and general uncertainties inherent in the assessment reflected.

Response may be direct or indirect. A secondary poisoning of raptors due to accumulation of pesticide residues in their prey, or the effects of harvesting on fish community structure, would be considered indirect adverse effects. The Science Advisory Board's Subgroup on Decisionmaking in the Face of Uncertainty has identified some more subtle indirect effects which are normally understood by scientists before they become apparent to the public, and also before changes are likely to be irreversible. It is the scientist's responsibility to sound an alarm when the probability of any of these effects is in question. A stressor may be of special concern not because its characteristics pose a current risk, but because its pathways or potential host responses do, which are not immediately apparent. For example, although a chemical pollutant appears to be isolated from the biosphere in its current state, economic activities of future generations may unwittingly release it. Furthermore, adverse human reactions may occur only at sensitive points in the human life cycle, for example, the embryonic period, whereas responses are studied for the adult organism. The ecological response may be environmental collapse due to a cascade effect in the environment or food web, whereas the exposure-response examination focuses on the first link in the chain of consequences. Detailed considerations for expanding the list of indirect adverse effects of concern to society are detailed in Appendix I. These adverse effects entail further identification of stressors, detailing uncertainties, and communicating the potential harm to risk managers and decisionmakers, even if the indirect effects are difficult to predict. There is an ethical imperative to explore the long-term consequences of our activities. Social scientists must be involved more directly in the predicting of human interference and/or interruption of predicted behaviours.

The third element of the risk assessment process is exposure assessment. This aspect has taken on a large role in the past few years, in that it focuses on the populations or subpopulations that the data indicate may be particularly exposed. The potential routes of exposure from particular pathways and sources must be identified and the uncertainties and relative importance of the assumptions, exposure models and confidence in the data must be described.

It is now well recognized that exposures vary widely by habitat, niche, food web and other considerations. Humans disperse their foods globally, and exposures to toxics may occur many miles away from the point of harvesting. Species with limited range and who are locally dependent on air, water and food make good sentinels for the local ecosystem. For humans, lifestyle, hobbies and non-traditional uses of the environment also must be considered when assessing exposure. The average adult exposure may prove inadequate to describe the significant response of children or Native people. Social science can identify and quantify these factors. Exposure assessment may be influenced by temperature, acidity, humidity or other factors influencing bioavailability and/or uptake.

Exposure assessment can incorporate the physical, chemical, pharmacokinetic and metabolic data for a chemical mixture polluting an ecosystem. It must consider chemical interactions, environmental transformations (fate, transport and degradation) of complex molecules. Examining biomarkers and biomonitoring data where available and revisiting the list of "adverse effects" of public and scientific concern are all recent developments to expand and clarify this part of the risk assessment process. Biomarkers serve as indicators of exposure and help to elucidate pathways and effects.

Risk characterization is the process of combining and integrating the information and analyses derived from the three stages of risk assessment to describe the likelihood that humans or the ecosystem will experience any forms of toxicity or biophysical harm associated with the hazard. The major components of the risk are presented, along with quantitative estimates, where appropriate, to give a combined and integrated view of the evidence.

Although impact on human health is still a major focus of risk characterization, there is increased recognition that environmental sustainability is essential to human survival and an important end point for human planning. Risk characterization describes the nature and often the magnitude of risk to the ecosystem, including any uncertainties expressed in understandable terms to decisionmakers and the public. Further extension of the risk characterization definition provided in the 1983 U.S. National Academy of Science report is needed to focus on uncertainty, to facilitate expression of risks in management-relevant terms (including valuation), and to emphasize the importance of communication between scientists and managers. Risk characterization synthesizes the results of technical analyses and expresses them in a form suitable for valuation studies or other policy analyses carried out as part of risk management.

Formal analysis of uncertainty is another major subject needing improvement in ecological risk assessments of all types. The "uncertainties" discussion group at the IJC workshop in 1993 identified three general categories of uncertainty that affect all types of risk assessments:

• Measurement uncertainties, e.g. low statistical power due to insufficient observations, difficulties in making physical measurements, inappropriateness of measurements, and natural variability in organic and population response to stress
• Conditions of observation, e.g. spacio-temporal variability in climate and ecosystem structure, differences between natural and laboratory conditions, and differences between tested or observed species and species of interest for risk assessment
• Inadequacies of models, e.g. lack of or knowledge concerning underlying mechanisms, failure to consider multiple stresses and responses, extrapolation beyond the range of observations, and instability of parameter estimates.

From this discussion, and the report summarized in the IJC 1993 workshop on risk assessment, the strengths and weaknesses of this methodology for decisionmaking are clear. The strengths are: (1) risk assessment is designed to organize scientific information specifically for decisionmaking and policy purposes and is rapidly becoming reproducible in its applications; (2) because it is capable of formalizing uncertainties, it can be quantitative in those areas where other approaches are largely qualitative (i.e. where the pathways by which laboratory outcomes are observed can also be expressed in field conditions); (3) it is designed to be prospective and as broad and interdisciplinary as the problem. By design, it complements valuation studies and implementing cost-benefit studies when those are appropriate.

Some weaknesses limit the prospects for applying risk assessment: (1) its applications are generally data intensive, expensive and long-term in nature; (2) it can appear to be dissociated from risk management in the absence of full communication and transparency o the process by which judgements are reached; and (3) it can give the impression that physical and biological sciences are the sole determinants of policy, often in practice neglecting the social sciences and ethics. As in all human undertakings, risk assessment involves human values in the choice of adverse outcomes considered, numbers and types of indirect outcomes included interpretations of data used and other parameters. By making the process as clear as possible to everyone, these value judgements are more apparent to decisionmakers and to the public.

Scientific or Expert Panels

Because decisionmaking should rest on "good science," an interdisciplinary expert panel may be needed to judge the quality and appropriateness of research, mediate arguments between scientists, evaluate negative studies (i.e. those which find no association between an exposure in question and a response of concern) and generally clarify the scientific basis of certain projected findings. It is normal for scientists to disagree, and the inability of a particular research project to reject a null hypothesis is not unusual when dealing with a rare biological effect. These situations are, however, disconcerting for decisionmakers and courts.

Scientific disputes may be a stepping stone to a refined understanding of the phenomenon, to identifying new research needs, to broader recognition of conditions causing effects, or to focusing attention on previously unnoticed pathways, biological mechanisms or biophysical effects. Scientific understanding of our complex world is now expanding into non-linear models and chaos theory, a world less comforting to decisionmakers than that of predictable linear models. Yet these new models appear more realistic to the biophysical interactions of the ecosystem. As with other methods, the use of expert panels has strengths and weaknesses. Generally, scientific panels have greater flexibility than risk assessment as a method for recognizing and accommodating information on the ecosystem impacts in a specific location. Risk assessment looks more to effects of hazards on an idealized ecosystem. The principal strength of expert panels lies in the finding's comprehensiveness and relative immediacy. Unlike either inference of causality or risk assessment, an expert panel can review all relevant information at one point in time, evaluate the uncertainty, and reach findings appropriate for the decision process. It also can identify biomonitoring appropriate to the situation and nature of the uncertainties, which could be implemented when the decision is taken to guard against untoward or unexpected effects. Thus, it can set a course "for now" with proper safeguards for subsequent review of policy and early warning of potential negative outcomes. The weaknesses of the expert panel lie in its tendency to deal with uncertainties qualitatively rather than quantitatively, and with the inherent nature of the selection process. The same findings may not be arrived at by another panel. Reproducibility is the most powerful element of the two quantitative approaches discussed previously, and although it is not lost entirely from an expert panel, it is not an inherent part of the process.

Once the scientific process, inference of causality, risk assessment or expert panel has determined the scientific basis of the decisionmaking together with uncertainties, other disciplines contribute to the decision within the constraints of the legal, social and moral systems of society. Collecting and integrating this further input into decisionmaking can be formal or informal. In Canada, it is evolving as Risk Acceptance. The public tolerance for risk is modified by past experience, by other risks they are currently carrying, by their knowledge, by socio-economic aspects of the problem and by the community's ethical beliefs. Some aspects of this societal input into the decisionmaking process are discussed under the headings of civic science and prudent avoidance. Both are societal attitudes brought to decisionmaking, and which colour the risks which are seen as acceptable and ethical for a particular population at this time in its history.

More formalized inputs into the assessment process, which are currently recognized in Canada, include briefs on the social equity implications of the proposal, the socio-economic impact of the proposal on culture, jobs and future plans of the community, and cumulative impact assessment which analyzes prior harm experienced by the community being asked to assume the risk, or other risks currently being carried by the community. The reality of Risk Acceptance was recognized in the U.S. when setting radiation standards for the people of the Republic of the Marshall Islands who were returning to their Atoll, contaminated earlier by nuclear testing in the Pacific. Radiation Standards more restrictive than those in the United States were used because of the prior damage the population had suffered from nuclear exposure.

While risk characterization tends to be more universal, risk management addresses a particular pr