special section perspectives of the scientiﬁc community on … · 2003-10-13 · sciences and...

SPECIAL SECTIONPerspectives of the Scientific Communityon the Status of Ecological Risk AssessmentEdited by M. POWER*Department of Agricultural EconomicsUniversity of Manitoba,Winnipeg, ManitobaR3T 2N2, Canada

S. M. ADAMSEnvironmental Sciences DivisionOak Ridge National Laboratory,1

Oak Ridge, Tennessee 37831-6038, USA

DEBATE ABSTRACT / Views from a wide variety of practic-ing environmental professionals on the current status of eco-logical risk assessment (ERA) indicate consensus and diver-gence of opinion on the utility and practice of riskassessment. Central to the debate were the issues ofwhether ERA appropriately incorporates ecological and sci-entific principle into its conceptual paradigm. Advocates

argue that ERA effectively does both, noting that much of thefault detractors find with the process has more to do with itspractice than its purpose. Critics argue that failure to vali-date ERA predictions and the tendency to over-simplify eco-logical principles compromise the integrity of ERA and maylead to misleading advice on the appropriate responses toenvironmental problems. All authors felt that many improve-ments could be made, including validation, better definitionof the ecological questions and boundaries of ERA, im-proved harmonization of selected methods, and improve-ments in the knowledge base. Despite identified deficien-cies, most authors felt that ERA was a useful processundergoing evolutionary changes that will inevitably deter-mine the range of environmental problems to which it can beappropriately applied. The views expressed give ERA a cau-tious vote of approval and highlight many of the criticalstrengths and weaknesses in one of our most important envi-ronmental assessment tools.

Charlatan or Sage? A Dichotomy of Views onEcological Risk Assessment

M. Power1 and S. M. Adams.2 1Department of Agricultural Economics,University of Manitoba, Winnipeg, Manitoba, R3T 2N2, Canada;2Environmental Sciences Division, Oak Ridge National Laboratory,(managed by Lockheed Martin Energy Research Corporation for theUS Department of Energy under contract number DE-AC05-96OR22464), Oak Ridge, Tennessee 37831-6038, USA.

Global climate change, reductions in biodiversity,and the potential implications of pesticide and toxicchemical releases have all raised public awareness ofecological issues in the last decade. Established toxicol-ogy, ecology, and fisheries journals have devoted increas-ing space to discussion of impact related issues. No lessspectacular has been the profusion of new journalsdedicated to various aspects of assessing the impacts ofanthropogenic activities on the environment. Theseinclude Environmental Modeling & Assessment, Humanand Ecological Risk Assessment, Ecotoxicology, and the

Journal of Aquatic Ecosystem Stress and Recovery. Central tomany of these journals, their readers, and their authorsare issues relating to the use, abuse, and development ofthe practice of ecological risk assessment.

The respectability and fashion of risk assessment as ameans for evaluating scientific information on thepotential adverse environmental effects of physical andchemical stressors was sanctified in 1992 with therelease of the Framework for Ecological Risk Assess-ment by the US Environmental Protection Agency (USEPA 1992). Broadly speaking, ecological risk assessmentcan be viewed as an exercise in environmental problemsolving. It is a systematic means of assessing the state ofecological resources and determining remediative pri-orities for the myriad of potential ecological problemshuman action has caused. More precisely, ecologicalrisk assessment is defined as the use of available toxico-logical and ecological information to estimate theprobability of an unwanted ecological event occurring(Wilson and Crouch 1987).

Regulatory and scientific interests have both playedimportant roles in the development of risk assessmenttechniques. Regulatory interest has focused on legalissues surrounding the need to complete assessmentsand the design of management protocols. Science hasdefined the kinetics and potential impact of stressors

KEY WORDS: Ecological risk assessment; Ecology; Probability

1Managed by Lockheed Martin Energy Research Corporation for theUS Department of Energy under contract number DE-AC05-960R22464.

*Author to whom correspondence should be addressed.

Environmental Management Vol. 21, No. 6, pp. 803–830 r 1997 Springer-Verlag New York Inc.

through the completion of hazard, effects, and expo-sure assessments. Regulatory interest, resulting in legis-lation, has forced decisions on the use and disposal ofcontaminants. Accordingly, as Bartell and others (1992)have argued, it is prudent to develop and evaluatescientifically defensible methods that can be employedby environmental decision makers and regulators. Un-doubtedly, a general framework for ecological riskassessment would have heuristic value, despite theunique scientific questions and regulatory issues likelyto be raised in each case.

To its advocates risk assessment is a rigorous form ofassessment that uses formal quantitative techniques toestimate probabilities of effects on well-defined end-points, estimate uncertainties, and partition the analysisof risks from discussions concerning their significance,or choice, of appropriate remediative action (Suter1993). To its detractors, however, risk assessment is littlemore than a jumble of technical jargon designed toexpedite the degradation of natural environments.Slightly less extreme views characterize it as a means offacilitating the acceptance of the views of the technicalelite by the public under the guise of scientific objectiv-ity (Lackey 1994). Clearly, there is much that divides thetwo views of risk assessment. This, in part, may reflectthe fact that scientists and analysts are not effectivelyconveying ecological options to decision makers or thepublic. Confusion leads to mistrust, and mistrust toridicule. If risk assessment techniques are ultimately tobe successful and contribute to the continued refine-ment of environmental guidelines, then scientificallycredible methods must be developed and debated. Therate at which the methodology is developing, anddiscussions concerning the utility of its applicationwithin the wider context of social decision making,suggest there is need and merit in critically assessing thecurrent state of the art in ecological risk assessment.

Permeating much of the discussion on risk assess-ment is the technocratic view of objectivity. This viewsrisk assessment as objective in the sense that differentrisks may be evaluated using a standard approach.Furthermore, it claims that, at least at the stage ofcalculating the probabilities associated with hazards andtheir effects, risk assessment is completely objective,neutral, and value free. Many, however, still view riskanalyses as being value laden (Shrader-Frechette 1991).Assessors must make value judgements about whichdata to collect, how to simplify available facts intosimple models, how to extrapolate because of un-knowns, how to select the statistical tests to be used, howto select sample size, and which exposure–responsemodel to employ.

Even within the scientific community there is discus-sion concerning the confidence that can be placed inobtained estimates. Incomplete understanding of eco-logical systems combined with the uniqueness of indi-vidual systems (Loucks 1985) implies much uncertainty.If, however, the ultimate goal of risk assessment ispredictability, then there must be a good understandingof all possible ecological outcomes and probabilitiesassociated with any perturbation. Although practitio-ners are quick to claim that an important feature of riskanalysis is the explicit consideration of uncertainties inthe analyses, questions remain about how the unknowncan be adequately expressed as a probability in the finalestimate of effect. Calow (1994) maintains that we stilldo not know enough about ecological systems to be ableto identify what it is we want to protect about them andhence, what we should be measuring. Ludwig andothers (1993) point out that scientific understanding ishampered by lack of controls and replicates to the pointwhere each new problem involves learning about a newsystem.

We are left, then, to question a central tenet ofecological risk assessment: probability. Probability istypically defined as the portion of times a given out-come is observed in a series of identical, repeatableexperiments. Without probability, the distinction be-tween risk assessment and other assessment methods(e.g., environmental impact assessment) becomes murky.Without clear knowledge of what we should measure,how do we select or define endpoints? And withoutendpoints, or replication, how do we define probability?The fact that available data indicate a low frequency ofobserved effects does not prove a low probability ofoccurrence if each observation cannot be viewed as theproduct of an identical, repeatable experiment. With-out repeatability, the statistical foundations of riskassessment appear weak and the scientific foundationsare almost certainly questionable. Protocols whose re-sults cannot be readily replicated will inevitably laborunder the same suspicions as the alchemists of old.

The above observations raise the obvious questionsas to what lack of knowledge about ecological systemsand lack of replications imply for ecological risk assess-ment. Since much of the information incorporated intoassessments is demonstrably not of acceptable probabi-listic character, is ecological risk assessment abusingstatistics with wanton disregard for the principles andassumptions of the practice? Furthermore, can theproblem be remedied, given the definitional impossibil-ity of duplication in ecological systems? If not, do wecondemn the practice of risk assessment as a charlatanpseudoscience intent more on the enunciation of frame-works than on improving scientific practice? Or do we

M. Power and S. M. Adams804

regard ecological risk assessment as delivering the toolsnecessary to describe, priorize, and ultimately addressthe many complicated environmental issues that liebefore us?

Literature CitedBartell, S. M., R. H. Gardner, and R. V. O’Neill. 1992.

Ecological risk estimation. Lewis Publishers, Boca Raton,Florida, 252 pp.

Calow, P. 1994. Ecotoxicology: what are we trying to protect?Environmental Toxicology and Chemistry 13:1549.

Lackey, R. T. 1994. Ecological risk assessment. Fisheries 19:14–18.

Loucks, O. L. 1985. Looking for surprise in managing stressedecosystems. Bioscience 35:428–432.

Ludwig, D., R. Hilborn, and C. Walters. 1993. Uncertainty,resource exploitation and conservation: lessons from his-tory. Science 260:17–36.

Shrader-Frechette, K. S. 1991. Risk and rationality: Philosophi-cal foundations for populist reforms. University of Califor-nia Press, Berkeley, California, 312 pp.

Suter, G. W., II. 1993. Ecological risk assessment. LewisPublishers, Boca Raton, Florida, 538 pp.

US EPA (United States Environmental Protection Agency).1992. Framework for ecological risk assessment. EPA/630/R-92?001. Washington, DC.

Wilson, R., and E. A. C. Crouch. 1987. Risk assessment andcomparisons: an introduction. Science 236:267–270.

Science and Subjectivity in the Practice ofEcological Risk Assessment

P. Calow1 and V. E. Forbes.2 1Department of Animal and Plant Sciences,University of Sheffield, Sheffield S10 2TN UK; 2Department of LifeSciences and Chemistry, Roskilde University, P.O. Box 260, DK-4000Roskilde, Denmark.

As highlighted in the introduction to this debate,much of the disagreement surrounding risk assessmentdepends on whether it is viewed as inherently anobjective or a subjective process. Objective means set-tling issues using scientific procedures of critical analy-sis and carefully controlled observation, and subjectivemeans settling issues by an appeal to human values.However, it is our main thesis that presenting the debateas polarized between these extremes obscures the poten-tial value that both objectivity and subjectivity can addto the risk assessment process. In particular, this paperfocuses on the following points: (1) risk assessment, as aprocedure, can be defined unambiguously and objec-

tively; (2) scientific procedures can be used to providethe information and understanding (criteria) neededto carry out risk assessment if it is known a priori what isbeing protected; (3) defining what it is that we want toprotect may not (and perhaps should not) be entirely,or even mainly, something that can be decided on thebasis of scientific criteria especially as far as ecologicalsystems are concerned; and (4) risk assessment repre-sents the application of science in situations wherethere is much uncertainty. Estimating the statisticalconfidence with which effects can be detected (givenfixed experimental designs) can be a substantial aid tothe risk assessment.

Defining Risk Assessment Objectively

Risk is defined as the likelihood that the potential tocause harm inherent in some substance, process, oractivity is realized in some ‘‘real-world’’ situation. As faras commercial chemicals and pesticides go, estimatingrisk involves combining an understanding of theirpotential to cause harm to a target (hazard identifica-tion) with an understanding of the likelihood andextent of that harm being realized (fate and exposure).The elements in this kind of risk assessment are illus-trated in Figure 1. The goal of risk assessment in thiscontext is to determine those situations in which thelikely exposure concentration overlaps with the concen-trations likely to cause biological effects.

Clearly, the most important requirement for a riskassessment is knowing what it is that we want to protect

KEY WORDS: Extrapolation; Risk quotients; Ecosystem health; Eco-system service; Statistical errors

Figure 1. Risk assessment of chemicals compares likely expo-sure and effect distributions. The variability in each derivesfrom biological sources (effects) and the physico-chemicalcomplexity of the environment (exposure). Though both arerepresented as symmetrical distributions, they need not be so.A probability of effect is computed from the extent of overlapbetween the two distributions.

Status of Ecological Risk Assessment 805

and hence calculate a risk for, i.e., that we have a clearlydefined target and clearly defined features of the targetthat we want to keep in some preferred state (Calow1994). Harm is defined in terms of a shift away from thepreferred state. For humans, the preferred state isstraightforwardly quantified in terms of morbidity andmortality.

The situation is much less straightforward for ecologi-cal systems because of the amount of uncertaintyassociated with the selection of preferred states. Inpredicting the nature and extent of harm to ecosystems,uncertainties arise from errors, ignorance, and thestochastic nature of natural systems. Risk assessmentsattempt to quantify these uncertainties, either from firstprinciples or from observation, and package them intoprobability statements of effect. Commonly a few, orpossibly none, of the uncertainty elements can bespecified, and they remain more or less hidden in therisk assessment. Consequently, risk assessments are of-ten expressed in terms of risk quotients. These give onlygeneral indications of the extent to which ecologicalharm is likely to occur as most of the uncertaintyremains concealed.

Difficulties of Defining Targets forEcological Systems

For individual organisms it is relatively straightfor-ward to define optimum states. Organisms exist asrecognizably organized and coherent systems, are persis-tent, and have specific properties and processes thatcontribute to this organization. Because individual or-ganisms generally act as the units of natural selectionand are composed of cooperating biochemical, physi-ological, and behavioral systems that have evolvedhomeostatic control, differences in fitness among organ-isms can be used as criteria for defining optimum states(Sibly and Calow 1986). The criteria used for assessingrisks to individuals (e.g., humans) from commercialchemicals are survival (an important component ofDarwinian fitness) and the properties and processesthat contribute to it. Emphasis is thus appropriatelyplaced on the effects of chemicals on individual perfor-mance or health. Identifying risk assessment criteria isless straightforward for ecological systems for they areless obviously organized and coherent, cannot generallybe considered as units of selection, and cannot unam-biguously be described by terms such as ‘‘health’’(Forbes and Forbes 1994, Calow 1995). As Suter (1993)and others have pointed out, ecosystems do not necessar-ily have clear boundaries, they do not have consistentstructures from one individual example to the next,

they do not develop in a consistent and predictablemanner, and they do not have mechanisms like theneural and hormonal systems of organisms to maintainhomeostasis. In short, ecosystems are not superorgan-isms. Furthermore, the relationship between ecosystemprocesses and the organisms of which they are com-posed remains unclear, although several testable hypoth-eses have been proposed (Lawton 1994).

Despite the difficulties in identifying appropriateecosystem targets, in practice ecological risk is oftendefined in terms of the probability of: (1) mortality/morbidity in a fraction of a population; (2) extinction ofa species or fraction of a species; or (3) loss of a certainfraction of ecosystem process and/or ‘‘service.’’ Thereare two possible views on the extent to which thesetargets can be defined objectively: the ecosystem healthand the ecosystem services paradigms. In the ecosystemhealth paradigm, ecological systems have a coherenceand integrity that depend upon certain states. It is forecology to define these states and for ecological riskassessment to be expressed in terms of them (Calow1995, cf. Rapport 1989). In the ecosystem serviceparadigm, ecological systems do not have intrinsiccoherence (Calow 1995) but what human society getsfrom them does depend on their structure and function(Ehrlich and Wilson 1991). The ecosystem health para-digm requires that ecology and ecologists play a domi-nant role in not only carrying out risk assessments butin defining the criteria by which risks are evaluated. Theecosystem service paradigm suggests that social needsand values are as important as scientific ones and soemphasizes the need for interaction between scientistsand society at large in defining ecological risk criteria.

Whereas both of these positions may be valid indifferent circumstances, deciding which of them ismore useful for any given risk assessment will depend,among other things, on the ecological level beingaddressed. Thus, if the requirement is to protect particu-lar populations, the criterion of persistence is impor-tant, and the critical variables and the values that theymust take for long-term stability are defined by thetheory of population dynamics. On the other hand,deciding which species should be protected and whatlevels of ecosystem processing are desirable may not beso decisively defined even if we understood enoughabout the structure and functioning of ecosystems andthe relationship between the two. The reason is that, asdiscussed above, ecosystems do not operate as unitarywholes but rather as dynamic collections of speciespopulations without goal states, except insofar as theyare defined in terms of our human needs and desires.


Ecological Risk Assessment Is Pragmatic

There are two major areas of ecological uncertaintyin the practice of risk assessment as presently con-ducted. The first is the uncertainty in predicting effectson organisms in field situations from their responses incontrolled laboratory test systems. The second is theuncertainty about the roles of different species inecosystems and the point at which damage to individualspecies or populations can be considered as damage tothe community or ecosystem. Given targets, a riskassessment can be carried out objectively on the basis ofscientifically defined criteria and procedures. If weknow what population, species, combination of species,or ecosystem process we want to protect, then inprinciple we can use the scientific method to definewhich variables are important and within which limitsthey should be maintained. However, uncertaintiesarise out of our rudimentary understanding of ecologi-cal structures and processes and the extent to whichappropriate states can, and indeed should, be definedin ecological terms.

Accordingly, the current approach to ecological riskassessment is to define circumstances likely to be gener-ally protective for an ecosystem based on generalizedecological responses such as survivorship and fecundityin representative systems. The most common approachinvolves applying arbitrary uncertainty factors to thelowest observed effect concentrations [to calculate pre-dicted no-effect concentrations (PNECs)] and usingworst-case assumptions to estimate predicted environ-mental concentrations (PECs). A risk assessment is thencarried out by dividing PEC by PNEC. A ratio of lessthan 1 does not give a precise probability of effect, onlyan indication that the likelihood of effect is very low.The approach, therefore, masks some of the uncertainty.

Another approach for extrapolating effects of chemi-cals from laboratory test systems to field situations is tofit toxicity data for a few (e.g., 3–11) laboratory testspecies to an assumed distribution function (e.g., log-normal, log-logistic, or log-triangular) and use theresulting curve to estimate the chemical concentrationat which most (usually 95%) of the species in anecosystem will be protected (Stephan and others 1985,Aldenberg and Slob 1991, Wagner and Loekke 1991).In other words, it provides the basis of a probabilisticrisk statement and gives the impression of uncoveringsome of the uncertainty. However, like the quotientapproach, the distribution-based extrapolation modelsinvolve a number of, as yet, untested assumptions suchas that the sensitivities of laboratory test species providean unbiased measure of the sensitivity distribution ofspecies in natural communities, that species interac-

tions can be ignored, and that protecting speciescomposition protects ecosystem processes. Comparisonof the predictions of the distribution-based extrapola-tion models with those derived from employing arbi-trary uncertainty factors indicates that the former donot appear to result in significantly improved riskassessments (Forbes and Forbes 1994).

Avoiding False Negatives as Wellas False Positives

The design and analysis of scientific experiments hastraditionally focused on minimizing the incidence offalse positives (i.e., of concluding that there is an effectwhen, in fact, there is none; in statistical terminology, ofmaking a type I error). Thus significance criteria aretypically set very low and a relatively large effect isrequired before it can be detected statistically. However,to maximize the probability of protecting the environ-ment, it is more appropriate to guard against falsenegatives (i.e., concluding that there is no effect when,in fact, there is; making a type II error). Experimentaldesigns and statistical analyses need to be adjustedaccordingly when minimizing the probabilities of falsenegatives (Forbes and Forbes 1994). An extreme ver-sion of this strategy suggests that in environmentalprotection decisions the precautionary principle shouldreplace risk assessment altogether (Wynne and Mayer1993). This is both unrealistic and unnecessary. Precau-tion can be used in the application of risk assessment,for example, in establishing levels of statistical signifi-cance to take account of type II errors. Improving thestatistical power of test designs to detect existing effectsso that both type I and type II error rates are minimizedcan greatly strengthen the risk assessment process(Cranor 1993).

The Present System of Ecological RiskAssessment Is Not Satisfactory

Perhaps the biggest weakness, then, in the presentpractice of risk assessment is in defining the criticalquestions that need to be addressed. Part of this is achallenge for science to define natural systems and theirnatural states, but part is a more general challenge ofdefining what it is about these natural systems that wevalue. We need more explicit interaction between scien-tists and society to clarify these questions (Calow 1997).Once the questions are refined, scientists ought to beable to develop risk assessment procedures that areboth rigorous and transparent. However, this will not beachieved by measuring as many variables in as many


ecosystems as we can, but rather by defining clearhypotheses that can be tested rigorously in well-designed research programs. Although ecosystems intheir entirety may be unique and hence unrepeatable inspace and time, many of their component parts can bereplicated and can, therefore, lend themselves to experi-mentation and statistical analysis. Given our presentlack of knowledge, it would be wise to develop flexiblerisk assessment approaches that are responsive to newinformation. As we learn more about ecosystems andabout what we want and/or need to protect, bothscientific objectivity and informed value judgementshave key roles to play in improving the practice ofecological risk assessment.

Literature CitedAldenberg, T., and W. Slob. 1993. Confidence limits for

hazardous concentrations based on logistically distributedNOEC data. Ecotoxicology and Environmental Safety 25:46–63.

Calow, P. 1994. Ecotoxicology: What are we trying to protect?Environmental Toxicology Chemistry 13:1549.

Calow, P. 1995. Ecosystem health—a critical analysis of con-cepts. Pages 33–41 in D. J. Rapport, C. L. Gaudet, and P.Calow (eds.) Evaluating and monitoring the health oflarge-scale ecosystems. NATO ASI Series I, Vol. 28. Springer-Verlag, Berlin.

Calow, P. 1997. Management decisions before assessment inconsidering environmental risks for commercial chemicals:some observations from standardisation of ecotoxicologicaltests. In R. Bal and W. Halffman (eds.), The politics ofchemical risks: Scenarios for a regulating future. KluwerAcademic Publishers, Dordrecht, The Netherlands (in press).

Cranor, C. F. 1993. Regulating toxic substances. OxfordUniversity Press, Oxford, 252 pp.

Ehrlich, P. R., and E. O. Wilson. 1991. Biodiversity studies:science and policy. Science 253:758–762.

Forbes, V. E., and T. L. Forbes. 1994. Ecotoxicology in theoryand practice. Chapman and Hall, London, 247 pp.

Lawton, J. H. 1994. What do species do in ecosystems? Oikos71:367–374.

Rapport, D. J. 1989. What constitutes ecosystem health?Perspectives Biology Medicine 33:120–132.

Sibly, R. M., and P. Calow. 1986. Physiological ecology ofanimals. Blackwell, Oxford.

Suter, G. W., II. 1993. A critique of ecosystem health conceptsand indexes. Environmental Toxicology Chemistry 12:1533–1539.

Stephan, C. E., D. I. Mount, D. J. Hanson, J. H. Gentile, G. A.Chapman, and W. A. Brungs. 1985. Guidelines for derivingnumeric national water quality criteria for the protection ofaquatic organisms and their uses. PB85-227049. US Environ-mental Protection Agency, Duluth, Minnesota.

Wagner, C., and H. Loekke. 1991. Estimation of ecotoxicologi-cal protection levels from NOEC toxicity data. Water Research25:1237–1242.

Wynne, B., and S. Mayer. 1993. How science fails the environ-ment. New Scientist 138:33–35.

Ecological Risk Assessment: Use,Abuse, and Alternatives

R. T. Lackey. National Health and Environmental Effects ResearchLaboratory, US Environmental Protection Agency, 200 SW 35th Street,Corvallis, Oregon 97333, USA.1

Debates over the role of ecological risk assessmentshould start with agreement on a definition. Much ofthe confusion and divisiveness over risk assessmentapplied to ecological problems is caused by using thesame terms but attributing different meanings to them.Indeed, in the classical applications of risk analysis(automobile, life, and health insurance; industrial fail-ures; natural disasters; accident prevention; etc.), thedefinition, if not the practice, of risk assessment isdifferent from that typically used in ecological riskassessment. The concept of ecological risk assessment Irefer to here is best defined along the lines of ‘‘. . . a wayof examining risks so that they may be better avoided,reduced, or otherwise managed’’ (Wilson and Crouch1987); ‘‘. . . a process to evaluate the likelihood ofundesirable effects on ecological receptors from expo-sure to one or more stressors . . . ’’ (Regens 1995); and‘‘. . . a series of questions directed to the available datato analyze the expected risk’’ (Patton 1995). There aremany other risk assessment concepts, but these arebeyond the discussion here: human health risk assess-ment, comparative risk assessment, engineering riskassessment, environmental risk assessment, risk commu-nication, risk regulation, risk reduction, risk allocationor justice issues, and a suite of decision making para-digms.

Opinions on the use of ecological risk assessment inpublic policy analysis are diverse; they range frompositive: ‘‘. . . scientifically credible evaluation of theecological effects of human activities’’ (Suter 1993) tosceptical: ‘‘. . . one more hurdle on the road to apermit’’ (Webster and Connett 1990) to dismissive: ‘‘. . .risk assessment is a sham . . .’’ (Merrell 1995). Themiddle ground is populated by a disjointed array of

1The views expressed here do not necessarily reflect those of the USEnvironmental Protection Agency or any other organization.

KEY WORDS: Ecological risk assessment; Risk decision making;Risk management; Ecological consequence assess-ment


opinions because the debate over the proper role ofecological risk assessment defies the simplistic categori-zation of right versus left, conservative versus liberal,technocratic versus democratic, and green versus bal-anced use. Neither the debate nor the tool is new, butthe intensity of the debate has increased as ecologicalrisk assessment has become the policy tool of choice insome organizations (Lipton and others 1993, Lackey1994, 1995, Regens 1995).

Some critics contend that risk assessment is ‘‘. . .deeply flawed and subject to abuse’’ (Montague 1995).Its use to help resolve public policy on human healthissues is equivalent to ‘‘premeditated murder’’ (Merrell1995). Some people will die or suffer disease by a policydecision when society makes a decision as to how manydeaths will be tolerated for the benefits obtained. Whenused in making policy on ecological issues, the users ofrisk assessment have accepted a form of ecologicaltriage, administratively deciding which individuals, popu-lations, or species will live and which will die (Montague1995).

Another type of criticism is the contention thatecological risk assessment as currently practiced isnothing more than ‘‘the paradigm of human health riskassessment, laying on an underlying, unsophisticated,ecological veneer’’ (Karr 1995). While not morallyrepugnant to such critics, many current applications ofecological risk assessment are inappropriate. Further,such a position questions the assertion that ecological‘‘health’’ is analogous to human ‘‘health’’ and, thus,that the approaches and tools of human health riskassessment can be easily adapted to ecological problems(Menzie 1995).

Ecological risk assessment is not without strongsupporters (Suter 1993). Generally, supporters tend tocome from regulatory agencies and much of the regu-lated community. Some governmental agencies arestrongly supportive, to the point of implementing poli-cies and guidelines (Patton 1995). Much of the scien-tific and engineering infrastructure has embraced theapproach. ‘‘How-to’’ training courses and symposia areplentiful. Books and technical papers explain in detailthe technical intricacies of conducting ecological riskassessments (Suter 1993, Molak 1996).

Other advocates of ecological risk assessment, al-though far from effusive, acknowledge a useful role:‘‘Risk assessment is a very flawed tool, but, given thecurrent process of policy formulation, it has a legitimaterole to play’’ (Funke 1995). In an ideal application,ecological risk assessment ‘‘. . . should be a purelytechnical analysis driven by scientifically acquired data,and free from social bias’’ (Fairbrother and others1995).

What is the appropriate use of ecological risk assess-ment? When is its use inappropriate? How is it misused?What are the alternatives? Is it underused, and would itsincreased use bring rationality to public policy? I willargue here that ecological risk assessment: (1) has alegitimate, appropriate, but limited role in science,policy analysis, and policy implementation; (2) is oftenmisused, and in some circumstances abused, bothnaively or intentionally; and (3) is not the only ap-proach and others should be seriously evaluated.

Appropriate Use

Risk assessment has been used effectively in manyfields as an aid in decision making. It is used to estimatethe likelihood of an event (i.e., automobile fatality,flood, nuclear accident, etc.), clearly recognized asadverse. Its typical use in decision making with regard toecological issues is similar: estimating the likelihood of acertain, defined event occurring (e.g., what are likelymortality consequences to biota of the use of a particu-lar chemical?). The key requirement is that the conse-quence be adverse by definition, which enables theanalyst to conduct the risk assessment (Bartell andothers 1992, Lackey 1994, 1996a). In classical riskassessment the adverse consequence is easy to identify: anuclear accident is universally accepted as adverse, as isan automobile crash, a skiing accident, a heart attack, oran airplane crash. The analog in ecological risk assess-ment is less clear.

To be technically tractable and credible, the riskproblem must be defined in fairly narrow terms. Evenwhen defined in fairly narrow terms, the analysis may betechnically complex and require sophisticated scientificinformation. Often the narrowing results from legisla-tive policy mandate [e.g., Comprehensive Environmen-tal Response, Compensation, and Liability Act and itsimplementation (Lipton and others 1993, Friant andothers 1995)]. The risk problem then becomes rela-tively simple analytically [e.g., one or a few chemicalsare the stressor causing effects on a few biologicalcomponents; the effects, if present, are adverse bydefinition (Regens 1995)]. Simplification, of course,begs the question of whether analysis leads to goodpolicy options, but it does give the analyst a toe hold onwhat is desired or adverse (Friant and others 1995).Another way to state this perspective is that risk assess-ments assume a definition of health conditions. (Funke1995).

Although beyond the scope of the debate here, anobvious potential role of risk assessment is to helpallocate scarce resources. Risks need to be managed inour personal lives, companies, and government. How


does one compare risks? Potentially, risk assessmentoffers an approach for comparing, rather than measur-ing, individual risks. Such an approach has been used bythe EPA Science Advisory Board (US EPA 1990). Thefour ecological problems with highest risk were habitatalteration, decrease in biological diversity, stratosphericozone depletion, and global climate change. Risks suchas herbicides, pesticides, toxic chemicals, and aciddeposition were medium-risk problems. Oil spills,groundwater pollution, radionuclides, acid runoff, andthermal pollution were relatively low-risk problems. Anobvious use of risk assessment would be as one tool tohelp allocate research and regulatory efforts from lowerto higher risk problems. Risk, of course, is only onefactor in allocating resources, so ecological risk assess-ment should be only one element in the decision-making process.

Whether ecological risk assessment is used to addressrelatively narrow, technical questions [e.g., toxicologi-cal effects of a chemical on a particular biotic compo-nent (Friant and others 1995)] or to allocate scarcegovernment resources (i.e., comparative risk assess-ment), it is critical to recognize that risk assessment ismerely a tool in the decision-making process. At best,and if used properly, it is a tool that can assist inpresenting the likely consequences of various decisionalternatives (Woodhouse 1995).

Abuse

Concerns about abuse in ecological risk assessmentoften revolve around the contention that the tool canbe used to support a predetermined policy position(Merrell 1995). The metaphor often used to illustratethis concern is that of the tortured prisoner. A torturedprisoner will confess to any crime; the confessions areonly limited by the creativity and persistence of theprison guards. The allegation about risk assessment andits practitioners is: given enough creativity, virtually anypolicy position can be supported by risk assessment.The most common allegation is that the policy ques-tions are formulated in a way that will produce virtuallyany result and that the result has the aura of scientificacceptability (Montague 1995). Those who know themost about how to manipulate the procedures controlthe discourse, the questions asked, and how they areanswered (Funke 1995, O’Brien 1995).

Another potential abuse is that technocrats andpoliticians will define risk problems in ways that can besolved technically but have little real relevance to thepublic policy issue. The metaphor often used is that of arisk assessor looking for his lost keys under the streetlamp. Although the keys were most likely lost far from

the street light, the risk assessor laments that he haslittle chance of finding the keys in the dark so why wastetime looking there. Although this makes a humorousstory, ecology is complex and our understanding islimited. There is a strong tendency to define problemsin ways that we can handle scientifically, even thoughthe formulation may be policy irrelevant (O’Brien1995).

A potential abuse of ecological risk assessment is toapply the same analytical tool to every problem: if youronly tool is a hammer, every problem must be a nail. Ifevery ecological policy problem is viewed from a riskperspective, then it is not surprising that technocratswill try to force a fit. The useful role of ecological riskassessment is to help solve fairly narrow, well-definedtechnical questions, not to answer larger, more complexpublic policy questions.

One of the most serious types of misuse by techno-crats is to substitute their values and priorities for thoseof the public or their elected representatives (Websterand Connett 1990, Menzie 1995). In philosophicalterms, this is illustrated by shifting the scientific ‘‘is’’ tothe political ‘‘ought.’’ In science there are no ‘‘oughts.’’Animals, plants, and ecosystems are neither good norbad, better or worse, or healthy or sick unless a valuecriterion is applied. ‘‘Risk’’ has no definition in ecologyunless someone defines what is adverse (or healthy).Whether the introduction of wheat, horses, or zebramussels to North America is good or bad depends onthe value criteria applied, not the personal opinions ofscientists and risk assessors.

Alternatives

One obvious alternative to ecological risk assessmentis a modification to drop the idea of risk. Some havereferred to this as benefits analysis, where the desiredbenefits are selected by the political process (Principe1995). Others refer to consequence analysis, whichsimply assesses ecological consequences without defin-ing good or bad, adverse, ecological health, or risk(Lackey 1996b). Eliminating the concept of risk inecological policy problems reduces the number ofvalue-based choices scientists and assessors must make;thus more choices are reserved for democratic pro-cesses through accountable decision makers. In somecases it may be that analysts are actually conductingsomething closer to ecological consequence analysisthan ecological risk analysis.

Another, related approach is ‘‘ecological alternativesassessment’’ (O’Brien 1995, Pagel 1995). Alternativesassessment steps back even further from risk assessmentand focuses on the questions being asked. Critical


public policy questions would not be buried in technicalanalysis. As with other analytical tools, whoever controlsthe questions asked in risk assessment constrains thepolicy options under consideration (Funke 1995).

The old concept of benefit–cost analysis, whichsuffered, justifiably, many of the same criticisms nowleveled against ecological risk assessment (Schrecker1991), is potentially a viable alternative to ecologicalrisk assessment. Benefit–cost analysis is much closer to adecision-making framework than is risk assessment.Although, it is an illusion that public policy questionscan be reduced to a single metric of value (money), thebasic concept of trade-offs has appeal. Although fraughtwith analytical difficulty, benefit–cost analysis is closer tothe kind of information decision makers actually needfrom technocrats, but it is subject to the same distortionby improper use of personal values as is ecological riskassessment.

Conclusion

Over the past few decades there has been a diversearray of tools and paradigms offered as the solution toecological policy and decision-making problems. Isecological risk assessment just another tool that willfollow the fate of computer-based modeling, geo-graphic information analysis, management by objec-tives, total quality management, management reengi-neering, Delphi, and organizational reinvention? Eachburst on the scene and was advocated by some with nearreligious zeal, only to fall from favor in lieu of and bereplaced by another, newer tool. Each did eventuallyassume a useful role as a tool appropriate for someproblems under some circumstances but not as anoverall panacea (Lackey 1996a,b).

Ecological risk assessment is a tool that might helpresolve some kinds of public policy questions, althoughpolicy questions must be fairly narrowly defined andexplicitly described to be analytically tractable. Thisdoes not mean that they are simple problems, merelyanalytically tractable. The most common use of ecologi-cal risk assessment will continue to be with questionswhere the definition of ecological adversity is providedby legislation, policy, or arbitrarily assumed by theanalyst. Its potential for addressing more complex, andrelevant, ecological policy questions has yet to be fullyevaluated.

A serious allegation is that ecological risk assessment(along with many other technocratic tools) provides aroute to subvert the democratic process (e.g., how doesthe public decide which ecological changes are adverseand which are beneficial?). The very nature of theprocess requires the analyst to make many value-based

decisions, hence the charge that the process is elitist byits very nature (O’Brien 1995). In fact, the decision touse risk assessment is a heavily value-laden decision.Technical expertise cannot substitute for values andpriorities in ecological risk assessment; these are issuesof policy, not science.

Ecological systems are complicated and the ecologi-cal consequences of decisions are often difficult for theecologist to grasp, much less those without this exper-tise. Furthermore, analysts and scientists involved inecological risk assessment are using a tool that most ofthe public does not comprehend; thus the danger ofmisuse through miscommunication, unintentional orotherwise, is great. There is no scientific or ecologicalimperative in risk assessment; it is only a tool to helpthose charged with making decisions evaluate options.

Those of us involved in ecological research andassessment should remember that risk assessment is thelatest of a large number of tools and approaches thathave played on the scientific and management stagewith great fanfare. These tools and approaches haveeach evolved to be useful in certain, clearly definedcircumstances. Ecological risk assessment is undergoinga similar transformation.

Literature CitedBartell, S. M., R. H. Gardner, and R. V. O’Neill. 1992.

Ecological risk estimation. Lewis Publishers, Chelsea, Michi-gan, 252 pp.

Fairbrother, A., L. A. Kapuska, B. A. Williams, and J. Glicken.1995. Risk assessment in practice: Success and failure.Human and Ecological Risk Assessment 1:367–375.

Friant, S. L., G. R. Bilyard, and K. M. Probasco. 1995.Ecological risk assessment—is it time to shift the paradigm?Human and Ecological Risk Assessment 1:464–466.

Funke, O. C. 1995. Limitations of ecological risk assessment.Human and Ecological Risk Assessment 1:443–453.

Karr, J. R. 1995. Risk assessment: We need more than anecological veneer. Human and Ecological Risk Assessment1:436–442.

Lackey, R. T. 1994. Ecological risk assessment. Fisheries 19:14–18.

Lackey, R. T. 1995. The future of ecological risk assessment.Health and Ecological Risk Assessment 1:339–343.

Lackey, R. T. 1996a. Is ecological risk assessment useful forresolving complex ecological problems? Pages 525–540 inD. J. Stouder, P. A. Bisson, and R. J. Naiman (eds.), Pacificsalmon and their ecosystems: Status and future options.Chapman and Hall, New York.

Lackey, R. T. 1996b. Ecological risk assessment. Pages 87–97 inV. Molak (ed.), Fundamentals of risk analysis and riskmanagement, CRC Press, New York.

Lipton, J., H. Galbraith, J. Burger, and D. Wartenberg. 1993. Aparadigm for ecological risk assessment. Environmental Man-agement 17:1–5.


Menzie, C. A. 1995. The question is essential for ecological riskassessment. Human and Ecological Risk Assessment 1:159–162.

Merrell, P. 1995. Legal issues of ecological risk assessment.Human and Ecological Risk Assessment 1:454–458.

Molak, V. 1996. Fundamentals of risk analysis and risk manage-ment. CRC Press, New York, 472 pp.

Montague, P. 1995. Making good decisions. Rachel’s Environ-ment and Health Weekly, No. 470, 30 November.

O’Brien, M. H. 1995. Ecological alternatives assessment ratherthan ecological risk assessment: Considering options, ben-efits, and dangers. Human and Ecological Risk Assessment1:357–366.

Pagel, J. E. 1995. Quandaries and complexities of ecologicalrisk assessment: Viable options to reduce humanistic arro-gance. Human and Ecological Risk Assessment 1:376–391.

Patton, D. E. 1995. The U.S. Environmental Protection Agen-cy’s framework for ecological risk assessment. Human andEcological Risk Assessment 1:348–356.

Principe, P. 1995. Ecological benefits assessment: A policy-oriented alternative to regional ecological risk assessment.Human and Ecological Risk Assessment 1:423–435.

Regens, J. L. 1995. Ecological risk assessment: Issues underly-ing the paradigm. Human and Ecological Risk Assessment1:344–347.

Schrecker, T. 1991. Risks versus rights: Economic power andeconomic analysis in environmental politics. Pages 265–284in D. C. Poff and W. J. Waluchow (eds.), Business ethics inCanada, 2nd ed. Prentice-Hall, Scarborough, Ontario,533 pp.


US EPA (United States Environmental Protection Agency).1990. Reducing risk: Setting priorities and strategies forenvironmental protection. Science Advisory Board, UnitedStates Environmental Protection Agency, Washington, DC,SAB-EC-90-021.

Webster, T., and P. Connett. 1990. Risk assessment: a publichealth hazard? Journal of Pesticide Reform 10:26–31.

Wilson, R., and E. A. C. Crouch. 1987. Risk assessment andcomparisons: An introduction. Science 236:267–270.

Woodhouse, E. J. 1995. Can science be more useful in politics?The case of ecological risk assessment. Human and EcologicalRisk Assessment 1:395–406.

Ecological Risk Assessment: An Input forDecision-Making

C. J. Van Leeuwen. Ministry of Housing, Spatial Planning and theEnvironment, P.O. Box 30945, 2500 GX The Hague; and ResearchInstitute of Toxicology, Utrecht University, P.O. Box 80176, 3508 TDUtrecht, The Netherlands.

In the European Union (EU), ecological risk assess-

ment (ERA) is a well-developed tool used primarily asan input into the decision-making process governingchemical regulation. It is entrenched in legislationpertaining to new and existing chemicals, includingpesticides. It is applied in waste management and waterquality control decisions. In short, ERA is an importantelement in the environmental decision-making pro-cesses. For critical environmental decisions, however,the need to compare risks and benefits requires ERA beonly one of many elements considered in the decision-making process. Legislative, political, and social factors;research uncertainties, and the technical feasibility ofoptions for risk reduction (Carnegie Commission 1993)must also be considered before a final decision isreached. Whether ecological risks will ultimately gainpriority over other aspects of the decision-making pro-cess cannot be said at this point. Nevertheless, it is clearthat the most important development in environmentaldecision making in the last decade has been theinclusion of ERA by regulatory managers in decisionmaking.

ERA: A Drama-Driven Emergence

Although regulatory and scientific interests haveplayed important roles in the development of ERA,neither drove its emergence. Instead, accidents andgrowing public awareness of environmental problemsprovided the impetus to both science and regulation inthe development of ERA. Well-known examples fromEurope include: sharp declines in bird populationscaused by persistent organic pollutants (including pesti-cides), surface waters covered with foam (nondegrad-able detergents), calamities in Seveso, Italy (dioxins),and massive fish kills in the river Rhine (endosulfan andorganophosphate pesticides). These events triggeredthe development of legislation on pesticides, end-of-pipe emission standards, requirements for the rapidprimary degradation of detergents, and internationalenvironmental treaties for the Rhine and North Sea,respectively.

ERA was thus born in an era of rapid industrializa-tion, where the effects of pollution could not remainunnoticed. The public could see, smell, taste, hear, andsometimes feel the pollution. The resulting public andpolitical awareness of environmental degradation andits possible consequences created the need for legisla-tion and international cooperation out of which ERAgrew. ERA was, accordingly, a drama-driven process thatevolved as a reactive solution to obvious pollutionproblems. Now, however, ERA is used in a proactivemanner to prevent the unacceptable ecological effects

KEY WORDS: Ecological risk assessment; Generic ERA; EUSESmodel


associated with the production, use, and disposal ofchemicals.

Further Development of ERA

The occurrence of accidents and incidents is unpre-dictable. That ERA developed in response to the soci-etal and political needs triggered by such events isnatural. In the 1960s and 1970s, industry’s response torequests to reduce pollution stress was defensive. Theneed for environmental legislation was often denied.Questions were raised about the need to protect speciesand ecosystems, and cleaner production was portrayedas a threat to economic growth.

Viewpoints, however, have changed dramatically.Table 1 details some of the perceptual changes thathave occurred over the last 25 years. Industry hasbecome more proactive. Initiatives are now taken tomeet perceived consumer environmental needs. Height-ened environmental consciousness suggests the inclu-sion of pollution prevention measures in process optimi-zation helps, rather than hinders, corporate profitability.As a result, ERA and LCA (life-cycle assessment) havebecome important instruments in the development andevaluation of chemical products and processes.

In part the changes in societal views are reflected in

the numerous developments with respect to ERA thathave occurred within the OECD and the EU. In the EU,legislative instruments for new chemicals (the seventhamendment to Directive 67/548/EEC), existing chemi-cals (EC Council Regulation 793/93), plant protectionproducts, i.e., agricultural pesticides (914/414/EEC)and biocides (in preparation), have been developed[see Vermeire and Van Der Zandt (1995) for a descrip-tion of these frameworks]. In Europe ERA was, and is, apragmatic response to the need to mitigate and controlthe most obvious impacts of pollution. It is important torealize that ERA procedures have evolved from longdiscussions between the chemical industry, memberstates, and the European Commission. ERA has evolvedfrom communication and perhaps is communication.

ERA and Science Policy Decisions

Calow (1994) stated that we still do not know enoughabout ecological systems to be able to identify what it iswe want to protect about them and, hence, what weshould be measuring. While correct from a scientificviewpoint, from a risk management point of view thestatement is naive. It is not ecotoxicologists who shouldbe determining the targets and extent of protection.Both are matters for policy. Scientists have played, andshould continue to play, a crucial role in the develop-ment and improvement of ERA techniques. However, asthe developments of the last decades have shown, it issocietal views that provide the crucial force behind thedevelopment of environmental protection measures,not scientific investigation. Further research directedtoward elucidating what needs to be protected is not theissue, especially when world-wide biodiversity is ex-pected to decrease by approximately 25% in the nextfew decades (McNeely 1990). What is important is thatERA is a risk management tool. Our efforts should beconcentrated on making it acceptable to those whoapply it now, rather than deferring to the need for yetmore knowledge. Making ERA acceptable to an interna-tional community of academia, industry, and regulatorybodies will require a combination of both political andscience policy decisions (Milloy and others 1994). Forexample:

1. Determination of appropriate ERA protocols re-quires scientific input. Ultimately, however, the selec-tion of methodologies will involve a political decisionon the acceptable trade-off between scientific precision,industry cost, and realistic time frames.

2. Although measurement endpoints require scien-tific input, decisions about valued ecosystem compo-nents and assessment endpoints depend critically onrisk managers, politicians, and society. In the end

Table 1. Changing perceptions of health andenvironmental problems and their solution

1970 1995

PerceptionsSectoral (air, surface,

water) problemsparamount

Multiple media problems insoil, sediments and groundwater

Local pollution issues Diffuse pollution issuesLimited economic damage

caused by pollutionEconomic losses from

pollution known to belarge

Limited environmentalresponsibility on thepart of industry

Environmental responsibilitya leading industrialprinciple

SolutionsProtection of human

health and well-beingProtection of ecosystem

health, human health,economic well-being

Use of end-of-pipetechnologies

Use of integratedapproaches to pollutionprevention, remediation

Pollution prevention athreat to economicwelfare

Pollution prevention apremise and opportunityfor economic growth

Use of legislation andregulation to controlpollution

Use of fiscal incentives andnegotiated agreements

ERA framework absent ERA part of legislation


decisions regarding assessment endpoints define themeasurement endpoints employed in an ERA.

3. Although various methods for exposure and ef-fects assessments exist, all invariably contain explicitand implicit science-based policy choices pertaining totechnique and interpretation (Commission of the Euro-pean Communities 1996). In the end the nature of theERA methodology relies on decisions taken by riskmanagers. Risk assessors, therefore, have an importantrole in communicating the uncertainties inherent intheir own policy-based methodology choices to riskmanagers.

4. While the outcome of an ERA may be quantified,the categorization of the outcome as a low, medium, orhigh risk is ultimately a matter for societal debate. Thisis a dynamic process and views may change over time(see Table 1).

Generic Nature of ERA in the EU

The cumulative effect of social and political choiceon ERA is not necessarily detrimental to its practice. Aglance at the EU guidance documents for risk assess-ment (Commission of the European Communities 1996)indicates it is comprised of the following elements: astandardized minimum data set (the so-called base set)that contains both short-term toxicity and ecotoxicitydata, basic physicochemical data, use information, andimport/production data. A standardized realistic worst-case emission scenario is used by applying emissionscenarios for use categories based on use-specific emis-sion patterns and emission factors. Predicted environ-mental concentrations (PECs) are determined for localand regional situations using multimedia exposuremodels. These models operate by simplifying environ-mental media—air, water, soil, sediment, and biota—into homogeneous compartments and then trackingmovement, degradation, and accumulation of chemi-cals from compartment to compartment. Geographic,hydrological, and climatic variability are excluded fromthe models as standardized environments are used forall exposure calculations. All factors and methodologiesare clearly described. Estimation methodologies and acritical assessment of their limitations are given forphysicochemical parameters, (bio)degradation, sorp-tion, and (eco)toxicity data. The model also containsmodules for characterizing the risks of occupational,consumption, and indirect environmental (via drinkingwater, fish, plants, milk, and meat) exposure.

The shortcomings of this generic approach for ERAare obvious. It has almost nothing to do with ecology.Numerous assumptions and arbitrary decisions aremade to arrive at a risk characterization for both

humans and ecosystems. These include assumptionsconcerning equilibrium partitioning, instantaneous mix-ing in the compartments of the exposure model,simplistic representations of aquatic and terrestrialecosystems, and consideration of only a few biomagnifi-cation routes. Despite the uncertainties introduced withthese assumptions, the greatest uncertainties are prob-ably not related to what we do not know about ecology,but to what we do not know about emission and fateprocesses. For example, little is known about actualchemical use patterns or their compartmentalization orspeciation after release. Product registers can onlypartly overcome this problem. For all practical pur-poses, the enormous variations in climatic (e.g., tempera-ture and precipitation), hydrologic, and geographic(e.g., soil type) conditions and techniques used toreduce emissions necessitate the use of average values inthe completion of ERAs. Under these conditions ERAmethods can do little but focus on the generic assess-ment of risk. While such assessments may not be suitedto site-specific problems, the process is more easilyadapted to the environmental, scientific, and politicalconditions found in other regions or nations. Site-specific assessments require the development of moresophisticated models, increased data detail, and greatercosts. Their results lack generality and cannot easily beadapted to conditions found outside the bounds of theproblem and site for which they were developed.

As Aristotle pointed out, ‘‘It is the mark of aninstructed mind to rest easy with the degree of precisionwhich the nature of the subject permits and not to seekan exactness where only an approximation of the truthis possible.’’ As long as ERA is carried out in a standard-ized manner, it is possible to compare the generic risksof many chemicals and their substitutes. With genericERA it is also possible to predict the environmentalimpact of proposed risk-reduction measures. Even insituations where risks cannot be precisely determined,because of the inability to include all site-specificfactors, generic ERA approaches allow gross compara-tive risks to be quantified. In doing so, generic ap-proaches make a valuable and pragmatic contributionto aiding risk managers and society in controlling theimpacts of chemical use.

A European uniform system for the evaluation ofsubstances (EUSES model) has been developed for theevaluation of new and existing chemicals. EUSES is asimple, straightforward, and transparent set of riskassessment methodologies that recognizes the availabil-ity of limited data sets, the accuracy and variability of thedata, and the uncertainties associated with many of themethodological assumptions. The advantages of a stan-


dardized approach for risk assessment, both ecologicaland human, are threefold:

1. Risk assessments are accompanied by completetransparency of methods, assumptions and uncertain-ties. Assessments become predictable and their accep-tance will increase, leading to increased cooperationand sharing of the burden.

2. Conditions are created for regular methodologicalimprovements and adaptations on the basis of bothscientific and legislative developments.

3. Time and money can be more effectively spent onpriority chemicals for which in-depth ERA and expertjudgement are required.

ERA and Capacity Building

ERA is becoming an increasingly important issue,particularly in view of the rapid industrialization ofmany national economies. Rapid growth is often accom-panied by increasing levels of pollution. Without properregulation of production and consumption patterns,rapid growth is likely to have large environmentalimpacts. Lessons from the ‘‘do-nothing’’ decades of thepast have taught developed countries that pollutionprevention is cheaper than cleanup. The high costsinvolved in cleanup operations (e.g., polluted soils,aquatic sediments, or dump sites) provide painfulreminders of the past when dilution, adsorption, orleaching were used as an excuse for not taking preven-tive measures. The ‘‘out-of-sight, out-of-mind’’ policyled to severe environmental deterioration. Despite avail-able financial resources, restoration is out of the ques-tion in many instances. Strengthening the capabilitiesand capacities of less-developed nations for risk assess-ment is, therefore, a necessary prerequisite to theiravoiding a similar fate. To teach them not do to what wehave done is probably our greatest challenge in the areaof chemicals control. For example, Agenda 21 from the1992 United Nations Conference on Environment andDevelopment (UNCED) places great emphasis on theenvironmentally sound management of chemicals(United Nations 1992). ERA promises to allow us to putaction behind those words. However, to properly accom-plish the task, risk assessment must continue to evolve.In that regard, among the important developments forthe future of ERA will be the following:

1. Periodic updating of guidance documents due tothe rapid development of risk assessment theory andmethods. It is essential that this proceeds as a result ofadequate communication between industry, nationalgovernments, and international bodies, such as theEuropean Commission. Significant contributions from

the scientific community in this process will remainessential.

2. Priority should be placed on refining the ERAmodules or data and/or estimates with the largestassociated uncertainties. This assumes that uncertaintycan be reduced through additional research.

3. Currently only a few endpoints and exposureroutes are considered within the context of ERA.Further development, refinements and validation ofmodels and modules for assessing additional endpointsand exposure routes is accordingly necessary.

4. The emission and exposure of chemicals through-out their life cycles needs further attention. Industryshould play an important role here, as their responsibil-ity should not stop once a chemical is sold or exported.

5. Cost-effective approaches to expediting the riskassessment process are necessary as many existingchemicals have yet to be assessed for their ecologicalrisks. The clustering of chemicals on the basis of modesof toxic action, use pattern, etc., may provide promisingavenues of investigation. Furthermore, the develop-ment of structure–activity relationships (SARs) andquantitative structure–activity relationships (QSARs) toovercome data gap problems remains essential.

6. To harmonize the generic risk assessment ofchemicals, international agreement must be reached onbase-set requirements for new chemicals, existing chemi-cals, pesticides, and biocides. The worldwide accep-tance of the OECD minimum premarketing data set fornew chemicals is an essential step.

Despite the need for change, risk assessment isclearly a valuable methodology. It delivers the toolsnecessary to describe, priorize, and pragmatically ad-dress the complicated environmental issues that lieahead. With appropriate international coordination,generic approaches to risk assessment, transferable tothe rapidly developing Third World nations, can helpavoid the concomitant increases in pollution stressinvariably associated with economic activity.

Literature CitedCalow, P. 1994. Ecotoxicology: What are we trying to protect?

Environmental Toxicology and Chemistry 13:1549.

Carnegie Commission. 1993. Risk and the environment. Im-proving regulatory decision making. Carnegie Commissionon Science, Technology, and Government, New York, 150 pp.

Commission of the European Communities. 1996. Technicalguidance documents in support of the Commission Direc-tive 93/67/EEC on risk assessment for new substances andthe Commission Regulation (EC) No. 1488/94 on riskassessment for existing substances. Commission of theEuropean Communities, Brussels, Belgium.

McNeely, J. A. 1990. The sinking ark. The worldwide loss ofbiodiversity. Presentation at the symposium on flora and


wildlife under chemical pressure in Arnhem (The Nether-lands), 9–10 October 1990. International Union for Conser-vation of Nature and Natural Resources, Gland, Switzerland.

Milloy, S. J., P. S. Aycock, and J. E. Johnston. 1994. Choices inrisk assessment. The role of science in the environmentalrisk management process. Regulatory Impact AnalysisProject, Washington, DC, 270 pp.

United Nations. 1992. Environmentally sound management oftoxic chemicals including prevention of illegal internationaltraffic in toxic and dangerous products. Agenda 21, chapter19. United Nations Conference on Environment and Devel-opment, Rio de Janeiro, Brazil.

Vermeire, T., and P. Van Der Zandt. 1995. Procedures ofhazard and risk assessment. Pages 293–337 in Risk assess-ment of chemicals: An introduction. C. J. Van Leeuwen andJ. H. Hermens (eds.), Kluwer Academic Publishers, Dor-drecht, The Netherlands, 374 pp.

Truth and Validation in Ecological RiskAssessment

D. A. Holdway. Department of Applied Biology and Biotechnology,Faculty of Biomedical and Health Sciences, Royal Melbourne Instituteof Technology, GPO Box 2476V, Melbourne, Victoria 3001, Australia.

Extensive use has been made of the term ‘‘ecologicalrisk assessment’’ in recent years. It has a comfortingkind of ring to it and permits politicians and environ-mental managers to persuade a sceptical and oftensuspicious public that some major project or anotherhas been scientifically assessed and judged to haveminimal, little, or no risk. It is rare for an ecological riskassessment to be undertaken to disprove the nullhypothesis, since neither industry nor governmentgenerally wants to prove that significant risks exist. Thenorm for ecological risk assessments is rather to attemptto prove the null hypothesis, that is, to prove no risk.Risk assessments are thus fraught with type II errors(accepting as true a false null hypothesis) while spend-ing the majority of their efforts in minimizing type Ierrors (rejecting a true null hypothesis). This problemis a very serious statistical one by itself, even when onlyworking with single-species toxicity testing (Holdway1992), much less when involving the enormous difficul-ties of measuring and predicting the behavior of com-plex ecosystems.

In the introductory paper for this debate, the point ismade that a central tenet of ecological risk assessment isprobability, defined as the portion of times a given eventis observed in a series of identical, repeatable experi-

ments (Power and Adams 1997). Since major projectsthat require ecological risk assessments are rarely if evercompletely replicated elsewhere, and since individualecosystems are generally unique, this aspect of repeat-ability is obviously going to be compromised. However,if empirical evidence is gathered in sufficient quantity toaddress appropriately constructed critical questions,then it should be possible to construct a body ofinformation that would support or disprove any signifi-cant predictions made by ecological risk assessments.Given that we have had qualitative environmental im-pact statements for some 30 years or more, semiquanti-tative ecological risk assessments for roughly 10–15years, and involved some thousands of projects world-wide, there should be an enormous literature contain-ing the follow-up validation and monitoring data foreach of these assessments, providing a large empiricaldata set for assessing the strengths and failings of eachtype of impact/hazard/risk assessment. This should bethe case, but few if any validation follow-up studies havebeen conducted.

The Validation Requirement

The vast majority of projects have completed theimpact/hazard/risk assessment as a stand-alone effortfor project approval, usually without an ongoing pre- orpostoperational monitoring program. When monitor-ing has been required, there has generally been norequirement for monitoring information to be corre-lated with preproject predictions. There is thus nosignificant body of information available to determinethe accuracy of impact predictions, irrespective ofwhether or not the predictions were qualitative, semi-quantitative, or quantitative. Were the required environ-mental regulations sufficient to prevent significant mea-surable damage to the environmental attributes of valuein question? If not, how substantial was the actualmeasured damage and what factors were the likelycauses of the observed effects? If wrong, why were thepredictions wrong? Where is the literature providingthe data to permit empirical predictive modeling ofecosystem risks? Only when a detailed and appropri-ately planned validation program that addresses thecritical predictions for a large number of major ecologi-cal risk assessments is undertaken will we begin to get anidea of how well our pseudoquantitative (many wouldargue pseudoscientific) ecological risk assessments areactually doing in predicting the occurrence or absenceof impacts.

One major weakness that exists in any sort of ecologi-cal risk assessment is the lack of prediction validation.Another weakness lies in the simplistic assumptions and

KEY WORDS: Ecological risk assessment; Validation; Hazard assess-ments; Quantitative modeling


gross oversimplifications required to create the presentgeneration of ecological risk assessment models. Theseassumptions generally include some or all of: (1) singlespecies acute toxicity tests that are used to representpopulation impacts; (2) prediction of community ef-fects using data from experiments utilizing inappropri-ate time and spatial scales; (3) use of single-substancetoxicity tests to address complex mixture toxicity; (4)assumptions of simple additive effects in complex mix-tures of toxicants; (5) assumptions regarding the impor-tance of indirect effects and abiotic modifying factors;and (6) lack of appropriate data to permit the separa-tion of altered habitat from altered chemical status, justto name a few.

Quantitative Modeling Weaknesses

What is even more insidious is the move towardsmodels that are extremely user friendly and that do notexplicitly reveal the major assumptions being madeunless one takes the time and effort to read theappendices [e.g., the Dutch soil ecological risk assess-ment model (Denneman and van den Berg 1993)].Here we have the problem of a semiquantitative (atbest) model, full of critical and often highly simplisticassumptions, being disguised to appear far more authori-tative and accurate (e.g., two- to three-place decimalthreshold values) than is the case in reality. The oldadage ‘‘garbage in–garbage out’’ is as true today as itwas when the expression was coined. The only differ-ence is that today’s garbage often appears more palat-able because of attractive packaging.

I believe the move towards quantitative ecologicalrisk assessments stems from the desire of industry andgovernment to provide a legal and management frame-work that will produce unambiguous, legally defensibleassessments divorced from subjective scientific judge-ments. This, of course, is based on the erroneous beliefthat good science is always nonjudgemental and quanti-tative in nature. Nothing could be further from thetruth. Science and scientists are products of society. Theinherent biases and beliefs of the time will permeateand direct scientific endeavor and output. The self-correcting nature of science means, however, that suchbiases and any mistaken directions or hypotheses willnot stand the test of time. Eventually they will bereplaced by new ‘‘politically correct’’ hypotheses. Thusthe movement towards quantitative ecological risk assess-ment does not mean that previous approaches arenecessarily wrong, only that they have lost favor with thepowers that direct funding in these areas, i.e., govern-ments and industry. They are no longer accepted asconventional wisdom.

The numbers produced by most quantitative ecologi-cal risk assessments are generally unvalidated and inmany cases highly misleading. How can one put aprobability of occurrence on nonreplicated and uniqueevents for which one does not understand the complexcausal mechanisms involved (e.g., possible indirecteffects of a toxicant on population and communitystructure) or even the final endpoints or parametersthat are important? Is a change of species within a givenlevel of productivity important? Is it possible to realisti-cally quantitate the ecological relevance of such achange? How does one interpret the ecological mean-ing of a multitude of such changes on multiple trophiclevels within an ecosystem? Munkittrick and McCarty(1995) argue that the differences in spatial and timescales between population and community changes andsingle species effects make single-species toxicity testresults of very doubtful value for predicting impacts athigher levels of biological organization, a view shared byothers (Adams and others 1997, Depledge and Fossi1994). Certainly, we can define specific arbitrary objec-tives to be adhered to (e.g., a #10% loss of speciesrichness, #10% acute mortality, #20% decline in popu-lation, etc.), but we are still no closer to understandingwhat the generic meaning is (if any) of such changes tothe function of ecosystems as a whole.

Truth in Hazard Assessments

Regulators and politicians want to simplify biologicaland ecological processes to single numbers for the sakeof management. Such simplifications may be com-pletely irrational and meaningless both ecologically andstatistically. Nevertheless, they may be promoted byscientists as politically viable options. Are we not thensimply producing politically correct science rather thanecologically relevant science? We should not encourageour regulators and politicians to believe that we knowthe answers to questions that have not even beenappropriately conceived much less answered. Who canreally tell what a 5% or 20% loss of one species means tothe ultimate productivity and stability of an ecosystem?It may be irrelevant, but it might also be highlyimportant. If a professional judgement is called for, Iwould rather that judgement be made by an experi-enced ecotoxicologist sufficiently convinced of his orher beliefs to openly author the opinion. The idea thatsome simplistic computer program can be run anony-mously to produce a numerical result for regulatorypurposes, without understanding model limitations andassumptions and without the personal credibility of anamed professional being on the line, is very disturbing.


It is certainly possible to predict chemical transportand fate using reasonably validated models that providefor estimates of error (Suter 1995, Rand 1995). Biologi-cal exposure to chemicals can even, on occasion, beestimated using calibrated biomarkers (referred to asbioindicators by McCarty and Munkittrick 1996), al-though they do not generally represent very useful toolsfor risk assessment (Holdway 1996). However, it is notthe environmental chemistry that provides the greatestchallenge for ecological risk assessment. It is the ecologi-cal and biological aspects of ecological risk assessmentthat can not be easily modeled. Many of the new andgenerally highly simplistic computer models imply thisis not the case. Who needs to know about energy flow,trophic status, habitat alteration, abiotic or biotic modi-fying factors, and all the unknown variables involved inunderstanding ecosystems. Simply enter your eight orten literature LC50s and watch the computer provide‘‘unambiguous environmental regulation’’ by generat-ing to three significant digits ‘‘safe’’ exposure concentra-tions. Yet we only know the names of 10% of theorganisms present in some ecosystems and have littleinformation on their relative abundance (e.g., thepresent situation for inshore Australian marine ecosys-tems).

Detailed ecological information, i.e., population andcommunity level information, is generally difficult toobtain, can delay potential projects by months to years,and costs significantly more money while not necessarilyproviding unambiguous answers. This is not the type ofinformation generally sought by either project propo-nents or regulators. Quantitative ecological risk assess-ment marks the beginning of a new era in environmen-tal regulation: the ‘‘ecotoxicologist on a disc’’ era. Sincethe general move of ecological risk assessment in thefuture will be towards the use of more user-friendly,mathematically complex computer models, the averagerisk assessor will soon be little more than a computerkeyboard operator/clerical administrator.

I do not believe we should excessively simplifyecological science in order to cater to the needs ofregulators, lawyers, and politicians. The fact remainsthat few if any of our ecological risk assessment predic-tions are being validated today. Consequently, how canwe have any statistical or ecological confidence in thepredictions or the methods generating them? How is itpossible for the probabilities to be accurate if we haveno understanding of how often we have failed todetermine impacts or have succeeded in predictingeffects? Without a comprehensive retrospective investi-gation of the past predictions and ultimate outcomeswithin a variety of time scales, we are simply not able to

provide meaningful probability-based quantitative eco-logical risk assessments.

Conclusions

What is the solution to ecological risk assessment? Inmy opinion, the best solution we can presently offer is toundertake semiquantitative hazard assessments thatevaluate the potential of deleterious effects occurring inan ecosystem given the presence of defined toxic agentsor pollutants. Such assessments require extensive use ofexpert judgement, often from a number of expertsrepresenting various environmental disciplines and willbe open to alternate interpretation. The application ofsignificant safety or uncertainty factors to all toxicitythreshold numbers will be required both to protect theecosystems being assessed from type II errors, resultingfrom our lack of understanding of indirect populationand community effects and ecosystem dynamics, and toprovide conservative overall margins of error. Further-more, all assessments of significance should be followedup with appropriately designed environmental monitor-ing programs to test the truth of environmental risk(hazard) assessment predictions.

A concerted international effort needs to be made tocollect and collate impact or risk assessments contain-ing explicit predictions relating to significant projectsthat have proceeded and for which biological/ecologi-cal monitoring programs were put in place. These dataneed to be analyzed and evaluated to determine thetypes of predictions made and their relative accuracy.Only when a significant data base has been created canwe hope to have biological and mathematical modelersattempt to estimate the occurrence probabilities ofvarious ecological impact categories.

It is quite likely that some types of hazards are farmore amenable to accurate prediction than others andthat the approach eventually adopted by diligent envi-ronmental regulators will, of necessity, be site or project-type specific. These are no easy short cuts in ecologicalrisk assessment, only simple questions requiring diffi-cult and often complex answers. It is our responsibilityto ensure that regulators and politicians understandboth the complexity and uncertainty of any answers thatare provided for the ecological risk questions they haveasked. To do anything less is to greatly mislead thepublic and risk undermining the public’s overall confi-dence in the use of science to address importantquestions.

Literature CitedAdams, S. M., K. D. Ham, and R. F. LeHew. 1997. A framework

for evaluating organism responses to multiple stressors:


Mechanisms of effect and importance of modifying ecologi-cal factors. In J. J. Cech and B. W. Wilson (eds.), Multiplestresses in ecosystems. Lewis Publishers, Boca Raton, Florida.(in press).

Denneman C., and R. van den Berg. 1993. Exposure assess-ment—methods used for derivation of risk concentrations.Paper 13 in Ecological risk assessment from theory topractice. Conference proceedings, 6–8 October 1993, Mel-bourne, Australia. ISBN 0-7306-2860-4.

Depledge, M. H., and M. C. Fossi. 1994. The role of biomark-ers in environmental assessment (2) Invertebrates. Ecotoxicol-ogy 3:161–172.

Holdway, D. A. 1992. Uranium mining in relation to toxicologi-cal impacts on inland waters. Ecotoxicology 1:75–88.

Holdway, D. A. 1996. The role of biomarkers in risk assess-ment. Human and Ecological Risk Assessment 2:263–267.

McCarty, L. S., and K. R. Munkittrick. 1996. Environmentalbiomarkers in aquatic toxicology: Fiction, fantasy or func-tional? Human and Ecological Risk Assessment 2:268–274.

Munkittrick, K. R., and L. S. McCarty. 1995. An integratedapproach to aquatic ecosystem health: Top-down, bot-tom-up or middle-out? Journal of Aquatic Ecosystems Health4:77–90.

Power, M., and S. M. Adams. 1997. Ecological risk assessment:A dichotomy of views. Environmental Management (this issue).

Rand, G. M. (ed.). 1995. Fundamentals of aquatic toxicology,2nd ed. Taylor and Francis, Washington, DC, 1125 pp.

Suter, G. W., II. 1993. Introduction to ecological risk assess-ment for aquatic toxic effects. Pages 893–816 in G. M. Rand(ed.), Fundamentals of aquatic toxicology, 2nd ed. Taylorand Francis, Washington, DC, 1125 pp.

Controversies in Ecological Risk Assessment:Assessment Scientists Respond

Glenn W. Suter II and Rebecca A. Efroymson. Environmental SciencesDivision, Oak Ridge National Laboratory, Oak Ridge, Tennessee37831-6038, USA.

The following are responses by two scientists en-gaged in the practice of ecological risk assessment(ERA) to some recent criticisms of the field. In ouropinion, many of the critics are setting up and knockingdown straw men for purposes of advocacy. Some areattempting to hold ERA to standards of certainty thatare impossible to meet, and others simply mistake thevalues of particular agencies and organizations that useERA for an inherent feature of the practice. A completerebuttal would require far more space than is allowed.We believe, however, that the following points areparticularly important.

False Dichotomies

Most of the critics formulate their argument in termsof a dichotomy, ERA versus something else. However,they do not agree on what the alternative is, and somedo not even propose an alternative. We believe that afundamental dichotomy underlying many of the criti-cisms is between those who would impose a burden ofproof on ERA and those who estimate effects using theERA paradigm.

Advocates of strict environmental regulation take theposition that safety should be proven before actions areallowed. This position has been termed the ‘‘precaution-ary principle’’ (MacGarvin 1994). Critics who assumethis position attack ERA because it makes judgementsabout effects of actions and their potential acceptabilitywithout sufficient information to prove safety. For ex-ample, Pagel, who takes as his slogan ‘‘knowledgeabletrespass or none at all,’’ points out the infinite complex-ity of ecological systems, and invokes the chaos theorycliche of a flap of a butterfly’s wing causing a storm toargue that ERA does not and can not demonstrate safety(Pagel 1995).

Opponents of strict regulation take the position thatunacceptable injury should be proven before actionsare restricted. This position is frequently described as‘‘good science’’ because it demands the same highstandard of proof for anthropogenic effects that arerequired for proving a hypothesis in pure science (Suter1996). That is, it requires rejecting the hypothesis of noeffect with 95% confidence.

ERA does not demand that all actions be provenharmless with high confidence, and it does not requirethat effects be demonstrated with 95% confidence.Rather it uses a variety of quantitative and logicaltechniques to analyze the evidence and reach a judge-ment based on the weight of evidence (Suter 1993,Suter and Loar 1992, US EPA 1992). The weight ofevidence standard is unacceptable to those who requireproof. For example, Cooper (1995) advocated goodscience and rejected the weight of evidence standard asbeing used by lawyers with no interest in the truth.Cooper misunderstands the legal system. It is judgesand juries who weigh evidence and they do seek thetruth.

We argue that the weight of evidence standard is, inmost cases, the appropriate standard for risk assessors. Acentral tenet of risk assessment is that the assessor is atechnical consultant to the risk manager. The riskmanager is the individual responsible for making thedecision, including deciding what degree of confidenceis needed concerning safety. For a risk assessor torequire either proof of safety or proof of harm would be

KEY WORDS: Ecological risk assessment; Probability; Values; Bur-den of proof; Benefits assessment; Alternatives; Riskmanagement


to drastically bias the analysis. Such biases are appropri-ate only when mandated by law or regulation as in theUS Federal Insecticide, Fungicide, and RodenticideAct.

The Role of Values

A variety of criticisms of ERA relate to the role ofvalues. The most easily dismissed criticisms are the callsfor risk assessors to adopt a particular set of valuesadvocated by the critic. Most risk assessors realize that ifthey adopt an advocacy position and design theirassessments to support that position, they will lose theircredibility. Assessors are advised by some critics to‘‘retain emotion’’ and ‘‘treat the earth as their child’’(O’Brien 1995), reject humanistic values (Pagel 1995),and take a consciously spiritual approach to their work(Hayakawa 1995). If they did so, they would openthemselves to charges such as those of US Congres-sional Representative H. Chenoweth that environmen-tal protection violates the antiestablishment clause ofthe US Constitution because it is based on a religionwhich places the interests of trees before humans (citedin The Washington Post, 5 February 1996, p. A19). Riskassessors are not elected or appointed, so their valueshave no particular standing in a democracy. Like anycitizen they may vote, write letters, etc., but they may notuse their positions as scientific experts to make publicpolicy by biasing their analyses.

A more subtle argument involving values is that manyof the decisions made by risk assessors are value-laden,so they are making unconscious value judgements andtherefore they are functioning as risk managers (Craw-ford-Brown 1995). Specifically, Crawford-Brown (1995)argues that risk assessors should consider what shouldbe valued and why we should value it. That is clearly arisk management decision, and in the conventionalERA paradigm it is described as such (US EPA 1992).The risk manager is included in the problem formula-tion process, the planning phase of the risk assessment,specifically so that value-laden decisions about how toperform the assessment are made by the individualswho have the authority to address them.

An even more subtle argument is that there is noclear distinction between facts and values, and there-fore the division of risk assessment from risk manage-ment is artificial and untenable (Schrader-Frechette1985, Schrader-Frechette and McCoy 1994). To thosewho make this philosophical argument, even decisionsabout model forms or experimental methods are valuejudgements. However, in practice these are a verydifferent sort of value judgement than deciding whetherto allow the northern spotted owl to go extinct in order

to preserve the jobs of loggers. It is not necessary todistinguish value judgements from facts, or to keepscientists from making any value judgements. Rather,one must distinguish value judgements that the publicwould recognize as such, and would not want to leave toscientists, from those that a scientist can make with aclear conscience. The distinction is difficult in theabstract, but we have found it to be relatively simple inpractice. For example, the choice of an area or transectmethod for vegetation surveys may be made on the basisof efficiency of effort. That choice is value laden in thesense that efficiency is a value not a fact. However, noremedial project manager or citizens advisory councilwants the choice of sampling methods left to them sothat they can choose on some basis other than effi-ciency. Rather, they want to decide whether vegetationwill be assessed and what types of changes in vegetationare potentially significant.

Probability and Risk

The relationship between uncertainty, probability,and risk is another source of controversy. The editors ofthis forum present the argument that the scientificfoundations of ERA and presumably all other riskassessments are weak because risk incorporates theconcept of probability but does not define it on thebasis of observed frequencies from ‘‘identical repeat-able experiments’’ (Power and Adams 1997). This is anextremely narrow definition of probability that is not inconcordance with most concepts of risk (Morgan andothers 1985, Suter 1990). It ignores most of frequentiststatistics and all of Bayesian statistics and other nonfre-quentist methods of deriving probabilities. It is impos-sible or ethically unacceptable to base most risk assess-ments on the frequency of outcomes in identicalrepeated experiments (e.g., create identical waste sitesor crash identical tankers). Therefore, if we take thearguments presented by Power and Adams (1997) toheart, we must either restrict ourselves to the few caseswhere we can subject replicates of all of the systems thatare at risk to the agent of interest or ignore theuncertainties and pretend that we know the answerswith certainty. We argue that risk assessors must acknowl-edge that, because hazards have uncertain conse-quences that must be estimated, they cannot limitthemselves to a concept of probability that is irrelevantto the problem.

ERA versus Alternatives Assessment

The consideration of alternative options provides abasis for choosing the best action rather than determin-


ing whether the proposed action is acceptable (O’Brien1995, Pagel 1995). However, Pagel (1995) and O’Brien(1995) are mistaken in suggesting that ERA is not orcannot be comparative. For example, ERA is commonlyused as one of several means of comparing alternativeremedial choices: which sites carry greater potentialecological risks, or which proposed remedial actions areassociated with risks of greater geographic extent andduration than other scenarios, including the no-actioncase. ERA’s estimates of effects on well-defined assess-ment endpoints are a much simpler basis for compari-son than criteria that cross endpoint and scale (Suterand others 1995). In addition, both federal and stateagencies in the United States are beginning to usecomparative risk assessment methods to prioritize theirenvironmental programs.

ERA Versus Benefits Assessment

Benefits assessment is advocated as preferable to riskassessment in ‘‘large ecological units’’ (Principe 1995)or in evaluations of ecological alternatives (O’Brien1995). We agree that benefits should be consideredprior to decision making, but not because of a limita-tion specific to ERA at large scales (see below). It issomewhat ironic to practitioners of retrospective (reme-dial) ERA that advocates for the environment promotebenefits assessment, with roots in utilitarian cost–benefit analysis (Principe 1995). Historically, NationalEnvironmental Policy Act (NEPA)-driven environmen-tal impact assessment, to which ERA owes a philosophi-cal and methodological debt, arose because promotersof federal projects touted benefits but often overlookedenvironmental, health, or social risks. Advocates ofprojects seldom ignore benefits, including those thataccrued to ecological systems. Finally, one of the majorfunctions of retrospective ERA is to direct the remedia-tion and restoration of damaged ecosystems, the veryecological relief that critics such as O’Brien (1995)seek.

ERA Versus Management of Hazards

An alternative approach to the risk paradigm forenvironmental problem solving is to shift the relativeemphasis from assessment to management (Wood-house 1995). We might call this the natural disasterparadigm, where predictions are perceived to havelimited utility for reducing adverse consequences. Witha risk management emphasis, experts can deal withenvironmental uncertainties through ‘‘intelligent trialand error’’ (Woodhouse 1995). This choice is dominantwhen the political and economic stakes are high (e.g.,

containment for biotechnology or double-hulled tank-ers for petroleum transport) and when the cost of ERAis high because the knowledge base does not permitmuch extrapolation from case to case. In this manage-ment-heavy scenario, risk assessors are relegated to thepostdecision position of characterizing undesirable sideeffects and estimating effectiveness of an action (Wood-house 1995). We argue that management actions shouldnot be selected by exclusively political means. Rather,risks should be estimated so that managers can selecttrials on the basis of estimated risk, leading to fewererrors.

Not Enough Is Known

It is argued that ecological risks, particularly at largegeographic scales or at higher levels of organization, aredifficult to quantify (Regens 1995), to characterizegenerally (Principe 1995), or to predict at all (Pagel1995). However, some ecosystem level responses such asthe degree of eutrophication induced by a phosphorussource are routinely predicted with reasonable confi-dence. Even relatively simple toxicity tests can be usedto predict which effluents will change community com-position (Dickson and others 1992). Further, most ofthe risk assessments that are currently performed arefor contaminated sites where effects can be measured byvarious biological survey techniques, supplemented bymeasures of ambient exposure and tests of ambienttoxicity. ERA may be unable to quantify the value-ladenconcept of ecosystem health (Regens 1995), but we areoften able to estimate aspects of ecosystem condition.

It is surprising that these indictments of the knowl-edge base lead to condemnations of ERA rather than toarguments for research. If asked where most ERAscould be improved (aside from assuring basic compe-tence), we would point to failures to obtain site-specificinformation. For example, ERA practitioners substitutedata on estuarine benthic organisms for freshwater,total concentrations of contaminants in soil for bioavail-able concentrations, or generic soil-to-plant uptakefactors for site-specific factors. ERA should be held to abest-available knowledge standard, and society shouldreasonably expect that standard to improve with time.

Conclusions

ERA is not a perfect basis for environmental manage-ment, nor does it claim to be a complete basis. It issimply a conceptual paradigm for providing timelytechnical consultancy to environmental decision mak-ers without hiding uncertainties or introducing theanalyst’s values. By distinguishing risk management


from risk assessment and assigning them proper roles,ERA does not completely eliminate the blurring of factsand values, but it is a considerable improvement overpurely technocratic or purely political approaches.ERA’s insistence on pursuing estimation under uncer-tainty will not satisfy those who demand proof of safetyor of injury, but it is preferable to paralysis or decisionsbased purely on political pressures. ERA acknowledgesand estimates uncertainty even when the bases forquantification are incomplete because ignoring uncer-tainty can result in worse predictions. Knowledge isnever adequate, and the world will not stop whileadequate knowledge is sought.

Literature CitedCooper, W. E. 1995. Risks of organochlorine contaminants to

great lakes ecosystems are overstated. Ecological Applications5:293–298.

Crawford-Brown, D. J. 1995. The ethical basis of environmen-tal risk analysis. Pages 255–265 in C. R. Cothern (ed.),Handbook for environmental risk decision making: Values,perceptions, and ethics, CRC Press, Boca Raton, Florida,408 pp.

Dickson, K. L., W. T. Waller, J. H. Kennedy, and L. P. Ammann.1992. Assessing the relationship between ambient toxicityand insteam biological response. Environmental Toxicologyand Chemistry 11:1307–1322.

Hayakawa, E. 1995. Human spirituality in the workplace andits relationship to responsible environmental decision-making. Human and Ecological Risk Assessment 1:416–422.

MacGarvin, M. 1994. The precautionary principle and thelimits of science. Helgolander Meeresuntersuchungen 49:1–16.

Morgan, M. G., M. Henrion, S. C. Morris, and D. A. L. Amaral.1985. Uncertainty in risk assessment. Environmental Scienceand Technology 19:662–667.

O’Brien, M. H. 1995. Ecological alternatives assessment ratherthan ecological risk assessment: Considering options, ben-efits, and hazards. Human and Ecological Risk Assessment1:357–366.

Pagel, J. E. 1995. Quandaries and complexities of ecologicalrisk assessment: Viable options to reduce humanistic arro-gance. Human and Ecological Risk Assessment 1:376–391.

Power, M., and S. M. Adams. 1997. Ecological risk assessment:A dichotomy of views. Environmental Management (this issue).

Principe, P. P. 1995. Ecological benefits assessment: A policy-oriented alternative to regional ecological risk assessment.Human and Ecological Risk Assessment 1:423–435.

Regens, J. L. 1995. Ecological risk assessment: Issues underly-ing the paradigm. Human and Ecological Risk Assessment1:344–347.

Schrader-Frechette, K. S. 1985. Risk analysis and scientificmethod. D. Reidel Publishing, Dordrecht, The Netherlands,232 pp.

Schrader-Frechette, K. S., and E. D. McCoy. 1994. How the tailwags the dog: How value judgements determine ecologicalscience. Environmental Values 3:107–120.

Suter, G. W., II. 1990. Uncertainty in environmental riskassessment. Pages 203–230 in G. M. Von Furstenberg (ed.),Acting under uncertainty: Multidisciplinary conceptions.Kluwer Academic Publishers, Boston, Massachusetts.

Suter, G. W., II. 1993. Ecological risk assessment. LewisPublishers Boca Raton, Florida, 538 pp.

Suter, G. W., II. 1996. Abuse of hypothesis testing statistics inecological risk assessment. Human and Ecological Risk Assess-ment 2:331–347.

Suter, G. W., II, and J. M. Loar. 1992. Weighing the ecologicalrisks of hazardous waste sites: The Oak Ridge case. Environ-mental Science and Technology 26:432–438.

Suter, G. W., II, B. W. Cornaby, C. T. Hadden, R. N. Hull, M.Stack, and F. A. Zafran. 1995. An approach for balancinghealth and ecological risks at hazardous waste sites. RiskAnalysis 15:221–231.

US EPA (United States Environmental Protection Agency).1992. Framework for ecological risk assessment. EPA/630/R-92/001, Washington, DC.

Woodhouse, E. J. 1995. Can science be more useful in politics?The case of ecological risk assessment. Human and EcologicalRisk Assessment 1:395–406.

Ecological Risk Assessment: ProgressingThrough Experience or Stalling in Debate

S. M. Bartell. SENES Oak Ridge, Inc., Center for Risk Analysis, 102Donner Drive Oak Ridge, Tennessee 37830, USA.

It is deja vu all over again. The validity and efficacy ofecological risk assessment (ERA) have been recentlyquestioned in discussions in a special issue of Humanand Ecological Risk Assessment [Vol. 1(4), October 1995].Proponents and opponents of the process have ren-dered their opinions and positions in various symposiaand workshops, as well as in the popular and technicalliterature. To the extent that such deliberations im-prove and advance the concepts and methods that formthe foundation of ERA, these kinds of activities are useful.

This brief essay attempts merely to point out severalperceived strengths and limitations of ecological riskassessment gleaned from assessment experience. Theintent is not to fuel unproductive debate but to suggestthat ecological risk assessment follows simply, yet impor-tantly, as a conceptual and methodological extension oftraditional environmental assessments (i.e., NEPA envi-ronmental impact assessments). Throughout this entirediscussion, however, we should not lose sight of theultimate objective, to provide scientifically defensiblequantitative ecological and environmental inputs to

KEY WORDS: Ecological risk assessment; Environmental decisionmaking; Levels of organization; Baseline environmen-tal model


informed decision making, independent of terminol-ogy, the creation of new disciplines, and unproductivedebate.

Perceived Strengths

With a recognition and emphasis on an honestaccounting of uncertainty, ecological risk assessmentprovides an opportunity and offers regulatory incentiveto attack the complexities of real-world environmentalproblem solving and decision making. Ecological riskassessment opens the door for sophisticated applica-tions of systems analysis and ecological modeling, inpart, because the foci of ERA are precisely thosemiddle-number ecological systems that defy simpleanalytical or brute-force statistical description (Allenand Starr 1982, Weinberg 1975). The cleverness comesin using the ERA process to identify the critical simplifi-cations that translate seemingly intractable environmen-tal problems into tractable ones, using available ecologi-cal methods and models (i.e., knowledge).

Ecological risk assessment provides at least oneoperational advance over traditional NEPA-driven assess-ments, namely, the explicit consideration of uncer-tainty. This uncertainty results from: the imperfectcharacterization of ecological disturbances (e.g., toxicchemicals, habitat alteration, exotic species introduc-tions), an incomplete understanding of ecological re-sponses to disturbance, typically sparse site-specificdata, measured spatial and temporal variability, ofteninadequate sampling design, and errors in samplecollection and processing. Uncertainties were, of course,always present in the traditional NEPA assessmentprocess. However, the nature of the assessment and theprescribed reporting format for a NEPA environmentalimpact statement minimized the explicit examinationof uncertainties and consigned their presentation andevaluation to easily ignored appendices. In ERA, ex-plicit consideration of uncertainty includes identifying,quantifying, and propagating uncertainties through theanalysis and reporting their consequences for riskestimates. This aspect of risk assessment might in itselfjustify the effort and validate the process of ERA.

A related strength of ecological risk assessment isthat it is, in a sense, self-diagnostic. That is, uncertain-ties can be exploited using methods of sensitivity and/oruncertainty analysis to identify the key contributors toestimated risk. Limited resources can then be efficientlyallocated to provide the greatest reduction in assess-ment uncertainty per unit investment of time andmoney. This procedure can be repeated until a riskmanagement decision becomes evident or until it isrecognized that uncertainties cannot be reduced fur-

ther without substantial investment in acquiring newknowledge.

Insistence on a probabilistic context for ERA pro-vides the opportunity and capability to introduce ecologi-cal risks on quantitative scales commensurate with othercomponents of comprehensive environmental assess-ments, particularly those involving economics and engi-neering. In fact, the continued evolution and increasedsophistication of ecological risk assessment might ben-efit from an infusion of the more rigorous mathemati-cal and statistical formulations of risk existing in engi-neering disciplines (e.g. Haimes and Stakhiv 1989,1990).

A probabilistic framework for ecological risk alsoaffords an opportunity for ecological assessment toenter formally into decision analysis, including consider-ation of risks, benefits, and costs. Risk assessment can bebroadly viewed as decision making under uncertainty(Rubenstein 1975). Formal decision models can thenbe used in examining the implications of selectingamong decision alternatives. For example, in risk-basedremediation, the economic consequences and effective-ness of different cleanup technologies for reducing riskcould be described quantitatively and informed choicessubsequently made. As a result, risk assessment derivesits significance in the context of decision making(Kaplan and Garrick 1981).

Perceived Limitations

Ecological risk assessment, as currently conceivedand practiced, is not the final solution for environmen-tal problem solving. Increasing the effectiveness of ERAwill require successfully addressing several limitations inthe current assessment process. These limitations in-clude the imprecise nature of environmental legisla-tion, incomplete communication of the regulatorydecision-making process, oversimplified ecological con-cepts, assessment methods of unknown performance,and an accumulation of jargon that unfortunatelydisconnects ERA from the risk assessment methodsmore formally established in other fields (e.g., Helton1993, Kaplan and Garrick 1981). This brief exposition,however, cannot address all of these stated concerns.

Practicing the current paradigm (i.e., US EPA 1992)is made difficult by the absence of a clearly statedenvironmental baseline or a heuristic for defining orselecting the baseline. This difficulty arises in part fromthe many unspecified notional reference environmentsand different underlying human environment modelsthat influence the ecological characteristics used todefine or identify references for ERA (e.g., Holling1986). Human perceptions of the natural world will


influence the definition of reference environments, theselection of endpoints, and consequently, the effective-ness of any assessment. In the context of currentenvironmental regulation, some constant environmentand a corresponding desire for maintaining the ecologi-cal status quo seem to characterize the implicit model.However, ecological risks cannot be easily or convinc-ingly assessed in relation to such a static model ofnature. The significance of decreased productivity orextinction cannot be judged on purely ecologicalgrounds. In adopting the status quo as the frame ofreference, ecological entities are not accorded an inher-ent ecological value. Instead, value accrues only whenecological entities are identified as resources.

Several limitations were identified above as compo-nents of uncertainty in assessments. In a very pragmaticsense, many of these uncertainties prove difficult, ornearly impossible, to quantify given commonly encoun-tered budget and/or time limitations. There is a corre-sponding concern that uncertainty in ERA can paralyzethe decision-making process or be used as an excuse forindecision.

ERA can benefit from continued efforts to interjectmodern quantitative ecology to the risk assessmentprocess. Progress in ERA has been curtailed by unprofit-able debates concerning, for example, simple versuscomplex models, population versus ecosystem effects,ecology versus toxicology. Consider, for example, ‘‘lev-els of organization’’ in ERA. This reflects a lack ofecological sophistication in concept and method. Eachlevel in fact corresponds to a particular ecological pointof view (i.e., model) for describing the same naturalworld. These levels have been used to construct oversim-plified nested models of nature: landscapes that encom-pass ecosystems consisting of communities made ofpopulations of individuals. Independent of the argu-able utility of this simplistic model, the advances inecological understanding achieved through decades ofbasic study of each different level offer unique andpotentially powerful concepts and measurements thatshould be incorporated into the development andapplication of ecological risk assessment. Moreover,competent attempts at integrating across levels (e.g.,Allen and Starr 1982, Allen and others 1984, Holling1986, O’Neill and others 1986, King 1991) should beexplored for their relevance in assessing ecologicalrisks.

One of the dangers implied in the current state ofaffairs is that ERA might increase in rigor and sophistica-tion, but decrease in relevance. The continued refine-ment and improvement of concepts and methods usedto assess ecological risks is necessary and justifiable forcomplying with legislation designed to protect the

environment from the impacts of toxic chemicals.However, capabilities for assessing ecological risks willbe developed fully and applied with the greatest societalbenefit only when ERA becomes an integrated compo-nent of an overall rational plan for environmentalmanagement.

Conclusions

One challenge in realizing the potential of ecologi-cal risk assessment lies in interjecting modern ecologyand the environmental sciences into the decision-making process and regulatory arena. Here ecologicalprinciples are often poorly understood or poorly com-municated, and social, economic, and political consider-ations enter unequally into environmental decisionmaking. To contribute effectively, ecological risk asses-sors must focus their capabilities and resources onecological entities that are defensible from a scientificviewpoint and of vital interest to stakeholders anddecision makers alike. Successful application of ERAdemands the best from science and the best fromdecision makers.

ERA represents an important next step in a continu-ing journey from NEPA toward increasingly scientifi-cally defensible environmental decision making. Ecologi-cal risk assessment is neither sage nor charlatan. It issimply a process. The effectiveness of this process inadvancing the art and science of assessing environmen-tal impacts depends in large part on the capabilities,training, intentions, and integrity of its practitioners.Ecological risk assessment should not become merely anew name for traditional environmental impact assess-ment. Ecological risk assessment should neither beoversold nor undersold. It is not an environmentalpanacea or a regulatory silver bullet. Ecological riskassessment cannot transform an absence of informationand understanding into informed decision making.However, ecological risk assessment can force the ex-plicit identification, consideration, and incorporationof uncertainties into the assessment process. Used withtechnical competence and integrity, ERA interjects anhonest use of typically sparse information and incom-plete ecological understanding into environmental deci-sion making.

Presumably, increased capabilities in assessing eco-logical risks will result from emphasizing strengths andsurmounting limitations. Such progress will likely ensuein proportion to actual assessment experience, ratherthan as the result of debate. Keep talking, but moreimportantly, get to work.


Literature CitedAllen, T. F. H., and T. B. Starr. 1982. Hierarchy: Perspectives

for ecological complexity. University of Chicago Press,Chicago.

Allen, T. F. H., R. V. O’Neill, and T. W. Hoekstra. 1984.Interlevel relations in ecological research and management:Some working principles from hierarchy theory. USDAForest Service General Technical Report RM-110. RockyMountain Forest and Range Experiment Station, Fort Col-lins, Colorado.

Haimes, Y. Y., and E. Z. Stakhiv (eds.). 1989. Risk analysis andmanagement of natural and man-made hazards. AmericanSociety of Civil Engineers, New York, 353 pp.

Haimes, Y. Y., and E. Z. Stakhiv (eds.). 1990. Risk-baseddecision making in water resources. American Society ofCivil Engineers, New York, 331 pp.

Helton, J. C. 1993. Risk, uncertainty in risk, and the EPArelease limits for radioactive waste disposal. Nuclear Technol-ogy 101:18–39.

Holling, C. S. 1986. The resilience of terrestrial ecosystems:local surprise and global change. Pages 292–317 in W. C.Clark and R. E. Munn (eds.), Sustainable development ofthe biosphere. International Institute for Applied SystemsAnalysis, Laxenburg, Austria.

Kaplan, S., and B. J. Garrick. 1981. On the quantitativedefinition of risk. Risk Analysis 1:11–27.

King, A. W. 1991. Translating models across scales in thelandscape. Pages 479–517 in M. G. Turner and R. H.Gardner (eds.), Quantitative methods in landscape ecology.Springer-Verlag, New York.

O’Neill, R. V., D. L. DeAngelis, J. B. Waide, and T. F. H. Allen.1986. A hierarchical concept of ecosystems. Princeton Uni-versity Press, Princeton, New Jersey, 186 pp.

Rubenstein, M. F. 1975. Patterns of problem solving. Prentice-Hall, Englewood Cliffs, New Jersey, 544 pp.

US EPA (United States Environmental Protection Agency).1992. Framework for ecological risk assessment. EPA/630/R-92/001, Washington, DC.

Weinberg, G. M. 1975. An introduction to general systemsthinking. John Wiley & Sons, New York, 279 pp.

Assessing the Current Status of EcologicalRisk Assessment

S. M. Adams1 and M. Power.2 [1] Environmental Sciences Division, OakRidge National Laboratory,3 Oak Ridge, Tennessee 37831-6038, USA;[2] Department of Agricultural Economics, University of Manitoba,Winnipeg, Manitoba, R3T 2N2 Canada.

Risk assessment has developed from the specifics ofusing available toxicological and ecological information

to estimate the probabilities associated with unwantedenvironmental outcomes (Wilson and Crouch 1987) tobecome an important environmental policy instru-ment. As Bartell (1997) has noted, risk assessmentrepresents an important step in the journey frombroadly based, descriptive environmental impact assess-ments toward increasing the scientific defensibility ofenvironmental decision-making. Lackey (1997) con-cludes that much of the excitement surrounding riskassessment stems from the fact that it is the most recentin a succession of tools aimed at scientific environmen-tal management. It seems appropriate, therefore, tohave sought current views on risk assessment fromecotoxicologists, risk assessors, and environmental regu-lators on the acceptability and utility of ecological riskassessment as a tool for describing, ranking, and address-ing the complicated environmental issues that lie be-fore us.

Views about the details of risk assessment necessarilydiffer, but as the debate has shown, there is wideagreement on the need to apply available scientific toolsto the problem of minimizing the environmental conse-quences of human action. A point–counterpoint sum-mary of the views expressed by the discussants on acommon set of risk assessment issues is presented in theecological risk assessment issues matrix of Table 1. Thematrix focuses on 12 issues raised in common by thediscussants and gives a synthesis of their respective viewson risk assessment. Issues other than those detailed inthe matrix were discussed by authors, but not by asufficient number to warrant their inclusion in Table 1.From among the issues raised, we have chosen thesubset representing the greatest extent of disagreementas the focus of our own discussion. These issues are alsothose that we believe will be most pertinent to determin-ing the future success or failure of ERA as an environ-mental decision-making tool. Included in the selectedlist of critical issues are validation, sources of ambiguity,limitations, ecology, appropriate use, and most pressingdemands.

Validation

The traditional application of ERA has been predic-tive in nature and focused on the localized effects of aparticular stressor (Suter 1993). In that sense ERA is alogical extension of descriptive environmental impactsassessments (EIAs). Despite the long history of interestin predicting the effects of action, Holdway (1997)notes the paucity of information available to determinethe predictive accuracy of assessment methodologies.Given the predilection for modeling (Holdway 1997,Van Leeuwen 1997), the result is curious, because the

KEY WORDS: Ecological risk assessment; Validation; Ecology; Limi-tations

3Managed by Lockheed Martin Energy Research Corp. for the U.S.Department of Energy under contract number DE-AC05-96OR22464.


modeling literature itself has been quite clear on theneed for validation. There is no assurance that modelsthat provide the best fit to available sample data will bethe best predictors of future behavior. As a conse-quence, the statistical literature requires the validationof model predictive accuracy and bias in all circum-

stances (Montgomery and Peck 1982). Insofar as mod-els of ecological systems offer the potential for under-standing the consequences of human action, therebyallowing us to minimize the consequences of ouractions, the importance of predictive validation cannotbe understated (Power 1993).

Table 1. Ecological risk assessment issues matrix

IssueCalow and

Forbes Holdway van Leeuwen Lackey BartellSuter andEfroymson

Validation Many assumptionsremain untested

ERA predictionsnot validated

Evaluation of ERAapplicabilityincomplete

Many assessmentmethods are ofunknown value

Objectivityversussubjectivity

ERA objectivelydefinable

ERA necessarilysubjective

Subjective choiceused as requiredbut ERAstransparent anddefined

Subjective in thesense that valuecriteria used tobound ERAs

Subjective becausehuman valuesinfluencebaselinedefinitions

Objective, butsubjectiveprobabilitiesallowed

Managementor science

Both but moreinteractionrequired

ERA driven bymanagementnot science

ERA is a pragmaticpollutionmanagementprocess relianton science

ERA is a decision-making andresourceallocation tool

Science derivingsignificance inthe context ofdecision-making

ERA is science andprovides a basisfor management

Quantitativemethods

Selected methodsoften maskuncertainty

Too reliant onunvalidatedmodels andparameters

Must develop‘‘generic’’approachesusing ‘‘average’’values

Help identifyestimate,propagateuncertainties inall analyses

When used mustbe accompaniedby uncertaintyanalysis

Sources ofambiguity

Difficulties withdefining ERAproblemboundaries

Implications oftoxicity tests forecosystemsunknown

Generic approachremovesambiguity,makes ERAmethodstransparent

Terminologyrelated

Incompletecommunicationof theregulatorydecision process

Alternativeapproaches

Semiquantitativehazardassessment byexperts

SARS and QSARsto overcomedata gaps

Benefits orconsequenceanalysis

Traditional EISpossible but lessdesirable

No alternative is asappropriate asERA

Consistencyof view

Two paradigms:ecosystemhealthecosystemservices

Required only toease regulatoryprocesses

Emphasis on issuesvaries with time

Many diverse viewsexist

Appropriateuse

At population andcommunitylevels whenuncertaintyprevails

As currentlydefined, never

Generic ERA bestand adaptableto manyproblems

As an aid todecision makingfor narrowlydefined issues

To formally infuseecologicalassessment intodecision making

For choosingbetweenalternativeactions

Limitations Not valid forecosystems

Imposed bynumerousassumptions

Little known aboutpatterns ofchemical useand dispersal

Cannot addresslarge, complexpublic policyissues

Jargon and theimprecisenature ofenvironmentallegislation

Value criteria Specified a prioriby both scienceand the public

Injected bymodelers andthe politicallycorrect

Bounded bypolitical andsocial factors

Must be defined byboth scienceand the public

Required to judgeecologicalsignificance

A function of riskmanagement

Ecology Ecology concernsdominate underthe ecosystemhealth paradigm

Ecologicalcomplexity notreflected byERA

ERA has little to dowith ecology

Ecologicalconcepts oftenoversimplifiedin ERA

Site-specificecological detailincluded

Most pressingdemand

Defining thequestions to beasked

Validation Improvedmethodologicalharmonizationandcommunication

Further definitionof when andwhere ERA isappropriate

Lack of clearlystatedenvironmentalbaselines

Improvement ofthe knowledgebase

Summaryassessment

In ERA thequestions arenot clearlydefined

ERA is amisleading shortcut

ERA is a pragmaticdecision-makingtool forenvironmentalregulation

ERA is appropriatein limitedcircumstancesforenvironmentalmanagement

ERA is aconceptual andmethodologicalextension oftraditional EIS

ERA is aconceptualparadigmproviding timelyadvice toenvironmentaldecision makers


In addition to concerns about validation, Holdway(1997) argues that ERA predictions are likely to bemisleading given our incomplete understanding of thecomplex causal mechanisms governing ecological inter-actions. Lackey (1997) is less sanguine, noting that thepotential for ERA to address complex ecological policyquestions has yet to be fully evaluated. Calow andForbes (1997) and Bartell (1997) caution that many ofthe assumptions and techniques used in ERA remainuntested and of unknown value. Although advocates ofERA have made much of its scientific basis, they remainominously silent on the issue of validation. The ERAparadigm does not explicitly highlight the need forvalidation. Instead, practitioners are assumed to foreseethe need. The evidence, however, suggests that they donot. The fact that there is no validation is often excusedby ERA supporters on the grounds that it is the practice,and not the paradigm, that is at fault on the validationscore. It is not sufficient to impute that it is the practice,rather than the paradigm, that is flawed when the latterjustifies the former.

In its defense Bartell (1997) suggests ERA wouldbenefit from an infusion of the rigorous mathematicaland statistical formulations of risk extant in the engineer-ing disciplines. Among the rigors infused must be anexplicit requirement for the conduct of validationstudies. The methodologies used as part of ERA aresurrogates for actual experience or experimentationwith a specific ecosystem. Thus it is important toestablish the credibility of employed techniques toensure that assessors, regulators, and the public havesufficient confidence in the predictions generated byERAs (Van Horn 1971). Having the benefit of the EIAexperience, ERA should not repeat the mistake of itspredecessor in failing to validate its methods andpredictions. Science demands that predictive claims besubstantiated, and ERA cannot legitimately claim ascientific basis without living up to that same standard.Clearly an important future challenge for ERA will be toinstall confidence in both the paradigm and the tech-niques coopted by its application.

Sources of Ambiguity

To some extent the validation issue is confounded bythe ambiguities extant in procedural applications ofERA. While Suter and Efroymson (1997) admit particu-lar applications of the ERA paradigm have led many toconfuse a feature of ERA practice with ERA, or mixvalues with facts, they find such distinctions easily madein practice. Others would disagree. Lackey (1997)points to a diffuse set of similar paradigms and terminol-ogy as being one source of ambiguity about the nature

and practice of risk assessment. Bartell (1997) addsincomplete communication between regulatory deci-sion makers and risk assessors and the imprecise natureof environmental legislation to the list of ambiguitieshampering our ability to define decision alternativesand the best means of selecting among them. Calow andForbes (1997) reiterate the claim in ecological terms,noting the difficulties with defining ERA problem bound-aries and determining what it is about the environmentwe actually want to protect.

The list of factors contributing to misunderstandingsabout ERA suggests more than mere confusion over theparticulars of its practice. They are fundamental exposi-tional or practical weaknesses of the paradigm which,more than anything, underpin the dichotomy of viewsexpressed by Holdway (1997) on the one hand, andSuter and Efroymson (1997) on the other hand. Al-though one can sympathize with the appeal to refineERA through experience, rather than debate (Bartell1997), with limited resources we must be cognizant ofthe need to think before we act. Far from being anappeal to act only when sufficient information is avail-able, this is an appeal to wisely use what information isavailable. For complex, intractable environmental prob-lems constructive debate is surely as legitimate a tool ofaction as toxicity testing.

The lack of a coherent view on the extent and natureof ERA presents obvious problems for validation. Validat-ing predictions of questionable value is undoubtedly avain use of limited scientific resources. The genericEuropean approach (Van Leeuwen 1997) attempts toremove much of the ambiguity associated with mandateand boundary issues by standardizing ERA within acomputerized modeling environment. Although clearlytransparent, the approach does not overcome the tech-nical issues involved in extrapolating laboratory toxicitytest results to the complexities of the environment(Holdway 1997). This suggests that the lack of predic-tive validation is itself one of the largest sources ofambiguity in ERA. Without the ability conferred byvalidation to quantitative arbitrate between competingmethods, methodological choice is left to the subjectiveselection of the assessor and rancorous, unproductivedebate ensues.

Limitations

Limitations arising from the numerous assumptionsand imprecise nature of much of the knowledge basepertaining to the application of ERA were generallyrecognized by the discussants. As a result, emphasis wasplaced on attempts to increase the scientific credibilityof ERA by addressing identified limitations and reduc-


ing ambiguities to the maximum extent possible. Sev-eral authors argued that a major limitation of ERA wasthe generic assessment approaches that have beendeveloped without focusing on clearly defined prob-lems or target areas with well defined features. VanLeeuwen (1997) believes that ERA has been mainlygeared to function in a generic assessment mode.Lackey (1997), Calow and Forbes (1997), and Bartell(1997), however, advocate the use of ERA for address-ing well-defined technical questions, including establish-ment of environmental baseline conditions, rather thanas a means of addressing complex public policy ques-tions. Defining clear hypotheses that can be rigorouslytested in well-designed research programs should bethe basis of scientifically credible ERA programs. Whenused in generic assessments, ERA is too easily adaptedto environmental, scientific, and political conditionsand becomes more an instrument of environmentalpolicy than science.

Generic approaches have little to do with ecologyand are fraught with numerous assumptions and valuejudgements (Van Leeuwen 1997). Site-specific assess-ments require the development of sophisticated mod-els, use large amounts of data, and involve high costs.There are, therefore, disadvantages in using ERA onboth a generic and site-specific basis. Defining an ERAproblem on a narrow scale makes its solution lesstransferable, but it also makes the assessment moreecologically relevant by requiring fewer assumptionsand/or scientific judgements. Because the nature of theERA process invariably requires the analyst to makemany value-based decisions, even in the most focused ofstudies, limiting application to focused and well-definedissues will help reduce the dependence of ERA onassumptions and value judgements and increase itsscientific credibility.

Another major concern addressed in this series ofpapers is that ERA is typically fraught with type II errorsor that it strives to prove the null hypothesis of no risk.This problem is a serious statistical one when workingonly with single-species toxicity tests (Holdway 1997).When the difficulties of measuring and predicting thebehavior of complex ecosystems are included, it be-comes an even more serious statistical problem. Tomaximize the probability of protecting the environ-ment, experimental designs and statistical analysis mustbe adjusted to minimize the probabilities of falsenegatives (Forbes and Forbes 1994). Improving thestatistical power of test designs to detect existing effectsand minimize type I and II errors can only strengthenthe risk assessment process (Cranor 1993).

Ecology

One of the major concerns voiced by several authorswas the general lack of ecological realism and sophistica-tion in the conception and practice of ERA. Of neces-sity, ecosystem models must be relatively simple indesign, yet in ERA many models have been oversimpli-fied (Bartell 1997, Holdway 1997). Calow (1994) statedthat we still do not know enough about ecologicalsystems to be able to identify what it is we want to protectand, hence, what we should be measuring. Holdway(1997) believes that as a result ecological science hasbeen oversimplified in order to cater to the needs oflegislators, politicians, and attorneys. In support of thisposition, Lackey (1997) points out that a serious misuseof the ERA process has been to substitute politicalvalues and priorities for those of public concern, thusshifting the emphasis of ERA from the scientific to thepolitical arena.

Questions have also been raised relative to the issueof making decisions of ecosystem risk based on general-ized ecological responses such as the survivorship andfecundity of representative species. Typically, damage toindividual organisms or populations is used to extrapo-late to effects at the community or ecosystem level.Because of our rudimentary understanding of ecosys-tem structure, function, and processes, extrapolatingacross multiple-trophic levels is a major limitation inERA. Reasonable attempts should be made at integrat-ing across multiple levels (Bartell 1997) and interpret-ing the ecological meaning of change at one particularlevel to effects at multiple trophic levels (Holdway1997). In the same context, assessing the ecologicalrelevance of some predetermined and arbitrary percent-age loss of a species (e.g., 10%) may be somewhatunrealistic because we do not currently understandwhat a partial species loss, or even a total species loss,ultimately means to the productivity, stability, or fitnessof the entire ecosystem.

If the observable loss of a species, or population, isthe ecological endpoint on which decisions aboutecological harm are based, then the use of an arbitrarypercentage loss defeats the predictive purposes of ERA.For example, a 10% loss in a population has to beobserved before risk can be assigned. When the risk isknown, the harm has already occurred. The assessment,therefore, loses much of its predictive function simplybecause its conclusions have become largely retrospec-tive in nature. In reality, individual organisms becomesublethally stressed and their physiological systems com-promised before more serious effects are manifested atthe higher levels of biological organization. For ex-ample, the biochemical and physiological systems of


organisms become compromised before changes inreproductive and population attributes are observed. IfERA is to be truly predictive in nature, assessmentapproaches must incorporate the ability to address thesublethal changes in organisms that function as early-warning signals or indicators of impending environmen-tal damage. Unfortunately, most ERAs currently fail tomeet this crucial need.

Appropriate Use

Views on validation, ambiguities, limitations, andecology invariably influenced the opinions of authorson where and when ERA should be used. Holdway(1997), having argued that ERA was neither appropri-ately validated nor ecologically sophisticated, con-cluded that as currently defined ERA should never beused. In its place semiquantitative hazard assessmentsthat evaluate the potential for deleterious effects occur-ring in an ecosystem given the presence of defined toxicagents or pollutants were suggested. Lackey (1997),noting the tendency to apply ERA to every environmen-tal problem at hand, called for a clear definition of thecircumstances under which ERA should be used as adecision-making tool. Bartell (1997) viewed ERA as acredible means of infusing ecological information intothe wider social environmental decision-making pro-cess. Calow and Forbes (1997), however, restricted theuse of ERA to population and community-level relatedproblems dominated by uncertainty.

Although advocates offer ERA as a panacea forenvironmental decision making, most authors clearlyargued that ERA should be applied only in limitedcircumstances. The split over circumstances occurredlargely between those who viewed ERA as a genericprocess capable of screening, ranking, and expeditingthe solution of environmental problems (e.g., VanLeeuwen 1997), and those who viewed ERA as mostpotentially useful for addressing site-specific decisions(e.g., Bartell 1997, Lackey 1997). Future methodologi-cal developments will ultimately determine which viewprevails. For example, improved, validated modelingapproaches would clearly enhance the effectiveness ofERA as a screening tool, while the infusion of moreecological principles into the practice of ERA is likely toincrease the specificity of its conclusions, thereby limit-ing its use to narrowly defined questions. It is difficult topredict into which role ERA will ultimately evolve. That,in part, will depend on the enthusiasm and efforts ofboth its detractors and proponents. What is clear,however, is that ERA will not effectively fulfill either roleif further conceptual and methodological develop-ments do not occur.

Most Pressing Demands

Despite divergences in opinion regarding the utilityof risk assessment, there was a consensus on the need tofurther develop and refine the approach. The specificrecommendations of each author varied. Calow andForbes (1997) focused on the need to clarify theecological questions being asked. Van Leeuwen (1997)stressed improvements in the harmonization and com-munication of methodological developments within theassessment and regulatory communities. Predictive vali-dation and the selection of environmental baselines,however, undoubtedly represent the greatest challengesto ERA. Without confidence in the accuracy of thepredictive techniques used and knowledge of the base-line conditions against which to compare predictedchanges, ERA cannot hope to either effectively predictthe consequences of human action or judge the prob-able significance of its predictions. And without theseabilities, ERA will almost certainly fail to convince itssceptics or live up to its promise of injecting scientifi-cally sound advice into the environmental decision-making process.

Conclusions

The opinions offered by discussants in this debate donot allow us to either condemn risk assessment as apseudoscience or to treat it as a fully adequate tool fordescribing, ranking, and addressing complicated envi-ronmental problems. ERA has undergone significantchange since it first appeared little more than a decadeago. Revision of the US EPA risk assessment guidelines,numerous methodological conferences, and specialtybooks all point to considerable scientific and regulatoryinterest in both the practice and development of ERA.Risk assessment will undoubtedly evolve to meet theconcerns of its critics on many issues (e.g., validation,improved definition of problem boundaries), but onmany issues the concerns of its critics will not easily bemet (e.g., demonstration of ecological relevance, ex-trapolation between levels of biological organization),which will serve to limit the functional role of ERAwithin the wider activity of environmental decisionmaking and management.

In point of fact there are no methods that evaluate orpredict the status of ecological systems that are withoutlimitations. Every approach has its unique set of advan-tages and limitations. Furthermore, every approachwill, to some extent, share characteristics in commonwith other approaches aimed at a similar, genericobjective. We have indicated some of the agreed onstrengths and limitations of ERA in relation to theecological processes we seek to understand and man-


age. Recognizing the strengths and weaknesses of anyparadigm, ERA included, is the best means of attempt-ing to delineate its appropriate use and reinforce itsassociated scientific credibility. As we learn more aboutecosystems and what it is we want or need to protect inthem, scientifically credible methods of assessment willhave increasingly important roles to play in improvingthe practice of environmental management. This im-plies that ERA, as it predecessors have done, will undergo asignificant transformation as it establishes its place in thecollection of methods we use to provide timely technicaladvice to environmental decision makers.

Literature CitedBartell, S. M. 1997. Ecological risk assessment: progressing

through experience or stalling in debate. EnvironmentalManagement (this issue).

Calow, P. 1994. Ecotoxicology: What are we trying to protect?Environmental Toxicology and Chemistry 13:1549.

Calow, P., and V. E. Forbes. 1997. Science and subjectivity inthe practice of ecological risk assessment. EnvironmentalManagement (this issue).

Cranor, C. F. 1993. Regulating toxic substances. OxfordUniversity Press, Oxford, 252 pp.

Forbes, V. E., and T. L. Forbes. 1994. Ecotoxicology in theoryand practice. Chapman and Hall, London, 247 pp.

Holdway, D. A. 1997. Truth and validation in ecological riskassessment. Environmental Management (this issue).

Lackey, R. L. 1997. Ecological risk assessment: Use abuse andalternatives. Environmental Management (this issue).

Montgomery, D. C., and E. A. Peck. 1982. Introduction tolinear regression analysis. John Wiley & Sons, New York,504 pp.

Power, M. 1993. The predictive validation of ecological andenvironmental models. Ecological Modelling 68:33–50.


Suter, G. W., II, and R. A. Efroymson. 1997. Controversies inecological risk assessment: Assessment scientists respond.Environmental Management (this issue).

Van Horn, R. L. 1971. Validation of simulation results. Manage-ment Sciences 17:247–258.

Van Leeuwen, C. J. 1997. Ecological risk assessment: An inputfor decision-making. Environmental Management (this issue).

Wilson, R., and E. A. C. Crouch. 1987. Risk assessment andcomparisons: an introduction. Science 236:267–270.


special section perspectives of the scientiﬁc community on … · 2003-10-13 · sciences and...

Documents