Risk characterization is the final stage of an ecological risk assessment in which results of exposure and effects analyses are integrated to provide decision-makers with a risk estimate—the probability of adverse effects of exposure to a chemical stressor—and its associated uncertainty. A decision-maker does not want to make a decision on the basis of a belief that a pesticide is unlikely to yield an adverse effect and discover afterwards that it did yield an adverse effect. That is often referred to as avoiding a Type II error. For example, if the US Environmental Protection Agency (EPA) proposes registering a pesticide with a specific label use, it needs to know how much confidence there is that doing so will lead to the desired outcome, such as reduction in the abundance of the target species, and not result in jeopardy to a listed species. It is most useful if the risk estimate and its associated uncertainty are expressed in a quantitative manner—for example, “there is a 20% ± 10% probability of a 25% reduction in the population growth rate as a result of this action.”
In addition to generating a quantitative risk estimate, risk characterization includes a narrative discussion (termed the risk description) that includes discussion of data gaps, lack of knowledge, natural variability, and other factors that might influence confidence in the risk estimate. The discussion can be viewed as a weight-of-evidence description in which the strengths and weaknesses of each assumption and each type of data used in the risk assessment are discussed. At the risk assessor’s discretion, the narrative might be summarized in a table that lists all the lines of evidence and their various weights that are scored on the basis of relevance, degree of quantification, variability, and robustness of the data analysis (see, for example, Linkov et al. 2009; Exponent 2010). The discussion provides guidance to the decision-maker about which aspects of the risk assessment are more reliable, where there are greater unknowns, and how natural variability or lack of knowledge might hinder the development of a more accurate estimate of risk.
There are many practical methods for combining the results (with their associated uncertainties) of exposure and effects analyses to provide an estimate of risk and the confidence in it. Two broad approaches have been used; one is a
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 148
5 Risk Characterization Risk characterization is the final stage of an ecological risk assessment in which results of exposure and effects analyses are integrated to provide deci- sion-makers with a risk estimate—the probability of adverse effects of exposure to a chemical stressor—and its associated uncertainty. A decision-maker does not want to make a decision on the basis of a belief that a pesticide is unlikely to yield an adverse effect and discover afterwards that it did yield an adverse ef- fect. That is often referred to as avoiding a Type II error. For example, if the US Environmental Protection Agency (EPA) proposes registering a pesticide with a specific label use, it needs to know how much confidence there is that doing so will lead to the desired outcome, such as reduction in the abundance of the target species, and not result in jeopardy to a listed species. It is most useful if the risk estimate and its associated uncertainty are expressed in a quantitative manner— for example, “there is a 20% ± 10% probability of a 25% reduction in the popu- lation growth rate as a result of this action.” In addition to generating a quantitative risk estimate, risk characterization includes a narrative discussion (termed the risk description) that includes discus- sion of data gaps, lack of knowledge, natural variability, and other factors that might influence confidence in the risk estimate. The discussion can be viewed as a weight-of-evidence description in which the strengths and weaknesses of each assumption and each type of data used in the risk assessment are discussed. At the risk assessor’s discretion, the narrative might be summarized in a table that lists all the lines of evidence and their various weights that are scored on the basis of relevance, degree of quantification, variability, and robustness of the data analysis (see, for example, Linkov et al. 2009; Exponent 2010). The discus- sion provides guidance to the decision-maker about which aspects of the risk assessment are more reliable, where there are greater unknowns, and how natu- ral variability or lack of knowledge might hinder the development of a more accurate estimate of risk. There are many practical methods for combining the results (with their as- sociated uncertainties) of exposure and effects analyses to provide an estimate of risk and the confidence in it. Two broad approaches have been used; one is a 148
OCR for page 148
Risk Characterization 149 deterministic concentration-ratio approach, which compares point estimates of exposure and effect concentrations, and the other is a probabilistic approach, which evaluates the probability that exposure to a chemical will lead to a speci- fied adverse effect at some future time. The latter is technically sound, and the former is ad hoc (although commonly used) and has unpredictable performance outcomes. EPA uses the concentration-ratio approach for its assessments. In biological opinions on salmon, the National Marine Fisheries Service appears to favor a probabilistic approach that is based on population modeling. The Fish and Wildlife Service seems not to use a quantitative approach, either concentra- tion-ratio or probabilistic, for risk characterization. CONCENTRATION-RATIO APPROACH The concentration-ratio approach, which is commonly used by EPA for Step 1 and 2 assessments (see Figure 2-1), does not estimate risk (the probability of an adverse effect) itself but rather relies on there being a large margin be- tween a point estimate of the most likely maximum pesticide environmental concentration and a point estimate of the lowest concentration at which a speci- fied adverse effect might be expected (EPA 2004). The superficial attraction of this approach is that one feels confident that a decision will not lead to an ad- verse effect (that is, a Type II error will be avoided) if sufficiently large margins are used. There is a belief that the larger the margin between the estimated expo- sure and the response threshold, the greater the certainty (or the smaller the un- certainty). The flaws in that approach are that it does not account for the proba- bility of an adverse effect before worst-case assumptions are applied and that it does not calculate how the use of the assumptions modifies that probability. Given that approach, decision-makers do not know what the probability of an adverse effect is, but they hope that they can assume (or be reassured) that it is small. However, such an assumption is not reliable. If they or their constituen- cies have doubts, the common response is to widen the margin with additional conservative assumptions, including addition of specific uncertainty factors or more stringent, and possibly implausible, exposure scenarios. However, simply widening the gap indefinitely might lead to decisions that limit pesticide use to a greater extent than is intended by policy and will not meaningfully express the underlying probability of an adverse effect. For pesticides, as evaluated by EPA, the concentration ratio is quantified in the form of a risk quotient (RQ) that might be less or greater than some speci- fied level of concern (LOC). However, an RQ is not actually a risk estimate in that it provides no information about the probability of an adverse effect. Thus, although an RQ of 10 is several times higher than most numerical LOCs, there is no fixed relationship between RQs and the probability of an adverse effect on a listed species. Therefore, it is not possible to determine what an RQ of 10 means with respect to a possible adverse effect on a listed species. Nor is there a fixed relationship for comparing the difference between, for example, RQs of 10 and
OCR for page 148
150 Assessing Risks to Endangered and Threatened Species from Pesticides 100 with respect to the probability of an adverse effect. Theoretically, an RQ of 100 means a greater probability of an adverse effect than an RQ of 10, but one cannot determine whether the difference in probability between the two RQs is substantial or negligible or whether the final error associated with the risk esti- mate is appropriate for the management needs. Thus, although RQs are often used by EPA for Step 2 assessments that might trigger later, more refined and focused assessments for listed species, the committee concludes that RQs are not appropriate for assessments for listed species or indeed for any application in which it is desired to base a decision on the probabilities of various possible outcomes. Furthermore, the committee con- cludes that adding uncertainty factors to RQs to account for lack of data (on formulation toxicity, synergy, additivity, or any other aspect) is unwarranted because there is no way to determine whether the assumptions being used sub- stantially overestimate or underestimate the probability of an adverse effect. The committee has not been asked about and is not commenting on policy decisions about what level of risk is acceptable or how conservative the agencies should be in establishing an “acceptable” risk level when considering jeopardy to listed species. PROBABILISTIC APPROACH Risk is defined as the probability of an adverse effect (Burmaster 1996). Thus, natural tools for quantifying and analyzing risk are probability, statistics, and the algebra of random variables, and an alternative to the deterministic con- centration-ratio approach is a probabilistic one. In the probabilistic approach, the probability that a decision will lead to an adverse effect is calculated from the available information and then used to support an informed decision (again, the committee is purposefully refraining from a discussion of what an “acceptable” probability of risk might be). The probabilistic approach requires integration of the uncertainties (from sampling, natural variability, lack of knowledge, and measurement and model error) in the exposure and effects analyses by using probability distributions, rather than single point estimates, for uncertain quanti- ties (EPA 2001). The distributions are then integrated mathematically to calcu- late the risk as a probability and its associated uncertainty in that estimate. Ulti- mately, decision-makers are provided with a risk estimate that reflects the probability of exposure to a range of pesticide concentrations and the magnitude of an adverse effect (if any) to the exposures that answers the fundamental ques- tion, What is the probability that registration of this pesticide will lead to a spec- ified adverse effect on a listed species or its critical habitat? Implementing a probabilistic approach requires three primary actions on the part of a risk assessor: (1) Describe uncertain variables with distributions and recognize that not all variables in a model or an analysis need be treated this way. The task can be
OCR for page 148
Risk Characterization 151 made considerably more tractable if only variables identified as key drivers via a sensitivity analysis are defined by distributions. The methods and problems in fitting or otherwise deriving the distributions from data are not discussed here because a large literature is available on these topics (see, for example, EUFRAM 2006; Warren-Hicks and Hart 2010). However, the models or meas- urements used to estimate exposure concentrations are capable of providing re- sults as distributions, and results of the multispecies toxicity testing that is al- ready part of the registration process could be expressed as discrete exposure- response distributions or combined into a species sensitivity distribution. (2) Propagate the uncertainty through to distributions of exposure and ef- fect by using one of several calculation methods. The most readily accessible of these (in terms of software and experience) are Monte Carlo analysis (including second-order methods), probability-bounds analysis, and Bayesian methods (Warren-Hicks and Hart 2010) (see Chapter 2 for recommendations of method selection). (3) Integrate exposure and effect estimates to calculate risk. Aldenberg et al. (2001) have shown that a variety of risk-estimation methods calculate the same probability that a stated exposure concentration will produce a specified adverse effect given a specific exposure-response relationship. Such methods include discrete summation for expected risk (Cardwell et al. 1999), ecological risk overlap plot (Van Straalen 2002), numerical integration of risk-distribution curves (Parkhurst et al. 1996; Solomon and Takacs 2001; Warren-Hicks et al. 2001), and various area-under-the-curve (AUC) methods, such as exceedance profile plots (ECOFRAM 1999ab; Giesy et al. 1999; Solomon and Takacs 2001), cumulative profile plots (Aldenberg et al. 2001), and cumulative distribu- tion functions of risk estimates (Aldenberg et al. 2001; EUFRAM 2006). The area under the joint probability curve is considered as a numerical measure of the risk to a species posed by a chemical stressor (Giddings et al. 2005), a value that a decision-maker would seek to minimize. The committee has concluded that EPA and the Services can begin the tran- sition now from concentration ratios to established, scientifically defensible statis- tical-inference methods for propagating uncertainties in exposure and effect through to a risk estimate for both individual receptors (Step 2) and populations of receptors (Step 3). The committee recognizes the pragmatic demands of the pesti- cide registration process and encourages EPA and the Services to consider proba- bilistic methods that have already been successfully applied to pesticide risk as- sessments (Odenkirchen 2003 [EPA’s Terrestrial Investigation Model v 2.0]; Giddings et al 2005; Warren-Hicks and Hart 2010), have otherwise appeared fre- quently in the technical literature, are familiar to many risk-assessment practition- ers, can be implemented with commercially available software, and are most readily explicable to decision-makers, stakeholders, and the public. The committee also notes that transitioning to a probabilistic approach can begin with simple reg- istrations (for example, pesticides for use on a few crops or in a small geographic area) and will not require that all variables be immediately represented with prob-
OCR for page 148
152 Assessing Risks to Endangered and Threatened Species from Pesticides ability distributions (that is, sensitivity analyses can be used to identify key param- eters that are important to represent as probability distributions). CONCLUSIONS Inclusion of uncertainty factors to account for lack of various data is unwarranted because there is no way to determine whether the assumptions be- ing used substantially overestimate or underestimate the probability of adverse effects. RQs are not appropriate for risk assessments or for any application in which it is desired to base a decision on the probabilities of the various possible outcomes. EPA (for Step 2 assessments) and the Services (for Step 3 assessments) should use established, scientifically defensible, statistical methods to calculate risk as a probability to assist decision-makers’ understanding of the potential consequences of their decisions. A number of existing probabilistic methods have been shown to be ap- plicable and practical for ecological risk assessments that involve pesticides. The transition from concentration-ratio to probabilistic approaches can begin now, starting with simple registrations, focusing on a small set of sensitive key parameters, and drawing on the considerable literature and guidance on probabilistic approaches. REFERENCES Aldenberg, T., J.S. Jaworska, and T.P. Traas. 2001. Normal species sensitivity distribu- tions and probabilistic ecological risk assessment. Pp. 49-102 in Species Sensi- tivity Distributions in Ecotoxicology, L. Posthuma, G.W. Suter, II, and T.P. Traas, eds. Boca Raton: CRC Press. Burmaster, D.E. 1996. Benefits and costs of using probabilistic techniques in human health risk assessments—with an emphasis on site-specific risk assessments. Hum. Ecol. Risk Assess. 2(1):35-43. Cardwell, R.D., M.S. Brancato, J. Toll, D. DeForest, and L. Tear. 1999. Aquatic ecologi- cal risks posed by tributyltin in United States surface waters: Pre-1989 to 1996 data. Environ. Toxicol. Chem. 18(3):567-577. ECOFRAM (Ecological Committee on FIFRA Risk Assessment Methods).1999a. ECOFRAM Aquatic Report. Office of Pesticide Programs, U.S. Environmental Protection Agency, Washington, DC. May 4, 1999 [online]. Available: http:// www.epa.gov/oppefed1/ecorisk/aquareport.pdf [accessed Nov. 21, 2012]. ECOFRAM (Ecological Committee on FIFRA Risk Assessment Methods). 1999b. ECOFRAM Terrestrial Draft Report. Office of Pesticide Programs, U.S. Envi- ronmental Protection Agency, Washington, DC. May 10, 1999 [online]. Avail- able: http://www.epa.gov/oppefed1/ecorisk/terrreport.pdf [accessed Nov. 21, 2012]. EPA (U.S. Environmental Protection Agency). 2001. Risk Assessment Guidance for Superfund: Volume III - Part A. Process for Conducting Probabilistic Risk As-
OCR for page 148
Risk Characterization 153 sessment. EPA 540-R-02-002, OSWER 9285.7-45. Office of Emergency and Remedial Response, U.S. Environmental Protection Agency, Washington, DC [online]. Available: http://www.epa.gov/oswer/riskassessment/rags3adt/pdf/rags3 adt_complete.pdf [accessed Nov. 21, 2012]. EPA (U.S. Environmental Protection Agency). 2004. Overview of the Ecological Risk Assessment Process in the Office of Pesticide Programs, U.S. Environmental Protection Agency: Endangered and Threatened Species Effects Determina- tions. Office of Prevention, Pesticides and Toxic Substances, Office of Pesti- cide Programs, U.S. Environmental Protection Agency, Washington DC. Janu- ary 23, 2004 [online]. Available: http://www.epa.gov/oppfead1/endanger/con sultation/ecorisk-overview.pdf [accessed Aug. 29, 2012]. EUFRAM. 2006. Concerted Action to Develop a European Framework for Probabilistic Risk Assessment of the Environmental Impacts of Pesticides. Fifth Framework Programme, Quality of Life and Management of Living Resources, European Commission, Brussels [online]. Available: http://www.eufram.com/ [accessed Nov. 21, 2012]. Exponent. 2010. Report on Guidance for a Weight of Evidence Approach in Conducting Detailed Ecological Risk Assessments (DERA) in British Columbia. Prepared for The Ministry of Environment, British Columbia, by Exponent, Inc., Belle- vue, WA. June 2010 [online]. Available: http://www.sabcs.chem.uvic.ca/woe. html [accessed Aug. 2012]. Giddings, J.M., T.A. Anderson, L.W. Hall, Jr., A.J. Hosmer, R.J. Kendall, R.P. Richards, K.R. Solomon, and W.M. Williams. 2005. Atrazine in North American Surface Waters: A Probabilistic Aquatic Ecological Risk Assessment. Pensacola, FL: SETAC Press. 392 pp. Giesy, J.P., K.R. Solomon, J.R. Coats, K.R. Dixon, J.M. Giddings, and E.E. Kenaga. 1999. Chlorpyrifos: Ecological risk assessment in North American aquatic en- vironments. Rev. Environ. Contam. Toxicol. 160:1-129. Linkov, I., D. Loney, S. Cormier, K. Satterstrom, and T. Bridges. 2009. Weight-of- evidence evaluation in environmental assessment: Review of qualitative and quantitative approaches. Sci. Total Environ. 407(19):5199-5205. Odenkirchen, E. 2003. Evolution of OPP's Terrestrial Investigation Model Software and Programming to Meet Technical/Regulatory Challenges. Presentation at EUro- pean FRamework for Probabilistic Risk Assessment of the Environmental Im- pacts of Pesticides Workshop, June 5-8, 2003. Bilthoven, the Netherlands [online]. Available: http://www.epa.gov/oppefed1/ecorisk/presentations/eufram _overview.htm [accessed Mar. 28, 2013]. Parkhurst, B.R., W. Warren-Hicks, T. Etchison, J.B. Butcher, R.D. Cardwell and J. Vo- loson. 1996. Methodology for Aquatic Ecological Risk Assessment. RP91- AER-1 1995. Water Environment Research Foundation, Alexandria, VA. Solomon, K.R., and P. Takacs. 2001. Probabilistic risk assessment using species sensi- tivity distributions. Pp. 285-314 in Species Sensitivity Distributions in Ecotoxi- cology, L. Posthuma, G.W. Suter, II, and T.P. Traas, eds. Boca Raton: CRC Press. Van Straalen, N.M. 2002. Threshold models for species sensitivity distributions applied to aquatic risk assessment for zinc. Environ. Toxicol. Pharmacol. 11(3-4):167- 172. Warren-Hicks, W.J., and A. Hart, eds. 2010. Application of Uncertainty Analysis to Eco- logical Risks of Pesticides. Pensacola, FL: SETAC Press. 228 pp.
OCR for page 148
154 Assessing Risks to Endangered and Threatened Species from Pesticides Warren-Hicks, W.J., B.R. Parkhurst, and J.B. Butcher. 2001. Methodology for aquatic ecological risk assessment. Pp. 345-382 in Species Sensitivity Distributions in Ecotoxicology, L. Posthuma, G.W. Suter, II, and T.P. Traas, eds. Boca Raton: CRC Press.