ES/RP 531 Fundamentals of Environmental Toxicology Fall 2003ESRP531 Lect 24 ProbEcoR.doc Page 1 of 8December 8, 2003Lecture 24: Ecological Risk Characterization—Probabilistic PerspectivesI. Probabilistic Ecological Risk CharacterizationA. One major conundrum with characterizing risk to ecosystems (or populations,communities, etc. within) is that there is a tremendous diversity of species toprotect. Thus ERA has challenges not faced by assessment of risk for humanhealth (which can rely on “simple” animal models—i.e., rats and dogs).B. To characterize risk, there are essentially two methods—deterministic andprobabilistic1. Quick review of deterministic risk characterizationa. Divide the level of exposure (a point estimate) by a toxicological endpoint(NOEL or LC50) for the most sensitive species tested.b. The ratio is judged acceptable (or as EPA says, below their Level ofConcern, LOC) or not; i.e., whether the deterministic risk characterizationis acceptable or not is risk management.c. For ex., with an assessment of pesticide risk for causing adverse effects inan endangered species, the ratio of exposure to LC50 should be 0.05 (i.e., a20-fold safety factor).C. Probabilistic Techniques1. Combine the distribution of possible exposures (modeled environmentalconcentrations and from real environmental measurements) with thedistribution of acute and chronic effects endpoints for various individualspecies (= species sensitivity distributions).D. Species sensitivity distributions are databases of how different species and/or taxarespond to a contaminant.1. The distributions usually represent the fraction of all the known tested speciesresponding to a given toxicity endpoint (i.e., the dose-response point estimate,LC50 or NOEC).2. The species sensitivity distributions (SSD) for responses to chlorpyrifos areshown below in Figures 1, 2, and 3(data from van den Brink et al. 2002, inSpecies Sensitivity Distributions in Ecotoxicology, Posthuma et al. (ed.), CRCPress, pp. 155-193).3. The SSD curves are derived from logistic regression modeling.4. Note that the y axis is the proportion of all the animals or plants tested in anygroup that exceed a given toxicity endpoint (in the case of Figure 1 and 2, theendpoint is the EC50 or LC50).ES/RP 531 Fundamentals of Environmental Toxicology Fall 2003ESRP531 Lect 24 ProbEcoR.doc Page 2 of 8Figure 1: Lab-based SSD curves for chlorpyrifos by groups of aquatic organisms.5. Note in Figure 2 below that the logistic curve for lab-based EC50’s ispredictive of the field-based EC50’s.a. Note that the experiment that generated the field data were an outdoorditch microcosm rather than a natural aquatic system.6. Also note that the distribution of field and lab based EC50’s were onlycomparable when Arthropods, rather than all invertebrates were compared.7. Arthropods are extraordinarily sensitive to chlorpyrifos, which is anorganophosphate insecticide.Figure 2. SSD curves for chlorpyrifos; comparison of laboratory based EC50determinations and a semi-field exposure.ES/RP 531 Fundamentals of Environmental Toxicology Fall 2003ESRP531 Lect 24 ProbEcoR.doc Page 3 of 8Figure 3. SSD curves for the acute toxicity and chronic toxicity of the fungicidecarbendazim. The arrow lines indicate concentrations actually tested in amicrocosm test 3.3 µg/L (first line, left to right), 33 (2nd line), 100 (3rd line),330 (4th line), and 1000 (5th line). The NOEC for chronic exposure bydirect and indirect effects was 3.3 µg/L, corresponding to a fractionaffected of about 25%.E. The distribution of concentrations can be superimposed on the SSD.1. In Figure 4 below, I’ve illustrated such an imposition, but instead of showingthe SSD as a logistic curve, I’ve plotted it as a normal distribution.2. The distribution of hypothetical data illustrated in Figure 4 can be changed toa cumulative frequency distribution as shown in Figure 5.3. The cumulative frequency distribution can be linearized as shown in Figure 6from Solomon et al. 2000. Probabilistic risk assessment of agrochemicals inthe environment. Crop Protection 19:649-655.))a. An example of this probabilistic ecological risk characterization isillustrated for atrazine. (Figure 7)b. Using the log distribution of residues measured for the herbicide atrazine,and the log distribution of the LC5 (as well as the LC50), probabilisticecorisk of atrazine was characterized by Solomon et al (1996) [Solomon,K. R., D. B. Baker, R. P. Richards, K. r. Dixon, S. J. Klaine, T. W.LaPoint, R. J. Kendall, C. P. Weisskopf, J. M. Giddings, J. P. Giesy, L. W.Jr. Hall, and W. M. Williams. 1996. Ecological risk assessment ofatrazine in North American surface waters. Environ. Toxicol. Chem.15(1):31-76.]c. This method enabled a calculation of the margin of safety given thedistribution of environmental residues of atrazine and the distribution ofspecies responses.ES/RP 531 Fundamentals of Environmental Toxicology Fall 2003ESRP531 Lect 24 ProbEcoR.doc Page 4 of 8F. What is acceptable risk?1. One benchmark that the EPA is considering for acceptable ecological risk isthat no more than the 10th percentile of species would be exposed tocontaminant residue levels exceeding the LC5 (lethal concentration to 5% ofthe test population).2. In the graph represented in Figure 4, the overlapped area would have to besmaller than a certain guideline to be acceptable.3. Remember, however, what is acceptable is a risk management decision, not atestable hypothesis. However, Figure 9 below illustrates an approach to usinga joint probability curve (aka exceedance frequency curve) to determine whatis acceptable. Figure 8 illustrates how the joint probability curve is derivedfrom the logarithmic plots of the cumulative frequency distributions forexposure and toxicity endpoints.Figure 4. Overlap of distribution of exposure values (i.e., concentrations or doses)and distribution of toxicity values (suitable toxicological endpoint such asLC50, LC5, or NOAEC).Figure 5. Derivation of cumulative frequency distribution (right hand side) fromfrequency distribution (left hand side) of overlapping exposure and toxicitydistribution functions (from Solomon, K., J. Giesy, and P. Jones. 2000.ES/RP 531 Fundamentals of Environmental Toxicology Fall 2003ESRP531 Lect 24 ProbEcoR.doc Page 5 of 8Probabilistic risk assessment of agrochemicals in the environment. CropProtection 19:649-655.)Figure 6. Linearization of cumulative frequency