Emerging Technologies for Environmental ... - ACS Publications

Article pubs.acs.org/est

Emerging Technologies for Environmental Remediation: Integrating Data and Judgment Matthew E. Bates,*,† Khara D. Grieger,‡ Benjamin D. Trump,§ Jeffrey M. Keisler,⊥ Kenton J. Plourde,∥ and Igor Linkov*,† †

Environmental Laboratory, Engineer Research and Development Center, U.S. Army Corps of Engineers, 696 Virginia Road, Concord, Massachusetts 01742, United States ‡ RTI International, 3040 East Cornwallis Road, Research Triangle Park, North Carolina 27709, United States § School of Public Health, University of Michigan, 1415 Washington Heights, Ann Arbor, Michigan 48109, United States ⊥ College of Management, University of Massachusetts Boston, 100 Morrissey Boulevard, Boston, Massachusetts 02125, United States ∥ Contractor to U.S. Army Corps of Engineers, SOL Engineering Services, 696 Virginia Road, Concord, Massachusetts 01742, United States S Supporting Information *

ABSTRACT: Emerging technologies present significant challenges to researchers, decision-makers, industry professionals, and other stakeholder groups due to the lack of quantitative risk, benefit, and cost data associated with their use. Multi-criteria decision analysis (MCDA) can support early decisions for emerging technologies when data is too sparse or uncertain for traditional risk assessment. It does this by integrating expert judgment with available quantitative and qualitative inputs across multiple criteria to provide relative technology scores. Here, an MCDA framework provides preliminary insights on the suitability of emerging technologies for environmental remediation by comparing nanotechnology and synthetic biology to conventional remediation methods. Subject matter experts provided judgments regarding the importance of criteria used in the evaluations and scored the technologies with respect to those criteria. The results indicate that synthetic biology may be preferred over nanotechnology and conventional methods for high expected benefits and low deployment costs but that conventional technology may be preferred over emerging technologies for reduced risks and development costs. In the absence of field data regarding the risks, benefits, and costs of emerging technologies, structuring evidence-based expert judgment through a weighted hierarchy of topical questions may be helpful to inform preliminary risk governance and guide emerging technology development and policy.

■

contaminated sites.1,2 In light of this, we propose the use of qualitative subject expert interviews alongside iterative decision analysis to serve as an first step for assessing the risks and benefits of emerging technologies until more quantitative data can be developed. Here, we specifically examine two emerging technologies that have been proposed as interesting alternatives to conventional remediation: nanoremediation and synthetic bioremediation. Nanoremediation involves the application of reactive engi-

INTRODUCTION There is currently a significant lack of quantitative field data and rigorous analysis regarding the use of emerging technologies in environmental remediation. Despite these limitations, policymakers and regulators need to make ongoing technology funding and governance decisions. Decisions based on the uncertain risks and benefits of emerging technologies must be considered in the context of the cost and inefficiencies of conventional remediation methods. Making timely and appropriate long-term remediation decisions in the absence of abundant data is also challenging for environmental engineers, researchers, and site managers. Furthermore, the stakes are largethe United States federal government is estimated to spend several hundred million dollars per year remediating © 2015 American Chemical Society

Received: Revised: Accepted: Published: 349

February 21, 2015 November 17, 2015 November 18, 2015 November 18, 2015 DOI: 10.1021/acs.est.5b03005 Environ. Sci. Technol. 2016, 50, 349−358

Article

Environmental Science & Technology

Figure 1. Decision-analytic framework showing primary and secondary criteria relevant for choosing a remediation technology. Different scientific experts provide judgments based on knowledge of scientific data, which are represented as evaluation scores for each technology for each criterion. Relative importance weights integrate individual criteria scores into a total relative technology suitability score.

technology development and use27,28 by providing a method to compare emerging and existing technologies and determine whether pursuing new solutions is likely to provide benefits worth the costs involved. Paired with supplemental analyses (such as scenario analysis and value of information (VoI)29), MCDA can help guide and prioritize research areas with regard to the potential impacts of emerging technologies, including synthetic biology.30 Using this framework, stakeholders can provide input on the decision criteria used while experts evaluate different policy or technological alternatives based on how they are expected to perform with respect to the provided criteria.15,30 MCDA uses weights and normalization to aggregate estimated technology performance scores across multiple criteria. Where the weights or scores are uncertain, as with emerging technologies, distributions can be fit to the estimates and Monte Carlo simulations can identify the probability of one technology outperforming another. In cases where it is difficult to reliably differentiate preference among the decision criteria, the related stochastic multiobjective acceptability analysis (SMAA)31 method uses Monte Carlo simulations to explore the entire weight space and produce an acceptability index showing which of the myriad potential weighting schemes would make each alternative seem most preferred. A key difference between uncertain MCDA and SMAA is that the former explores uncertainty within a defined space reflecting differences in expert or decision maker preferences and beliefs, whereas the latter does not use information from experts or decision makers to inform criteria preferences (weighting). MCDA can be used with any mix of quantitative and qualitative inputs, and using MCDA with subjective inputs from expert interviews creates limitations. As such, this combination is best suited to provide preliminary indicators to guide policymaking and decision making in initial stages of the process (when quantitative data may be most lacking), rather than to concretely prescribe a path for ultimate governance and technological deployment. Even a preliminary assessment using MCDA can be helpful to innovators and stakeholders identifying general areas of potential risk and benefit relevant for making early risk governance or technology advancement decisions. MCDA has emerged as an important way to quantitatively evaluate alternative actions in a transparent and reproducible manner, noting appropriate data-quality qualifications, and has been recommended by several high-level reports to support emerging technology decisions.32−34 In the context

neered nanomaterials, such as nanoscale zerovalent iron (nZVI), to degrade environmental contaminants.3 Synthetic bioremediation involves applications of synthetic biology that take advantage of natural metabolic pathways of engineered bacteria to reduce or degrade environmental pollutants by metabolizing them.4,5 Despite the theoretical promise of these emerging techniques, concerns have been raised regarding the environmental, health, and safety risks of engineered nanomaterials6−8 and the risks and ethical considerations of synthetic biology.9,10 In the case of synthetic biology, concerns have also been raised regarding the potential for synthetic organisms to breed, mutate, or exchange genetic information with other organisms, although there is a lack of experimental data to support or refute these concerns at this time.11−14 To date, the use of nanoremediation has occurred in limited field-scale applications15,16 and the potential benefit and extent of synthetic bioremediation in at-scale applications is largely unknown. The classical risk assessment methods commonly used to support risk governance and other risk management decisions are less suitable for application to emerging technologies. This is because of their intensive data requirements and their difficulties working across multiple sources of uncertain, subjective, and highly divergent information. Classical methods are not well-suited for integrating risk data with other uncertain and potentially subjective criteria relevant for supporting technology development decisions. Decision-makers are left with a limited number of tools to evaluate potential trade-offs between emerging risks, costs, and benefits17,18 and for justifying technology development decisions. Given the limitations of traditional risk assessment frameworks, other tools and methods to assess the complex and uncertain risks and benefits of emerging technologies are needed and have been proposed.19 One alternative method applicable to emerging technologies with only sparse or subjective data is multi-criteria decision analysis (MCDA),20−24 which is well-suited to accommodate the often fragmented modes by which potential risks and societal benefits associated are conveyed.25,26 By enabling the integration of expert judgment and stakeholder values with available data, MCDA can lend insights to decision-making guiding emerging technology development and deployment. Advantages of this approach include its transparency, focus, and the iterative process it facilitates for constructing better technology policies. MCDA supports early decisions about 350

DOI: 10.1021/acs.est.5b03005 Environ. Sci. Technol. 2016, 50, 349−358

Article


international organizations; and professors of environmental and human health risks, contaminant remediation, biotechnology and genetic engineering, nanotechnology, emerging technology policy, and related fields at major research universities and represent diverse geographic regions, organizations, professional positions, and expertise (Table 2).

of environmental remediation, our work demonstrates how MCDA can facilitate the integration of expert judgment and scientific data to compare the suitability of emerging vs conventional technologies.

■

METHODS We implement an MCDA framework to integrate expert judgment and scientific data for evaluating three remediation technologies: synthetic bioremediation (SR), nanoremediation (NR), and conventional ex situ mechanical remediation (CR). The following sections provide an overview of the MCDA framework used, details regarding the expert selection and interview process, handling of data, implementation of the MCDA model, Monte Carlo simulations, and analysis of generated results. Decision-Analytic Framework. MCDA generally consists of six main steps: (1) outline objectives, (2) identify alternatives under consideration, (3) develop criteria and metrics that provide specificity to the objectives, (4) measure performance of alternatives with respect to those criteria and metrics, (5) weight the relative importance of the different criteria, and (6) synthesize these data to identify the most preferred alternative.19,20 Within the family of MCDA, we specifically use a linear-additive multi-attribute value theory model for this analysis, as described in the section Monte Carlo Simulation and MCDA Implementation. The evaluation of the three alternative remediation technologies is structured through a decision-analytic framework where technology performance is evaluated across multiple criteria. Each technology is scored in terms of its expected risks, benefits, and costs. These three primary criteria are subdivided into pairs of secondary criteria that provide greater specificity for scoring anticipated technology performance (Figure 1). Total suitability scores for each technology as a remediation alternative come from aggregating scores across criteria and subcriteria based on weights that reflect the relative importance of each criterion in the decision. Subject Matter Expert Sampling and Interviews. Expert Selection. A total of 119 subject matter experts from academia, government, and industry were identified and invited to participate in the study based on their publication records, invitations to talk at relevant conferences, and recommendations from other experts (Table 1). Nineteen completed the expert elicitation process in full (a 16% completion rate) with both written responses and follow-up oral conversations. Responding experts included directors of university risk and synthetic biology centers; the chief scientific officer of a national agency; a technology director for a major research consortium; researchers from national agencies, industry, and

Table 2. Distribution of Experts Completing Interviews, by Professional Field and Geographic Location field academia government industry location USA China U.K. Austria Canada Denmark

field

North America Europe Asia Oceania

contacted

percentage (%)

74 9 36 contacted

62 8 30 percentage (%)

73 38 4 4

62 32 3 3

percentage (%)

13 4 2 completed

68 21 11 percentage (%)

11 3 2 1 1 1

58 16 11 5 5 5

While the number of completed responses is modest, each of the respondents reflected during the interview that they were comfortable discussing the implications of risk and benefits for environmental remediation using each of these technologies. This is contrasted to many of the contacted experts whom did not feel confident discussing all three technological alternatives. Due to the niche nature of the emerging technologies, acquiring a much larger pool of participating experts whom all felt comfortable discussing these remediation options is likely to be difficult. The current sample of experts is sufficient for drawing interesting conclusions regarding the potential net gains to be had by investing in emerging technologies for remediation purposes and demonstrates how MCDA can be applied to integrate these types of data for emerging technologies. Expert Interviews. Expert elicitation was conducted with a mixed-mode design. Those experts that were contacted by email completed an electronic survey followed by a telephone interview to discuss and clarify responses. When respondents were available in-person, the survey and interview were completed during a single session. An explanation of the decision-analytic framework was given and participants approved it as a valid approach for comparing environmental alternatives. Two types of judgment data were collected about the remediation technologies under consideration: (1) relative importance of the criteria that comprise the framework and (2) scored performance of technologies for each criterion. In order to condition experts to make quantitative judgments, they were first asked to assign ordinal ranks to the primary criteria (risks, benefits, and costs) and secondary criteria (human health risk, environmental health risk, effectiveness, timeliness, development cost, and deployment cost) with respect to environmental remediation. They were next asked to quantify the relative importance (weight) of the criteria by evaluating how relatively more important the higher ranked criteria were than the lower ranked criteria, following the Min10 method that has been shown to have good reproducibility and consistency.35 Next, technology scoring questions were asked to elicit their evidence-based judgment on the expected performance of the eventual, full-scale use of each remediation technology across the six secondary criteria. Again, rank order questions were

Table 1. Distribution of Experts Invited, by Professional Field and Geographic Location

academia government industry location

completed

351


Article


Table 3. Benchmarks for the Secondary Criteria Defining the Ranges Over Which Technology Scores Were Elicited in the Interviews secondary criterion human health risk environmental health risk effectiveness timeliness development cost deployment cost/acre

minimum score (0)

maximum score (100)

where 0 indicates no human health risk resulting from remediation where 0 indicates no environmental health risk resulting from remediation where 0 indicates completely ineffective remediation where 0 indicates instant remediation effect

where 100 indicates existing human health risk without remediation (greater than 100 for additional risk) where 100 indicates existing environmental health risk without remediation (greater than 100 for additional risk) where 100 indicates perfectly effective remediation (all scores should be in this range) where 100 indicates the time needed for natural recovery cost estimated 2012 USD without benchmarks cost estimated 2012 USD without benchmarks

asked first to condition the experts to make quantitative judgments and then quantitative scoring questions were asked in the context of identified benchmarks common to all experts (Table 3). The secondary criteria scores for human health risk, environmental health risk, remediation effectiveness, and remediation time frame were scored on a scale from 0 to 100 points using direct weighting and the identified benchmarks, and the cost scores were provided in monetary units. Data Preparation. Technology performance scores and criteria weights for each interviewee were normalized in order to facilitate direct comparison among results and provide insights at the secondary criteria, primary criteria, and total technology suitability levels (see Supporting Information for details). These normalized values were used in the decision analysis drawing directly from the expert responses. In addition to analyzing the individual interviewee responses, a layer of uncertainty was added to the analysis in recognition that experts differ and that the full pool of knowledgeable experts is broader than those who provided judgment for this study. Distributions were fit to the data, and the variation in observed expert responses was used to simulate a broader range of weights and scores at each level. Inputs to the MCDA were then also sampled from the simulated data. This method of augmenting the interviewee responses with possible inputs from a broader range of expert views is useful for exploring the sensitivity of the analysis results and for investigating the probability density of expected technology scores and rank order. It can also help categorize uncertainty in the decisionanalytic recommendations for technology policy related to potential variability in expert perspectives and expectations. Distribution Fitting. Statistical methods were used to develop distributions of best fit for each set of normalized criteria weights and technology scores. The small number of discrete data points for weight/score estimates and the bounded nature of the data (0−1) necessitated the use of triangular distributions (standard procedure based on least mean squared error produced results that did not properly reflect the data). Triangular distributions were fitted as follows. First, we denoted the minimum, maximum, and mean values of the interview data for each factor: xmin, xmax, and xmean. We then fit a triangular distribution with lower bound tmin, upper bound tmax, and mode tmode. This fitted distribution is intended to reflect the range of observed values but should not preclude the possibility of values outside the observed range. Thus, we defined a provisional minimum ymin = xmin/2 halfway between the theoretical minimum of 0 and the observed minimum, and a provisional maximum ymax = 1 − ((1 − xmax)/2) halfway between the theoretical maximum of 1 and the observed maximum. We then calculated a provisional mode for a

triangular distribution of ymode = 3xmean − ymin + ymax. If the range was not too strongly determined by outlier data points, the provisional values sufficed for the parameters of the triangular distribution, but otherwise we shifted the end point where the outlier was present so that it was possible to have a triangular distribution that preserves the observed mean, the theoretical minimum, and the theoretical maximum: • If ymode > 3xmean − ymin (i.e., ymode > ymax), we assign y′min = 3xmean − 2ymax. • If ymode < 3xmean − ymax (i.e., ymode < ymin), we assign y′max = 3xmean − 2ymin. • Finally, we assign tmin = max(ymin,y′min), tmax = min(ymax,y′max), and tmode = 3xmean − tmax − tmin. Monte Carlo Simulation and MCDA Implementation. Using the triangular distributions fit for each input weight and score (see the Supporting Information, Table 1, for distributions), a Monte Carlo simulation was run where each attribute, secondary criterion, and primary criterion drew a value from its distribution. This process was repeated 10 000 times, and for each set of samplings the MCDA was implemented independently, resulting in a distribution for each alternative’s score across each secondary criterion and primary criterion, and for the total technology suitability score summarized across all simulations. For both the individual interviewee responses and in the Monte Carlo simulations, the total technology suitability score was computed via a linearadditive multi-attribute value theory (MAVT) implementation of MCDA.20,21 Here, Total Scorea = ∑wa,i f i(xa,i), where xi is a reported or simulated score for criterion i for alternative a, f( ) is a normalizing function to convert raw scores to a 0-to-1 scale (see the Supporting Information for details), and w is the weight for criterion i. To confirm the accuracy of the simulation, 10 instances of 10 000 iterations were compared, and the means and standard deviations across instances were observed to be constant to the third decimal place.

■

RESULTS AND DISCUSSION

Technology Scoring Based on the Individual Expert Interviews. Total technology suitability was evaluated in this study by aggregating each technology’s scores across all criteria via weights applied at the secondary and primary levels. The technology scores and criteria weights provided by or derived from the experts during interviews are summarized below. With respect to technology performance at the secondary criteria level, 50% of experts ranked nanoremediation (NR) as posing a greater expected human health risk than synthetic bioremediation (SR), but ranked SR as being more risky for environmental health than NR. In both cases, conventional remediation (CR) is viewed as least risky (Figure 2e,f). In terms 352


Article


Figure 2. Results of direct analysis of interviewee response data showing the proportion of experts that ranked each technology in either first, second, or third place with respect to each of the six secondary criteria (a−f) related to technologies for environmental remediation. (Note that all values are normalized to follow the convention that a rank of first indicated greatest suitability.)

Figure 3. Analysis of interviewee response data depicting the percent of experts whose aggregated subcriteria weights and scores effectively rank each technology in either first, second, or third place at the primary criteria level.

of potential benefits, the experts ranked SR first for both remediation effectiveness and timeliness (as reported by 74 and 66% of respondents, respectively), followed by NR and CR (Figure 2g,h). The experts anticipated SR to have the highest development cost, followed by NR and then CR, but also to have the lowest eventual deployment cost (as ranked by 74 and 76% of respondents, respectively), followed by NR and CR (Figure 2i,j). Consideration of respondent score distributions suggests that efforts to reduce expected development costs and environmental and human health risks can further differentiate SR from competing technologies (Figure 2e−j). Subsequent analysis of these interview results shows considerable agreement between experts with respect to the relative ranking of criteria and alternatives. The same method of ranking was used to summarize expert responses at the primary criteria level. The greatest proportion of respondents favored SR in terms of expected future benefits

(70%), but CR in terms of reduced risks (58%) and overall costs (56%) (Figure 3). With respect to criteria weighting, the interviewees showed considerable interexpert agreement regarding the relative importance of the primary criteria, identifying risks as most important (as agreed on by 79% of respondents), followed by benefits in second place (61%) and costs last (82%). Among the secondary criteria, the experts generally agreed that safeguarding human health is more important than environmental health, that ultimate technology effectiveness is more important than remediation timeliness, and that technology development and deployment costs share similar importance. Two distinct groups emerge when the secondary criteria weights are viewed through the lens of their respective primary criteria weightshuman and environmental health risks and remediation effectiveness are clearly identified as relatively 353


Article


Figure 4. Global weights from the expert interviews at the secondary criteria level. These result from multiplying the local secondary criteria weights by their respective primary criteria weights.

Figure 5. Probability density of primary criteria weights (a) simulated from the distributions that were fit to interviewee responses and (b) directly from interviewee responses.

preference in this context for first considering technology risks, then benefits, and last costs. The Monte Carlo simulations then combine the sampled weights with unweighted technology scores in order to yield probability densities of weighted technology scores. The simulation results identify weighted score distributions that are broader (more uncertain) for human health than for environmental health risks, and show CR risks as lowest, SR and NR environmental health risks as nearly equivalent, and NR human health risks as higher than SR human health risks (Figure 6a,b). The simulations show expected preference for SR over the other technologies in terms of both remediation effectiveness and time frame, with overlapping distributions of NR and CR effectiveness scores (Figure 6c,d). As expected, the simulations favor CR in terms of expected development costs, but show similar distributions of SR and NR expected development cost. They favor SR over CR and then NR in terms of expected deployment cost (Figure 6e,f).

more important than technology deployment and development costs and remediation time frame (Figure 4). In terms of total technology suitability, interviewee results show an expectation for SR and CR to outperform NR. On a 0−1 preference scale, aggregated respondent suitability scores for these technologies were 0.62 for SR, 0.61 for CR, and 0.53 for NR, representing a 17% difference in mean between NR and SR. Technology Scoring Based on the Monte Carlo Simulations. To draw conclusions from the results of the Monte Carlo simulation we primarily use probability densities, descriptions of the chance that a given value would be sampled at random for a continuous result variable. Probability density results for the simulated weights (which were sampled from distributions fit to the interviewee responses) (Figure 5a) can be compared with raw interviewee-reported ranking of weights (Figure 5b). In both cases, the results show overlap between the importance of the different criteria but demonstrate a clear 354


Article


Figure 6. Simulation results showing probability density of weighted technology scores for each of the six secondary criteria. Because they are weighting at the primary criterion level, only secondary criteria score distributions within each of the risk, benefit, and cost criteria are directly comparable.

Figure 7. Distributions of scores resulting from Monte Carlo simulations that sample and aggregate individual weights and scores from score probability curves fit to interview data. (a) Probability density of relative total technology suitability scores. (b−d) Disaggregated simulation results showing score probability densities for the three primary criteria. (Note that all values are normalized to follow the convention that higher scores represent greater preference.)

In terms of total technology suitability, the simulations rank SR first 47% of the time, CR first 43% of the time, and NR first only 10% of the time (Figure 7a). Results at the primary criteria level (Figure 7b−d) show SR outranking NR for benefit (and

somewhat for cost) reasons and CR outranking NR for risk and cost reasons. The simulations show the greatest difference in expected scores between SR and CR/NR benefits, the next greatest difference in scores between CR and SR/NR risks, and 355


Article


Figure 8. Sensitivity of alternatives to change in model input parameters. The range of each bar shows variation of mean resulting technology suitability scores with changes of each input at the 5 and 95% percentiles of each input distribution (with all other inputs held constant). 356



■

distinct but overlapping distributions for the three technology cost scores. Sensitivity Analysis. Sensitivity analysis of the simulation model suggests that efforts to refine expert estimates should first focus on the technology scores for human health, environmental health, and remediation effectiveness, and that the model is generally more robust to changes in expert reported weight than changes in score (Figure 8). In the absence of field data regarding the risks, benefits, and costs of emerging technologies, structuring evidence-based expert judgment through a weighted hierarchy of topical questions is helpful for informing preliminary risk governance and guiding emerging technology development and policy. Our expert interviews and subsequent analysis demonstrate this decision-analytic approach and suggest that, in the context of environmental remediation, synthetic biology remains a promising emerging technology. Despite the uncertainties surrounding its efficacy and potential risk to human and environmental health, synthetic biology was expected by the experts to offer significant improvement to conventional environmental remediation technologies and processes in terms of effectiveness, timeliness, and deployment costs if fully developed and realized. Conversely, nanotechnology was expected to offer only slight advantages in some areas. This analysis demonstrates that, for important cases involving high uncertainty and a lack of field data, decision aids such as MCDA may offer a much needed first step in the risk analysis process to aid stakeholders with funding and governance decisions that will be further refined and improved upon as quantitative data becomes more readily available.

■

REFERENCES

(1) Superfund: EPA’s Estimated Costs to Remediate Existing Sites Exceed Current Funding Levels, and More Sites Are Expected to Be Added to the National Priorities List; Highlights of GAO-10-380; United States Government Accountability Office: Washington, DC, 2010. http:// www.gao.gov/products/GAO-10-380. (2) Formerly Utilized Sites Remedial Action Program Update 2012; U.S. Army Corps of Engineers: Washington, DC, 2012. http://www.usace. army.mil/Portals/2/docs/Environmental/FUSRAP/FUSRAP_ Stakeholder_Report_2012_Final.pdf. (3) Diallo, M. S.; Fromer, N. A.; Jhon, M. S. Nanotechnology for sustainable development: retrospective and outlook. J. Nanopart. Res. 2013, 15 (2044), 1−16. (4) Purnick, P. E.; Weiss, R. The second wave of synthetic biology: from modules to systems. Nat. Rev. Mol. Cell Biol. 2009, 10 (6), 410− 422. (5) Wu, Z.-Y.; Li, C.; Liang, H.-W.; Zhang, Y.-N.; Wang, X.; Chen, J.F.; Yu, S.-H. Carbon nanofiber aerogels for emergent cleanup of oil spillage and chemical leakage under harsh conditions. Sci. Rep. 2014, 4, 4079. (6) Grieger, K. D.; Fjordbøge, A.; Hartmann, N. B.; Eriksson, E.; Bjerg, P. L.; Baun, A. Environmental benefits and risk of zero-valent iron nanoparticles (nZVI) for in situ remediation: Risk mitigation or trade-off? J. Contam. Hydrol. 2010, 118 (3−4), 165−183. (7) Maynard, A. D. A decade of uncertainty. Nat. Nanotechnol. 2014, 9 (3), 159−160. (8) Batley, G. E.; Kirby, J. K.; McLaughlin, M. M. Fate and risk of nanomaterials in aquatic and terrestrial environments. Acc. Chem. Res. 2013, 46 (3), 854−862. (9) Gutmann, A.; Wagner, J. New Directions: The Ethics of Synthetic Biology and Emerging Technologies; The Presidential Commission for the study of Bioethical Issues: Washington, DC, 2010. http:// bioethics.gov/synthetic-biology-report. (10) Church, G. M.; Elowitz, M. B.; Smolke, C. D.; Voigt, C. A.; Weiss, R. Realizing the potential of synthetic biology. Nat. Rev. Mol. Cell Biol. 2014, 15 (4), 289−294. (11) Dana, G. V.; Kuiken, T.; Rejeski, D.; Snow, A. A. Four steps to avoid a synthetic-biology disaster: assess the ecological risks of synthetic microbes before they escape the lab. Nature 2012, 483 (7387), 29−29. (12) Hoffman, E.; Hanson, J.; Thomas, J. The Principles for the Oversight of Synthetic Biology; Friends of the Earth: Washington, DC, 2012. (13) Caruso, D. Synthetic Biology An Overview and Recommendations for Anticipating and Addressing Emerging Risk; Center for American Progress: Washington, DC, 2008. Available at http://www. scienceprogress.org/wp-content/uploads/2008/11/syntheticbiology. pdf. (14) Benner, S. A.; Sismour, A. M. Synthetic biology. Nat. Rev. Genet. 2005, 6, 533−543. (15) Karn, B.; Kuiken, T.; Otto, M. Nanotechnology and in situ remediation: A review of the benefits and potential risks. Environ. Health. Perspect. 2009, 117 (12), 1823−1831. (16) Mueller, N. C.; Braun, J.; Bruns, J.; Cerník, M.; Rissing, P.; Rickerby, D.; Nowack, B. Application of nanoscale zero valent iron (NZVI) for groundwater remediation in Europe. Environ. Sci. Pollut. Res. 2012, 19 (2), 550−558. (17) Hristozov, D. R.; Gottardo, S.; Critto, A.; Marcomini, A. Risk assessment of engineered nanomaterials: A review of available data and approaches from a regulatory perspective. Nanotoxicology 2012, 6 (8), 880−898. (18) Linkov, I.; Bates, M. E.; Canis, L. J.; Seager, T. P.; Keisler, J. M. A decision-directed approach for prioritizing research into the impact of nanomaterials on the environment and human health. Nat. Nanotechnol. 2011, 6 (12), 784−787. (19) Grieger, K. D.; Linkov, I.; Hansen, S. F.; Baun, A. Environmental risk analysis for nanomaterials: Review and evaluation of frameworks. Nanotoxicology 2012, 6 (2), 196−212.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.5b03005.

■

Article

Additional data and details related to the analysis and results presented in this article (PDF)

AUTHOR INFORMATION

Corresponding Authors

*Phone: (978)318-8795; email: [email protected]. mil. *Phone: (978)318-8197; email: [email protected]. Author Contributions

I.L. envisioned the concept; M.E.B., K.D.G., and J.M.K. developed the framework; M.E.B. and B.D.T. performed the interviews; M.E.B., K.J.P., and J.M.K. analyzed the data; all authors collectively wrote the manuscript and have given approval to the final version of the manuscript. The authors thank Margaret Kurth and Matt Wood for their comments. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This study was funded by the Engineer Research and Development Center of the U.S. Army Corps of Engineers and by RTI International. It has been approved for publication by authority of the Chief of Engineers, U.S. Army Corps of Engineers. 357


Article

Environmental Science & Technology (20) Multiple Criteria Decision Analysis: State of the Art Surveys; Figueira, J., Greco, S., Ehrgott, M., Eds.; Springer Science & Business Media: New York, 2005; Vol. 78. (21) Belton, V.; Stewart, T. Multiple Criteria Decision Analysis: An Integrated Approach; Kluwer Academic Publishers: Boston, MA, 2002. (22) Linkov, I.; Bates, M. E.; Canis, L. J.; Seager, T. P.; Keisler, J. M. A decision-directed approach for prioritizing research into the impact of nanomaterials on the environment and human health. Nat. Nanotechnol. 2011, 6, 784−787. (23) Linkov, I.; Satterstrom, F. K.; Kiker, G.; Seager, T. P.; Bridges, T.; Gardner, K. H.; Rogers, S. H.; Belluck, D. A.; Meyer, A. Multicriteria decision analysis: A comprehensive decision approach for management of contaminated sediments. Risk Anal. 2006, 26 (1), 61− 78. (24) Linkov, I.; Welle, P.; Loney, D.; Tkachuk, A.; Canis, L.; Kim, J. B.; Bridges, T. Use of multicriteria decision analysis to support weight of evidence evaluation. Risk Anal. 2011, 31 (8), 1211−1225. (25) Linkov, I.; Satterstrom, F. K.; Steevens, J.; Ferguson, E.; Pleus, R. C. Multi-criteria decision analysis and environmental risk assessment for nanomaterials. J. Nanopart. Res. 2007, 9 (4), 543−554. (26) Linkov, I.; Seager, T. P. Coupling multi-criteria decision analysis, life-cycle assessment, and risk assessment for emerging threats. Environ. Sci. Technol. 2011, 45 (12), 5068−5074. (27) ter Meulen, V. Time to settle the synthetic controversy. Nature 2014, 509 (7499), 135. (28) Expert Elicitation Task Force White Paper; U.S. Environmental Protection Agency: Washington, DC, 2011. Available at http://www. epa.gov/stpc/pdfs/ee-white-paper-final.pdf. (29) Back, P.; Rosén, L.; Norberg, T. Value of information analysis in remedial investigations. Ambio 2007, 36, 486−93. (30) Linkov, I.; Moberg, E. Multi-Criteria Decision Analysis: Environmental Applications and Case Studies; CRC Press: Boca Raton, FL, 2012. (31) Lahdelma, R.; Hokkanen, J.; Salminen, P. SMAA − Stochastic multiobjective acceptability analysis. Eur. J. Oper. Res. 1998, 106 (1), 137−143. (32) Committee on Decision Making Under Uncertainty; Board on Population Health and Public Health Practice; Institute of Medicine. Environmental Decisions in the Face of Uncertainty; National Academies Press: Washington, DC, 2013. (33) Committee on Improving Risk Analysis Approaches Used by the U.S. EPA, National Research Council. Risk and Decisions: Advancing Risk Assessment; National Academies Press: Washington, DC, 2009. (34) National Nanotechnology Initiative. Strategy for NanotechnologyRelated Environmental, Health, and Safety Research; U.S. National Science and Technology Council: Washington, DC, 2008. (35) Bottomley, P. A.; Doyle, J. R. A Comparison of three weight elicitation methods: good, better, best. OMEGA 2001, 29 (6), 553− 560.

358
