Integrated Environmental Assessment and Management — Volume 10, Number 3—pp. 323–324 © 2014 SETAC

323

RESPONSE TO O'REILLY ET AL. (2014) Judy L Craney yMinnesota Pollution Control Agency, St Paul, Minnesota, USA

(Submitted 18 March 2014; Accepted 15 April 2014)

* To whom correspondence may be addressed: [email protected] Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/ieam.1539

classified as pyrogenic because they are derived from combustion sources (Boehm 2006). Designating PAHs as derived from petrogenic sources is appropriate when the concentrations of alkylated PAHs exceed the parent PAHs and when 4 to 6 ring PAHs are absent or present at low concentrations (Boehm 2006). This is evident when the parent and alkyl homologues appear as a bell‐shaped curve in histogram plots of PAH environmental data (Boehm 2006). Although alkylated PAHs may, in general, be less useful for forensic evaluations of urban sediments, measurement of these compounds is very important for evaluating ecological risk to aquatic invertebrates through the United States Environmental Protection Agency’s (USEPA) Equilibrium Partitioning Sediment Benchmark Toxic Units model (Burgess 2009). O’Reilly et al. did not discuss the limitations of using principal components analysis (PCA) to compare PAHs in sediments to potential source materials. There was no mention of the minimum number of samples needed to use PCA. This is important because PCA creates optimized linear combinations of variables that tend to over fit the data, and inflated error rates may occur when sample size is too small (Osborne and Costello 2004). The example case study used by O’Reilly et al. relied on a small number of samples from multiple sources in the PCA diagrams. For example, Figure 2 presented 19 data points from 3 potential PAH source types, and Figure 5 showed only 15 data points. Because PCA is a large sample technique, it is not well‐suited for small sample sizes that contribute to higher incidences of Type II errors (Osborne and Costello 2004). Fifty samples are generally regarded as a minimum total sample size for PCA analysis. Comfrey and Lee (1992), for example, concluded that confidence in PCA analysis based on 50 samples was very poor, 100 samples was poor, 200 samples was fair, 300 samples was good, 500 samples was very good, and analysis based on 1000 or more samples was excellent. Furthermore, Johnson et al. (2004) point out that PCA may provide a false sense of security to users and that more sophisticated goodness‐of‐fit diagnostics than percentage variance must be used in environmental forensic analyses. The use of PCA in the example case study provided by O’Reilly et al. is simply not supported by a sufficiently robust data set to derive statistically meaningful conclusions. O’Reilly et al. discuss the use of receptor models as a more rigorous forensic technique than PCA to quantify multiple sources of PAHs in sediment data sets. They recommend using both a forward‐calculating mixing method, such as the USEPA’s chemical mass balance (CMB) model, and a backward calculating unmixing method, such as the USEPA’s Unmix model, for comparing the results of multiple lines of evidence in complex studies. However, the use of both types of receptor models is not feasible for most sediment investigations. Although the CMB model has the advantage that it can be run on one to many samples, the Unmix model requires hundreds of samples. The Unmix model uses bootstrapping, and the fit diagnostics guidance for the model describes the number of bootstrap attempts to obtain 100 feasible solutions for data set size categories of greater than 600 samples, 400 to 600 samples,

Letter to the Editor

DEAR SIR: Recently, O’Reilly et al. (2014) provided a review of environmental forensic techniques for source characterization and identification of pyrogenic polycyclic aromatic hydrocarbons (PAHs) in sediments. They also provided an example case study pertaining to refined tar‐based sealers (commonly known as coal tar‐based sealants [CT‐sealants]) to support their contention that CT‐sealants are not a significant source of PAHs in urban sediments and thus should not be the subject of source control policies. Several government jurisdictions in the United States have banned the sale and use of CT‐sealants, including the states of Minnesota and Washington. In addition, the Minnesota Pollution Control Agency determined CT‐sealants are a major source of PAHs in storm water pond sediments from the Minneapolis‐St. Paul metropolitan area (Crane 2014). Furthermore, Mahler et al. (2012) provided a review of scientific studies that verified the migration of PAHs from CT‐ sealcoated surfaces into the environment, including lake sediment. Accordingly, it is useful to discuss a few of the limitations and errors in O’Reilly et al.’s approach to PAH forensic work and government source control policy. In the Introduction, O’Reilly et al. made several inappropriate generalizations regarding how government agencies regulate and respond to PAH contamination in the environment. The authors observe correctly that PAHs are generally regulated in the United States as toxic substances, but they incorrectly classify all PAHs as potential carcinogens when only a subgroup of PAHs is classified as such (MDH 2013). O’Reilly et al. did not survey US federal and state government agencies for how they make decisions regarding PAH‐contaminated sediments, which would have been useful. For example, state agencies do not share a uniform policy regarding benchmark guidelines or regulatory‐based numeric criteria that trigger either investigation or corrective interventions based on PAH concentrations and/or benzo[a]pyrene equivalents in sediment; there is also no consistent, nationwide list of carcinogenic PAH compounds shared by all state environmental and health agencies. O’Reilly et al. also emphasize the need for government scientists to communicate forensic results to lay audiences, but this is standard practice for any complex environmental issue. Government agencies routinely rely on communication specialists to help with the technical transfer of information to stakeholders and the public. Later on in O’Reilly et al.’s paper, they mistakenly confuse this technical transfer and education of research results by government scientists to a wider audience as advocacy. O’Reilly et al. suggest the use of alkylated PAHs in forensic evaluations to determine petrogenic (i.e., oil‐based) sources of PAHs. However, PAHs in urban sediments are frequently

324

and less than 400 samples (Norris et al. 2007). It is important to note that the Unmix model was designed for use in air modeling studies where large numbers of samples are typically obtained (Norris et al. 2007); large data set requirements are unusual in most sediment studies of PAHs. Furthermore, the CMB model was developed initially as an air model (Coulter 2004), and O’Reilly et al. discuss its early application to the source apportionment of PAHs in sediment samples by researchers like Li et al. (2003). These early applications were conducted before the emergence of CT‐ sealants as an important source of PAHs in some urban waterways. In Figure 1 of O’Reilly et al., they used data from Li et al. to show the fractional PAH source contributions from 9 CMB model operations representing a total of 422 model runs. However, the CMB model performance statistics were violated for some of Li et al.’s (2003) runs in each model operation. It would have been more appropriate for O’Reilly et al. to use only those model results that met the CMB model requirements for acceptability when comparing the source apportionment of different model operations. Some of the variability in the Li et al. (2003) work noted by O’Reilly et al. may be due to skewness from unacceptable model runs, as well as varying the number of individual PAHs from between 2 and 14 compounds. As noted in Crane (2014), a reexamination of Li et al.’s (2003) CMB modeling would be interesting if a CT‐ sealant source was included in the analysis because CT‐sealants are widely used in the Chicago area. Van Metre and Mahler (2010) determined that CT‐sealant dust comprised 45.8% and 70.5% of the modeled PAH concentrations from lake sediments collected from 2 Chicago suburbs. The authors made a few incomplete or incorrect comments about CMB model assumptions and collinearity. O’Reilly et al. indicated that “some deviations can be tolerated” in regard to the CMB model assumptions, which seemed to downplay Coulter’s (2004) statement that “Fortunately, CMB can tolerate reasonable deviations from these assumptions, though these deviations increase the stated uncertainties of the source contribution estimates.” O’Reilly et al. were incorrect about the conflict between the first and third CMB model assumptions; they should have referred to the first and fourth CMB model assumptions. O’Reilly et al. indicated that CT‐sealant is not the only source with a similar profile, but they did not identify specifically other sources with PAH profiles similar to CT‐sealants. In contrast, Coulter (2004) indicated that it is impossible to decide a priori whether a set of source profiles is collinear or not since ambient data uncertainties and relative levels of source contributions vary from sample to sample. A subsequent evaluation of the eligible space collinearity displays of the CMB model runs published in Crane (2014), indicated no collinearity was observed between CT‐sealant sources and other source types, including traffic tunnel air, gasoline vehicle particulate emissions, and pine wood soot particles (all of which were based on measured data for these common, urban sources of PAHs). The Source Elimination option was activated in each CMB model run performed by Crane (2014), and use of this option affects the fit obtained by effectively removing collinear sources (Coulter 2004). O’Reilly et al. fault previous research by Van Metre and Mahler (2010) for not including a negative control (i.e., excluding CT‐sealant) in their CMB model runs. However, this is not a practice recommended by Coulter (2004) for proper operation of the CMB model. Nonetheless, Crane (2014) included a negative control model scenario and found significantly better model performance in four parameters when CT‐

Integr Environ Assess Manag 10, 2014—JL Crane

sealants were included in the CMB model. For sediments analyzed from 15 storm water ponds in the Minneapolis–St. Paul metropolitan area (none of which were located near historical manufactured gas plants), CT‐sealants comprised about 67%, on average, of a total suite of 12 PAHs included in the CMB model (Crane 2014). The model results were quite robust, demonstrating the applicability of this model to environmental forensic evaluations of PAHs in sediment samples. Regrettably, O’Reilly et al. provide little new understanding of environmental forensic work for pyrogenic PAHs; their work is unlikely to alter current policy in the state of Minnesota regarding regulation of PAHs and CT‐sealants. Minnesota’s current statewide ban on the sale and use of CT‐sealants is expected to provide a long‐range, positive contribution toward reducing PAH concentrations in storm water pond sediments. The intent is to reduce the economic burden on municipalities responsible for periodically removing and disposing of sediment from their storm water ponds (Crane 2014). Furthermore, the 1990 Minnesota Toxic Pollution Prevention Act established a state policy encouraging pollution prevention to eliminate or reduce hazardous and toxic pollutants at the source. Pollution prevention research by the Minnesota Pollution Control Agency regarding CT‐sealants is being shared with other Great Lakes states and provinces facing similar economic and environmental challenges. This information is available at: http://www.pca.state.mn.us/ahx9qrk.

REFERENCES Boehm PD. 2006. Polycyclic aromatic hydrocarbons (PAHs). In: Morrison RD, Murphy BL, editors. Environmental forensics: Contaminant specific guide. New York (NY): Elsevier. 313–337 p. Burgess RM. 2009. Evaluating ecological risk to invertebrate receptors from PAHs in sediments at hazardous waste sites. Cincinnati (OH): US Environmental Protection Agency. EPA/600/R‐06/162F. Comfrey AL, Lee HB. 1992. A first course in factor analysis. 2nd ed. New York (NY): Psychology Press. 430 p. Coulter CT. 2004. EPA‐CMB 8.2 users manual. Research Triangle Park (NC): US Environmental Protection Agency. EPA 452/R‐04‐11. Crane JL. 2014. Source apportionment and distribution of polycyclic aromatic hydrocarbons, risk considerations, and management implications for urban stormwater pond sediments in Minnesota, USA. Arch Environ Contam Toxicol 66:176–200. Johnson GW, Ehrlich R, Full W. 2004. Principal components analysis and receptor models in environmental forensics. In: Murphy BL, Morrison RD, editors. Introduction to Environmental Forensics. Burlington (MA): Elsevier Academic Press. 461–515 p. Li A, Jang J‐K, Scheff PA. 2003. Application of EPA CMB8.2 model for source apportionment of sediment PAHs in Lake Calumet, Chicago. Environ Sci Technol 37:2958–2965. Mahler BJ, Van Metre PC, Crane JL, Watts AW, Scoggins M, Williams ES. 2012. Coal‐ tar‐based pavement sealcoat and PAHs: Implications for the environment, human health, and stormwater management. Environ Sci Technol 46:3039–3045. [MDH] Minnesota Department of Health. 2013. Guidance for evaluating the cancer potency of polycyclic aromatic hydrocarbon (PAH) mixtures in environmental samples. St. Paul (MN): Minnesota Department of Health. Norris G, Vedantham R, Duvall R, Henry RC. 2007. EPA Unmix 6.0 fundamentals & user guide. Washington (DC): US Environmental Protection Agency. EPA/600/R‐ 07/089. O'Reilly KT, Pietari J, Boehm PD. 2014. Parsing pyrogenic polycyclic aromatic hydrocarbons: Forensic chemistry, receptor models, and source control policy. Integr Environ Assess Manag 10:279–285. Osborne JW, Costello AB. 2004. Sample size and subject to item ratio in principal components analysis. [cited 2012 March 12]. Available from: http://PAREonline.net/getvn.asp?v¼9&n¼11. Pract Assess Res Eval 9(11). Van Metre PC, Mahler BJ. 2010. Contribution of PAHs from coal‐tar pavement sealcoat and other sources to 40 U.S. lakes. Sci Total Environ 409:334–344.

Response to O'Reilly et al. (2014).

Response to O'Reilly et al. (2014). - PDF Download Free
67KB Sizes 1 Downloads 4 Views