JUser: :_load: Unable to load user with ID: 291

Aroclor Reconstruction and Identification: A Comparison of Statistical Methods Used at the Portland Harbor Superfund Site
« All Publications


Polychlorinated biphenyls (PCBs) are carcinogenic chemical compounds that were manufactured for a wide variety of industrial products from the 1930s to the 1970s and are often the primary driver remediation of sediment sites. By their nature, sediment sites act as a collector of PCB-related impacts from multiple sources, complicating equitable allocation of responsibility for remediation. Identification of individual sources of PCBs to sediment sites is often complicated due to the limitations of the commonly-used USEPA Method 8082 for PCB Aroclor identification; sediment transport processes; ongoing degradation of PCB mixtures that confounds interpretation of congener data, and the widespread former use of mixed Aroclors in industrial products.

Analysis of PCB congeners by USEPA Method 1668 provides a more robust data set for source identification. Various statistical techniques have been utilized to identify distinct sources of contamination; characterize chemical signatures associated with distinct sources of contamination; apportion PCB mixtures into constituent mixtures of chemicals; and allocate contamination to different sources. We provide an overview and comparison of the techniques that have been used along with a post-hoc discussion on how these methods can be used most effectively utilized.


The statistical techniques that have been used at Portland Harbor to identify sources of contamination can be grouped into the following categories:
· Grouping techniques are designed to identify chemically distinct sources of contamination and characterize the chemical signature associated with each source. Many different types of grouping techniques exist. Thus far, principal component analysis (PCA) and K-medoids cluster analysis have been used at Portland Harbor. Other similar grouping techniques, such as positive matrix factorization (PMF), exist.
· Apportioning techniques separate mixtures of contamination into individual constituent. Multiple linear regression (MLR) and Bayesian Mixture Modeling have been used to apportion PCB contamination in Portland Harbor. Thus far, the focus has been on apportioning PCB mixtures into constituent Aroclors; however, these techniques could be used to apportion contamination into constituent sources if the chemical signatures associated with each source were known.

Here, we compare techniques from the same category using numerical methods and outline the types of inference that can be derived from each category of technique. Discussion of the level of effort and quality assurance required is also summarized.

Results/Lessons Learned.

Techniques such as MLR provided relatively high confidence identification of Aroclors 1242, 1248, 1254, and 1260, which are the predominant Aroclors in Portland Harbor. Aroclors 1016, 1221, and 1232 were not identified with the same level of confidence, commensurate with the lower frequency of detection (and lower manufactured volumes) of these Aroclors. The results generally agreed with K-medoid clustering performed by others, where Aroclor 1016, 1242, and 1248 were treated as a group distinct from 1254 and 1260 (combined with 1262). These techniques can be used in sequence to map and quantify contamination coming from individual sources. After source specific chemical signatures have been identified using clustering techniques and PCB contamination has been partitioned by source at the sample level using an apportioning technique, interpolation techniques can be used in estimating the degree of source specific contamination between sampling points. The inference derived from these methods is far more limited when these techniques are used in isolation.

Publication Summary

  • Geosyntec Authors: James G.D. Peale, Jordy Bernard, Brian Webb
  • All Authors: James G.D. Peale (Geosyntec Consultants, Portland, OR, USA), Jordy Bernard, and Brian Webb
  • Title: 2023 Battelle Sediments Conference
  • Event or Publication: Event
  • Practice Areas: Sediment assessment and remediation, Contaminated site assessment and clean up
  • Citation: Battelle's International Conference on the Remediation and Management of Contaminated Sediments at the JW Marriott in Austin, Texas, on January 9 through 12, 2023
  • Date: January 9 through 12, 2023
  • Location: JW Marriott in Austin, Texas
  • Publication Type: Platform Presentation