Resolving Polychlorinated Biphenyl Source Fingerprints in Suspended Particulate Matter of San Francisco Bay

Abstract

The presence of PCB contamination in San Francisco Bay has been documented, but the number of sources, their chemical composition, and their geographic/temporal distribution are poorly understood. A self-training pattern recognition technique, polytopic vector analysis is used to determine those parameters from PCBs adsorbed on the particulate fraction of surface waters. Five chemical fingerprints (end-members) were resolved. Four were consistent with published Aroclor patterns. Aroclor 1260 was observed throughout the estuary, in all cruises, with highest proportions observed in Coyote Creek, a tributary of the South Bay. A pattern that matches typical Aroclor 1254 was observed in all cruises but was in generally higher abundance in spring 1995. A second Aroclor 1254 pattern, consistent with an atypical Aroclor 1254 batch described in the literature, was observed in moderate proportions in the three 1996 cruises. Aroclor 1248 was present in significant proportions in only one cruise (cruise 12: July 1996) but was the dominant fingerprint in the Central Bay samples collected at that time. End-member 5 did not match published Aroclor source patterns. Its composition exhibits high proportions of the metabolism-resistant congeners PCB-138 and PCB-153. The source of this pattern is not known, but we hypothesize that it may be due to sewage inputs in the Bay or from atmospheric inputs.

Integrated Forensics Approach to Fingerprint PCB Sources in Sediments using Rapid Sediment Characterization (RSC) and Advanced Chemical Fingerprinting (ACF)

“…RSC can suggest "where" potential sources are located by following concentration gradients while ACF can confirm "what" potential source fingerprints are present. As described in previous PCB forensic studies (Johnson et al, 2000;Magar et al,2005), both source fingerprints as well as weathering patterns (due to differences in solubility, dechlorination, etc.) can often be discerned using these statistical analyses.…”

“…PCBs also behave quite differently in the environment and many factors need to be considered in addition to the most common weathering factors; interpretation methods need to be modified and new data interpretation and analysis considerations need to be developed (Johnson et al, 2006). Emerging PCB fingerprinting techniques have successfully been applied at a few sites in the US (Johnson et al, 2000;Emsbo-Mattingly and Durell, 2003;Magar et al, 2005;Johnson et al, 2007), but there is a need to more fully develop, demonstrate, and validate the utility of fingerprinting PCB contamination. Additional discussions and comparisons for some analytical techniques (both RSC and ACF) are presented in Section 8, and comparisons for some statistical interpretation techniques (receptor models) are shown in Section 6.…”

“…Much of the previous work has focused on regional scale problems and not the finer scale issues seen at individual DoD sites. DoD sites are typically on the size of acres, whereas other previous work has concentrated on much larger scales (hundreds of miles in San Francisco Bay (Johnson et al, 2000) or the Delaware River TMDL (Rodenburg et al, 2010)). Our demonstration sites were selected to represent the scale of typical DoD sites and address the types of questions posed at DoD sites.…”

“…Development of numerical methods to determine these parameters has been a major goal in environmental chemometrics and receptor modeling for more than 25 years (Hopke, et al, 2006;Johnson, et al, 2007). Multivariate receptor models have been used increasingly for the characterization of sources and alteration patterns of chlorinated organic compounds in sediments of complex environmental settings (Jarman, et al, 1997;Johnson, et al, 2000;Barabas, et. al., 2004;Imomoglu, et al, 2002;Magar, et al, 2005;Bzdusek, et al, 2006).…”

“…literature reported dechlorination methods - Bedard and Quensen, 1995). We discuss PVA in this effort, because it is a method known well by the authors, and has been applied extensively in PCB forensics applications (Jarman, et al, 1997;Johnson, et al, 2000;Magar, et al, 2005;Johnson, et al, 2007). We also have access to other receptor modeling methods (ALS, PMF, UNMIX -see discussion of their applications in Johnson, et al, 2007) and we ran comparative analyses using all these methods for a performance assessment test in Section 6.…”