PCBs in Great Lakes sediments, determined by positive matrix factorization

Abstract

Previously determined PCB concentrations in 10 dated sediment cores from Lakes Michigan, Huron, Erie, and Ontario are analyzed by positive matrix factorization in order to find characteristic congener patterns including signs of anaerobic dechlorination. Three or four factors are sufficient to describe the PCB data for each lake. All four lakes are dominated by Aroclors 1248A, 1248G, 1254A, and 1254G. Aroclor 1260 is only a significant factor in Lake Michigan prior to 1975, reflecting in part a usage patterns of heavy chlorinated Aroclors early and less chlorinated Aroclors in the phase-out years in the 1970s. Dated records of factors or sources indicate clear PCB concentration maxima for Lakes Erie (1981) and Ontario(1968), while redistribution of PCBs in the less contaminated Lakes Michigan and Huron has occurred after 1980. Using a single data matrix for all four lakes provides a common basis and possibility to examine low degrees of dechlorination, while individual data matrices for each lake provide more accurate results and better separation of factors. Lakes Ontario, Michigan, and Huron undergo dechlorination via reactions such as 66(24-34) → 25(24-3) and 18(25-2) → 4(2-2) consistent with reactions H' + M, while Lake Erie appears to be dominated by 18(25-2) → 4(2-2) and 53(25-26) → 19(26-2) that are typical for processes M + Q.