A 9-year water dataset from the Houston Ship Channel (HSC) was analyzed to understand partitioning in polychlorinated biphenyls (PCBs) and polychlorinated dibenzo-p-dioxins/polychlorinated dibenzofurans (PCDD/Fs). Total PCBs had more mass as dissolved (74%) whereas total PCDD/Fs did not (11%). Generally, the limited number of PCDD/Fs (only 2378 substituted) explained these differences though differences in chemical behavior beyond log K ow also likely influence partitioning. The particular fractionation seen in the HSC also seemed related to a wide variation in particulate organic carbon (POC)/dissolved organic carbon (DOC) ratio (0.42-180%). Published and unaltered linear free energy and linear solvation energy relationships for DOC, POC, and particulate black carbon (BC) resulted in predictions that were at best 27% (PCB) and 25% root-mean-square error (RMSE) (PCDD/F) partition fraction compared to observed (using estimated BC/POC fractions of 10 and 25%, respectively). These results show, at least in light of the uncertainties in this data (e.g., precise fraction of BC), that a 25% accuracy in model prediction of operationally dissolved or suspended fraction for any one PCB or PCDD/F congener is the best prediction that may be expected. It is therefore recommended that site-specific data be used to calibrate most any water column-partitioning model if it is to be expected to describe what actually occurs in field conditions.
Keywords: Black carbon; Chemical partitioning; Dioxins; Houston ship channel; Linear free energy relationships (LFERs); Linear solvation energy relationships (LSERs); Organic carbon; Polychlorinated biphenyls.