Quantitative evaluation of dechlorination signatures in contaminated sediments using modified polytopic vector analysis with uncertainty assessment

Risk-based sediment management decisions require the delineation of contaminants and sources, an assessment of fate processes and the quantification of uncertainty. Quantitative evidence for biochemical transformation processes in field situations is important for the development of alternative strategies of managing contaminated sediments even for persistent halogenated compounds. Given extensive concentration measurements, a procedure to extract indicators of dehalogenation activity from concentration data is highly desirable. This research represents the first quantitative method to achieve this goal by integrating polytopic vector analysis (PVA), a multivariate technique used to resolve chemical fingerprints, with geostatistical techniques of uncertainty assessment and with laboratory-derived information about dehalogenation processes. The traditional algorithm is constrained to positive fingerprint (end-member) components and cannot resolve dechlorination fingerprints, i.e. the net change of dioxin and furan congeners (positive and negative values). The algorithm was modified to isolate compositional changes due to dechlorination from the dioxin pattern in a separate end-member. Using artificial data sets, for which the composition and sample contribution of all end-members are known, the dechlorination fingerprint was reproduced with a root mean square error of 28–41%. The dechlorination end-member's contribution to total variability (set at 4.0 and 10%, respectively) was overestimated by a factor of 1–5. When applied to dioxin concentration data from Passaic River sediment cores, the procedure identified a dechlorination end-member indicating an increase of 2,3,7,8-tetraCDD at the expense of 1,2,3,4,6,7,8-heptaCDD, with an average sample contribution to total variability of 3 ± 1%. However, at 33 ± 25%, dechlorination was the second most important contributor to 2,3,7,8-tetraCDD concentrations (after 2,4,5-T production, 60 ± 30%), which translated to an average of 1.2 μg/kg of 2,3,7,8-tetraCDD per sample. Uncertainty with respect to the signature of dechlorination was minor except for the 1,2,3,4,7,8-hexaCDD component. The variability contribution of dechlorination is the most uncertain result, with a 1–5-fold overestimation indicated by artificial data analysis. Model uncertainty depends on the number of end-members in the model, which end-member is modified, and the contribution of dechlorination to variability. The quantitative integration of the spatial distribution of fate processes in a layered information system applied to contaminated sediments helps focus resources for effective decision-making.