The purpose of this Triad-based, high-resolution field investigation was to define patterns and determine source(s) of polychlorinated biphenyls (PCBs) in a reach of the Lower Duwamish River, Seattle, Washington, and to generate data to be used to guide Remedial Investigation/Feasibility Study (RI/FS) and Corrective Measures Study (CMS) decisions. The data assisted EPA in assessing responsibility for the contamination. At the outset, the regulated parties involved were averse to collecting sediment data; the issue of responsibility was contentious because the PCBs appeared to be within collocated plumes. The program is associated with two regulatory authorities, the Resource Conservation and Recovery Act (RCRA) and the Comprehensive Environmental Response, Compensation and Liability Act (CERCLA).
The CSM for PCB transport in the river indicated possible up-river transport of contaminated sediments by seasonal estuarine counter-currents to normal river flow. The downstream party disagreed with EPA, and a nearly 2-year dispute occurred. Up-river, the presence of PCBs was known, but their distribution was not understood in relation to other sources. Through careful systematic project planning (SPP), a real-time measurement system, and a dynamic work strategy (DWS), the project team was able to collect data to resolve the PCB patterns and to permit EPA to make a "boundary" decision concerning the influence of two different source areas. Immunoassay analyses (IA) of sediments from shallow borings were used to direct placement of deeper cores. Because of the legal/regulatory nature of the final decision, all samples were "split" into Aroclor samples (for off-site laboratory analysis) and IA samples (for in-field analysis), resulting in a high-density collaborative data set (97 stations) that permitted a DMA for the IA in a complex urban sediment mixture. This data set was evaluated retrospectively to determine the false positive error [FPE] and false negative error [FNE] rates of the IA kits relative to risk-based regulatory thresholds. Based on the investigation findings, a second source of upstream PCBs was delineated relative to bankside features.
As a result of this project’s high-density data set as well as data generated by other sources, EPA’s sediment boundary decision, finalized in September 2005, has now been accepted by the regulated parties.
|Site Name||Boeing Company Plant 2|
|Site Regulatory ID||Site ID# 110000489463/RCRA# 1092-01-22-3008(h)|
|Site Name||Jorgenson Forge Corporation (as part of the Lower Duwamish Superfund Site)|
|Site Regulatory ID||WA0002329803|
|Project Lead Organization||U.S. Army Corps of Engineers (USACE), on behalf of EPA|
|Project Lead Type||EPA Lead|
|Regulatory Lead Program||RCRA Corrective Action (Boeing Plant 2), Superfund Remedial (Jorgenson Forge)|
|Triad Project Status||Field Program Completed|
|Reuse Objective Identified||No|
The present investigation was intended to break the stalemate and advance EPA’s decision on boundaries and responsibilities for impacts to the river. The patterns of PCB contamination known at the outset of the study indicated both surface hot-spots and complex depth trends. The USACE proposed use of a high-resolution, Triad-based approach during Technical Project Planning meetings with EPA, and was able to get additional support from EPA’s Office of Superfund Remediation & Technology Innovation (OSRTI), Technology Integration & Information Branch (TIIB) to test the application of the IA technology. The US Navy SPAWARSYSEN Laboratory in San Diego, CA, provided off-site IA support because it was more cost-effective than on-site. The project phasing consisted of 1 week in the field; one week to permit IA analysis (with turnaround within 3 days of sampling including transport time to the laboratory), reporting, visualizations, and evaluation; and 4 days of coring. Fixed-base laboratory samples were analyzed for PCBs by EPA laboratories, and, because of busy laboratory schedules, took 4.5 months to complete. The latter was longer than had been planned.
Three PCB mixtures (Aroclors) were present as detectable quantities in the sediment: Aroclors 1248, 1254, and 1260. Regulatory limits for PCBs in Washington to protect ecological receptors are normalized to the organic carbon (OC) content of the sediments. The conversion is: milligrams per kilogram-OC (mg/kg-OC) = mg/kg-dry/(OC*1000). Thus, a value of 130 micrograms per kilogram (µg/kg) total PCB at 1% carbon is 13 mg/kg-OC. PCB concentrations exceeding 12 mg/kg-OC represent a potentially significant exposure. Values greater than 65 mg/kg-OC require cleanup to 12 mg/kg-OC or lower. Results were compared to these ranges when comparing the technologies.
The investigation sought to do the following:
The total number of split samples analyzed by both the IA kits and the laboratory was 97. Seventeen samples (17.5%) were misclassified by the IA kits based on FNEs and 6 samples (6.2%) were misclassified based on FPEs in relation to both the lower and higher risk-based standards described above. Total misclassification error of the kits was thus 23.7%, with approximately twice as many under-estimates as over-estimates. The differential sensitivities ("cross-reactivities") of the IA for the Aroclor composition of the samples were also assessed. The IA is calibrated for Aroclor 1254, with published relative sensitivities of 117% to Aroclor 1248 and 64% to Aroclor 1260. These published sensitivities are available at Strategic Diagnostics, Inc.
Overall, approximately 6% (1/3) of the FNE appears to have occurred due to cross-reactivities. A second contribution to the error appears to be within-sample heterogeneity. Six replicate samples (from the same core) were submitted for Aroclor analysis; the mean Relative Percent Difference (RPD) was 66% and the maximum was 106%, which represent significant heterogeneity. A third contribution is the quantity of water in the sample; greater amounts of water reduce the efficacy of the methanol extractant in the kits. The latter two sources of error may be reduced in the field by thorough sample mixing and decanting excess water prior to extraction. Post-sampling interpretation frameworks (e.g., use 10 mg/kg-OC instead of 12 mg/kg-OC as the decision point) were applied to the data to see whether the FNE could be reduced without affecting FPE: generally speaking, decreasing the FNE increases the FPE in a proportionate manner.
During the field event, a short-term (3-day) visualization of the site data showed widespread surface contamination between 12 and 65 mg/kg-OC with 3 mid-reach stations >65 mg/kg-OC. Two sub-surface maxima were noted between Outfalls 8 and 9, at the northern (down-river) end of the reach. The southern or up-river end of the reach was not contaminated at depths below 1 foot. When the OC data were available (10-12 days following the field event), these patterns were unchanged except for slight reduction in the intensity of the deeper hot-spots. When the Aroclor data were available (4.5 months following the field event), the mid-reach surface stations were confirmed to be >65 mg/kg-OC, and the number of contiguous stations with such exceedances increased. Also, the subsurface hot-spots were confirmed. More subsurface stations in the southern reach were between 12 and 65 mg/kg-OC than with the IA tests. These differences are consistent with a systematic under-estimation of PCBs by the IA tests. However, the pictures are remarkably similar with respect to the major features. Plots were created with GIS to emulate conceptual dredge plans based on both data sets. In the Aroclor dredge plan, the estimated depths were 2-7 feet; in the IA plan, 3-6 feet. A difference plot shows that the IA correctly estimated (<2 feet difference) about 70% of the project area.
Based on the results of the study, the project team concluded that the IA data were sufficient by themselves to address the questions raised in the CSM, namely that separate PCB releases associated with the bank along the Jorgensen Forge property had occurred and had led to subsurface contamination. However, the test somewhat underestimated the absolute extent of the surface contamination.
The team further concluded that:
The team’s project objectives were to:
The Triad team developed specific data collection objectives in support of the project objectives, which included:
The Triad-based, high-resolution approach offered a cost-effective approach to guide the collection of information necessary to ensure that vertical and horizontal patterns could be discerned with confidence, so that decisions regarding the location and influence of PCB sources could be made. Investigation costs and mobilizations were optimized by near-real-time selection of key "boundary" samples. In addition, the strengths and weaknesses of the IA kits as applied to sediments were quantified.
There were two principal benefits of using a Triad-based, high-resolution approach for this investigation. First, IAs were used in conjunction with data visualizations (using the Spatial Analysis and Decision Assistance [SADA] freeware package) to place new samples at locations of greatest utility in defining hot-spots and determining depth-wise patterns in sediment cores. Second, the retrospective data analysis showed the utility of the IA kits in overall uncertainty management. Due to the high-density collaborative (IA and laboratory) data set, it was possible to look at the overall decision uncertainty both with and without IA screening. Based on the results of this investigation, the IA may be used both to guide field coring and to "triage" core depth-intervals to more rapidly complete a sediment investigation.
The incremental cost of the field IA program was approximately $24,000, or 16% of the overall investigation cost of $147,000. Cost savings were not the emphasis of this study because 100% of the IA samples were subjected to off-site laboratory analyses in order to generate a collaborative data set. However, the activity demonstrated that the IA in conjunction with site visualization can accelerate nature and extent characterizations for PCBs in sediments. The low relative cost and high data generation rate for the kits further showed how a more accurate CSM can be developed through high data densities and in-field decision-making.
A retrospective analysis of the patterns available at various stages of the investigation (short- and intermediate-term with partial fixed-base laboratory data, and after all data were in hand) suggests that the key CSM questions were answered largely by the field method; that is, patterns were discernible that supported both upstream flow and bank-origin collocated plumes. Numerous prior phases of sampling had been performed that had failed to identify the boundaries of contamination. This sampling finally answered the major CSM questions in a single mobilization largely using field-based methods, and further sampling activities will not be required prior to the design phase of site response.
Systematic project planning (SPP) was accomplished by EPA, USACE, Washington Department of Ecology (Ecology), the Boeing Company, and Jorgensen Forge Corporation. Due to the two regulatory programs and contentious issues, however, EPA and the USACE were the principal planners. The physical CSM, primarily a physical transport prediction, was presented to the regulated parties.
The initial CSM for the Boeing/Jorgensen river reach area was based on historical information and data from previous investigations. RCRA constraints required EPA to make a decision as to limits of responsibility of the Boeing Company; that is, a "boundary decision." Acoustic Doppler Radar data indicated seasonal reversing flows near the riverbed, which EPA believed could transport and deposit PCBs up the river from a downstream source as well as the expected upstream-to-downstream direction. PCB inputs were identified on the Boeing Company property at several outfalls. To refine the CSM, however, contributions from upriver sources were required; a few up-river stations had previously indicated PCBs elevated above risk-based levels of concern.
EPA Method 4020 IA kits showed promise in saving investigation costs while achieving an acceptable level of data density over the study area. Although these kits had been used for sediment screening in US Navy projects in Hunters’ Point, California, and also to triage samples in Dyes Inlet in Puget Sound, they had not been used in regulatory decisions, and their published sensitivities are higher than used in sediment remediation. During the SPP process, it was perceived by the project team that prior to such use, a DMA was needed. Thus, this investigation program was designed to provide information for the DMA.
Concurrently with the DMA, the Boeing Company and Jorgensen Forge Corporation accomplished focused sampling in portions of the project reach. After the IA DMA results were entirely in hand and evaluated, USACE, Ecology, and the Boeing Company cooperated on data visualization to support the decision. These data include values not shown in the embedded graphics linked to from this Profile, which feature just the Triad data set.
From a non-collaborative beginning, the joint picture developed through the project was suitable to achieve concurrence by the stakeholders on the revised CSM.
The project phasing consisted of:
The results were transmitted by email and ftp between San Diego and Seattle. EPA and USACE met to discuss results as they were mapped. The physical CSM suggested that PCB sources to sediments would be distributed in a "plume" both downstream and upstream from the source of release. The distance that the plume travels by riverine transport processes would be determined by the strength and longevity of the source as well as current strengths and critical velocities for the sediments. Deeper, significantly elevated PCB concentrations would suggest a local, intense source. Distribution of PCB particles by the river currents would "smear" hot spots along the river surface, but likely could not create new ones given the topology of the river bed. Therefore, horizontal patterns were examined in the first project sampling phase, using the 10-cm (biological compliance interval) and 1-2 feet depth interval. Based upon results of the first set of Ias, a second sampling event (as part of the same mobilization) targeted several intensely-contaminated areas with deeper (up to 4 foot) cores to determine vertical patterns. Additional surface samples were also taken to adaptively "fill" possible hot spot boundaries.
This entire project is essentially a DMA for IA in a moderate-scale real-world activity. Collaborative data for the Ias were from EPA regional laboratories which, because of throughput issues, were slow to provide results. A high-density of IA data was collected in a short time frame by the SPAWARSYSEN laboratory. A detailed calibration and QA/QC program was implemented to assist in the real-time evaluation of kit performance. Data management and data communications were not problematic because a period of field "down-time" had been programmed into the contract to receive and review the initial results. Team members consisting of EPA, Herrera Associates, SPAWARSYSEN, and USACE were able to participate in the expedited data review and decision-making process (for additional core sampling locations) while the field crew was mobilized. Following receipt of the voluminous Aroclor data, spreadsheets and a database were used to manage them along with the immunoassay data. Eventually, the Aroclor data were prepared for submission to the regional sediment database, SEDQUAL. Additional details concerning the performance of the kits are presented in the Project Results and Outcomes and Data Quality Assessment fields.
TQRS not prepared
Quality control protocols for the immunoassay kits included:
Requirements for these QC protocols were met by the SPAWARSYSEN analytical team.
|U.S. Army Corps of Engineers (USACE). 2005. Use of the Triad Approach to Characterize PCB in a Riverine Sediment Site. Presentation at AEHS Sediment and Soils Conference. March. (720 KB)|
To update this profile, contact Cheryl T. Johnson at Johnson.Cheryl@epa.gov or (703) 603-9045.