Fate and Transport Modeling of Sediment Contaminants in the New York/New Jersey Harbor Estuary
by Robin E. Landeck Miller, Kevin J. Farley, James R. Wands, Robert Santore, Aaron D. Redman, and Nicholas B. Kim
HydroQual, Inc., 1200 MacArthur Blvd., Mahwah, NJ 07430 T: 201-529-5151 F: 201-529-5728
Sediment contamination in the NY/NJ Harbor estuary has adversely affected both disposal costs and disposal options for material dredged from the Harbor. In response to this problem, the Port Authority of New York and New Jersey and several state agencies, through a bi-state dredging agreement, formed the Contamination Assessment and Reduction Project (CARP). One feature of CARP was the development of a series of numerical models that serve as both diagnostic and predictive tools. The CARP numerical models include hydrodynamic, sediment transport, organic carbon production, contaminant fate and transport, and bioaccumulation models. These models describe the causal link between external sources of contaminants and ambient concentrations of multiple contaminant classes in water, sediment, and biota of the Harbor. The external sources include tributary headwaters, sewage treatment plants, urban runoff, combined sewer overflow, atmospheric deposition, and landfill leachate. The model domain includes the Passaic River south of Dundee Dam and contiguous waterways such as the Hackensack River, Newark Bay, Kill van Kull, Arthur Kill, Raritan Bay and River, Upper NY Bay, East and Harlem Rivers, Jamaica Bay, Long Island Sound, and New York Bight. The contaminant classes considered include PCBs, dioxin/furans with 2,3,7,8 substitutions, organochlorine pesticides related to DDT and chlordane, PAHs, and the metals cadmium, mercury, and methyl mercury. After several years of development and calibration, the CARP models now diagnose how much of observed Harbor contamination results from current loadings versus legacy contamination still remaining in the system. Further, the CARP models have been used to forecast expected future contamination levels achievable through natural attenuation and a combination of natural attenuation and various reductions of current loadings and/or removal and remediation of in-place sediments. The modeling approach and application applied under CARP serves as an excellent case study for other urban estuaries and ports. Although developed specifically for the NY/NJ Harbor, the CARP model kinetic formulations are easily transferable to other systems. Model source codes are available upon request from the Hudson River Foundation at www.carpweb.org. Some of the novel features of the CARP model include mechanistic mercury methylation kinetics and the inclusion of a eutrophication model.
Key Words: numerical model; contaminated sediments; contaminant fate and transport; bioaccumulation; Passaic River; NY/NJ Harbor; dioxin; PCB; dredged material management; Contamination Assessment and Reduction Project (CARP)
Sediment contamination in the NY/NJ Harbor estuary has adversely impacted both disposal costs and disposal options for material dredged from the Harbor. In response to this problem, the Port Authority of New York and New Jersey and several state agencies, through a bi-state dredging agreement, formed the Contamination Assessment and Reduction Project (CARP). CARP's objectives, originally set out by the Hudson River Foundation in its July 2000 request for development of numerical models, include quantifying the relative importance of specific contaminants in dredged material both today and in the future, and quantifying how dredged material quality will change over time if specific loadings are left as they are or reduced. Other goals include prioritizing actions needed to reduce the fraction of dredged material that is contaminated from 70% in 2000 to less than 15%. These dredged material–related goals coincide with objectives of the NY/NJ Harbor Estuary Program (HEP), and include attainment of water quality standards and maintenance of a healthy ecosystem (NY/NJ Harbor Estuary Program 1996).
To achieve these goals, CARP undertook an ambitious data collection and numerical modeling effort. Data collection (see www.carpweb.org) included contaminant measurements in ambient Harbor water, sediment, and biota and in contaminant loading sources over a four-year period from 1998 to 2002. This effort was implemented by several state and federal agencies and academic institutions. Numerical modeling was necessary to define possible cause-and-effect relationships between sources of contamination and observed levels of contamination, to bridge temporal and spatial gaps in the data collected, and to predict the response of water, sediment, and biota contaminant levels to changes in contaminant loadings. The numerical modeling effort is further described below.
After several years of development and calibration, CARP models now diagnose how much observed Harbor contamination results from current loadings and how much is legacy contamination still remaining in the system. Further, the CARP models have been used to forecast expected future contamination levels achievable through natural attenuation or through a combination of natural attenuation and various reductions of current loadings and/or removal and remediation of in-place sediments. The model development and application processes are described below along with key results.
The CARP numerical models include hydrodynamic, sediment transport, organic carbon production, contaminant fate and transport, and bioaccumulation models. These models account for the contribution of external sources of contaminants, such as tributary headwaters, sewage treatment plants, urban runoff, combined sewer overflow, atmospheric deposition, and landfill leachate, to ambient concentrations of contaminant classes in Harbor water, sediment, and biota. The contaminant classes considered include PCBs, dioxin/furans with 2,3,7,8 substitutions, organochlorine pesticides related to DDT and chlordane, PAHs, and the metals cadmium, mercury, and methyl mercury (HydroQual 2007c).
The relationship between the various CARP numerical models and the various media considered (i.e., air, water, sediment, and biota) are shown in Figure 1.
Separate contaminant fate and transport kinetics were developed for hydrophobic organic, metal, and methylmercury contaminants. Each CARP model required both detailed forcing information, which was based upon analysis of CARP data, and specification of model constants and coefficients, based on literature values and values used for similar project areas, when site-specific data were lacking (HydroQual 2007c).
Hydrodynamic transport modeling for CARP (HydroQual 2007a) required applying a previously calibrated and validated hydrodynamic transport model, one used in the System Wide Eutrophication Model (SWEM) (Landeck Miller and St. John 2006), to the CARP 1998–2002 data collection period. This model (Blumberg et al. 1999) is based on the Estuarine, Coastal, and Ocean Model (ECOM) (Blumberg and Mellor 1987) source code. It is driven by measuring water level, meteorological forcing, and spatially and temporally varying surface heat flux and freshwater fluxes from the numerous rivers, wastewater treatment plants, combined sewer overflows, land runoff, and landfills that enter the NY/NJ Harbor Estuary, Long Island Sound, and the New York Bight. The hydrodynamic model solves a coupled system of differential, prognostic equations describing conservation of mass, momentum, heat, and salt.
Skill assessments of the performance of the hydrodynamic model under 1998–2002 conditions were made using data collected by CARP and other agencies in ongoing, routine monitoring programs.
Sediment Transport and Organic Carbon Production Modeling
HydroQual's effort on the CARP sediment transport/organic carbon production model (HydroQual 2007b) is one of the first attempts to apply a sediment transport model to a domain as large and complex as the NY/NJ Harbor-Bight-Sound complex. Because field data for the calibration were limited, the model was initially developed using simplified formulations and a set of geographically constant coefficients to describe the relevant processes of settling and resuspension. Spatial variations in settling (based on variations in salinity and fluid shearing rates), resuspension (based on consolidation in sediment), and bottom shear (based on wind waves) were then adopted to provide a better description of sediment transport throughout the CARP model domain. This sequential process of adjusting model coefficients and providing a physical justification for the adjustments is an important aspect of model calibration.
In addition to developing and calibrating a new sediment transport model for the Harbor-Bight-Sound complex, HydroQual's effort incorporated the newly developed and calibrated sediment transport model into the previously calibrated and validated SWEM organic carbon production model, effectively forming a new combined sediment transport and organic carbon production model, Sediment Transport SWEM (ST-SWEM). This necessitated both verifying that the original calibrations/validations of the organic carbon production model from SWEM had not been destroyed when the sediment transport model formulations were incorporated and assessing the ST-SWEM organic carbon production model performance using data collected by CARP and other agencies during the 1998–2002 period.
The CARP sediment transport model development effort included hourly to daily specification of suspended sediment, organic carbon, and nutrient levels in the NY/NJ Harbor based on data that were comprehensive in representing various loading source types but were limited in temporal frequency. However, flow measurements were available at much greater temporal frequency than suspended sediment or particulate organic carbon (POC) measurements. Accordingly, historically observed relationships between suspended sediments/POC loadings and river flow under both baseline and storm conditions were used to estimate suspended sediment and POC loadings. A similar approach to that described in HydroQual 1996 was followed.
Contaminant Fate and Transport and Bioaccumulation Modeling
The CARP contaminant fate and transport and bioaccumulation models (HydroQual 2007c) originate from a simpler mathematical model of the long-term behavior of PCBs in the Hudson River Estuary (Thomann et al. 1989) and an integrated model of organic chemical fate and bioaccumulation in the Hudson River Estuary (Farley et al. 1999, 2006), collectively called the Thomann-Farley model. Some of the technical advantages of the CARP contaminant fate and transport and bioaccumulation models over the Thomann-Farley model are: better spatial resolution of contaminant hot spot and dredging areas, vertical resolution of the water column to capture estuarine two-layer flow dynamics (represented in ten vertical depth layers), open boundaries away from the zone of influence of NY/NJ Harbor contaminant loads, inclusion of the Historic Area Remediation Site (HARS) within the model domain, a mechanistic consideration of hydrodynamic transport, suspended sediment and organic carbon through fully-linked sub-models, incorporation of kinetics for a broader range of hydrophobic organic contaminants, incorporation of kinetics for metal contaminants including mercury methylation/demethylation processes, and inclusion of additional species in bioaccumulation calculations (i.e., polychaete worms, clams, striped bass, white perch, American eel, and blue crab). Also, the Thomann-Farley model did not have the benefit of the comprehensive ambient and loading source data collected by CARP.
The water quality model source code underlying the Thomann-Farley model is WASTOX (Connolly and Thomann 1985, Connolly 1991). The water quality model source code underlying the CARP contaminant fate and transport and the sediment transport/organic carbon production sub-models is Row Column Aesop (RCA). Both RCA and WASTOX originate from the Water Analysis Simulation Program (WASP) developed by Hydroscience (HydroQual's predecessor firm) in the 1970s (DiToro et al. 1981, DiToro and Paquin 1984). RCA code has been used to develop numerous models outside of the NY/NJ Harbor region.
The principal attributes of the RCA source code include:
- RCA is a general purpose code used to evaluate myriad water quality problem settings. The user is able to customize an RCA subroutine to address water quality issues that are specific to a given water body.
- RCA formulates mass balance equations for each model segment for each water quality constituent or state-variable of interest. These mass balance equations include all horizontal, lateral, and vertical components of advective flow and diffusive/dispersive mixing between model segments; physical, chemical, and biological transformations between the water quality variables within a model segment; and point, nonpoint, fall-line, and atmospheric inputs of the various water quality variables of interest.
- The partial differential equations which form the water quality model, together with their boundary conditions, are solved using several mass conserving finite difference techniques.
CARP contaminant fate and transport model kinetics, collectively referred to as RCATOX, include separate routines for hydrophobic organic, divalent metal and methyl mercury contaminant groups. CARP bioaccumulation model kinetics within RCATOX include calculations of both Biota Accumulation Factors (BAFs) and Biota Sediment Accumulation Factors (BSAFs) from site-specific data, as well as more detailed steady-state and time variable mechanistic equations which help explain the behavior of observed BAFs and BSAFs at several pelagic and benthic trophic levels.
Significant aspects of the CARP contaminant modeling include development of contaminant loadings from CARP data and development of site-specific, three-phase partition coefficients for the hydrophobic organic contaminants with temperature and salinity dependencies. The development of metal speciation and mechanistic mercury methylation kinetics within the CARP model is state-of-the-science.
The calibration process for the CARP contaminant fate and transport model involved a current conditions calibration of CARP data collected between 1998 and 2002 for 10 PCB homologs, 17 dioxin and furan congeners with 2,3,7,8 substitutions, 22 PAH compounds, six DDT-related chemicals, five chlordane-related chemicals, and the metals cadmium, mercury, and methyl mercury. PCB homologs rather than PCB congeners were modeled for practical reasons (schedule, budget, etc.); however, the CARP contaminant fate and transport model could be applied to PCB congeners. The calibration process also included a hindcast verification for 137Cs, 2,3,7,8-TCDD, and several PCB homologs in which model simulations were started in 1965 and carried foward to 2002. For 137Cs, the historical loadings were well known. For 2,3,7,8-TCDD and the PCB homologs, reasonable estimates were made of historical loadings. Hindcast model results were compared to data from dated sediment cores.
An important aspect of CARP model development was the involvement of a Model Evaluation Group (MEG). The CARP MEG worked with the Hudson River Foundation (HRF) to select HydroQual as the CARP modeling contractor. The MEG was also involved in frequent and ongoing peer review of every aspect of the CARP model development and application process. MEG review comments are incorporated in the appendices of HydroQual 2007a–c.
Results and Discussion
While each of the CARP numerical models has its own set of independent and instructive results, the discussion here will focus on the results of applying the various CARP models in series. The development and calibrations of hydrodynamic and sediment transport/organic carbon production were sufficiently detailed to force the contaminant fate and transport and bioaccumulation models. The current conditions calibrations for the CARP hydrophobic organic and metals models, by virtue of the calculations reproducing measured contaminant concentrations, demonstrate that source, sink, transport, and transformation terms were correctly represented. The hindcast verification exercise, by virtue of the calculations reproducing measured contaminant concentrations over several decades, demonstrates that the model has the time dynamic correct for exchange processes between the water column and sediment bed. Correct representation of this time dynamic is critical for projecting future conditions.
Contaminant Fate and Transport
At an interim point in the CARP hydrophobic organic contaminant (HOC) model calibration process, a sensitivity analysis was performed on the assignment of initial contaminant concentrations in the sediment bed. For purposes of these sensitivity calculations, the initial sediment bed concentrations of 10 PCB homologs and 17 dioxin/furan congeners were set to zero rather than the concentrations interpolated (using the Laplacian inverse distance method) from observed data. "Clean bed" model simulations were carried out for 96 years. The "clean bed" analysis provided useful, albeit preliminary, information with management implications for the time behavior of the system and for separating observed sediment bed contaminant concentrations into those associated with either current-day or legacy source inputs. The latter separation is achieved by subtracting "clean bed" CARP model simulation results from the interpolated current concentration measurements, as described below in the discussion of Figure 2.
The "clean bed" analysis shows that the time to steady state, or the time required for the contaminant in the sediment bed to increase to a quasi-steady concentration in response to a continuous loading source, is less than 96 years for most areas in the core of the NY/NJ Harbor as represented by the CARP model computational grid. Residence time of contaminants in the system and the time to steady state are based on exchange between the water column and sediment bed. For most of the CARP model domain, the particle mixing rate controls this exchange. One exception is the majority of the East River, where shear stresses are extremely high and suspended solids concentrations are relatively low. In major portions of the East River, almost no particle accumulation or particle exchange occurs, and the dissolved mixing rate controls exchange between the water column and sediment bed. When and where diffusive mixing controls this exchange, particularly for high Kow compounds, time to steady state can be upwards of one hundred years. If any of the mixing processes were overestimated, the CARP model would not be able to predict elevated contaminant concentrations in the sediment bed.
From the perspective of a sensitivity calculation, the results from the "clean bed" analysis show that:
- For most of the CARP domain, the time for the top 10 cm of the sediment bed (represented by ten layers in the model) to reach steady state is approximately 30 years.
- Some areas at the outer fringes of the CARP model domain required 100 years or longer to reach steady state.
- The "clean bed" analysis results indicate that the "memory" of harbor sediments to past contaminant loads is likely around 30 years (since it takes that long for sediments to reach steady state with current continuous loadings) and that the assignment of initial contaminant conditions in the sediment bed is critical in computing long-term responses.
It is noted that, in addition to mixing rates, sediment transport and "estuarine trapping" also impact the time to steady state (HydroQual 2007c). Over a 96-year simulation, particulate phase contaminants are continually resuspended, transported (oftentimes by bottom waters moving in a net landward direction), and redeposited, and this process further impedes the loss of contaminants to the ocean.
In addition to interesting sensitivity results, the "clean bed" analysis results also have management implications. They show how much of the contaminant concentrations observed in surficial bed sediments today can be accounted for by current contaminant loading sources. Calculated contaminant concentrations in the sediment bed are presented in a color concentration map format in three ways (Figure 2): based on interpolated data, based on the results of a CARP model "clean bed" simulation, and based on the subtraction of interpolated data and "clean bed" simulation results. The interpolated data represent the contamination due to both current and legacy sources of contamination. The "clean bed" simulation results represent contamination due only to the current loading sources included in the CARP model. Subtracting the "clean bed" results from the interpolated data indicates the contribution to sediment contaminant concentrations from loading sources not included in the CARP model, interpreted to be legacy sources. Example diagrams are presented here as Figure 2 for 2,3,7,8-TCDD. For the majority of contaminants, the observed contaminant concentrations in the sediment bed cannot be explained by the current day loadings and are therefore likely due to historical sources. This is particularly evident for some of the higher chlorinated PCB homologs, 2,3,7,8-TCDD, and 1,2,3,4,7,8-HxCDF.
Use of alternative interpolation schemes for developing estimates of the observed concentrations in surficial bed sediments today from measurements would not produce concentrations similar to "clean bed" outputs. "Clean bed" results were subsequently confirmed when the final CARP model was used to perform a loading source component analysis. In this analysis, each loading source category (e.g., sediment initial conditions, tributary head-of-tide, runoff, sewage treatment plants, combined sewer overflows, atmospheric deposition, etc.) was activated individually in the model to isolate its impact on contaminant concentrations in water, sediment, and biota throughout the system over a simulated 32 years. Like the "clean bed" results, the loading component results indicate that legacy sediments are a major component of observed contamination, particularly for 2,3,7,8-TCDD. The component results further demonstrate that of the current loading sources, runoff and head-of-tide appear to be most important. Loading component results suggest that over time, overall contaminant levels in surficial sediments will drop and the surficial sediment contamination will become less attributable to legacy sources.
The calculation of BAFs and BSAFs using measured chemical concentrations in various organisms (i.e., zooplankton, white perch, striped bass, mummichog, American eel, winter flounder, blue crab, clams, and worms) sampled by CARP and corresponding exposure concentrations from the water column and sediment, either based on data or extracted from the CARP contaminant fate model, produced interesting results, particularly for the accumulation of hydrophobic organics in worms.
Regarding this bioaccumulation, we specifically observed (1) a variation in BSAFs for PCB homologs as a function of KOW, and (2) significant differences in BSAFs for discrete locations within the Inner (i.e., Arthur Kill, Newark Bay, and Upper Bay) and Outer (i.e., Long Island Sound, Jamaica Bay, and Sandy Hook) Harbor. These homolog and spatial patterns are shown in Figure 3. Given the serious implications this observed difference in BSAFs could have for setting sediment cleanup targets based on desired dredged material endpoints, we evaluated possible reasons for the spatial differences using a mechanistic steady-state bioaccumulation model and bioenergetic parameters from the literature.
The observed spatial variation in BSAFs could be modeled or explained by: (1) differences in chemical assimilation efficiencies (presumably attributable to the quality of the worms' food supply) for the Inner and Outer Harbor sites, (2) differences in respiration or growth rates (presumably associated with environmental stressors) for the Inner and Outer Harbor sites, or (3) the presence of predatory worms at Outer Harbor sites. Based on our sensitivity modeling alone, it is not possible to definitively ascribe the differences in BSAFs to a single cause or to some combination of the three. Our analysis of NY/NJ Harbor PCB bioaccumulation data presented by others (Meador et al. 1997) confirms that there is a geographic difference in BSAFs for NY/NJ Harbor worms. Further, since these data (Meador et al. 1997) included only a single worm species, they rule out the possibility that differences in worm populations or the presence of predatory worms could explain the differences in BSAFs observed at Inner and Outer Harbor sites. Lastly, there is some evidence that there are differences in growth and possibly other bioenergetic behaviors between Inner and Outer Harbor worms (Rice et al. 1995).
The exercise of developing, calibrating, and applying the CARP models thus far has led to many important conclusions regarding the contamination of sediments and dredged material in the NY/NJ Harbor, describing both conditions today and what might be expected in the future. It is clear that both field data and corresponding model results are critical in evaluating contaminant fate and bioaccumulation behavior of dredged material. Tidal resuspension and estuarine circulation cause long-term trapping of particle-bound contaminants in the NY/NJ Harbor. As a result of this trapping and the persistence of many of the contaminants, contamination of surficial sediments is due to both current and historical sources.
The "clean bed" analysis suggests that historical sources were much larger than current sources for most contaminants, and that if NY/NJ Harbor sediments were to undergo remediation, current sources would likely produce some surficial recontamination, but not to the extent of current observed levels of contamination. Preliminary component-run findings confirm these conclusions in that the observed sediment initial conditions are a major component of contamination. These results indicate that over several decades, contamination in surficial sediments will decline, as current sources of contamination are smaller than legacy sources. Of the current sources of contamination, runoff and head-of-tide appear to be most important for many contaminants.
CARP bioaccumulation data and modeling results suggest that varying levels of contamination throughout the NY/NJ Harbor may be affecting bioaccumulation behavior of benthic organisms. Additional data need to be collected to support these findings.
It is likely that even more conclusions will emerge as existing CARP model outputs are studied further and additional model simulations are performed. It is intended that the CARP models will be used by USEPA Region 2 and the States of NY and NJ for TMDL analyses and by the interagency Regional Sediment Management Work Group (RSMWG) for strategic planning. CARP model results will help focus future TMDL, Superfund, and Restoration data collection and modeling efforts.
Lead Author Biography
Robin Landeck Miller is an associate at HydroQual, Inc. in Mahwah, New Jersey. Ms. Landeck Miller has more than twenty years of experience addressing water and sediment quality issues in the NY/NJ Harbor. She holds a Master of Science degree in Environmental Engineering and a Bachelor of Science degree in Biology, both from Manhattan College.
Funding for this work was provided by the Port Authority of New York and New Jersey through the Bi-State Dredging Agreement between the States of New York and New Jersey. This work was contracted through the Hudson River Foundation and is part of the Contamination Assessment and Reduction Project (CARP).