The New Bedford Harbor Superfund sitelong-term monitoring program (1993–2009)

William G. Nelson & Barbara J. Bergen

Received: 6 June 2011 /Accepted: 2 January 2012# Springer Science+Business Media B.V. (outside the USA) 2012

Abstract New Bedford Harbor (NBH), located insoutheastern Massachusetts, was designated as a marineSuperfund site in 1983 due to sediment contamination bypolychlorinated biphenyls (PCBs). Based on risks tohuman health and the environment, the first two phasesof the site cleanup involved dredging PCB-contaminatedsediments from the harbor. Therefore, a long-term mon-itoring program (LTM)was developed to measure spatialand temporal chemical and biological changes in sedi-ment, water, and biota to assess the effects and effective-ness of the remedial activities. A systematic, probabilisticsampling design was used to select sediment samplingstations. This unbiased design allowed the three seg-ments of the harbor to be compared spatially and tempo-rally to quantify changes resulting from dredging thecontaminated sediments. Sediment was collected at eachstation, and chemical (e.g., PCBs and metals), physical(e.g., grain size), and biological (e.g., benthic communi-ty) measurements were conducted on all samples. Thispaper describes the overall NBH-LTM approach andthe results from the five rounds of sample collections.There is a decreasing spatial gradient in sediment PCB

concentrations from the northern boundary (upper har-bor) to the southern boundary (outer harbor) of the site.Along this same transect, there is an increase in biolog-ical condition (e.g., benthic community diversity). Tem-porally, the contaminant and biological gradients havebeenmaintained since the 1993 baseline collection; how-ever, since the onset of full-scale remediation, PCB con-centrations have decreased throughout the site, and oneof the benthic community indices has shown significantimprovement in the lower and outer harbor areas.

Keywords Long-termmonitoring . Sediment . PCBs .

New Bedford Harbor . Superfund

Introduction

The introduction of anthropogenic contaminants into theenvironment has resulted in numerous aquatic areasthroughout the USAwhere sediment is contaminated topotentially harmful levels (USEPA 2004a). These sitesare in both freshwater and marine locations, and posevarious degrees of risk to both human health and theenvironment. The highest risk sites are identified on theEnvironmental Protection Agency's (EPA) National Pri-orities List (NPL) for cleanup under the ComprehensiveEnvironmental Response, Compensation, and LiabilityAct (CERCLA) and additional sites under the ResourceConservation and Recovery Act (RCRA) (USEPA2004a). Due to the costs and potential risks associatedwith contaminated site remediation, it is crucial that both

DOI 10.1007/s10661-012-2517-0

W. G. Nelson (*) : B. J. BergenOffice of Research and Development,National Health and Environmental Effects Laboratory,Atlantic Ecology Division,US Environmental Protection Agency,27 Tarzwell Drive,Narragansett, RI, USAe-mail: Nelson.William@epamail.epa.gov

Environ Monit Assess (2012) 184:7531–7550

/Published online: 28 January 2012

the environmental effects and effectiveness of remedialactivities be rigorously documented.

The need to implement effective monitoring pro-grams at contaminated sediment sites has been empha-sized many times (NRC 1997; Goldberg and Bertine2000; USEPA 2005). Most recently, the National Re-search Council (NRC 2007) conducted a review of con-taminated sites and indicated that (1) pre-remediationmonitoring is essential to develop baseline information;(2) monitoring should be supported with rigorous statis-tical analyses; (3) although measuring fish tissuecontaminant concentrations is important, site-relatedimpacts are possibly better assessed by examining ben-thic community condition; and (4) because of the ex-tended time frame associated with contaminatedsediment remediation (especially “mega-sites,” whereremedial costs are >$50 million), consistency and flexi-bility in monitoring are both critical to make usefulassessments. Even before these national calls for robustmonitoring programs, it was decided that because of thepotential cost of remediation at the New Bedford Harbor(NBH) Superfund site, a long-term monitoring (LTM)program would be essential to document the benefitsgained for the costs incurred; therefore, the NBH-LTMprogram was designed and first implemented in 1993(Nelson et al. 1996). The program incorporates all therecommendations called for in the 2007 NRC report andis described in this paper.

The NBH Superfund site, located in Bristol County,Massachusetts (MA), extends from the shallow northernreaches of the Acushnet River estuary, south through thecommercial harbor of New Bedford, MA, and into17,000 acres of Buzzards Bay (Fig. 1). Industrial andurban development surrounding the harbor has resultedin sediments becoming contaminated with high concen-trations of many pollutants, notably polychlorinatedbiphenyls (PCBs) and heavy metals (e.g., cadmium,chromium, copper, and lead) with contaminant gradientsdecreasing from north to south (Nelson et al. 1996). In1979, various limitations on fishing, shellfishing, andlobstering were implemented in NBH and portions ofBuzzards Bay due to elevated levels of PCBs in seafood.The site was proposed for the EPA's NPL in 1982 andwas finalized on the NPL in September 1983. EPA's site-specific investigations began in 1984, including pilotstudies on various remedial approaches (Nelson andHansen 1991).

Remedial activities in NBH have varied in size (i.e.,number of cubic yard dredged) and duration (Fig. 2). In

the area north of the Coggeshall St. Bridge [termed theupper harbor (UH), Fig. 1], several remedial activitieshave occurred in the subtidal area. The “Hot Spot”remediation removed sediments with PCB concentra-tions greater than 4,000 ppm and was completed in thefall of 1995 (Bergen et al. 2005). Limited dredgingoccurred between 2000 and 2004, and full-scale dredg-ing to remove “Operable Unit 1” sediments began in2004 in the UH. The area between the Coggeshall St.Bridge and the Hurricane Barrier [termed the lowerharbor (LH), Fig. 1] has several smaller areas identifiedfor remediation. Several locations from the HurricaneBarrier to the outer closure line [termed the outer harbor(OH), Fig. 1] are also being investigated currently forpossible remediation. Additional site history and infor-mation is available online at the NBH web site (http://www.epa.gov/nbh).

Prior to the 1994 dredging of the “Hot Spot” in theUH, it became evident that it would be necessary toanswer the basic question, “How can the effectivenessof the remedial actions taken in NBH best be docu-mented?” Based on that question, a comprehensiveLTM program was designed in 1993 to assess the over-all effectiveness of the remedial activities by document-ing the environmental benefits. There is no all-encompassing monitoring blueprint because Superfundsites can be quite different, both with respect to types ofcontaminants (e.g., PCBs, metals, and dioxins) andphysical characteristics (e.g., freshwater streams, andestuaries). However, certain elements are inherent inany well-designed monitoring program. These include(1) identifying the specific goals of the monitoring ef-fort, (2) selecting the most appropriate variables to mea-sure so that managers and/or scientists can determinewhether those goals are met, (3) designing a statisticallyrigorous program to provide quantitative assessments,(4) establishing quality assurance guidelines that areflexible enough to incorporate technological improve-ments (e.g., analytical advances and new metrics), and(5) ensuring that the results can be understood by andtransferred to a broad spectrum of stakeholders (e.g.,managers, scientists, and the public).

The primary goal of the NBH-LTM program is toassess the effectiveness of the remediation by quantify-ing spatial and temporal biological and chemicalchanges in different environmental compartments. Be-cause the primary focus of remedial activities in NBH isthe sediment (i.e., main repository for PCBs and othercontaminants), concentrations of PCBs (measured as 18

7532 Environ Monit Assess (2012) 184:7531–7550

congeners) and nine metals were quantified. Sedimenttoxicity tests were also conducted to ascertain short-termacute effects, and the benthic invertebrate communitywas characterized to assess the longer-term chroniceffects of contaminants on biota. Several other variables

such as sediment grain size and concentrations of totalorganic carbon (TOC) and acid volatile sulfide (AVS)were measured because they can affect chemical bio-availability and can be important when interpreting bi-ological effects.

1993 1995 1999 2004

Baseline “Hot Spot”Dredging

SedimentDredged

(yd3)

14,000(Upper)

0 20,600(Upper)4,000

(Lower)

159,000(Upper)

NoDredging

LimitedDredging

Full ScaleDredging

2009

Fig. 2 Timeline (not to scale) showing the relationship betweenthe five long-term monitoring collections (red), the remedialdredging activities (black), and the cubic yards of contaminatedsediment dredged (brown) during each collection interval. The

total amount of material dredged up to, and including, the 2009collection (about 200,000 cy3) represents approximately 20% ofthe total volume to be removed; another 710,000 cy3 are sched-uled to be remediated

New Bedford Fairhaven

UH

OH

LH

NBH 2 (Coggeshall St.)

NBH 4

NBH 5

Hurricane Barrier

Outer boundary

Fig. 1 Map of the NewBedford Harbor (NBH)Superfund site showing thethree segments of the harbor(UH, LH, and OH), with thehexagonal systematic sam-pling grid superimposed onit. Sediment samples werecollected at the center of eachhexagon for each of the fivecollections (1993, 1995,1999, 2004, and 2009). Alsoshown are the three musseldeployment stations(NBH-2, NBH-4,and NBH-5)

Environ Monit Assess (2012) 184:7531–7550 7533

In addition to the sediment, PCB tissue concentrationsin selected marine organisms from NBH are being mon-itored separately by the State of Massachusetts to assessthe risks of consuming fish and shellfish. However,directly linking some biota, such as fish, to specificremedial actions can be difficult because of their mobil-ity and exposure to offsite conditions (NRC 2007). Analternative approach is to use deployed organisms toquantify changes in PCB water column concentrations.Filter-feeding bivalves (e.g., mussels and oysters) “sam-ple” the water column almost constantly, thus integratingthe effects of tides and weather. Because of this, bluemussels have been used extensively for many years toquantify chemical pollution (Farrington 1983; Tanabeet al. 1987). Caged mussels have been used in severalprevious monitoring and research efforts in NBH, and anextensive data set has been compiled on PCB accumu-lation in bluemussels,Mytilus edulis (Bergen et al. 1993;Nelson et al. 1995). These studies demonstrated a goodrelationship between mussel PCB tissue concentra-tions and temporally integrated water column con-centrations. Therefore, PCB bioaccumulation in bluemussels was selected for the NBH-LTM program as away to monitor changes in PCB water column concen-trations over time.

With this framework in mind, this paper describes theNBH-LTM program established for the NBH Superfundsite to document the effectiveness of remedial activities.First implemented in 1993 (baseline), the program iscurrently ongoing with five sampling events completedto date, and it incorporates all the elements discussedabove and meets the criteria suggested by other groups,such as the NRC. It uses the general approaches of othermonitoring programs (e.g., EPA's Environmental Mon-itoring and Assessment Program; Paul et al. 1999), suchas a probabilistic sampling design as well as the “triad”approach (Chapman 1990) which combines chemistry,toxicity, and benthic community analysis. Finally, aswith many monitoring programs, questions arise as towhy certain endpoints were selected, how samplingstations are located, etc. For the NBH-LTM, a balancewas struck between the goals of the program, budgetaryrealities, and state-of-the-science available at the time.This overview paper describes the overall monitoringprogram and approach and presents selected data col-lected during the five sampling periods spanning theyears from 1993 to 2009. More comprehensive anddetailed analyses delving deeper into the specific datawill be presented in subsequent papers.

Methods

Station locations

To meet the goals of the NBH-LTM program, a proba-bilistic sampling design was used, and stations wereselected by applying a systematic hexagonal grid, thensampling as close as possible to the center of eachhexagon in 1993 (Fig. 1). Each subsequent NBH-LTMcollection used these original locations. Because differ-ent remedial activities are planned within the UH, LH,and OH areas, spatial and temporal changes within eachof these three segments, as well as the entire Superfundsite, needed to be quantified separately. To accomplishthis, the hexagon size in the sampling grid was adjustedto include approximately 30 stations per segment(Fig. 1); however, due to the shoreline configuration,and non-sampleable areas within each segment, the ac-tual number of stations was 27 in the UH, 29 in the LH,and 23 in the OH. The area of the individual hexagons ineach segment was 0.03, 0.11, and 2.2 km2, respectively,for the UH, LH, and OH segments.

Sediment sampling

Sediment collection procedures were identical at each ofthe sampling locations (USEPA 1995). Sediment wascollected using a Young-modified van Veen grab sam-pler (440 cm2 in surface area). At each site, numerousgrabs were taken to obtain an adequate amount of sed-iment for chemical and toxicological analyses. TheNBH-LTM program is designed to quantify changesover an extended period of time, especially changesresulting from remedial activities; therefore, only thetop 2 cm of these grabs were used in the composite forchemical analysis because these are most reflective ofcurrent sediment concentrations.

From each individual grab, a sub-sample was re-moved for AVS analysis. The remainder of the top2 cm from each grab at a given site was compositedand homogenized before sub-samples were taken forchemistry, sediment toxicity, and grain size analyses.Prior to laboratory analysis, PCB and TOC sedimentsamples were kept cold at 4°C, while the metals andAVS samples were frozen. Sediment for toxicity testingwas taken from the same composite, press-sievedthrough a 2-mm mesh stainless steel screen, thoroughlyhomogenized, and stored at 4°C until testing. Finally,

7534 Environ Monit Assess (2012) 184:7531–7550

approximately 100 ml of the composited sediment wereplaced in a polyethylene bag for grain size analysis.

Three additional grabs were collected and processedseparately at each station for benthic community analy-sis. These grabs were a minimum of 7 cm deep to ensureadequate sampling of the benthic community. Each grabwas processed according to the procedures described inReifsteck et al. (1993). A small core was taken from eachgrab for sediment grain size characterization, indepen-dent of the grain size analysis for the chemistry compos-ite described previously. The remainder of the grab wassieved through a 0.5-mm screen using a backwash tech-nique to minimize damage to soft-bodied animals. Sam-ples were preserved in buffered 10% formalin andseawater solution with rose bengal added as a vital stain.Two of the three samples were processed for taxonomicidentification; the third was archived.

A full suite of variables was measured at each stationfor the first four sampling events (Table 1); however, dueto funding constraints, and an evaluation of results fromthe previous four sampling events, the most recent roundof monitoring in 2009 did not include metals analysis orsediment toxicity (see “Discussion” section).

Sediment chemistry

A rigorous quality assurance project plan (QAPP) wasdeveloped to ensure data comparability from one sam-pling event to another and to be flexible enough to allowfor innovation in analytical techniques. The QAPP isperformance based, delineating the guidelines that mustbe met by participating laboratories (e.g., analysis ofcertified standard reference sediment, matrix spike re-coveries, field duplicate agreement, and blanks) but doesnot specify particular analytical methods. In the interestof brevity, the methods reported here are general, witheach year of the program having some slight modifica-tions to analytical techniques. However, the same qualityassurance standards have been met every year, and spe-cific details of the quality assurance standards promul-gated for the NBH-LTM program can be foundelsewhere (USACOE 1999). Other information andtechnical documents related to the site, including detailsof all analytical techniques used in each sampling year,can be found on the internet at http://www.epa.gov/nbh.

PCB analysis While there are multiple ways to quan-tify PCBs (e.g., Aroclor® mixtures, homologues, andindividual congeners), the sediments in the NBH-LTM

program were analyzed for 18 of the 209 PCB conge-ners. These are the same 18 congeners that are used inthe National Oceanic and Atmospheric Administration's(NOAA) National Status and Trends Program to assessmarine environmental quality (Calder 1986).

As mentioned above, exact PCB methods variedslightly among collection years; however, they are con-sistent with those found in Pruell et al. (1990) andUSEPA (2007). For PCB analysis, approximately 1 gof sediment was solvent extracted with acetone andmethylene chloride. Samples were cleaned by opencolumn chromatography with activated copper. Extractswere further cleaned by treatment with concentratedsulfuric acid, then solvent-exchanged to hexane, andanalyzed by gas chromatography with an electron cap-ture detector for the 18 individual PCB congeners listedin Table 1. Total PCB concentrations were calculated asthe sum of these 18 congeners. As part of this analysis, asecond gram of sediment was weighed, dried, andweighed again to determine moisture content.

Metals analysis Although sediment PCB concentra-tions are the primary contaminant of concern in NBH,metals are present in high concentrations in some areas.Because one goal of the remediation is to remove theseco-located contaminants, sediments were also analyzedfor nine metals, including copper (Cu), cadmium (Cd),lead (Pb), zinc (Zn), arsenic (As), selenium (Se), mercury(Hg), chromium (Cr), and nickel (Ni).

Metal concentrations were determined by ultrasoni-cating a mixture of approximately 5 g of sediment and2 M nitric acid. The resultant extract was analyzed on aninductively coupled plasma spectrophotometer (ICP) oran atomic absorption spectrophotometer (AA) asnecessary.

In addition, the sum of the molar concentrations offive divalent metals (Ni, Pb, Cd, Cu, and Zn) was used inconjunction with AVS concentrations to estimate biolog-ically available metals (Di Toro et al. 1990).

Acid volatile sulfide AVS was quantified as the amountof sulfide released upon extraction of the sediment witha cold 1M hydrochloric acid (HCl) solution. The sulfideliberated in this process was trapped and measured witha sulfide-specific electrode (Boothman and Helmstetter1992).

Total organic carbon The analysis of TOC consistedof first removing the sediment inorganic carbon from

Environ Monit Assess (2012) 184:7531–7550 7535

carbonates and bicarbonates by acid treatment. Organiccompounds were combusted and the resultant carbondioxide measured by direct non-dispersive infrared de-tection (USEPA 2004b).

Sediment toxicity

The test species used in this project is the euryhalinebenthic amphipod, Ampelisca abdita, and it has beenused previously in NBH (Otis and Averett 1988).Following EPA procedures (USEPA 1994), sedimentswere added to exposure chambers 1 day before theamphipods. The exposure chambers were 1 L canningjars with an inverted glass dish as a cover. Two hun-dred milliliters of control or NBH sediment wereplaced on the bottom of each of five replicates persediment sample and covered with approximately600 ml of filtered Narragansett Bay, Rhode Island(RI) seawater. Gentle aeration of the overlying waterand lighting were continuous during the test period.Twenty sub-adult amphipods (passing through a 1.0-mm but retained on a 0.71-mm screen) were addedrandomly to each replicate chamber. Amphipods were

exposed to test and control sediments for 10 days understatic conditions. Salinity of the overlying water rangedfrom 30 to 35 parts per thousand (ppt), and temperaturewas maintained at 20°C.

Sediment grain size distribution

Sediment grain size was determined based on the meth-ods outlined in ASTM D422-63 (ASTM 2007). Shellfragments and organic debris larger than 1 mm wereremoved, then the silt/clay fraction was separated fromthe rest of the sediment using wet sieving and pipetteanalytical techniques. The sediments were categorized asgravel (>2 mm), sands (>63 and <2 μm), and silt/clay(<63 μm) to determine the relative proportions of gravel,sand, and silt/clay.

Benthic community analysis

In the laboratory, macrobenthos were identified to lowestpractical taxonomic level (generally species) andcounted. A complete description of the methods can befound in the Environmental Monitoring and Assessment

Table 1 Suite of physical,chemical, and biologicalvariables collected at eachsediment sampling stationduring the NBH-LTMprogram

Physical Chemical Biological

Grain size PCB congeners Metals

Gravel CB008 Arsenic Sediment toxicity

Sand very coarse CB018 Cadmium Benthic community

Sand coarse CB028 Chromium

Sand medium CB044 Copper

Sand fine CB052 Lead

Sand very fine CB066 Nickel

Silt and clay CB101 Mercury

CB105 Selenium

CB118 Zinc

CB128

CB138

CB153

CB170

CB180

CB187

CB195

CB206

CB209

Total organic carbon

Acid volatile sulfide

7536 Environ Monit Assess (2012) 184:7531–7550

Program (EMAP) Laboratory Methods Manual (USEPA1995). All analyses and sample collections were per-formed in accordance with a strict Quality Assurance/Quality Control program (Valente and Strobel 1993).

As an initial step prior to quantitative analysis of thebenthic data, a rigorous evaluation of the five collectionswas conducted to ensure as much taxonomic compara-bility among the collections as possible. For example, aspecies identified as “X” in 1993 could have undergonea taxonomic name change and be called “Y” in 2009.Further, an animal identified to species 1 year mightonly be identified to genus another year due to itsphysical condition after sieving (e.g., small worms) orthe proficiency of the taxonomist. It was necessary alsoto select a specific level of taxa across all five collectionsto provide a consistent evaluation. For example, if onlyanimals identified to species were used, it would omitanything identified only to genus or family. This wouldleave out “difficult to identify” animals, such as marineoligochaetes, which may be present in large numbers inimpacted areas like NBH, and still be meaningful in anoverall benthic assessment. Therefore, it was decided toinclude all identified taxa when evaluating benthic con-dition (approximately 90% were identified to the spe-cies/genus level).

Four metrics were used to assess the status of thebenthic community. Taxonomic richness was determinedas the total number of unique taxa per station, cumulativefor both grabs per station. Dominance (i.e., evenness)was quantified by determining the number of taxa thatconstituted 75% of the total number of organisms withineach segment. This was done by (1) calculating the meannumber of individuals for each taxon at each station, (2)summing those values to obtain the total abundance foreach taxa within the segment, (3) dividing by the numberof stations per segment, (4) totaling the number of indi-viduals, (5) calculating what percent each taxon com-prised of that total, and (6) ordering the cumulativepercentages from highest to lowest until 75% wasreached. Two other indices, Shannon's H′ (Washington1984), a widely used diversity index, and the EMAPbenthic index (EMAP-BI) for the Virginian Biogeo-graphic Province (Paul et al. 2001), where NBH is locat-ed, were also calculated for each station. These fourmetrics were selected for use here because they eachevaluate different aspects of the benthic community;collectively, they provide a comprehensive suite to assessthe overall condition of the benthos throughout the entireSuperfund site.

Mussel PCB bioaccumulation

Detailed methods for collecting and deploying bluemussels are found in Nelson and Gleason (1996).Briefly, mussels (5–7 cm in length) were collectedfrom uncontaminated locations in either MA or RIand placed in polyethylene mesh bags. Four indepen-dent replicate bags with 25 mussels each weredeployed 1 m above the bottom at three sites: theCoggeshall St. Bridge (NBH-2), at the boundary be-tween the UH and LH; the Hurricane Barrier (NBH-4),at the boundary between the LH and OH; and at WestIsland (NBH-5), a reference site in Buzzards Bay(Fig. 1). After 28 days, the mussels were retrievedand frozen. Prior to analysis, mussels were thawed,shucked, and homogenized. Two grams of the homog-enate were extracted with acetonitrile and pentane,solvent-exchanged to hexane, and analyzed by gas chro-matography for the same 18 PCB congeners quantifiedin the sediments (Bergen et al. 1993). Since 1993,mussel deployments have occurred twice a year at eachstation, except monthly during the Hot Spot remedia-tion, for a total of 41 deployments.

Statistical analysis

Spatial interpolation Spatial interpolation of the PCBsediment data was conducted by kriging using theGeostatistical Analyst Extension for ESRI's ArcGISmapping and analysis software (Johnston et al. 2001).Krig parameters were optimized for each harbor seg-ment using the 1993 NBH-LTM PCB data. Theseparameters for the UH, LH, and OH segments, respec-tively, include (1) model type (“ordinary” for allthree), (2) lag size (140, 302, and 937), (3) numberof lags (13, 12, and 12), (4) search neighborhood type(“standard” for each), and (5) geometry (“circular” forall three). Once a completed model was validated for aharbor segment, it was applied to the 1995, 1999,2004, and 2009 data for that segment. The resultingdata were saved to ESRI GRID format and thenclipped to the outline of the appropriate segment.The value table (containing concentration and thenumber of cells with each concentration) was exportedto dbase format (.dbf) and used to generate contourmaps (Fig. 3).

Spatial and temporal analysis Statistical differences(p<0.05) among the harbor segments within a

Environ Monit Assess (2012) 184:7531–7550 7537

collection year and temporal changes from the 1993baseline sampling were tested using a repeated meas-ures analysis of variance, with an unstructured vari-ance–covariance matrix for collection years, to test forsegment (i.e., UH, LH, and OH) by year interaction.When the segment-by-year interaction was significant,individual harbor segments were tested for year differ-ences, and individual years were tested for segmentdifferences. Because of the wide range in PCB con-centrations from the UH to the OH, these values weretransformed (log10) prior to analysis; all other variableswere not.

Results

Chemical endpoints

PCBs The spatial PCB distributions for five ranges (0–1, 1–10, 10–50, 50–100, and >100 ppm), and the overallsurface weighed average concentration (SWAC) for eachsegment/year combination, are presented in Fig. 3. Thelargest source of PCBs to NBHwas in the northern-mostportion of the UH, and this is reflected in a distinct spatialgradient in PCB concentration from the UH to the LH tothe OH (Fig. 3). For example, in 1993, the UH SWAC

2009

4.8

61

0.2

2004

0.4

7.2

109

1999

0.4

7.2

139

1995

109

0.8

7.4

1993

5.1

75

0.2

UH

LH

OH

Collection YearH

arb

or

Seg

men

t

Fig. 3 Spatial distribution of the interpolated sediment PCBconcentration data for the UH, LH, and OH segments, foreach of the five long-term monitoring collections. The num-ber to the right of each segment is the surface-weightedaverage PCB concentration for that segment. Spatially, with-in a collection year, each segment is significantly different

from each other for all five collections. Temporally, within asegment, each mean PCB concentration was compared to the1993 baseline, with values that are significantly highershown in red, values not different in black, and significantlylower values shown in green

7538 Environ Monit Assess (2012) 184:7531–7550

PCB sediment concentration was 109 ppm, with signif-icantly decreasing concentrations in the LH (7.4 ppm)and OH (0.8 ppm). This spatial pattern of significantlydecreasing PCB concentrations from the UH to OH wasconsistent for each of the five LTM collections.

All PCB concentration temporal comparisons wererelative to the 1993 baseline collection. In the UH, therewas a significant increase in PCB concentration in 1995,which decreased back to the baseline concentration in1999, followed by a statistically significant (p<0.05)decrease in the 2004 and 2009 collections (Fig. 3). TheLH concentration decreased significantly in 2004 and2009 relative to the baseline concentration, and the OHconcentrations in 1995, 1999, 2004, and 2009 weresignificantly less than the baseline value.

Metals The overall spatial trend for the metals wassimilar to that for PCBs in that there was generally adecreasing gradient from the UH (i.e., highest meanconcentrations) to the OH (lowest mean concentrations;Table 2). For all metals (As and Se not presented inTable 2 because concentrations were low), the concen-trations were lowest in the OH when compared to theUH and LH. With the exception of Cd and Ni in 1995and Cd in 1999, all of the OH metal concentrations weresignificantly lower than the LH concentrations. The UH

concentrations were also significantly higher than theLH concentrations for Cr, Pb, Hg (except 1993), Ni,and Zn. For Cu (except 1995), the concentrations werenot significantly different between the UH and LH.

Temporally, all of the metals with the exception of Cdshowed significant increases in the UH and LH in 1995compared to 1993 (Table 2). In the 1999 collection, thepatterns began to diverge. Cadmium and Pb were backto baseline concentrations in both the UH and LH, whileHg and Ni were still significantly higher in both areas.Copper remained higher in the UH, then returned tobaseline in the LH, while Zn was significantly higherthan baseline in the LH. By 2004, Cd and Cr weresimilar to baseline concentrations in both segments,while Cu, Ni, and Hg were significantly higher thanbaseline in both areas. Lead and Zn were only higherin the LH. In the 1995 OH samples, all of the metalswith the exception of Cd, Cu, and Hg were statisticallyincreased. In 1999, they were all back to 1993 baselineconcentrations. In 2004, Cr, Pb, Ni, and Zn again dem-onstrated significant increases over baseline.

Sediment toxicity

The sediment toxicity data showed a consistent spatialresponse for each of the four sampling years, with

Table 2 Means and standard deviations (SD) for seven of the metals quantified during the NBH-LTM program for each harborsegment–collection year combination

Year Segment Cd Cr Cu Pb Hg Ni Zn

Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD

1993 UH 64.8 138.6 304.2 205.3 611.7 376.3 267.3 158.5 0.4 0.4 33.1 24.6 627.9 393.1

LH 12.4 38.6 189.2 175.8 454.2 460.6 129.1 93.2 0.4 0.3 10.6 6 258 141.3

OH 0.3 0.3 19 15.4 19.5 19.2 18.2 14 0.1 0.1 5.3 3.6 42.1 28.6

1995 UH 9.3 7.1 428.3 271.2 840.5 464.5 346.2 219.6 0.9 0.6 65.2 65 850 585.5

LH 2.2 1.8 234.9 210.1 574.2 544.8 161 105.8 0.6 0.4 20 12.2 320.1 175.5

OH 0.2 0.3 26.1 20.7 26.3 30.3 23.5 18.8 0.1 0.1 7.7 4.5 55.7 37.8

1999 UH 9.4 6 398.9 209.7 760.6 363.7 280.1 129.4 0.8 0.4 38.5 23.7 686.7 358.5

LH 1.9 1.2 211.9 175.6 638.5 927.5 138.2 87.8 0.6 0.3 18.3 21 372.9 303.7

OH 0.2 0.3 20.5 16.9 18.7 20 18.5 16.9 0 0 5.8 3.7 42.5 30.7

2004 UH 6.6 4.7 341.1 207.9 694.4 385.3 286.1 158.9 0.8 0.4 56.3 32.3 622.1 373.3

LH 1.9 1.9 203 158.4 535.6 614.8 152.5 89.1 0.5 0.3 31.3 31.2 366.7 264.8

OH 0.2 0.2 34.9 22.8 21.6 20.9 22 15.9 0.1 0.1 11.1 7 60.4 37.8

Spatial comparisons: means that are not significantly different (p>0.05) among harbor segments within a collection year are shown in italics;those that are different are not italicized (e.g., in 1993, Cu concentrations were not different between the UH and LH). Temporalcomparisons: means that are significantly different (p<0.05) from the 1993 baseline values for each harbor segment are shown in bold(e.g., mean Cu concentrations for 1995, 1999, and 2004 in the UH were significantly higher than the 1993 values)

Environ Monit Assess (2012) 184:7531–7550 7539

greatest mean survival in the OH, followed by the LHand the UH (Table 3). In all comparisons, these differ-ences were statistically significant (p<0.05), except in1993 where survival in the UH and LH were notdifferent.

The temporal comparisons to the 1993 baseline dataindicate that in both the UH and LH, survival was lowerin 1995 and 1999; however, 2004 was not significantlydifferent from 1993. In the OH, survival was signifi-cantly lower for 1995, 1999, and 2004 collections com-pared to the 1993 baseline collection.

Benthic community

The simplest benthic health measure presented here isrichness, or the number of taxa present at each station(Table 4). There was a consistent spatial gradient withsignificantly lower taxa per station in the UH, except for1995 when the UH and LH were the same. The OHexhibited significantly higher numbers of taxa than ei-ther the UH or LH in each of the five collections.Temporally, there was a significant reduction in UHrichness in 1999 and 2004 compared to the 1993 base-line value; however, the mean value was higher again in2009 and not different from 1993. The LH exhibited asimilar pattern with 2004 being significantly reducedcompared to 1993, but also rising in 2009. The OH taxanumbers were significantly lower in both 1999 and 2004,but were again comparable to 1993 levels by 2009.

A second method used to investigate the status ofthe benthic community was the number of dominant

species in each segment of the harbor. As an example,the results for the 1993 baseline sampling are shown inFig. 4, although a similar pattern was observed for theother 4 years also. A clear spatial gradient was ob-served in 1993 with the UH having three dominantspecies, the LH five, and the OH 26. Further, thedominant species in the UH was Streblospio benedicti,an indicator of impacted benthic condition. The dom-inant species in the LH was the opportunistic clam,Mulinia lateralis. The OH displayed a much morediverse benthic community with smaller numbers oflarger animals, indicative of a healthier condition.Temporally, the same spatial pattern for each of thethree harbor segments was observed for the other fourLTM collections.

Shannon's H′ index is a measure of benthic diversity(higher values indicate greater diversity) that incorpo-rates both abundance and evenness (Washington 1984).When applied to the NBH benthic data, a consistentspatial pattern was observed relative to the three seg-ments of the harbor (Table 4). For each of the fiveLTM collections, the UH was always significantlylower than the LH, which was always significantlylower than the OH. Temporally, in the UH, the col-lections in 1995 and 2004 were significantly higherthan 1993. In the LH, none of the collections weredifferent from 1993 except for 2009, which wassignificantly higher. Relative to the OH, the meanvalues in 1995, 1999, and 2004 were significantlylower than in 1993, but increased and were notdifferent again by 2009.

Table 3 Mean amphipod (Ampelisca abdita) survival (percent) for the 10-day sediment toxicity test

Year Segment Mean Survival (%) 55HU66HL1993 19HO

UH 22 LH 48 1995 OH 84 UH 2 LH 32 1999 OH 75

64HU86HL2004

OH 81 Spatial comparisons: within a year, values connected by a line are not significantly different, which only occurred in 1993. Temporalcomparisons: within a segment across years, bolded values are significantly different from 1993

7540 Environ Monit Assess (2012) 184:7531–7550

The final metric applied to the NBH benthic data wasthe EMAP-BI for the Virginian Biogeographical Prov-ince (Paul et al. 2001). This biodiversity index wasdeveloped to assess estuarine benthic condition fromCape Cod, MA to the mouth of Chesapeake Bay, VA.

It incorporates a salinity-normalized Gleason's D biodi-versity metric as well as the abundances of spionid poly-chaetes and salinity-normalized tubificid oligochaetes.Positive EMAP-BI values indicate a good condition,while negative values are indicative of impaired

Mitrella lunataPolycirrusRictaxis punctostriatusNephtyidaeBivalviaCrepidula planaNemerteaTellina agilisBrania wellfleetensisActeocina canaliculataParapionosyllis longicirrataMinuspio cirriferaMulinia lateralisCirratulidaeBoonea seminudaNephtys incisaScolelepis texanaMacoma tentaCrepidula fornicataTharyx acutusAcmira catherinae

TellinidaeNucula proxima

Haminoea solitariaMediomastus ambisetaOligochaeta

Mercenaria mercenariaMediomastus ambisetaOligochaeta

Streblospio benedictiMulinia lateralis

Gemma gemmaMulinia lateralis

Streblospio benedicti

0 100 200 300 400 500 600 700Mean # Animals/Station

UH

LH

OH

Fig. 4 Number of dominanttaxa (i.e., species or taxaconstituting 75% of the totalindividuals) within eachharbor segment for the 1993baseline sampling. The UHstations had greaterabundance of a few small,opportunistic species(three species), the LHslightly more taxa (five) witha few larger species, and theOH exhibited a much greaterdiversity (26 taxa). Thispattern, with minor variationsamong dominant species, wasrepeated for the 1995, 1999,2004, and 2009 collections

Table 4 The means of three benthic metrics (number of taxa, Shannon's H′, and EMAP-BI) for each harbor segment–collection yearcombination

Year Segment Number of Taxa Shannon’s H´ EMAP-BI UH 19 0.5 -4.2 LH 28 0.7 -0.6 1993 OH 62 1.2 2.5 UH 21 0.6 -3.7 LH 26 0.8 -0.4 1995 OH 57 1.1 2.4 UH 16 0.5 -2.8 LH 30 0.8 -0.5 1999 OH 54 1.0 2.2 UH 14 0.6 -2.3 LH 22 0.8 -0.1 2004 OH 42 1.1 2.1 UH 19 0.5 -3.7 LH 30 0.9 0.2 2009 OH 64 1.2 3.3

Spatial comparisons: means that are not significantly different (p>0.05) among harbor segments within a collection year are connected bysolid lines (e.g., mean taxa in UH and LH in 1995). Temporal comparisons: means that are significantly different from the 1993 baselinevalues for each harbor segment are shown in bold (e.g., the mean taxa in the UH in 1999 and 2004 were significantly lower than 1993)

Environ Monit Assess (2012) 184:7531–7550 7541

communities. Consistent with the other measures usedhere, there was a similar spatial pattern for the EMAP-BI;the UH exhibited the worst condition, as evidenced bythe large negative values (i.e., degraded condition) ob-served each collection year; the LH was significantlyimproved relative to the UH, with values near zero; andthe OH was always significantly highest with positivevalues, indicative of a good benthic community (Table 4and Fig. 5). Temporally, the EMAP-BI did not changesignificantly from 1993 in the UH during any of the foursubsequent collections. The LH showed a similar pat-tern until 2009, when the index switched to positive andwas significantly higher than 1993. The same patternoccurred in the OH, with each year similar to the 1993baseline until 2009, when a significantly higher valuewas obtained.

Mussel bioaccumulation

The means of total PCBs (as the sum of 18 congeners) inthe blue mussel tissue for the 41 deployments at the threestations are shown in Fig. 6a. There was a significantspatial gradient among stations, with an approximatefour-fold decrease in overall mean concentration be-tween stations NBH-2 (19 ppm) and NBH-4 (5.1 ppm)and an order of magnitude decrease between stationNBH-4 (5.1 ppm) and NBH-5 (0.5 ppm). These PCBdifferences among stations were maintained over time(Fig. 6b, note the difference in scales among stations).

Each station exhibited seasonal variability due to themussel reproductive cycle, when lipid-rich gametes(along with lipophilic organic contaminants such asPCBs) increase during the year, then decrease dramati-cally during spawning.

Discussion

In the five NBH-LTMcollections to date, a large amountof data were collected and analyzed. While data collec-tion can be relatively straightforward, a greater chal-lenge involves interpreting the results in the context ofthe two programmatic goals to assess remedial effectsand effectiveness. In this paper, remedial effects areevaluated based on spatial and temporal changes in theindividual measurements, relative to remedial activities.Remedial effectiveness, on the other hand, goes beyondmerely documenting changes and involves looking atthose changes in the context of site-specific or generalcriteria (e.g., sediment cleanup levels), and includes amore integrative assessment of all the data collectively(e.g., how do changes in chemical concentrations relateto biological metrics?). In addition, it must be remem-bered that the remediation of PCB-contaminated sedi-ment in NBH was only about 20% complete as of the2009 LTM collection. Therefore, this paper is not a finalevaluation of the overall NBH remediation; rather, it isan interim appraisal of the NBH-LTM program's ability

-5

-4

-3

-2

-1

0

1

2

3

4

UH

LH

OH

20091993 1995 1999 2004

*

*

EM

AP

-Ben

thic

Ind

ex

Collection Year

Fig. 5 Bar graph showing the mean EMAP benthic index valuefor each harbor segment (UH, LH, and OH) and collection year.Positive values indicate a “good” benthic community, while nega-tive values are considered “impaired.” Spatially, within a collection

year, there was a significant difference among harbor segments.Temporally, when compared to the 1993 baseline collection, therewas a significant increase (asterisks) in benthic condition in the LHand OH segments in the 2009 collection

7542 Environ Monit Assess (2012) 184:7531–7550

to document effects to date, as well as the effectivenessof those changes.

Remedial effects Overall, the results of NBH-LTM todate present a fairly consistent narrative relative to reme-dial effects. At the spatial scale of the entire Superfundsite, the PCB data demonstrate a clear decreasing gradi-ent in sediment contaminant concentrations from the UHto the LH to the OH during each of the five collectionevents (Fig. 3), and the biological data (e.g., benthiccommunity) show an increasing gradient along the sametransect (Fig. 5). These data can also be inspected at afiner spatial scale to document effects within a harborsegment relative to remedial activities. For example,there is a significant increase in the 1995 PCB concen-tration in the UH, which can be related to the “Hot Spot”dredging from 1994 to 1995 (Fig. 2), a massremoval operation to dredge sediments with PCB con-centrations >4,000 ppm. Figure 7 shows the PCB con-centrations at each UH station as well as their location.From the graph, it can be seen that therewere increases in

sediment PCB concentrations at some stations during the1995 collection (dashed red lines); however, this oc-curred only in those stations closest to the Hot Spotdredging area (Fig. 7b). These data show that there wasa localized effect, mostly likely due to resuspensionwhile dredging PCB-contaminated sediments with con-centrations ranging from 4,000 to 100,000 ppm. A sim-ilar pattern was observed in 2009, during which severalof the individual UH stations exhibited slight increases inPCBs in 2009 (Fig. 7a), presumably due to resuspendedPCBs during the full-scale dredging activities in the UH.Once again, these increases were localized to the imme-diate vicinity of the dredge area within the UH and hadless of an impact than the Hot Spot dredging (i.e., 2009UHmean PCB concentrationwas still significantly lowerthan the 1993 baseline value, Fig. 3).

The overall strategy used during the Hot Spot andsubsequent dredging was that localized, short-termincreases in sediment PCB concentrations were accept-able, as long as there was no significant PCB transport tothe LH andOH due to the dredging. The rationale for this

0

4

8

12

1988

S

1990

F

1992

S

1993

F

1994

June

1994

Sep

t

1994

Nov

1995

Jan

1995

Mar

1995

Jun

e

1995

Sep

t

1996

S

1997

S

1998

S

1999

S

2000

S

2001

S

2002

S

2003

S

2004

S

2005

S

2006

F

2007

F

0.0

0.4

0.8

1.2

0

10

20

30

40

1988

S

1990

F

1992

S

1993

F

1994

June

1994

Sep

t

1994

Nov

1995

Jan

1995

Mar

1995

June

1995

Sep

t

1996

S

1997

S

1998

S

1999

S

2000

S

2001

S

2002

S

2003

S

2004

S

2005

S

2006

F

2007

F

Mu

ssel

tis

sue

PC

B C

on

cen

trat

ion

(p

pm

)

‘93 ‘95 ‘99 ‘04 ‘09

(b)

Collection Date

0

5

10

15

20

25

NBH-2 NBH-4 NBH-5(N=48) (N=47) (N=44)

19 ± 1.0

5.1 ± 0.3

0.5 ± 0.03

(a)

Mussel Deployment Station

Mu

ssel

tis

sue

PC

B C

on

cen

trat

ion

(p

pm

)

NBH-4

NBH-5

NBH-2

Fig. 6 Blue mussel PCB tissue concentrations measured at threestations (NBH-2, NBH-4, and NBH-5). The overall means (a) forall deployment (1993–2009) are significantly different among thethree areas and reflect the dramatic gradient in water column PCBconcentrations. Within a station over time (b), there is similar

natural variability pattern at all stations, primarily due to themussel reproductive cycle; however, the differences in PCB tissueconcentration are maintained over the entire period (note differ-ence in scales among stations)

Environ Monit Assess (2012) 184:7531–7550 7543

approach was that almost the entire UH, including theHot Spot area, was to be dredged again during futureremedial activities to obtain final PCB cleanup concen-trations. An extensive operational water quality monitor-ing program was implemented to ensure this strategyworked (Bergen et al. 2005), and operational water col-umn monitoring has demonstrated that PCBs were notbeing transported to cleaner areas of the harbor. TheLTM data show that the UH PCB sediment concentra-tions were back to baseline in 1999 (Fig. 3). In addition,the LH and OH PCB concentrations from 1995 showedno increase from 1993, further demonstrating the local-ized nature of the dredging effects and lack of far-fieldPCB transport.

From a temporal perspective, the major observationfrom Fig. 3 is that there is a significant decrease (p<0.05) in average PCB concentration within each seg-ment in 2004 and 2009 compared to 1993. The remedi-ation timeline (Fig. 2) indicates that up to the 1999 LTM

collection, only the “Hot Spot” remediation had oc-curred; however, by the 2004 LTM collection, approx-imately 38,600 yd3 of PCB-contaminated sediment hadbeen removed as a result of the Hot Spot and otherlimited dredging activities. This removal correspondswith the significantly lower PCB concentrations seenin 2004 compared to the 1993 baseline data (Fig. 3);however, the LTM program is not intended to ascertainthe effect of any individual action. From the 2004 col-lection to the 2009 collection, another 159,000 yd3 ofcontaminated sediment were removed from the UH, andthe trend of significantly lower mean PCB concen-trations, compared to the baseline, has continued ineach segment of the harbor. Looking collectively atthe spatial and temporal changes in the LTM PCBsediment data, in the context of the goals to limittransport while removing significant amounts of con-taminated sediment that serve as the primary PCBsource for the entire Superfund site, it is clear that

Stations105 –126

Stations128 –155

Hot SpotDredge Area

1994-95

Full ScaleDredge Area

2004-09

(b)

0

100

200

300

400

500

600

700

1993 1995 1999 2004 2009

Mea

n P

CB

Co

nce

ntr

atio

n (

pp

m)

Collection Year

(a)

108

120

115

111411711

109105

125

117

105

125121

123

Fig. 7 Sediment PCB concentration data from the UH for eachstation and collection year (a). Also shown is a map of theUHwiththe location of each station relative to the dredging activities (b).The red numbered dashed lines show the stations located in the

northern portion of the UH, in close proximity to the dredgingareas; the black lines show those stations in the southern portion ofthe UH, further away from the remedial activities

7544 Environ Monit Assess (2012) 184:7531–7550

the dredging to date can be deemed to be effective,but still a work in progress.

The spatial PCB gradient observed in the sedimentdata also is reflected in the deployed blue mussel PCBtissue data (Fig. 6a) and has been observed for indige-nous ribbed mussel populations in NBH as well (Bergenet al. 2001). Previous studies in NBH have demonstratedthat blue mussels closely follow water column PCBconcentrations (Bergen et al. 1993) and that musselPCB bioaccumulation was more closely linked to stormevents than to any dredging activity (Bergen et al. 2005).Temporally, the entire data set shows that between 1993and 2009, no net change inmussel PCB bioaccumulationhad occurred, indicating that overall water column PCBconcentration had not been altered dramatically enoughto impact mussel PCB bioaccumulation (Fig. 6b). It isreasonable to expect that once remediation is completedthroughout the entire UH (i.e., PCB sediment concen-trations <10 ppm), water column PCB concentrationswill also decrease, leading to decreased mussel PCBtissue concentrations. Continued mussel deploymentswill track these changes to determine if this is thecase.

Remedial effectiveness It is very important to not onlyexamine the NBH-LTM results in terms of spatial andtemporal changes but also in the context of how effectivethese changes are in improving the overall environmen-tal quality in NBH. In order to make this assessment, thedata need to be evaluated in terms of both site-specificcriteria, as well as a more integrated approach such as thesediment quality triad (Chapman 1990), which examineschemistry, sediment toxicity, and benthic communitydata in a comprehensive manner. For the NBH-LTMdata set, a comprehensive table was generated to com-pare the site-specific PCB sediment criteria with othermeasured variables for which “threshold” values (dis-cussed below) were assigned (Table 5). For example, thePCB cleanup level for subtidal sediments established inthe 1998 NBH Record of Decision (http://www.epa.gov/region1/superfund/sites/newbedford/38206.pdf) is10 ppm in the UH and 50 ppm in the LH. For thepurpose of this analysis, a “presumptive” cleanup con-centration of 10 ppm was used for the OH (currently, theOH is undergoing human health and ecological riskassessments to establish a cleanup value). The meanPCB sediment concentration for each segment–collec-tion year combination was compared statistically to itscorresponding criterion value using a single sample t

test. If the mean was significantly worse than the crite-rion, the box in Table 5 was colored red, if not signifi-cantly different from it, yellow, and if significantlybetter, green. The bolded value in the colored boxesindicates the percent of stations within that segment–collection year combination that violated the criterion.Because a systematic grid was used in the samplingdesign, this number corresponds directly to the actualarea within a segment that violated the criterion (within afew percent, some hexagons partially overlap land so it isnot exactly 100%). This same approach (i.e., statisticaltest and box colors) was applied to several other varia-bles that do not have NBH site-specific criteria, but wereassigned “threshold values” based on the literature. Thevalues in the non-colored boxes are the means of addi-tional calculated values for which it is difficult to assign a“good” or “bad” value; however, they are included inTable 5 to help put the NBH data in context of othercommonly used metrics of sediment quality.

For the contaminant data, PCBs and metals are sep-arated to help identify relationships to correspondingbiological effects (Table 5). It is apparent that the PCBcriterion of 10 ppm in the UH was violated for each ofthe five collections, although the percent of area exceed-ing the criteria decreased during 2004 and 2009, con-sistent with the significant decreases in mean PCBconcentrations following remedial dredging activities(Fig. 3). With respect to the LH and OH, each of thesegment–collection year combination boxes is green,and the values are all zero, indicating that no stationsviolated the existing or presumptive criteria (althoughthis may change in the OH when the final PCB cleanupconcentration is set).

There are no NBH site-specific remedial criteria forsediment metals concentrations; however, it is possibleto estimate their effect by quantifying the amount ofbiologically available metals, evaluated as the differencebetween simultaneously extracted metals (SEM) andtotal AVSs in the sediments (Di Toro et al. 1990). A“threshold value” of SEM-AVS<0 is considered “good”because AVS is greater than SEM and, theoretically,metals are not available to cause toxicity. Table 5 showsthat at the spatial scale of a segment, bioavailable metalsshould not be a major toxicity problem in NBH for anysegment–collection combination (green boxes), al-though the bolded numbers in those boxes indicate thatthere were some areas within each segment that hadsediment metals concentrations that were potentiallytoxic. These data also indicate that as PCB remediation

Environ Monit Assess (2012) 184:7531–7550 7545

has continued in the UH, co-located bioavailable metalsconcentrations in the sediment have also been reduced(e.g., the UH area with SEM-AVS>0 has decreasedfrom 37% in 1993 to 11% in 2009). In fact, these datawere used to assess whether to collect metals data for theLTM 2009 collection as a cost-saving measure, which isdiscussed below.

To further put the chemistry data in perspective, themean effects range median quotient (ERM-Q) was cal-culated for both PCBs (sediment PCB concentrationdivided by effects range median, ERM) and metals(sediment As, Cd, Cr, Cu, Pb, Hg, Ni, and Zn concen-trations divided by their respective ERM value, thensummed at each station) and the values shown inTable 5.While the ERM value is not a site-specific criterion, ithas been used to identify potential sediment toxicity(Long et al. 1998). Values <1 are generally associatedwith minimal toxicity, while values above 1.6 are asso-ciated with significant toxicity. The metals ERM-Qvalues in the LH and OH were below 1, indicatingminimal effects, while those values in the UH wereborderline, suggesting that some toxicity might beexpected. In contrast, the PCB ERM-Qs were orders ofmagnitude higher than the metals, well above any

borderline values andwould be expected to be very toxicusing this screening metric.

The threshold value used here for the sediment tox-icity data is subjective because there is no absolutesurvival that is “good.” However, based on the experi-ence of many familiar with this test, a value of 80%survival or above is considered generally acceptable as“good” survival. Therefore, for the purposes of this pa-per, 80% survival or greater was considered good (i.e.,green), significantly lower was “bad” (i.e., red), and notdifferent from, but less than 80% (e.g., 78%) was con-sidered borderline and shaded yellow. Table 5 clearlyshows that both the UH and LH were considered badfor each of the years that sediment toxicity was evaluat-ed. In the OH, only the baseline year of 1993 would beconsidered good, while the other years were borderline.

The results of the amphipod sediment toxicity testwere not always consistent with the chemistry and ben-thic community data for the four collections. For exam-ple, looking at Table 5, the OH PCB and metalsconcentrations were consistently in the “good” category(green), as was the EMAP-BI, yet the sediment toxicityfluctuated between “good” to “borderline” (yellow) dur-ing the same time period. In addition, when looking at

Table 5 Summary table showing sediment contaminants, sediment toxicity, and benthic community indices relative to site-specificcriterion or threshold valuesa discussed in the text

Sediment Contaminants Sediment Toxicity Benthic Condition

Segment Year PCBs PCB

ERM-Q SEM-AVS Metals

ERM-Q EMAP-BI Shannon’s H´ Number of Taxa

UH

1993 78 592 37 1.7 74 96 0.5 19 1995 89 769 19 1.4 93 96 0.6 21 1999 81 567 19 1.2 100 100 0.5 16 2004 74 334 11 1.2 85 100 0.6 14 2009 74 393 93 0.5 19

LH

1993 0 46 19 0.7 52 74 0.7 28 1995 0 41 7 0.7 62 66 0.8 26 1999 0 42 28 0.7 100 62 0.8 30 2004 0 27 10 0.7 69 66 0.8 22 2009 0 29 45 0.9 30

OH

1993 0 4.6 13 0.1 4 0 1.2 62 1995 0 2.4 4 0.1 4 4 1.1 57 1999 0 2.1 9 0.1 48 0 1.0 54 2004 0 1.0 9 0.1 30 0 1.1 42 2009 0 1.3 0 1.2 64

Boxes colored red indicate that the mean value for each harbor segment–collection year combination is significantly (p<0.05) worsethan its criterion or threshold value, yellow indicates no significant difference, and green indicates significantly better. Bolded valueswithin the colored boxes are the percent of stations within each segment–year combination that violate the criterion or threshold value.Values in non-colored boxes (e.g., ERM-Q and taxa) are the segment–year means of other variables used to explain the responsesrelative to the criterion or threshold valuesa Site-specific criterion and threshold values:

PCBs010 ppm for UH and OH, 50 ppm for LH

SEM-AVS>0

Sediment toxicity<80%

EMAP-BI<0

7546 Environ Monit Assess (2012) 184:7531–7550

the percent survival within each segment (Table 3),sediment toxicity was much more variable than thebenthic metrics (Table 4).

Further, there was a lot of variability in mean survivalamong sampling years, with ranges of 53% in the UH(2–55%), 36% in the LH (32–68%), and 16% in the OH(75–91%). Finally, it is difficult to explain the cause ofthis variability in the context of remedial activities, es-pecially considering that the collection year with thelowest absolute survival values for each segment (i.e.,1999) had no remedial activities. The one consistentoverall pattern is that the harbor segments with the great-est variability and lowest survival correlated directlywith the existing PCB and metal contamination gradientfrom the UH to the OH.

A number of metrics were used to quantify the thirdcomponent of the sediment quality triad (i.e., benthiccondition). As with sediment toxicity, selecting a defin-itive “good/bad” criterion number is difficult and subjec-tive. For this paper, we selected the EMAP-BI metric toassign a threshold value to for several reasons: (1) it wasused extensively throughout the Virginian Biogeograph-ical Province (where NBH is located) to assess benthiccondition, and (2) negative EMAP-BI values are indica-tive of an impaired benthic community, and therefore it ispossible to assign a good/bad numerical assessment val-ue. In Table 5, values significantly greater than zero areconsidered good (i.e., green), significantly less than zeroare considered bad (i.e., red), and not different from zeroare borderline (i.e., yellow). The EMAP-BI demonstratesthat the benthic community in the UH was severelyimpacted over the entire time period, while the OH wasconsistently in the “good” category. The most interestingresponse for this metric is in the LH, where the index haschanged from bad in 1993 and 1995 to borderline in1999 and beyond. Further, the spatial extent of the bor-derline condition decreased to 45% in 2009, whichwas also a significant improvement from the 1993 base-line year (Fig. 5). For the purpose of putting thesenumbers into context, two other indices are presentedalso: Shannon's H′ and the number of taxa per station. Itis more difficult to assign a specific criterion or thresh-old value for these indices (e.g., are 28 species good and25 bad?); however, they both exhibited the same generalpattern of improving benthic condition from the UH toOH as did the EMAP-BI.

When examining all three components of the sedi-ment quality triad in Table 5 collectively, it is possible tomake some general conclusions for each harbor

segment. While the toxicity and BI components of thetriad do not define specific causative toxicants (e.g.,PCBs or metals), they do describe the impact fromexposure to the sum of whatever is in the sediments,and in the UH, this impact is significant. Since 2004,when full-scale remediation was initiated, the areaabove the PCB criterion value of 10 ppm in the UHhas been reduced (Table 5), and the mean concentrationin this segment has been significantly reduced (Fig. 3).Further, while metals may be a small contributor tosediment toxicity and impaired benthic condition inthe UH, they are dramatically overshadowed by thePCB effect as evidenced by the magnitude of theERM-Q for PCBs compared to that for metals. Theseresults are consistent with previous toxicity identifica-tion evaluation (TIE) work conducted in NBH by Ho etal. (1997), in which PCBs were also observed to be themajor contaminant contributing to toxicity. Becausemetals tend to be co-located with PCBs in the UH, thearea with SEM-AVS values greater than zero was alsoreduced as the PCBs were remediated. While the reme-diation appears to be having a positive effect relative tothe PCB criterion, the corresponding sediment toxicityand benthic community data demonstrate that this hasnot translated into significantly improved biological re-sponse to date. This is not surprising given that onlyabout 20% of the UH remediation is complete; however,as remediation continues and PCB levels decrease, thecurrent NBH-LTM program will document futurechanges.

The picture in the LH is more interesting in that asPCB sediment concentrations decreased, as occurred in2004 and 2009, the benthic community improved andwas significantly better in 2009 compared to the 1993baseline (Fig. 5). One reason for this is potentially due tothe reduced transport from the primary source of con-taminants in the UH. During the Hot Spot remediationin 1995, Bergen et al. (2005) determined that there is anet PCB transport from the UH to the LH under ambientconditions due to tidal overturn and weather events. Itfollows that as continued remediation reduces the over-all UH PCB sediment concentration, less will be trans-ported to the LH, and this reduction should contribute tobenthic community improvement in this harbor seg-ment. In addition, other factors may contribute to theimproved biological conditions in the LH. For example,navigational dredging has occurred at several locationsin the LH between 2005 and 2009, resulting in furtherPCB removal, and while the amount of PCBs removed

Environ Monit Assess (2012) 184:7531–7550 7547

in the LH was small (<5%) relative to the UH remedi-ation, it also may have contributed to observed benthiccommunity improvements. Further improvements areplanned for the LH in the future (Fort Point Associateset al. 2010), and the NBH-LTM will monitor the subse-quent changes.

The OH is currently undergoing a remedial investi-gation to determine whether additional remediation isrequired. The “presumptive” cleanup concentration of10 ppm used in this paper may change or isolated areasmay need to be remediated based on the ongoing humanhealth and ecological risk assessments. The NBH-LTMdata indicate that as a whole, the OH has very low PCBand metals concentrations, and the benthic communityis healthy.

As is common at most, if not all Superfund sites, thecosts associated with monitoring must be balanced withother remediation needs to maximize cleanup; therefore,flexibility in monitoring programs is a key to accom-plishing this end. This adaptive management approach(NRC 2003; Linkov et al. 2008) was used in the NBH-LTM, specifically for the 2009 collection. Under theadaptive management paradigm, the focus is on reach-ing a specified objective in a specified period of time,and not on establishing a set approach that is unchange-able (Gustavson et al. 2008). When the decision wasmade in the spring of 2009 to conduct another round ofmonitoring, all of the data from the previous years werecarefully evaluated to determine if some of the informa-tion was redundant, or otherwise less informative to theoverall purpose of the NBH-LTM (i.e., assessing effectsand effectiveness of the remediation). It was obviousthat the metals data were not that informative relative toassessing effects and effectiveness. In fact, when a cor-relation analysis was conducted with all the SEM-AVSand sediment toxicity data for all the years, no signifi-cant relation was observed (r0−0.04). In addition,SEM-AVS was not correlated with the two benthicindices (r00.17 and r00.19 for the EMAP-BI andShannon's H′ index, respectively). Based on these facts,it was determined to not include metals in the 2009NBH-LTM collection.

Further, the sediment toxicity was highly variableduring the course of the program. While the overallpattern was similar to the benthic condition data, theresponse was more variable and less correlated to con-taminant levels. For example, there was no correlationbetween sediment toxicity and metals (r0−0.04). Inaddition, sediment toxicity had a lower correlation with

PCBs (r0−0.42) than was exhibited for the EMAP-BIand PCBs (r0−0.86). Because sediments used fortoxicity testing were homogenized, which has beenshown to affect interstitial water PCB geochemistry(Burgess and McKinney 1997), this may have increasedvariability compared to the benthic community evalua-tions, which are more reflective of in situ longer-termresponses to environmental conditions. Based on thefact that the LTM is conducted at approximately 5-yearintervals, is concerned with longer-term changes, andthat the remediation is expected to proceed for an ex-tended period of time (>20 years) at current fundinglevels, it was decided that quantifying the benthic com-munity was more consistent with the goals of the pro-gram. As a result, sediment toxicity testing was alsodropped from the 2009 collection. This does not excludemeasuring metals or sediment toxicity again at somepoint in the future, or indicate that they cannot provideuseful information at another site. For example, it may bedesirable to document changes in sediment toxicity inthe UH once remediation is complete, which wouldrequire reinstating the sediment toxicity test at somepoint. However, based on the evaluation of the previousfour LTM collections, neither metals nor sediment tox-icity were monitored during the 2009 LTM collection.

Conclusions

In summary, there was a significant decreasing sedimentPCB gradient from the UH to the LH to the OH. Thisgradient was maintained temporally from baseline sam-pling in 1993 to the latest sampling in 2009; however,remedial activities in the UH between 1999 and 2009corresponded with significantly reduced PCB concen-trations in all three segments of the harbor in 2009. Thehigher contaminant concentrations along this gradientwere associated with negative benthic communityimpacts, increased sediment toxicity, and greater PCBbioaccumulation in deployed blue mussels. While thesebiological effects were evident over time in the benthiccommunity indices and mussel bioaccumulation, a sig-nificant increase in the EMAP-BI was observed in 2009in the LH and OH, indicating the possible beginning oflonger-term improvements.

As designed, the NBH-LTMprogram has been able toquantify both spatial and temporal effects relative to theremedial activities at the site. The spatial coverage issufficient to detect fine-scale localized effects (i.e.,

7548 Environ Monit Assess (2012) 184:7531–7550

individual stations) due to the dredging activities (e.g.,1995 UH dredging), and also adequate to determinelarger-scale segment-wide changes (e.g., PCB and bio-logical indices). The temporal coverage links chemicaland biological variables at appropriate time scales (i.e.,long-term benthic community changes do not relate withyesterday's dredging). To date, only a little over 20% ofthe scheduled remediation has occurred, and this isreflected in slow changes in the longer-term indicators,like the benthic community and bioaccumulation in bluemussels. However, in areas where PCB concentrationsare significantly lower (e.g., LH and OH), the benthiccommunity is beginning to improve (e.g., EMAP-BI).

Finally, this monitoring program has been flexibleenough to maintain a consistent level of sampling overa long time frame, but also adaptive enough to accom-modate inevitable changes during that same period. Forexample, the duration of activities at Superfund sites,especially expensive ones like NBH, can be very pro-tracted due to litigation, community concerns, and vari-able funding (e.g., at current funding levels, it isestimated that remediation in NBHwill not be completedfor 20+years). Further, the personnel involved in theseprojects typically change over time, including remedialproject managers (e.g., there have been four at this sitesince 1993) as well as the contractors that perform muchof the site characterization and remediation. In addition,new techniques that may be more informative and lessexpensive (e.g., analytical chemistry) need to be incor-porated, while at the same time be comparable to olderdata. We believe the NBH-LTM program is consistentwith these requirements and has been, and will continueto be, a capable approach to assess remedial effects andeffectiveness, and that other contaminated sites can ben-efit from the monitoring approach used in NBH.

Acknowledgements Since the inception of the NBH-LTM,dozens of people from various federal, state, local, and privateentities have contributed to this project. In particular, the authorsthank former NBH remedial project managers Gayle Garman andDavid Dickerson, Joseph Mackay of the US Army Corps ofEngineers, New England Division, Saro Jayaraman of AED formussel PCB analysis, James Heltshe, Computer Sciences Corpo-ration, for statistical analyses, and Chuck Audette, Raytheon, forbenthic community analysis. This manuscript greatly benefitedfrom technical discussions with Anne Kuhn, Glenn Thursby, andDaniel Campbell and the thorough editorial reviews of MohamedAbdelrhman, John Kiddon, and Walter Galloway of AED, as wellas an anonymous outside reviewer. Although the research de-scribed in this article has been partially funded by the US Envi-ronmental Protection Agency, it has not been subjected to Agencyreview. Therefore, it does not necessarily reflect the views of the

Agency. Mention of trade names or commercial products does notconstitute endorsement or recommendation for use. This is contri-bution number AED-11-034 of the Atlantic Ecology Division,National Health and Environmental Effects Research Laboratory,Office of Research and Development, US Environmental Protec-tion Agency.

References

ASTM. (2007). Standard test method D422 for particle-size anal-ysis of soils. West Conshohocken, PA: ASTM International.Available at: www.astm.org. doi:10.1520/D0422-63R07.

Bergen, B. J., Nelson, W. G., Mackay, J., Dickerson, D., &Jayaraman, S. (2005). Environmental monitoring of remedialdredging at the New Bedford Harbor, MA, Superfund site.Environmental Monitoring and Assessment, 111, 257–275.

Bergen, B. J., Nelson, W. G., & Pruell, R. J. (1993). The bioaccu-mulation of PCB congeners by blue mussels (Mytilus edulis)deployed in New Bedford Harbor, MA. Environmental Tox-icology and Chemistry, 12, 1671–1681.

Bergen, B. J., Nelson, W. G., Quinn, J. G., & Jayaraman, S. (2001).Relationships among total lipids, lipid classes and PCB con-centrations in two indigenous populations of ribbed mussels(Geukensia demissa) over an annual cycle. EnvironmentalToxicology and Chemistry, 20(3), 575–581.

Boothman, W.S., & Helmstetter, A. (1992). Determination ofacid-volatile sulfide and simultaneously-extracted metals insediments using sulfide-specific electrode detection. Standardoperating procedures. Narragansett, RI: US EnvironmentalProtection Agency, National Health and EnvironmentalEffects Research Laboratory, Atlantic Ecology Division.

Burgess, R. M., & McKinney, R. A. (1997). Effects of sedimenthomogenization on interstitial water PCB chemistry.Archives of Environmental Contamination and Toxicology,33, 125–129.

Calder, J.A. (1986). Marine environmental quality: NOAA's na-tional status and trends program. In Oceans '86 ConferenceRecord (pp. 1351–1354). Cat. No. 86CH2363-0. Washing-ton, DC: Marine Technical Society IEEE.

Chapman, P. M. (1990). The sediment quality triad approach todetermining pollution-induced degradation. Science of theTotal Environment, 97(98), 815–825.

Di Toro, D. M., Mahony, R. A., Hansen, D. J., Scott, K. J., Hicks,M. B., Mayr, S. M., et al. (1990). Toxicity of cadmium insediments: the role of acid volatile sulfide. EnvironmentalToxicology and Chemistry, 9, 1487–1502.

Farrington, J. W. (1983). Bivalves as sentinels of coastal chemicalpollution: The mussel watch. Oceanus, 26, 18–29.

Fort Point Associates, Inc., Apex Companies, LLC, Urban HarborsInstitute, University of Massachusetts Boston, & FXM Asso-ciates. (2010). New Bedford/Fairhaven Municipal HarborPlan 2010. Urban Harbors Institute Publications. http://scholarworks.umb.edu/uhi_pubs/2.

Goldberg, E. D., & Bertine, K. K. (2000). Beyond the musselwatch—New directions for monitoring marine pollution.Science of the Total Environment, 247, 165–174.

Gustavson, K. E., Burton, G. A., Francingues, N. R., Jr., Reible, D.D., Vorhees, D. J., & Wolfe, J. R. (2008). Evaluating the

Environ Monit Assess (2012) 184:7531–7550 7549

effectiveness of contaminated sediment dredging. Environ-mental Science and Technology, 42, 5042–5047.

Ho, K. T., McKinney, R. A., Kuhn, A., Pelletier, M. C., &Burgess, R. M. (1997). Identification of acute toxicants inNew Bedford Harbor sediments. Environmental Toxicologyand Chemistry, 16(3), 551–558.

Johnston, K., Ver Hoef, J., Krivoruchko, K., & Lucas, N. (2001).Using ArcGIS geostatistical analyst. Redlands, CA: ESRI(48 pp).

Linkov, I., Satterstrom, F. K., Kiker, G., Batchelor, C., Bridges, T.,& Ferguson, E. (2008). From comparative risk assessment tomulti-criteria decision analysis and adaptive management:Recent developments and applications. Environment Interna-tional, 32, 1072–1093.

Long, E. R., Field, L. J., & MacDonald, D. D. (1998). Predictingtoxicity in marine sediments with numerical sediment qualityguidelines. Environmental Toxicology and Chemistry, 17,714–727.

National Research Council (NRC). (1997). Building a foundationfor sound environmental decisions. Washington, DC: TheNational Academies Press. Library of Congress Catalog CardNumber 97-67451.

National Research Council (NRC). (2003). Environmental clean-up at Navy facilities: Adaptive site management. Washington,DC: The National Academies Press. Library of CongressControl Number 2003108132.

National Research Council (NRC). (2007).Dredging at Superfundmegasites: Assessing the effectiveness. Washington, DC: TheNational Academies Press. Library of Congress ControlNumber 2007931836.

Nelson, W. G., Bergen, B. J., Benyi, S. J., Morrison, G., Voyer, R.A., Strobel, C. J., et al. (1996). New Bedford Harbor long-term monitoring and assessment report: Baseline sampling.Narragansett, RI: US Environmental Protection Agency, Na-tional Health and Environmental Effects Research Laboratory,Atlantic Ecology Division. EPA/600/R-96/097.

Nelson,W. G., Bergen, B. J., & Cobb, D. J. (1995). Comparison ofPCB and trace metal bioaccumulation in the blue mussel,Mytilus edulis, and the ribbed mussel,Modiolus demissus, inNew Bedford Harbor, MA. Environmental Toxicology andChemistry, 14, 513–521.

Nelson, W.G., & Gleason, T. (1996). Deployment and re-trieval of caged bivalves for environmental monitoring.Standard operating procedures. Narragansett, RI: USEnvironmental Protection Agency, National Health and En-vironmental Effects Research Laboratory, Atlantic EcologyDivision.

Nelson, W. G., & Hansen, D. J. (1991). Development and use ofsite-specific chemical and biological criteria for assessingNew Bedford Harbor pilot dredging project. EnvironmentalManagement, 15, 105–112.

Otis, M.J., & Averett, D.E. (1988). Pilot study of dredging anddredged material disposal methods: New Bedford, Massachu-setts Superfund site. Superfund '88. Proceedings of the 9thNational Conference, Nov 28–30, Washington, DC.

Paul, J. F., Gentile, J. H., Scott, K. J., Schimmel, S. C., Campbell,D. E., & Latimer, R. W. (1999). EMAP-Virginian Province 4-year assessment (1990-1993). Narragansett, RI: US Environ-mental Protection Agency, National Health and Environmen-tal Effects Research Laboratory, Atlantic Ecology Division.EPA/620/R-99-004.

Paul, J. F., Scott, K. J., Campbell, D. E., Gentile, J. H., Strobel, C.S., Valente, R. M., et al. (2001). Developing and applying abenthic index of estuarine condition for the Virginian bio-geographic province. Ecological Indicators, 1, 83–99.

Pruell, R. J., Norwood, C. B., Bowen, R. D., Boothman, W. S.,Rogerson, P. F., Hackett, M., & Butterworth, B. C. (1990).Geochemical study of sediment contamination in New Bed-ford Harbor, Massachusetts.Marine Environmental Research,29(2), 77–101.

Reifsteck, D. M., Strobel, C. J., & Keith, D. J. (1993). EMAP-Estuaries 1993 Virginian province field operations and safetymanual. Narragansett, RI: US Environmental ProtectionAgency, Office of Research and Development, Environmen-tal Research Laboratory (172 pp).

Tanabe, S., Tatsukawa, R., & Phillips, D. J. (1987). Mussels asbioindicators of PCB pollution: A case study on uptake andrelease of PCB isomers and congeners in green-lipped mus-sels (Perna viridis) in Hong Kong waters. EnvironmentalPollution, 47, 41–62.

United States Army Corps of Engineers. (1999). Quality assur-ance project plan for long-term monitoring III at New Bed-ford Harbor Superfund site, New Bedford, Massachusetts.Prepared by ENSR Corporation for the US Army Corps ofEngineers, New England Division. Document number 9000-218-103.

United States Environmental Protection Agency. (1994). Methodsfor assessing the toxicity of sediment-associated contaminantswith estuarine and marine amphipods. Narragansett, RI: USEnvironmental Protection Agency, Office of Research andDevelopment, Environmental Research Laboratory. EPA/600/R-94/025.

United States Environmental Protection Agency. (1995). EMAP—Estuaries: Laboratory methods manual—Estuaries. Vol. 1.Biological and physical analyses. Narragansett, RI: US En-vironmental Protection Agency, National Environmental andEcological Effects Laboratory, Atlantic Ecology Division.EPA/620/R-95/008.

United States Environmental Protection Agency. (2004a).Updatedreport on the incidence and severity of sediment contamina-tion in surface waters of the United States, National SedimentQuality Survey. Washington, DC: US EPA, Office of Water.EPA-823-R-04-007.

United States Environmental Protection Agency. (2004b). Method9060—Total organic carbon. Nov 2004, rev01, 5 p. www.epa.gov/osw/hazard/testmethods/sw846/pdfs/9060a.pdf.

United States Environmental Protection Agency. (2005). Contam-inated sediment remediation guidance for hazardous wastesites. Washington, DC: US Environmental Protection Agency,Office of Solid Waste and Emergency Response. EPA-540-R-05-012.

United States Environmental Protection Agency. (2007). Method8082A—Polychlorinated biphenyls (PCBs) by gas chroma-tography, rev.01, 56 p. www.epa.gov/epaoswer/hazwaste/test/pdfs/8082a.pdf.

Valente, R. M., & Strobel, C. J. (1993). EMAP—Estuaries Virgin-ian province 1993 quality assurance project plan. Narragan-sett, RI: US Environmental Protection Agency, Office ofResearch and Development, Environmental ResearchLaboratory.

Washington, H. G. (1984). Diversity, biotic and similarity indices.Water Research, 18, 653–694.

7550 Environ Monit Assess (2012) 184:7531–7550