, 19:141

Bioaccumulation of polychlorinated biphenyls in juvenile chinook salmon (Oncorhynchus tshawytscha) outmigrating through a contaminated urban estuary: dynamics and application


    • Ecotoxicology and Environmental Fish Health Program, Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries ServiceNOAA
  • Gina M. Ylitalo
    • Environmental Assessment Program, Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries ServiceNOAA
  • Frank C. Sommers
    • Ecotoxicology and Environmental Fish Health Program, Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries ServiceNOAA
  • Daryle T. Boyd
    • Environmental Assessment Program, Environmental Conservation Division, Northwest Fisheries Science Center, National Marine Fisheries ServiceNOAA

DOI: 10.1007/s10646-009-0399-x

Cite this article as:
Meador, J.P., Ylitalo, G.M., Sommers, F.C. et al. Ecotoxicology (2010) 19: 141. doi:10.1007/s10646-009-0399-x


A field study was conducted to examine bioaccumulation of polychlorinated biphenyls (PCBs) for hatchery-raised and naturally reared (wild) ocean-type juvenile chinook salmon outmigrating through the Lower Duwamish Waterway (LDW), a contaminated urban estuary in Seattle, WA, USA. These results show differences in bioaccumulation of PCBs over time and space in this estuary, which may also occur for any contaminant that is distributed heterogeneously in this system. Highly mobile, outmigrating salmon accumulated ~3–5 times more PCBs on the east side of the LDW than fish on the west side, which is supported by an almost identical difference in mean sediment concentrations. The tPCB concentration data suggest that for most of the spring and early summer, juvenile chinook were likely segregated between the east and west side of the LDW, but may have crossed the channel later in the year as larger fish. Additionally, we used biota-sediment accumulation factors to assess the relative degree of bioaccumulation and explore these factors as potential metrics for predicting adverse sediment concentrations. These results highlight the importance of time and space in sampling design for a highly mobile species in a heterogeneous estuary.


PCBsBioaccumulationSalmonSpatial segregationToxicity guideline value


Even though polychlorinated biphenyls (PCBs) were banned in the United States in 1979, they persist at high concentrations in sediments and aquatic foodwebs. The influx of cleaner sediments over time was expected to accumulate and bury these contaminants below the biologically active zone; however, these compounds still occur at very high concentrations in surface sediment and are biologically available to biota.

The Green River flows northwest from the western flanks of the Cascade Mountains near Mt. Rainier and travels ~150 km to Elliott Bay near downtown Seattle, WA, USA. For the last 19 km the Green River is called the Duwamish River and for the final 9 km it is known as the Lower Duwamish Waterway (LDW; Fig. 1). At river kilometer (rkm) 0 the river splits into the East and West Waterways around Harbor Island for 2 km before entering Elliott Bay. The LDW is a marine-influenced urban estuary that has been the focus of intense studies due to its highly contaminated sediment and water. The average width of the LDW is ~130 m and the water depth ranges from 3 to 20 m; however, most of LDW is maintained at 10 m depth (mean lower low water) by dredging. Even though most of the natural habitat has been severely altered, off-channel areas (e.g., Slip 4 and Kellogg Island) and a narrow shallow-slope intertidal habitat can be found along the waterway where outmigrating salmon likely forage and can be collected.
Fig. 1

Map of the Lower Duwamish Waterway

Past work has documented that sediment and organisms in the LDW are contaminated with PCBs, PAHs, tributyltin, and other contaminants of concern (Varanasi et al. 1993; LDWG 2007). The entire LDW was listed as a Superfund site under the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) in 2001 and is currently progressing through the standard superfund remedial process. PCBs have been an important concern in the LDW for several years after they were discovered at high concentrations in sediment at several sites. We focused on PCBs because of elevated concentrations in the LDW, high potential for toxicity to juvenile salmon, low elimination rates in fish, and relative ease of assessing sediment and tissue concentrations.

Several salmonids including chinook (O. tshawytscha), coho (O. kisutch), chum (O. keta), and winter steelhead (O. mykiss), are raised in several hatcheries in this watershed and released every year. For most years, ~5–6 million fish have been released annually into the Green River and most of these (≈70%) are age 0+ (subyearling; age 0–1 year) ocean-type chinook (Sieler et al. 2002), which are protected in this watershed under the Endangered Species Act and were the target of our study. Juvenile chinook are released from three hatcheries on this system; however, 80% or more come from the Soos Creek hatchery. Additionally, ~1 million ocean-type chinook naturally rear (wild) in this system and also migrate through the LDW (Sieler et al. 2002). Since 2000, essentially all hatchery chinook released in this watershed have been marked by clipping their adipose fin. Because the error rate (bad clips) is generally low at ~4% (Ruggerone et al. 2006), this procedure has allowed us to distinguish hatchery from naturally reared fish with fairly high confidence. Juvenile salmonids migrate from relatively uncontaminated upstream waters into the Duwamish River and LDW during smoltification where they adjust to seawater, feed on relatively abundant invertebrates, and rear from a few days to several weeks before exiting to open water. The peak migration for age 0+ hatchery fish occurs from late May to mid June and wild fish are found in the Duwamish from mid January through late summer (Ruggerone et al. 2006).

The goal for this study was to examine PCB bioaccumulation in highly mobile, outmigrating juvenile salmon in this estuary, determine total amount accumulated, and examine the application of bioaccumulation factors to predict sediment concentrations that may result in adverse tissue concentrations. Our hypothesis was that juvenile chinook fish would migrate along the west or east bank of the river and reflect the contamination of each region. If fish freely crossed the waterway, the concentrations of PCBs and other contaminants in fish collected at Kellogg Island should be similar to the levels in fish collected at Slip 4. Small outmigrating salmonids tend to stay in shallow areas as they feed and migrate through an estuary (Healey 1991). On average, the west side of the LDW contains substantially lower concentrations of PCBs in sediment than those collected on the east side, which we hypothesized would be reflected in the amount bioaccumulated by the fish collected. Although not in our original design, we were also able to consider some temporal aspects of PCB bioaccumulation for juvenile salmonids because our sample dates spanned 11 weeks over late spring and mid summer.


The area of focus for this study is the lower Duwamish River occurring from the turning basin (rkm 7.6) to the confluence of the east and west waterways at the southern tip of Harbor Island (rkm 0; Fig. 1) and constitutes most of the marine influenced section of the Duwamish River. The surface area of intertidal and subtidal sediment in this section of river is ~142 ha (350 acres).

Fish sampling

Juvenile chinook were sampled from four locations in this river system. For the upstream sites in the Green River, fish were collected from the Soos Creek Hatchery on Big Soos Creek (a few km upstream of the confluence of the Green River and Big Soos Creek at rkm 54.4) for 3 years (2000–2002) usually before they were released in late May, except in 2002 when fish were sampled from the hatchery on 8 August. Naturally reared fish (wild) were also collected one year (2000) from a screw trap at rkm 55.6, which is upstream from the Soos Creek hatchery and confluence of the Green River and Big Soos Creek. These fish were acquired live from personnel of the Washington Department of Fisheries and Wildlife (WDFW).

On the west side of the LDW, we collected fish at Kellogg Island, which is a semi-natural area off the main channel at rkm 1.3. On the east side we sampled fish at Slip 4 (rkm 4.3), which is a 1.5 ha (3.6 acre) blind inlet off the main channel. Historically, we have observed large numbers of migrating salmon and other fish species at these two locations. We sampled at both LDW sites over 4 years (2000–2002 and 2004). For the year 2000, we sampled fish in late May; ~5 days after the last group of hatchery fish had been released from the Soos Creek hatchery. For subsequent years, we collected fish at these sites from late June to early August. We also analyzed two composite samples of juvenile coho collected at Slip 4 in 2002 to determine if the values for whole body and stomach concentrations were similar to those found for chinook.

A 100-m beach seine was used in the LDW for sample collections and all fish were kept alive in coolers until processing at our laboratory. Samples were frozen at −80°C until analyzed. Stomach contents were removed from all fish; therefore the whole-body concentrations represent only the PCBs that were assimilated. Whole fish were analyzed as individuals or composite samples, each containing from 3 to 10 individuals. Samples for stomach contents were almost always composites of material from several individuals.

Analytical determinations for OCs and lipid in tissue

Whole-body fish and stomach content samples were analyzed for organochlorines (OCs), including dioxin-like PCBs, other selected PCB congeners, by a high-performance liquid chromatography/ultraviolet photodiode array (HPLC/PDA) method (Krahn et al. 1994). Sample extractions were split for PCB and lipid analyses. Prior to sample cleanup, a 1 ml portion of each whole-body extract was removed for percent lipid analyses by thin-layer chromatography/flame ionization detection (TLC/FID) (Ylitalo et al. 2005b). Lipid classes were measured by FID, but are not reported here. Percent lipid values were calculated by summing the concentrations of all lipid classes determined for each sample.

A separate study compared the tissue concentrations from sample splits for our HPLC/PDA method (NOAA lab) and those obtained with high resolution gas chromatography/mass spectrometry (GC/MS; Axys Analytical Services LTD, Sidney, British Columbia, Canada). The results for 30 samples (four species, whole body and muscle, range of 5–300 ng/g) indicated close agreement between methods, although 80% of the GC/MS values were higher than those for the HPLC/PDA method (Sandie O’Neill and James West, WDFW, personal communication). The overall mean (SD) percentage difference among all samples was 24 (0.22)%, which is very low. These results are supported by other studies that have shown close agreement for summed PCB concentrations obtained by the HPLC/PDA and GC/MS methods for a wide range of marine biota (Krahn et al. 1994; Ylitalo et al. 2005a).

Quality assurance for HPLC/PDA method

A method blank and a National Institute of Standards and Technology (NIST) blue mussel Standard Reference Material (SRM 1974a or 1974b) sample were analyzed with each sample set containing 8–12 field samples as part of a performance-based quality assurance program (Sloan et al. 2006). Results obtained for SRMs were in excellent agreement with the certified and reference values published for these materials by the National Institute of Standards and Technology. In addition, the other quality control samples met established laboratory criteria. Duplicate analyses were conducted for 10% of the tissue samples, with relative standard deviations ≤30% for more than 80% of analytes detected in the samples. Method blanks contained no more than four analytes that exceeded four times the limit of quantitation (LOQ), unless the analyte was not detected in the associated tissue samples in the set. The percent recovery of the surrogate standard ranged from 70 to 105%.

Sediment concentrations

A separate study of 326 sediment samples for PCBs in the Duwamish estuary (Industrial Economics 1998) was used to analyze bioaccumulation in fish (Table 1). This study conducted a comprehensive analysis of PCBs in sediment over the entire Lower Duwamish Waterway (142 ha sampled) from the turning basin to rkm 0 that included our fish collection sites. Total organic carbon and PCBs were determined for each sample, which allowed determination of the organic-carbon normalized sediment concentrations (sedoc). The same method (HPLC/PDA) for PCB analysis described above for tissue was also used to quantify PCBs for these sediment samples. Of the sediment sites that were examined in detail, tPCBs from the LDW were mostly consistent with the Aroclor 1254 pattern or a mix of Aroclors 1254 and 1260 (>90% of samples).
Table 1

Concentrations of total polychlorinated biphenyls (tPCBs) in sediment

Regions and locations

Mean sediment (ng/g sed)

Mean sedoc (μg/g OC)


Sediment (ng/g sed)

Mean (SD)







West side

150 (20) 113

10.6 (1.5)








East side (to Slip 4)

500 (150) 95

33.5 (9.8)








Kellogg Island

190 (60) 35

8.9 (1.8)








Slip 4

1,200 (320) 42

88.8 (24.5)








East side—Slip 4 to opposite Kellogg Island

180 (40) 59

10.7 (1.6)








Values are mean and standard deviation (SD) for total PCBs in sediment and sedoc (organic carbon (OC) normalized values; μg total PCBs/g OC in sediment). Several percentile values are also shown for each region and location. All values determined with minimum unbiased estimator for a lognormal distribution. Following SD denotes the number of samples per mean value. Data from Industrial Economics (1998)

The waterway was divided into five cross-river sections (intertidal and subtidal for the east and west sides and the navigational channel). The demarcation between the subtidal areas and the channel was determined from navigation charts (Industrial Economics 1997). Within these major sections, numerous substrata were defined. A total of 90 substrata (nonoverlapping polygons of the sediment surface) were determined for the LDW. Some of the substrata represent discrete areas (e.g., slips, backwaters, non-continuous intertidal areas, outfalls, and seeps). The overall intent for this sampling scheme was the primary efficiency criterion of stratification designs that concentrations within strata are more homogeneous than concentrations over the entire study area (Industrial Economics 1997).

Sediment sample sites within substrata were determined randomly and spaced less than 100 meters apart. Of the 54 substrata selected for our analysis, the mean (SD) size was 1.42 (1.45) ha. The mean (SD) number of samples for all substrata from that study was 2.2 (1.7) per hectare and no one area was overly represented. Substrata in the navigation channel were not included because we assumed that juvenile chinook would not occur in that area of the LDW or interact with this benthic environment that is frequently disturbed by river flow, tidal flux, and vessels.

To determine the mean sedoc for the west side, all intertidal and subtidal samples from just north of the Turning Basin (rkm 7.6) to the southern tip of Harbor Island (rkm 0) were included. This value was used for the BSAF calculation for salmonids collected at Kellogg Island. Similarly, we choose all intertidal and subtidal sediment samples from just north of the Turning Basin to ~1,000 m north of Slip 4 on the east side for the BSAF equation for chinook collected at Slip 4. One sediment sample in Slip 4 was excluded because it was considered an outlier (Grubbs test, P < 0.0001). The tPCBs for this one sample was 25 μg/g, which was 50 times the mean value for all east side samples (n = 96) and was therefore not representative of values from this region. This hot spot represented a very small area and its inclusion would likely have skewed the BSAF values and conclusions. We also determined the sediment concentrations at the collection sites. For Kellogg Island, we included all inter- and sub-tidal sediment data from sampling sites around Kellogg Island and all sites ~1,000 m north and south of the island to calculate the mean sedoc. The sediment concentrations for Slip 4 were determined in a similar fashion including all sites in Slip 4 and those inter- and subtidal sites 1,000 m to the north and south of this area.

Most of the PCB sediment contamination occurs on the east side of the LDW in inter- and subtidal areas from the Turning Basin to Slip 4 and is substantially more contaminated than the west side (Industrial Economics 1998). We determined that 56% of the sample sites on the east side contained PCB sediment concentrations >100 ng/g dry wt, which was higher than that for the west side (25%). Because we did not sample fish downstream of Slip 4 on the east side of the river those sediment concentrations were not included. The mean concentration for all sub- and intertidal sediment samples between Slip 4 and Harbor Island (rkm 0) on the east side was determined to be much lower than the upriver portion of the east side and very similar to the mean determined for the entire west side of the LDW (Table 1). This area contained one sample that was 23 times higher than the mean value and 10 times higher than any other concentration. It was determined to be an outlier based on Grubbs test (P < 0.0001) and was excluded for the same reasons stated above for the one Slip 4 value. If included, the mean tPCB sediment concentration would be 220 ng/g dry wt. a 25% increase, which was considered an undue influence for one of 60 samples.

Determination of PCB accumulation in the lower Duwamish

We used a mass balance approach to determine the total ng of PCBs accumulated per fish (body burden, bb) collected in the lower Duwamish.
$$ {\text{PCB}}_{\text{bb}} = {\text{tPCB}}_{\text{ld}} \times {\text{WT}}_{\text{ld}} - {\text{tPCB}}_{\text{u}} \times {\text{WT}}_{\text{u}} $$
where PCBbb represents the total ng of PCBs accumulated, tPCBx denotes the concentration of total PCBs (wet weight), and WTx is the wet weight for each fish or composite mean sampled. Subscripts for x are as follows: ld denotes fish collected in the Lower Duwamish and u denotes upriver fish (hatchery or wild). For all hatchery fish collected in the LDW we used the hatchery-collected fish for the upriver concentration in Eq. 1 (tPCBu) and for all wild fish collected in the LDW we used the mean concentration of tPCBs measured in wild fish collected from the screw trap in 2000 (tPCBu).
Biota-sediment bioaccumulation factors (BSAFs) were calculated to highlight differences and similarities among species and sites. The following equation was used:
$$ {\text{BSAF}} = {\frac{{ [ {\text{tissue]/f}}_{\text{lip}} }}{{ [ {\text{sediment]/f}}_{\text{oc}} }}} $$
where foc is the fraction of organic carbon (g/g dry wt.) and flip is the fraction of lipid (g/g wet wt). For the collection year 2000, specific site and type (wild or hatchery) lipid concentrations were used. For all other years a mean lipid value of 1.0% was determined from all remaining data and used for the BSAF calculations for chinook.

We assumed that fish had an equal chance of visiting (temporally and spatially) each of the sediment sites that were used for these calculations. We also assumed that each tPCB sediment concentration was proportional to the tPCB concentration for water and prey in the immediate area around the sample and that accumulation was proportional to the OC normalized sediment concentration (sedoc). We calculated BSAFs using mean tissue and sediment concentrations, which we believe provided a better estimate of bioaccumulation than median values.

These BSAF values were used to determine a sediment concentration that would be expected to protect outmigrating juvenile salmon from adverse biological effects. This sediment quality guideline was calculated with Eq. 1 by solving for sedoc. For these calculations we used a mean whole-body lipid content of 1% wet weight (Table 2) and the 50th percentile for organic carbon (OC), which was 1.6% dry wt for each side of the waterway. We selected the PCB tissue toxicity guideline of 2.4 μg/g lipid for salmonids from Meador et al. (2002) for conversion to sediment values.
Table 2

Data for salmon collected in the Duwamish River and upstream



Wt (g)

Len (mm)

Lipid (%)


BSAF median

N {N tot}

Kellogg Island




Chinook W

4.4 (1.1)

76.5 (6.8)

1.6 (0.3) 4c

0.18 (0.01)


17 {31}



Chinook H

4.8 (0.2)

79.7 (0.3)

1.8 (0.1) 3c

0.21 (0.02)


3 {30}




Chinook W

5.4 (3.0)

84.5 (18.9)

0.82 (0.53)


4 {4}




Chinook W

12.1 (4.3)

106 (8.9)

0.35 (0.07)


35 {39}




Chinook H




1 {1}




Chinook H

12.3 (2.1)

111 (4.2)

0.89 (0.44)


6 {6}




Chinook W

10.7 (5.2)

100 (12.3)

1.1 (0.3) 7i

2.9 (1.3)


7 {7}



Chinook H



1.2 1i


1 {1}




Chinook H

9.8 (1.0)

102 (2)

0.9 (0.7) 3c

1.2 (0)


3 {9}



Chinook W





1 {3}

Slip 4




Chinook H

4.6 (1.0)

80.1 (5.6)

2.0 (0.1) 2c

0.30 (0.12)


7 {15}



Chinook W

3.4 (0.1)

69.5 (0.7)

0.25 (0.3)

2 {2}




Chinook W

3.5 (0.9)

72.3 (5.6)

1.1 (0.18)


12 {12}




Chinook W

12.7 (4.3)

107 (11.0)

0.90 (0.6)


5 {5}




Chinook H

5.0 (0.08)

82.7 (1.5)

0.55 (0.16)


3 {3}




Chinook H

12.7 (3.3)

109 (7.3)

0.53 (0.1)


4 {4}




Chinook W



0.9 (0.3) 2i





Chinook H



1.1 1i





Coho W

5.4 (0.7)

78.8 (4.5)

1.8 (0.1) 2c

0.8 (0.1)

2 {7}

Soos Creek




3.9 (0.8)

73.3 (5.5)

1.9 (0.4) 2c

14 {26}




2.2 (0.6) 3i




2.5 (0.07)

7 {7}




9.4 (0)

1.6 (1.2) 2i

2 {2}

Values shown as mean and standard deviation and determined with algorithms for lognormal distributions (Gilbert 1987) for all n ≥ 3. Type (W wild; H hatchery; M mix of both types). N is the number of samples for each mean and n total is the total number of fish measured for length, weight, PCBs and BSAFs. Sample sizes for lipids shown next to value. “i” indicates individuals and “c” indicates composite values (ci indicates a combination of composite and individual values). Composite samples contained 3–10 individuals

Toxicity equivalents

We calculated the sum of toxic equivalents (ΣTEQs) for dioxin-like (dl) PCBs for each sample. Each TEQ was determined by multiplying a dl PCB concentration with its toxicity equivalent factor (TEF) for fish, which was obtained from van den Berg et al. (1998). Our analytical method quantified the dl-PCB congeners 77, 105, 118, 126, 156, 157, 169, and 189. The other four dl congeners (81, 114, 123, and 167) were not quantified due to problems with coelution by interfering compounds. The TEQ levels calculated in the current study are conservative values because of the higher limits of detection of the HPLC/PDA system compared to the GC/MS method and they do not include the contributions from polychlorinated dibenzodioxins (PCDDs) or dibenzofurans (PCDFs). In addition, when the concentration of a dioxin-like PCB was below the LOQ, a value of zero for the specific congener was used in the calculation, which was more conservative than the commonly used value of one-half the LOQ. These below-detection values were not used because our LOQ was relatively high (0.03–0.4 ng/g wet weight for most samples), which was due to low sample weights (<4 g).

Statistical analysis

Most of the concentration data reported here were lognormally distributed, which is very common for such data (Gilbert 1987). Because lognormally distributed data are skewed, a minimum variance unbiased (MVU) estimator is more appropriate for computing statistics, such as the mean, variance, and quantiles. We used the MVU estimator algorithms in Gilbert (1987) for estimating the mean, variance, and quantiles (Eqs. 13.1, 13.2, and 13.24) for all log-normally distributed data (TEQs, BSAFs, and whole-body, stomach, and sediment concentrations). This MVU algorithm was not used when sample sizes were <3. We used SYSTAT 11 to construct cumulative distribution functions (CDFs), perform regression analysis, and to examine distributions. Statview 5.0 was used to perform Analysis of Variance (ANOVA) and post-hoc testing. After performing the ANOVA, a post-hoc examination of treatment means was conducted with Fisher’s Protected Least Significant Difference (PLSD) test. Log values for concentrations were used for ANOVAs and regressions. We also used Grubbs Test to examine datasets for statistical outliers. Standard deviation is shown to provide a measure of the range in data and standard error of the mean (SEM) was used to indicate variation about the mean.


PCBs in salmon

Juvenile chinook from upstream areas (hatchery and screw trap) contained very low levels of tPCBs, except for hatchery fish in 2001 (Table 3). Mean tPCBs concentrations in fish collected from Slip 4 were always higher than those collected at Kellogg Island. Although variability was observed among individuals, it was likely due to a range in time spent in the LDW (Fig. 2). The differences between wild and hatchery fish collected in the LDW were mixed. There were no significant differences between hatchery and wild fish collected at Slip 4 for all years combined. Concentrations of tPCBs in the hatchery origin fish collected from Kellogg Island were significantly higher than wild fish (P = 0.04) when all years were considered, which was mostly due to a pulse of upriver wild fish with low tPCBs in August 2001.
Table 3

Total PCB concentrations in juvenile salmon collected in the Duwamish River and upstream


Soos Creek hatchery

Soos Creek wild

Kellogg Island hatch

Kellogg Island wild

Slip 4 hatch

Slip 4 wild

Whole body

   24–31 May 2000

15 (1.1) 5i

7.8 (0.8) 14ci

40 (4) 3c

30 (1.3) 17 ci

203 (80) 7ci

131 (159) 2i

   25 June 2001

50 (2.4) 7i

24 1i

94 (56) 4i

185 (59) 3i

376 (60) 12i

   1 August 2001

94 (47) 6i

37 (7) 35 ci

177 (34) 4i

302 (195) 5i

   7–8 August 2002

10 (0.1) 2i

445 1i

302 (151) 7i

725 (375) 2i M

495 (78) 2c ¥

   29 July 2004

130 (0) 3c

180 1c

Stomach contents

Soos hatchery

Soos Creek wild

Kellogg Island mix


Slip 4 mix



23 1c

57 (21) 3c


247 (30) 3c



182 (138) 2c


445 (360) 2c



12 Ø

260 (-) 1c


760 (14) 2c ¥


Values are mean and standard deviation (SD) ng/g. Following SD denotes n observations per mean value; “i” means individuals and “c” means composite values (ci indicates a combination of composite and individual values). Whole-body composite samples contained 3–10 individuals. M is mix for origin and mostly hatchery fish. Stomach contents were removed from these fish and used for separate analysis as composite samples containing 5–30 individuals. Date shows when in-river fish collected. Soos Creek fish (wild and hatchery) collected 18 May to 1 June, except for 2002 (8 August). Chinook in all samples except for ¥, which was two composite samples (n = 3 and 4 individuals) of juvenile coho and one comp for stomach contents (770 ng/g). Ø hatchery food. All values as wet weight, except fish food as dry wt (wet wt. equivalent for fish food ≈ 2.7 ng/g)
Fig. 2

Cumulative frequency distribution for total PCBs in juvenile chinook. Log10 concentrations are plotted. Upper x-axis show arithmetic equivalents. Location and origin (hatchery or wild) shown. a Data for the year 2000. b Data for the year 2001

The tPCB values for the composite samples containing coho salmon were not different than those containing chinook from Slip 4 in 2002. The coho whole-body concentrations were 550 and 440 ng/g, which were lower than the mean value for the two individual chinook (725 ng/g). The stomach contents concentrations for the coho and chinook composite samples (one each) for 2002 from Slip 4 were essentially identical (750 and 770 ng/g), which is reflected in the mean value and low SD.

The temporal aspect of PCB bioaccumulation is also noteworthy. The fish collected in 2000 were sampled in late May, which was ~5 days after the last release of fish from the Soos Creek hatchery. Total PCB concentrations in both wild and hatchery fish for the year 2000 were relatively low compared to the other sampling periods, which occurred later in the summer (Fig. 2). The Kellogg Island fish contained substantially lower concentrations of tPCB than Slip 4 fish for the years 2000 (P < 0.005) and 2001 (P < 0.0001; Table 3; Fig. 2). For 2002, the differences were far less substantial (P = 0.12), which may have been due to larger fish that were able to cross the waterway. The highest tPCB concentrations for Kellogg Island fish occurred in the largest fish collected, which may be the result of an increased ability to cross the waterway from the east side. Excluding all fish with tPCB concentrations <15 ng/g (these were considered background levels), the correlation between fish weight and tPCBs for Kellogg Island fish (all years) was highly significant (P < 0.001) with an r2 = 0.50 (n = 59). There was no such correlation when all fish from Slip 4 were considered (P = 0.42, r2 = 0.02, n = 36). Additionally, any whole-body tPCB value over 400 ng/g in fish from Kellogg Island was determined to be a statistical outlier (P < 0.05) in Grubbs test, which supports the contention that larger fish (>15 g) collected at Kellogg Island did not accumulate most of their PCBs from the west side of the LDW.

Concentrations of tPCBs in stomach contents of juvenile chinook collected at Kellogg Island and Slip 4 were substantially elevated compared to stomach contents in upriver wild fish and hatchery food (Table 3). These values also show site and year differences that are consistent with those for whole-body tPCBs. An analysis of the ratio for tPCBs in whole-body juvenile chinook and stomach contents (wet weights) for site/year combinations were relatively consistent with a mean (SD) of 0.77 (0.40) n = 12.

For the 2001 hatchery fish, we had sufficient data to estimate a likely growth rate. Five fish were sampled from the hatchery (mean (SD) 2.5 (0.1) g) on 7 June 2001 and compared to hatchery fish collected 54 days later at Kellogg Island and Slip 4 in the LDW. The mean weight (SD) for those fish was 13.7 (4.6) g n = 10. Based on a simple growth equation the mean growth rate was determined to be 3.2% bw/day (range = 2.6–4.4% bw/day). Fish were released from the hatchery between 18 May and 11 June 2001, therefore these values represent the maximum growth rate. If we assumed that all of the fish collected were from the earliest date (18 May) the mean growth rate would be 2.4%; however, these fish would have been smaller at the time of release.

For each individual fish and composite sample we determined the amount of tPCB that was accumulated in the LDW, which is presented as a percentage increase in total body burden (Fig. 3). This plot shows the general trend of higher bioaccumulation for Slip 4 fish and compared to Kellogg Island fish. All fish exhibited a positive increase in the total amount of PCBs and most increases were substantial. For example, the median increase in total ng of PCBs for all juvenile chinook collected in this study was 11-fold, which is equivalent to a 1,000% increase.
Fig. 3

Increase in total polychlorinated biphenyls (PCBs) in juvenile chinook. Cumulative frequency plot shows the percent increase in total nanograms of PCB per fish for the years 2000 and 2001. Data are based on individual fish or mean values for composite samples and plotted as log10 values. Arithmetic values shown on top x-axis. Location, fish origin (wild or hatchery), and year of collection indicated in legend

The ΣTEQ values (PCBs only) for all salmonid samples were low exhibiting a mean (SD) of 0.012 (0.024) ng/g lipid. The relationship between tPCBs and ΣTEQs in juvenile salmonids was very strong (r2 = 0.90, n = 110) indicating that the concentration of tPCBs is a good predictor for the toxic potential from the dioxin-like congeners (Fig. 4).
Fig. 4

Regression of total PCBs and PCB TEQs. Values are log10 total PCB concentrations in whole body juvenile chinook salmon and the sum of toxic equivalent quotients (TEQs) for the dioxin like PCBs. Arithmetic equivalents shown on the upper x-axis and right y-axis. The equation is ΣTEQ = 3.39 + 1.03*tPCBs, all concentrations as log10 ng/g or pg/g lipid


Percent lipid content for whole-body juvenile chinook based on wet weight was similar for the years 2001–2004 but higher for the year 2000 (Table 2), which is consistent with the usual pattern of smoltification whereby fish lose lipid content as they transition to seawater (Brett 1995). The mean and SEM was 1.0% (0.1) for 16 individual and composite chinook samples collected over 2001–2004.


The P-values (n = 6) for all possible pair combinations for the year 2000 BSAFs from the PLSD multiple comparison test were high (P > 0.57) indicating no difference between regions or fish origin for this year (Table 2). The majority (74%) of all pairwise comparisons between year 2000 BSAFs and all other years were significantly different (n = 26). Fish collected for the years 2001–2004 were collected later in the summer, which provided potentially more time for bioaccumulation and higher BSAFs. Almost all comparisons among 2001–2004 BSAFs returned high P-values (P > 0.1), except for one low value for Kellogg Island wild fish for 2001.

Sediment guideline

We calculated the 50th, 90th, and 95th percentile sediment concentration associated with its respective BSAF for a given region for the years 2001–2004 (Table 4). These were calculated for all outmigrating juvenile salmon, except those from the year 2000 because of the short time spent in the lower Duwamish. If the year 2000 samples were included, the percentile values for the BSAFs would change slightly (e.g., 90th percentile, Kellogg Island = 1.4 and Slip 4 = 2.2) from the values presented in Table 4.
Table 4

Proposed sediment values to protect against adverse effects in the Lower Duwamish Waterway


Fish (tPCBs)


Sediment guideline (tPCBs)


μg/g lipid

μg/g OC

Sed ng/g

Kellogg Island/west side (n = 58)



















   Mean (sd)

88 (19)

0.81 (0.18)

Slip 4/east side (n = 26)



















   Mean (sd)

360 (75)

1.1 (0.2)

Mean, SD, and various quantile values (Qtiles) determined with equation for lognormal distribution in Gilbert (1987). All fish for a given region over years (except 2000) were combined (years 2001–2004). Equation 2 used to determine sedoc guideline values using BSAF and tissue guideline (2.4 μg/g lipid) for salmonids from Meador et al. (2002). Mean whole-body juvenile chinook lipid was 1% wet weight and 50th percentile for organic carbon (OC) for each side was 1.6% dry wt


PCBs in tissue

The variability in tPCB concentration in outmigrating juvenile chinook was high over time and space; however, a few distinct patterns were detected. These data show that fish on the more contaminated east side of the LDW accumulated far higher amounts of tPCBs than those collected on the west side. Even though some benthic areas on the west side of the LDW contain high concentrations of tPCBs, it appears that the overall average concentration for the different sides is the more important metric for determining bioaccumulation in this mobile species. Based on these observations we conclude that the outmigrating fish probably follow the shallow areas of one side of the waterway or the other and are not likely to cross the channel until later in the summer when they achieve a larger size. One study (Ruggerone et al. 2006) sampled the mid channel area of the LDW from December through February 2005 with a purse seine and found no young-of-the-year chinook (~1.5 g individuals) in this habitat.

The concentrations of tPCBs in fish collected in the year 2000 were on average lower (two to tenfold) than for fish sampled in other years. This lower tPCB trend was not apparent for the year 2000 Slip 4 hatchery fish, which was due to due one individual fish out of 15 that comprise the mean. Without that one value, the mean drops 38% (from 203 to 125 ng/g). These lower values for the year 2000 fish may have been due to the relatively short time for exposure due to recent releases from the main hatchery, increased competition for prey items, or a change in the composition of their prey. The low tPCB concentrations in hatchery fish for the year 2000 may have been caused by the limited time these fish were in the LDW; however, this does not explain the lower values for wild fish, which may have been in the system longer. A plausible explanation for these differences is the expected high degree of competition for prey items among all fish during peak migration of the hatchery fish, which is supported by the lower concentrations for stomach contents for the year 2000 fish. The large release of hatchery fish and subsequent potential competitive interactions among these fish in the Duwamish for scarce resources has been proposed by Nelson et al. (2004) and Ruggerone and Jeanes (2004). This peak in abundance is relatively short-lived because most of the hatchery fish spend little time in this estuary (Nelson et al. 2004).

The low values for 2001 (August) Kellogg Island wild fish were considered atypical due to a number of large fish with near background concentrations. Based on this observation it appears that some juvenile salmon may reside upriver for extended periods before migrating into the contaminated lower estuary. This was observed by Nelson et al. (2004) for both wild and hatchery fish collected at rkm 21 in late June. Interestingly, the percentage of wild fish with low tPCBs (<25 ng/g) for both sampling dates (June and August) at Kellogg Island was far higher (58%) than what we observed at Slip 4 (13%), indicating that these newly arrived fish likely migrated down the west side of the waterway or spent very little time in the LDW before collection at Kellogg Island.

Wild fish are present in the Duwamish as early as January (Ruggerone et al. 2006; Nelson et al. 2004) and show two peaks in abundance, late February/early March for the fry migrants and late May for the fingerlings (Nelson et al. 2004). Based on these data, it is possible that wild chinook may spend several weeks in contaminated areas of the Duwamish accumulating PCBs. As discussed by Thorpe (1994), residence time in an estuary for juvenile chinook is variable and generally a function of season, fish size, and type of estuary; however, 30–90 days is not unusual.

All juvenile chinook increased their total PCB load as they outmigrated through the Lower Duwamish Waterway. As tPCB concentrations increased, fish also increased in mass, which resulted in very high percentage increases in total PCB burden. Juvenile chinook in an estuary are capable of growing at rates of 3–5% body weight/day (Brett 1995; Healey 1991), which is consistent with our observed growth rates of ~3.2% bw/day for the 2001 fish and one study conducted in the LDW (Cordell et al. 2006). This very high rate of growth is due to a feeding rate of 12–20% body weight per day (Brett 1995), which is an important factor because these fish are likely accumulating contaminants at a high rate as a consequence of their high ingestion rate. The rate of prey consumption is an important kinetic parameter for any food web or bioaccumulation model.

One interesting observation is the percentage occurrence of wild versus hatchery fish in our collections. For the year 2000 the percentage of wild fish was 38%, which was most likely related to the recent releases of hatchery fish into the system. For the succeeding years, the percent occurrence of wild fish was far higher averaging 62%, including 1 year (2001) that averaged 83% wild fish. Studies have shown that hatchery reared fish will spend less time in the estuary than naturally reared fish (Levings et al. 1986), which is apparent from these data. This observation is important because we are more concerned with impacts to wild fish, including chinook salmon, under the Endangered Species Act than fish of hatchery origin. Due to the higher percentage of wild fish during the summer months and the higher levels of bioaccumulation observed for these fish compared to those earlier in the spring, the main focus should be on this group of fish that have spent several weeks in the estuary accumulating high levels of toxic compounds.

It is difficult to predict habitat usage by highly mobile, outmigrating juvenile chinook; however, we expected that a large percentage of fish would stay close to shore because of the generally higher abundance of prey and protection from predators. We believe that the higher tissue concentrations and relatively similar BSAFs for fish from the east versus west side of the waterway support this assumption of segregation within this system and indicate the need to consider appropriate geographic scales for bioaccumulation assessment for this (or any) fish species.

We found a very high correlation (r2 = 0.90) between total PCBs and PCB TEQ values that could be used for predictions of toxicity. A few fish were elevated (PCB TEQ > 0.05 ng/g lipid); however, most were below the mean 95th percentile species protection benchmark for lethal effects (0.39 ng/g lipid) proposed by Steevens et al. (2005). When other dioxin-like compounds are considered, chinook at this life stage, and other species in the LDW, may exhibit TEQ values that are high enough to elicit toxic responses. It is known that dioxin-like compounds can impair the immune system, inhibit growth, cause thymic atrophy, and act as endocrine disruptors (Giesy and Kannan 1998), each an important function for estuarine fish.


As expected, the BSAFs for the year 2000 were generally lower because fish were collected in the spring, which is likely due to a short time period for accumulation, type of prey items available, or competition leading to reduced dietary uptake. For the other years, some of the juvenile chinook samples exhibited BSAF values that were surprisingly high. Based on their growth rate, juvenile chinook likely have a high rate of dietary accumulation and therefore would accumulate high tissue concentrations relatively rapidly. It is possible for these fish to exhibit high levels of accumulation and relatively high BSAFs after several days to a few weeks in the LDW. Additionally, salmonids have a high rate of ventilation, therefore uptake from the water column via the gills could be an important pathway for contaminant accumulation (Meador et al. 2008). The relative similarity for chinook BSAFs between the two regions for a given year (Tables 2, 4) and the high P-values between matched Kellogg Island and Slip 4 samples indicates that our selection of sediment concentrations for the BSAF calculations was appropriate for this species. This is also supported by the data in Table 4. If we had selected the sediment concentration at the collection sites, the tPCB tissue concentrations should have be tenfold higher in fish from the east side of the LDW compared to those from the west side. Additionally, using those Sedoc values (8.9 and 88 μg/g OC) would have produced highly skewed BSAF values. Given the expected similar rates of ingestion and ventilation for these fish, plus a similar time frame for exposure, the BSAF values between the two sides of the LDW were expected to be similar.

Our intent was not to use BSAFs as an indicator of steady-state bioaccumulation or the theoretical bioaccumulation potential, but to allow for interconversion between tissue and sediment concentrations with the lowest achievable variance. The mean and various quantiles for the chinook BSAFs for both regions were relatively similar and varied by less than a factor of two, which was considerably less than the variability observed for whole-body tPCBs. We believe that many of these fish are far from steady state and that the rates of uptake (dietary and ventilatory) are the main factors controlling the levels of whole-body PCBs. For bioaccumulation, organismal lipid content is an important factor only for individuals at steady state and for chemicals that are not metabolized. While the numerator of the BSAF equation (lipid-normalized tissue concentrations) may not be an accurate indicator of bioaccumulation for fish in this study, we do consider the denominator (sedoc) to be a reasonable indicator of the bioavailable fraction from all sources available for uptake, which is primarily water and prey.

Determining a sediment guideline based on bioaccumulation

The determination of sediment concentrations that may result in adverse tissue concentrations can be accomplished with BSAF values (Meador 2006). For example, Meador et al. (2002) proposed that a tissue concentration of 2.4 μg tPCBs/g lipid was a protective tissue quality guideline (TQG) for salmonids. This TQG describes the 10th percentile of a variety of adverse biological responses for non-embryonic salmonids (fry to adult) that was compiled from several research studies. Using the BSAF (Eq. 2) and the TQG, we can solve for a sediment concentrations that should be protective against adverse effects. By examining the distribution of BSAF values observed in this study, we were able to determine sediment concentrations that could be used to protect a given percentage of the individuals. The values we provide in Table 4 would allow regulators to select appropriate percentile values that would be used to protect a given percentage of the population of outmigrant chinook salmon. For example, if the 90th percentile BSAF value was selected for chinook in the LDW, the sediment value to protect fish from bioaccumulating an adverse tissue concentration (≥2.4 μg/g lipid) would be 1.0 μg/g OC. The vast majority of juvenile chinook are from hatcheries and these fish move quickly through this estuary; however, it is the naturally reared juvenile chinook salmon that can spend considerable time in this system and likely accumulate high concentrations of PCBs and other contaminants that justifies this high percentage value.

The data we present here are just one example describing this application. Of course, several factors affect bioaccumulation and the BSAF, such as variable uptake and elimination rates, reduced bioavailability, reduced exposure, and insufficient time for sediment-water partitioning or tissue steady state. Because of these differences in bioaccumulation, a BSAF that is specific for a given estuary and species is recommended for a more accurate representation of bioaccumulation as a function of the above factors. Lipid content is also an important factor. Even though organismal lipid likely had little effect on the magnitude of bioaccumulation of PCBs for these fish (e.g., Stow et al. 1996), we believe that tissue lipids will be a factor in determining the toxic response. As proposed elsewhere (Lassiter and Hallam 1990), the lipid content of tissue controls the proportional availability of accumulated hydrophobic toxicants and therefore the magnitude of the toxic response, which is a factor we considered when developing the tPCB TQG for salmonids (Meador et al. 2002).

It is clear from these data that bioaccumulation of PCBs for a given area and time is highly variable. This is strong support for the importance of extensively sampling a given area at various locations and times to adequately characterize bioaccumulation, especially when considering population responses. These recommendations for other small estuaries include sampling in several locations, taking multiple samples over a species’ potential residence time, and using a probabilistic approach for characterizing tissue concentrations that may lead to adverse effects. Obviously, a few composite samples from one or two randomly selected locations at one time period would severely underestimate the bioaccumulation potential for juvenile salmon as they rear in an estuary to accumulate mass and lipid stores before their first winter in open water. Additionally, these data indicate the importance of reducing sediment concentrations to effect reduced tissue concentrations to levels that are expected to be safe for fish and their prey. Assessing bioaccumulation in an iterative fashion after multiple rounds of sediment cleanup will provide needed information that remediation efforts are effective.


We thank Sean Sol, Maryjean Willis, Mark Myers, O. Paul Olson, Gladys Yanagida, Dan Lomax, and Bernadita Anulacion for assistance with field collections, sample preparation, and analytical analyses. Jay Field and Lyndal Johnson provided several insightful comments on this manuscript.

Copyright information

© US government employee 2009