A detailed table with BCF values can be found in the Supplemental Data. Table 1 summarizes all valid data for fish, with a column with nonnormalized BCFs and a column with BCFs normalized to 5% lipids for those studies where lipid contents of the fish were reported. A box plot containing all valid data can be seen in Figure 2. For QS derivation purposes, the overall BCF is derived by first calculating the geometric mean for each species, and then taking the geometric mean of all species. Both the nonlipid-normalized and the lipid-normalized geometric mean BCF are 12 800 L/kg. Lipid normalization reduces the variability that is caused by differences in characteristics of the fish used in the experiments. Therefore, the lipid-normalized geometric mean value of 12 800 L/kg is considered most reliable.
Table 1. Summary of fish bioconcentration data for HCB
|Species||BCF (L/kg)||BCF 5% lipids||Remark||Reference|
|Gambusia affinis||3730||6020|| ||Chaiksuksant et al. 1997|
|Gambusia affinis||3780||6090|| ||Chaiksuksant et al. 1997|
|Gambusia affinis||3750||6050||Geomean|| |
|Gasterosteus aculeatus||22 100||40 900|| ||Egeler et al. 2001|
|Ictalurus punctatus||11 000||7450||Dietary study||Woodburn et al. 2008|
|Lepomis macrochirus||21 900|| || ||Veith et al. 1979|
|Oncorhynchus mykiss||12 100|| || ||Lu and Wang 2002|
|Oncorhynchus mykiss||16 700||29 800||Dietary study||Exxon Mobil 2005|
|Oncorhynchus mykiss||15 800||16 500||Dietary study||Exxon Mobil 2005|
|Oncorhynchus mykiss||10 800||22 500||Dietary study||Exxon Mobil 2005|
|Oncorhynchus mykiss||10 100||15 800||Dietary study||Exxon Mobil 2005|
|Oncorhynchus mykiss||22 200||13 700||Dietary study||Exxon Mobil 2005|
|Oncorhynchus mykiss||15 000||13 400||Dietary study||Exxon Mobil 2005|
|Oncorhynchus mykiss||5500|| || ||Veith et al. 1979|
|Oncorhynchus mykiss||19 500||23 800||Dietary study||Fisk et al. 1998|
|Oncorhynchus mykiss||13 200||18 600||Geomean|| |
|Pimephales promelas||26 700|| || ||Carlson and Kosian 1987|
|Pimephales promelas||21 400|| || ||Carlson and Kosian 1987|
|Pimephales promelas||22 500|| || ||Carlson and Kosian 1987|
|Pimephales promelas||17 700|| || ||Carlson and Kosian 1987|
|Pimephales promelas||20 200|| || ||Carlson and Kosian 1987|
|Pimephales promelas||16 600|| || ||Veith et al. 1979|
|Pimephales promelas||18 200|| || ||Veith et al. 1979|
|Pimephales promelas||17 800|| || ||Veith et al. 1979|
|Pimephales promelas||45 700|| || ||Veith et al. 1979|
|Pimephales promelas||16 200|| || ||Veith et al. 1979|
|Pimephales promelas||18 500|| || ||Veith et al. 1979|
|Pimephales promelas||12 200||8840|| ||Nebeker et al. 1989|
|Pimephales promelas||15 300||11 100|| ||Nebeker et al. 1989|
|Pimephales promelas||21 100||15 300|| ||Nebeker et al. 1989|
|Pimephales promelas||12 600||9130|| ||Nebeker et al. 1989|
|Pimephales promelas||13 300||9640|| ||Nebeker et al. 1989|
|Pimephales promelas||11 500||8330|| ||Nebeker et al. 1989|
|Pimephales promelas||20 700||15 000|| ||Nebeker et al. 1989|
|Pimephales promelas||93 800|| || ||Schuytema et al. 1990|
|Pimephales promelas||19 900||10 700||Geomean|| |
|Poecilia reticulata||15 700||14 500|| ||Könemann and van Leeuwen 1980|
|Poecilia reticulata||7660||9580||Dietary study||Clark and Mackay 1991|
|Poecilia reticulata||11 000||11 800||Geomean|| |
|Overall geometric mean||12 800||12 800||See text|| |
Figure 2. Box plots of all valid data for BCF, BAF, and TMF. Box denotes 25th percentile, median, and 75th percentile. Whiskers denote minimum and maximum values. A line is included on the BAF-BCF level to facilitate comparison between BAF-BCF and TMF.
Download figure to PowerPoint
As a comparison, the BCF for fish can be calculated using the linear relationship developed by Veith et al. (1979) as used in the WFD guidance (EC 2011): log BCF = 0.85 × log Kow −0.70. Using the log Kow of 5.73, the resulting BCF is 14 800 L/kg. This is in good agreement with the geometric mean experimental value.
For invertebrates that are not suitable for consumption, at least by humans, BCFs between 13 200 and 75 000 L/kg have been determined, for insects a BCF of 29 000 L/kg is available, and for oligochaetes BCFs range between 25 100 and 106 800 L/kg (see Supplemental Data; Appendix A). For the purpose of QS derivation, these data serve as circumstantial evidence but are deemed to be less relevant when good data for bioconcentration in fish are available. For other frameworks however (e.g., the assessment of persistence, bioaccumulation and toxicity [PBT]), these data are very relevant. For mussels, BCF values tend to be lower than for fish (see Supplemental Data).
A description of bioaccumulation studies including the rationale on validity is given in the Supplemental Data. Results of valid studies are summarized in Table 4, with a box plot in Figure 2. All reported BAFs are based on lipid-weights. Recalculated BAFs normalized to 5% lipids are also included in the table.
Table 4. Summary of valid BAF data for HCB, with geometric means for all species where applicable
|Species||Trophic position||BAF (L/kg) (lipid-weight)||BAF (L/kg) (normalized to 5% lipids)||Reference|
| Arrow worms (Sagitta elegans, Eukrohnia hamata)c,d|| ||2.7||6.3 × 105||31 700||Hallanger, Ruus, et al. 2011|
| Calanus finnmarchusc,d|| ||2||1.9 × 105||9600||Hallanger, Ruus, et al. 2011|
| Calanus glacialisd|| ||2||8.8 × 104||4400||Hallanger, Ruus, et al. 2011|
| Calanus hyperboreusc,d|| ||1.6||9.8 × 104||4900||Hallanger, Ruus, et al. 2011|
| Krill (mostly Thysanoessa inermis)c,d|| ||2.4||1.0 × 106||51 000||Hallanger, Ruus, et al. 2011|
| Pontoporeia affinis|| || ||4.0 × 106||200 000||Oliver and Niimi 1988|
| Themisto abyssorumc,d|| ||1.5||9.4 × 105||47 000||Hallanger, Ruus, et al. 2011|
| Themisto libellulac,d|| ||2.0||1.4 × 106||69 000||Hallanger, Ruus, et al. 2011|
|Fish|| || || || || |
| Alosa pseudoharengus||3.51|| ||1.9 × 106||95 000||Oliver and Niimi 1988|
| Boreogadus saidac,d||3.10||2.9||1.3 × 106||66 000||Hallanger, Warner, et al. 2011|
| Comephorus dybowskiic||3.44||3.86||6.7 × 106||333 000||Kucklick et al. 1996|
| Comephorus baikalensisc||3.29||3.96||6.1 × 106||305 000||Kucklick et al. 1996|
| Coregonus autumnalis migratoriusc||3.57||3.40||1.8 × 107||885 600||Kucklick et al. 1996|
| Cottus cognatus||3.37|| ||3.2 × 106||158 000||Oliver and Niimi 1988|
| Gadus morhuac,d||3.73||3.3||3.0 × 106||148 000||Hallanger, Warner, et al. 2011|
| Mallotus villosusd||3.15||2.5||9.5 × 105||48 000||Hallanger, Warner, et al. 2011|
| Melanogrammus aeglefinusc,d||4.09||2.8||2.4 × 106||120 000||Hallanger, Warner, et al. 2011|
| Osmerus mordaxc||3.00|| ||1.7 × 106||86 000||Oliver and Niimi 1988|
| Pollachius virensc,d||4.38||3.2||2.2 × 106||111 000||Hallanger, Warner, et al. 2011|
| Salmo trutta (muscle)||3.16||3.14||8.7 × 106||433 000||Catalan et al. 2004|
| Salmonids (Oncorhynchus kisutch, O. mykiss, Salvelinus namaycush, Salmo trutta)e||3.16–4.42|| ||2.3 × 106||115 000||Oliver and Niimi 1988|
|Geometric mean of all individual fish data except the composite salmonid sample|| || || ||238 000|| |
In the European substance data sheet for HCB (EC 2005), a BAF of 42 000 L/kg is used based on data for bream from the river Elbe. However, this value was deemed to be less reliable because it is based on muscle tissue wet weight instead of 5% lipid-normalized whole fish and was based on total concentrations in water (including suspended solids) instead of dissolved concentrations. Moreover, the trophic position of the species used is low (2.31 according to Van Riel et al. 2006 and 2.94 ± 0.37 according to http://fishbase.org) and the species is mainly benthivorous, which renders the value of the BAF less reliable.
The geometric mean of all individual valid 5% lipid-normalized BAFs is 238 000 L/kg. The worst-case BAF value is 2 116 000 for 3-year-old Coregonus autumnalis. Bioaccumulation factor measurements show a high variation of more than 1 order of magnitude. Bioaccumulation factors are expected to correlate with trophic level of the fish (e.g., Borga et al. 2012). In Figure 3, it is shown that for HCB this is also the case. When all individual BAFs are used, the relationship with trophic level is not significant (p = 0.06). However, if the high value for C. autumnalis is left out of the regression, the slope of the regression line differs significantly from 0 (p = 0.0057). When the residuals of the plot are tested for outliers, the value for C. autumnalis is a significant outlier (p < 0.01). Thus, when the high value of C. autumnalis is not included, the regression between BAF and trophic level becomes highly significant. The value of the slope of this regression line (on a log-basis) is 0.39, which represents a TMF of 2.47.
Figure 3. Influence of trophic level on BAFs (based on individual data from the references included in Table 4).The regression line with 90% confidence interval shows the dependence of the BAF on trophic level (log BAF = 0.39 TL + 4.00; p = 0.0057).
Download figure to PowerPoint
Although the fish listed in Table 4 do not differ much in trophic level according to http://fishbase.org, there are distinct differences in feeding strategies. For example, the food chain in Lake Ontario, where samples in the study by Oliver and Niimi (1988) originated, includes Cottus cognatus as a benthic predatory fish, Osmerus mordax and Alosa pseudoharengus as pelagic predatory fish, and salmonids as a top predator. From the study by Kucklick et al. (1996), Coregonus is also a salmonid that feeds on smaller fish. Thus, besides trophic level, feeding strategy may influence the BAF for HCB. Other factors, such as age, size, reproductive status, biotransformation efficiency, and omnivorous feeding may be important for the observed contaminant concentrations in an organism (Borga et al. 2012). For fish at higher trophic levels, it is observed that size or age affect bioaccumulation of the contaminant, and larger, slower-growing individuals typically contain higher organic contaminant levels than younger, faster-growing individuals (Borga et al. 2012 and references therein). For HCB however, the expected increase in BAF values with increasing age is not found for 3 fish species in the study by Kucklick et al. (1996). To the contrary, the reverse is more likely, although a high variation is shown (Figure 4). It should be added that the correlation between trophic level and age is also rather weak, although positively correlated. These results are confirmed by measurements of Ferrante et al. (2010) in European eel (Anguilla anguilla) in Italy, where a significant correlation with weight and length was observed for PCBs but not for HCB. Thus, the observed BAF values can be explained primarily by the trophic level of the fish, although the variation remains high.
Even at lower trophic levels (algae, small zooplankton), accumulation of HCB already far exceeds what is expected through equilibrium partitioning. For instance the BAF for amphipods and plankton normalized to 5% lipids was 107 000 L/kg in the study by Oliver and Niimi (1988). In the study by Kucklick et al. (1996) data for invertebrates were below the limit of detection. However, for the predatory amphipod Macrohectopus branicii, a BAF could be derived. If the limit of detection for the zooplankton is taken as an upper limit, the BAF values for these invertebrates normalized to 5% lipid weight can be supposed to be below 92,000 L/kg. With the trophic level for zooplankton approximately 2 and the trophic level of the predatory amphipod M. branicii around 2.5, an increase in BAF with a factor of at least 2 per trophic level would be deduced from these data. The same pattern was observed for PCBs in the same study, where in the group of fish no clear trend with trophic level was observed, whereas there was a significant relationship with trophic level if invertebrates were included.
These relatively high HCB concentrations at lower trophic levels cannot be explained easily with the usual theories on bioaccumulation. Possibly, the low water solubility of HCB (that is lower than expected from its Kow) and high crystal energy may cause high sorption to the body surface. Especially at lower trophic levels, with small individuals with a high surface area to body ratio, this may contribute significantly to the observed contaminant levels. This affects BAFs at higher trophic levels as well.
Another source of uncertainty for BAF values is the fairly large uncertainty surrounding the measurements of the aqueous concentrations, which are often very low in the field. Because BAFs for HCB only originate from 4 studies, no definite conclusions can be drawn on the influence of the uncertainty in water concentrations on BAFs. However, it should be stressed that the concentrations of HCB are rather consistent over the 4 studies in the range of 10 to 150 pg/L. This probably reflects the global distribution of this substance.
Comparing BCF, BMF, and BAF
Lower trophic levels (plankton, amphipods) already may have field derived BAF values far above observed laboratory BCFs for fish. Arnot and Gobas (2006) in their review also observed that BAFs from field samples can differ by up to several orders of magnitude from laboratory BCF measurements for some chemicals and explained this through trophic position, sediment–water disequilibria, and unique characteristics for different ecosystems. When comparing BCF values for fish multiplied by the BMF (12 800 × 3 = 38 400 L/kg) to the observed BAF values for fish (48 000–885 600), there appears to be a large difference between these 2 values. This can also be observed with the box plots (Figure 2), where a line on the level of BAF-BCF is included to facilitate the comparison with the TMF. It is shown in this graph that the TMF does not explain the observed difference between the BAF and the BCF. Thus, the assumption that the BAF equals BMF × BCF (or TMF × BCF) does not seem to hold for HCB. This assumption only works if BAF values for small fish and other aquatic species are comparable to the laboratory BCF data, and consequently the BAF for species with a higher trophic level only need to be corrected for one biomagnification step. For HCB, this is apparently not the case. It appears that the BAF values are almost a factor of 20 higher than the BCF values, whereas the increase per trophic level is only a factor of 3. This means that the number of trophic levels that should be taken into account for the biomagnification process is 2 to 3 instead of the single trophic level that is considered in the current methodology for risk assessment and QS derivation (that can also be deduced from the box plots in Figure 2). This is not that far-off because often food chains are longer than 3 trophic levels and fish may also feed on their own brood. Thus, it is not considered appropriate to use BCF × BMF values to recalculate the biota standards into water standards, because this methodology greatly underestimates field BAFs for HCB.
Calculation of QS for HCB in water
In the substance data sheet for HCB (EC 2005), QS for biota based on human consumption of fishery products (QSbiota,hh food) and based on secondary poisoning (QSbiota,secpois) of 9.74 µg/kg and 16.7 µg/kg respectively, are derived. According to the substance data sheet, these can be recalculated into water concentrations using a BAF of 42 000 L/kg, which results in a QSfw,hh food of 0.00023 µg/L (0.23 ng/L) and a QSfw,secpois of 0.00004 µg/L (0.04 ng/L).
The value of 42 000 L/kg for the BAF is, however, not based on an extensive literature search. It is deemed to be less reliable, because it is based on muscle tissue wet weight instead of 5% lipid-normalized whole fish, and refers to total concentrations in water (including suspended solids) instead of dissolved concentrations. Following the most recent guidance for derivation of QS under the Water Framework Directive (EC 2011), QS have to reflect dissolved concentrations.
Using the information acquired above, there are 4 options to implement QS for HCB. These are all discussed below. It should be noted that the option that is preferable from a scientific point of view, may not be the most desirable option from a policy maker's point of view. Besides this, these methods are based on the most recent EU methodology, and the resulting choices below have been made from a European perspective. In other states other methodologies may prevail, like for instance in the United States (USEPA 2000).
Use the QSbiota,hh food and QSbiota,secpois from the substance data sheet without any recalculation into water concentrations. This would involve the highest degree of certainty surrounding the value as such. However, monitoring biota for compliance checking with the QSbiota introduces a very high variability, as was demonstrated by the highly variable BAF values in this article, even when these BAF values originated from the same site. This means that for compliance checking of the monitoring data with the QSbiota still a high uncertainty remains. Moreover, monitoring of biota is difficult to harmonize (which age, size, which species, which trophic level), labor-intensive, and consequently costly. Besides that, it should be discouraged from an animal welfare perspective.
Use the QSfw,hh food
from the substance data sheet (EC 2005
), where a BAF of 42 000 L/kg is used (based on total concentrations in water). However, this BAF is deemed to be less reliable and underestimates the observed BAF values and no extensive literature search was performed. Because the BAF is based on total concentrations in water, the resulting QS (QSfw,hh food
= 0.23 ng/L and QSfw,secpois
= 0.4 ng/L) would also refer to total concentrations. However, according to the most recent guidance on EQS derivation (EC 2011
), QS should be based on dissolved concentrations.
Use the worst-case BAF of 885 600 L/kg for Coregonus migratorius autumnalis
(Kucklick et al. 1996
). This value is a geometric mean of BAF values of individual fish of this species with different ages; no age-dependency of the value could be shown, but there was a large variability among the values. Moreover, this value is highly influenced by one single value that can be considered to be an outlier when the regression with trophic level is taken into account (see before). Because the BAF values are based on dissolved concentrations, these QS (QSfw,hh food
= 0.011 ng/L and QSfw,secpois
= 0.019 ng/L) refer to dissolved concentration.
Use the geometric mean of all individual valid BAFs, 238,000 L/kg. Some uncertainties surround this value, because the variation among BAFs is high. The height of the BAF is relatively high (for comparison, the BAF calculated as BCF × BMF would have been 38 400 L/kg (see discussion above), but the BAF is more relevant for the field situation than the BCF × BMF. Also in this case, the BAF values and thus the resulting QS (QSfw,hh food = 0.041 ng/L and QSfw,secpois = 0.070 ng/L) are based on dissolved concentrations.
Use the BAF for fish that are most relevant for human consumption and secondary poisoning (trophic level 4 in Figure 1
). This can be done by calculating the BAF value at trophic level 4 from the regression of the relationship between BAF and trophic level (Figure 3
). This BAF value is 372 000 L/kg. Although individual BAF values vary considerably, this average BAF at trophic level 4 has a confidence interval ranging from 254 000 to 547 000. Because trophic level 4 is at the upper end of the data from which the regression was calculated, the resulting BAF value obtained for trophic level 4 is higher than the geometric mean of all available data (option 4, see above). Also the lower value of the confidence interval is higher than this geometric mean. Because the BAF for trophic level 4 is based on dissolved water concentrations, the QS resulting from this value (QSfw,hh food
= 0.026 ng/L and QSfw,secpois
= 0.045 ng/L) are also based on dissolved concentrations.
The differences in water-based QS between options 2 and 3 through 5 are not only caused by the height of the BAF used, but are also caused by a difference based on dissolved concentrations versus (higher) total concentrations. This difference is not easy to quantify, because it depends on the amount of suspended solids in the systems. However, in the most recent guidance on QS derivation (EC 2011) it is specifically stated that QS should be based on dissolved concentrations.
Final choice of QS
Regarding the final choice for the above options, option 2 is not preferable because of the less reliable BAF value used. From a scientific point of view, option 3 is also not preferable, because it is highly influenced by 1 single value that can be considered to be an outlier when the regression with trophic level is taken into account. Option 5 might be preferred over option 4, because the QS are based on larger fish (trophic level 4 in Figure 1) that are suitable for consumption and that represent the prey for predators, and for this trophic level the representative BAF values are significantly higher than the geometric mean BAF value over all trophic levels.
Compliance checking involves both uncertainties in the quality standard as well as in the monitoring data. Compliance checking by means of monitoring in water (options 2 through 5) has advantages over biota sampling (option 1) in terms of reproducibility, costs, animal welfare, and uniformity of sampling. However, recalculation of biota standards into water standards (options 2 through 5) introduces considerable uncertainties regarding the height of the BAF used and the resulting quality standard is thus more uncertain than the quality standard for biota. On the other hand, concentrations of HCB in biota that should be monitored to check compliance to these WFD requirements, show high individual variability as reflected by the variability of the BAF values that were considered in this study. Furthermore, the monitored biota should correspond to the same trophic level as the level the EQS refers to. This also introduces a lot of uncertainties, because HCB concentrations in biota are highly variable and will also depend on the species, age, and trophic level of the fish sampled, and there is no guidance on this point yet. These aspects will result in high uncertainty if biota are monitored instead of water.
Thus, the following “tiered approach” is suggested. The geometric mean BAF for the trophic level on which the QS is set (option 5) is used, leading to a critical water quality standard of 0.026 ng/L (based on dissolved concentrations). If this standard is then exceeded in the field, it could be considered to sample biota and make a weight of evidence analysis with both water and biota samples for compliance checking. Such an analysis could for example include an overview of site-specific BAF values for several types of species.
For the marine environment, the QS for human consumption of fishery products equals the freshwater QS for human consumption of fishery products. However, for secondary poisoning the marine top predators (like polar bears) should also be protected. This means that the QS for secondary poisoning in freshwater should be divided by an additional BMF (3) from fish to birds and mammals to account for this extra trophic level in the food chain. The new guideline for deriving EQS (EC 2011) points out that biota QS expressed as concentrations in biota should be divided by this BMF value as well, because monitoring of biota at the level of food for the marine top predators (e.g., seals) is considered inappropriate.
The routine limit of quantification (LOQ) for HCB in water is usually approximately 1 ng/L, with the limit of detection (LOD), depending on the laboratory, approximately a factor of 10 lower. This is significantly higher than the currently derived QS for the water column, even more so because the QS is now based on dissolved concentrations instead of (higher) total concentrations. This means that concentrations of HCB in the water column that are close to the freshwater QS cannot be measured directly (via liquid–liquid extraction) and may have to be measured using passive sampling devices. These are, however, not (yet) used in routine monitoring programs. If conventional methods are used, the QS may already be exceeded when concentrations are still below the detection limit. However, from the studies underlying the BAFs, it can be deduced that analyzing much lower concentrations than the routine LOQ should be possible, e.g., by using large volumes of water. These techniques yield sufficiently low LODs for quantification of the derived QS. However, the practical implication of this is that costs of routine monitoring programs would increase.