Processes Influencing Chemical Biomagnification

Chemosphere 154 (2016) 99e108

Contents lists available at ScienceDirect

Chemosphere

journal homepage: www.elsevier.com/locate/chemosphere

Processes influencing chemical biomagnification and trophic magnification factors in aquatic ecosystems: Implications for chemical hazard and risk assessment

* Donald Mackay a, Alena K.D. Celsie a, , Jon A. Arnot b, c, David E. Powell d a Chemical Properties Research Group, Department of Chemistry, Trent University, Peterborough, ON, K9L OG2, Canada b ARC Arnot Research and Consulting Inc., 36 Sproat Avenue, Toronto, ON, M4M 1W4, Canada c Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, ON, M1C 1A4, Canada d Dow Corning Corporation, Health and Environmental Sciences, Auburn, MI 48611, USA highlights

TMFs are highly dependent on a substance's KOW value. Biotransformation can greatly reduce TMFs. Increased sediment-to-water ratios can inﬂuence TMFs. Food web omnivory has only a small effect on TMFs. Spatial variability in water concentrations can have a profound effect on TMFs. article info abstract

Article history: Bioconcentration factors (BCFs) and bioaccumulation factors (BAFs) are widely used in scientific and Received 11 December 2015 regulatory programs to assess chemical hazards. There is increasing interest in also using bio- Received in revised form magnification factors (BMFs) and trophic magnification factors (TMFs) for this purpose, especially for 10 March 2016 highly hydrophobic substances that may reach high concentrations in predatory species that occupy high Accepted 10 March 2016 trophic level positions in ecosystems. Measurements of TMFs in specific ecosystems can provide Available online 31 March 2016 invaluable confirmation that biomagnification or biodilution has occurred across food webs, but their use Handling Editor: I. Cousins in a regulatory context can be controversial because of uncertainties related to the reliability of measurements and their regulatory interpretation. The objective of this study is to explore some of the Keywords: recognized uncertainties and dependencies in field BMFs and TMFs. This is accomplished by compiling a Trophic magnification factors (TMFs) set of three simple food web models (pelagic, demersal and combined pelagicedemersal) consisting of Food web model up to seven species to simulate field BMFs and TMFs and to explore their dependences on hydrophobicity Biomagnification (expressed as log KOW), rates of biotransformation and growth, sediment-water fugacity ratios, and e fi Octanol water partition coef cient extent of food web omnivory and issues that arise when chemical concentration gradients exist in Hydrophobic substances aquatic ecosystems. It is shown that empirical TMFs can be highly sensitive to these factors, thus the use of TMFs in a regulatory context must recognize these sensitivities. It is suggested that simple but realistic evaluative food web models could be used to extend BCF and BAF assessments to include BMFs and TMFs, thus providing a tool to address bioaccumulation hazard and the potential risk of exposures to elevated chemical concentrations in organisms at high trophic levels. © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

1. Introduction

* Corresponding author. Trent University, 1600 West Bank Drive, Peterborough, There is scientific and regulatory interest in understanding and ON, K9J 7B8, Canada. quantifying the fate, bioaccumulation, exposure and potential for E-mail addresses: [email protected] (D. Mackay), [email protected] adverse effects of chemicals released to the environment. Trophic (A.K.D. Celsie), [email protected] (J.A. Arnot), david.powell@dowcorning. biomagnification across a food chain or food web can considerably com (D.E. Powell). http://dx.doi.org/10.1016/j.chemosphere.2016.03.048 0045-6535/© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). 100 D. Mackay et al. / Chemosphere 154 (2016) 99e108 increase chemical concentrations, hence exposure and potential of the chemical, food web branching (or more properly termed risk, at higher trophic levels compared to concentrations at lower omnivory) and spatial variability in water concentrations. The trophic levels (Czub and McLachlan, 2004; Czub et al., 2008; Kelly equations used here and the parameterisations are deliberately et al., 2007). There is thus an incentive to understand and include simplistic and thus have less fidelity to real systems, but this these magnification processes in both hazard and risk assessment. simplicity is justified by the objective of clearly identifying the key Bioaccumulation assessments seek to identify chemicals with high factors that influence TMFs. This model is only intended to be potential for bioaccumulation and employ several metrics and applied to neutral, relatively hydrophobic organic chemicals and criteria as reviewed by several authors, notably Borga et al. not to ionogenic chemicals. There is no intention in this study to (2012a,b), Gobas et al. (2009), Burkhard et al. (2013), and others. define the absolute uncertainty in reported TMFs. Uncertainty is inherent whether the bioaccumulation data are fi measured from laboratory tests, through eld monitoring cam- 2. Methods paigns, or calculated using models. Mass balance models provide mechanistic insights into bioaccumulation processes (Thomann 2.1. Trophic position and magnification et al., 1992; Arnot and Gobas, 2006; Kelly et al., 2007; Barber, 2003; Barber, 2008; Walters et al., 2011; Kim et al., 2016). Food Trophic level (TL) or position occupied by an organism can be web mass balance bioaccumulation models also highlight the basic estimated from the biological enrichment of 15N by organisms relationships between various chemical and biological properties during each digestive event. Applying an enrichment factor of and ecological properties and processes and the metrics used to typically 3.4‰ per trophic level step enables the TL to be estimated. assess bioaccumulation and exposure (Mackay et al., 2013). Reliable The enrichment factor is subject to variability, especially at lower fi fi measurements and models foster con dence in scienti c knowl- TLs. The TL is essentially the number of trophic level steps or edge and in applying various sources of information for decision- transfers (n) that the food has experienced as it is consumed and making. transferred from the base of the food web (TL ¼ 1orTL1)to Standard test protocols have been developed for determining increasingly higher trophic levels (TL ¼ norTLn). fi bioconcentration factors (BCFs) and biomagni cation factors When written in terms of fugacity (f) biomagnification across a fi (BMFs) under well-de ned laboratory conditions (OECD, 2012). In food web may be described as: these tests the organism is only exposed to a chemical either from the water (BCF) or from the diet (BMF). In the environment, or- logfy2 logfy1 ¼ðTLx2 TLx1Þ$logTMF (1) ganisms are exposed to chemical from their surrounding environ- e ment (e.g., water) and from their diet. Although insightful, the BCF The term TLx2 TLx1 is the trophic level separation between two is not necessarily ecologically relevant for hydrophobic chemicals species on a linear food chain or the length of the food web under D because dietary exposure, and hence the potential for bio- consideration, which may be represented as TL (i.e. D ¼ e D ¼ magnification, is not included (Qiao et al., 2000; Thomann et al., TL TLx2 TLx1). In the situation where TL 1, Eq. (1) reduces to ¼ fi 1992). The bioaccumulation factor (BAF) includes all exposure TMF fy2/fy1, which by de nition is a BMF. The simplest case is a routes under environmental conditions and is often determined linear food chain in which omnivorous feeding does not occur and ¼ from monitoring data (Arnot and Gobas, 2006). Biomagnification each predator species (TL x) consumes only the prey species ¼ and trophic magnification are conveniently expressed in terms of (TL x 1) that occupies the trophic position that is one step below fugacity increases from prey to predator or throughout the food that of the predator. Thus if the ratio of predator to prey lipid web as advocated by Gobas et al. (2009), Burkhard et al. (2013) and normalised concentrations or fugacities (i.e., the BMF) is equal for others. Lipid normalised concentration can also be used instead of each increase in TL, then for a linear food chain of n species fn/f1 (n1) fugacity if it is assumed that the two quantities are proportional. equals TMF where f1 is the fugacity in the organism that oc- ¼ For each organism the log fugacity (or log of the lipid normalized cupies TL 1 and fn is the fugacity in the organism that occupies ¼ concentration) is plotted as a function of trophic position or trophic TL n. A plot of log f versus TL for a linear ideal food chain thus has ¼ level (TL) and the trophic magnification factor (TMF) is deduced a slope of log TMF ( log BMF) and an intercept at TL1 of log f1. D D from the slope, i.e. TMF is 10(slope) or 10( logf/ TL) (Mackay et al., Similarly, a plot of log f versus TL for a food web has a slope of log 2013). The TMF is thus a metric of the “average BMF” of the TMF and an intercept at TL1 of log f1. Essentially a BMF is the slope chemical in the sampled ecosystem. Field metrics may be accepted between two points (and calculated as the ratio of concentrations D ¼ as lines of evidence for bioaccumulation assessment; however, because TL 1) while a TMF is a slope involving all points. considerable variability is expected in fugacities (and concentrations) and hence BAFs, BMFs and TMFs. Recently, Conder et al. 2.2. Evaluative TMF model development (2012), Starrfelt et al. (2013), Burkhard et al. (2013), Franklin (2016), McLeod et al. (2014), and Kim et al. (2016) have discussed To advance understanding of food web bioaccumulation it is uncertainties inherent in the use of TMFs. Mass balance TMF desirable to analyse both monitoring observations and predictions models have been developed, applied and evaluated, e.g., (Walters or simulations from mass balance models. First, we introduce the et al., 2011). Kim et al. (2016) compiled a “multi-box” model similar concept of an ‘ideal’ food web, using the word ‘ideal’ in the same to that developed here but addressing the effect of spatial differ- sense as used in the Ideal Gas Law. That law is a relatively simple ences in water and organism concentrations. They also conducted mathematical expression of the behaviour of atoms or molecules in an uncertainty and sensitivity analysis. a dilute gaseous state that explains the effects of pressure, tem- The objective of this study is to explore some of the recognized perature, density and related transport phenomena such as diffu- uncertainties and dependencies in field BMFs and TMFs with a view sion using a model consistent with kinetic theory. It is to ensuring that BMFs and TMFs are applied rigorously in both acknowledged that the model fails under many conditions such as scientific and regulatory contexts. To accomplish this we apply high density or temperature, but it provides a sound basis for simple, well-accepted evaluative mass balance models to illustrate introducing correction factors such as compressibility. It explains in factors that influence biomagnification (specifically BMFs and broad terms the dominant phenomena using intuitively satisfying TMFs), including a chemical's octanolewater partition coefficient assumptions or concepts. By analogy, in this case the aim is to (KOW), biotransformation rates, the sediment/water fugacity ratio demonstrate the factors that influence BMFs and TMFs in ideal food D. Mackay et al. / Chemosphere 154 (2016) 99e108 101 webs. with no link to sediment, with phytoplankton at the base and with The concept of ideal or evaluative food webs is not novel. The simple defined predatoreprey relationships; species 2 having an KOW based Aquatic BioAccumulation Model (KABAM) model of exclusive diet of species 1, species 3 having an exclusive diet of 2, Gerber (2009) employs a food web comprising 6 aquatic species etc. The phytoplankton compartment is assigned, by definition, a TL and two terrestrial species. It is designed to enable the US EPA of 1 and TLs increase by integers of 1.0 to TL of 5 for trout. The user Office of Pesticide Programs to estimate pesticide concentrations in can address more complex omnivorous food webs by assigning tissues that may be consumed by humans. It contains reasonable dietary preferences to organisms 2 to 5 in the form of a feeding estimates of partitioning and kinetic parameters and uses mass matrix, an example being given in Fig. 1B. Fig. 1C is the benthic or balance equations selected from the generally accepted AQUAWEB demersal food web in which the user defines a TL such as 1.2 (or 1.5) model of Arnot and Gobas (2004). for the detritus reflecting its origin from living organisms. The other The ideal food chains and webs employed here and illustrated in species (amphipod, mysis and trout) have respective TL increases of Fig. 1 assume a sediment containing organic detritus (organic car- 1.0. Fig. 1D is a combination of the pelagic food chain and the bon) overlaid by a water column with both phases having constant, demersal food web and is regarded as being the most general case. homogeneous, defined concentrations of the subject chemical. The By defining dietary preferences the combined diet concentra- model aquatic ecosystem is populated by up to seven species as tion is deduced and the resulting predator composition and TL is listed in Table 1. In this context, ‘species’ should be interpreted as a calculated. Each dietary input automatically introduces a TL in- loose grouping of actual organisms with generally similar charac- crease of 1 but when there are multiple food sources the dietary teristics to the named organism. To include all actual species and uptake can be weighted by fractional dietary preferences resulting especially all life stages of the species, would result in a more in non-integer TLs. The models are thus simple but flexible and are complex model yielding results that are more difficult to interpret. regarded as being consistent with the original formulation of tro- It is suggested that since pelagic food webs may be fundamentally phic positions by Vander Zanden and Rasmussen (1996). Steady different in character from demersal or benthic food webs there is a state but non-equilibrium conditions are assumed unless stated need to address each separately but also in combination. otherwise, thus the uptake and loss rates and concentrations can be The structures of the food webs are given in Fig. 1. The simplest deduced algebraically. Clearly for regulatory purposes it is essential case in Fig. 1A, is a linear pelagic food chain or web of 5 organisms, to test the model assertions against empirical data for a defined ecosystem.

7 Lake Trout 2.3. Assignment of aquatic system properties 7 Lake Trout (TL=4.25) (TL=5) The water is assigned an arbitrary freely dissolved concentration 3 Smelt C of 1 mg/m (1 mg/kg) that is selected to be well below the sol- 6 W 6 Smelt (TL=3.5) ubility limit of the chemical. An arbitrary Z-value has been chosen (TL=4) in order to give a water fugacity of 1 Pa (log fugacity ¼ 0) enabling Mysis fugacities in organisms at higher TLs to have higher calculated but 3 3 Mysis (TL=3) arbitrary fugacities, thus revealing their equilibrium status relative (TL=3) to each other. No estimate is made of reduced bioavailability from Zooplankton real waters containing dissolved and particulate matter, i.e. 2 2 Zooplankton (TL=2) bioavailability is 100%. To allow for non-equilibrium between water (TL=2) and sediment, the sediment pore water dissolved concentration Phytoplankton CPW is defined as a multiple of the water concentration. A multiple 1 1 Phytoplankton (TL=1) of 1.0 implies sediment-water equilibrium. The concentration in (TL=1) 1B) Non-linear pelagic omnivorous food web the sediment solids CSS is calculated as 0.35CPWOCKOW mg/kg 1A) Linear pelagic food chain assuming dietary preference is 50% when where OC is the organic carbon content selected as 0.04 g/g, i.e. 4%, consuming two different organisms the coefficient 0.35 L/kg is the suggested octanol-equivalent sorp- 7 Lake Trout tion capacity of OC (Seth et al., 1999). The volumetric concentration (TL~4.3) in the sediment can be calculated from the sediment bulk density as can the corresponding biota-sediment accumulation factor Lake Trout Smelt (BSAF). The chemical's KOW and molar mass are defined by the user. 7 6 (TL~3.6) (TL=4.5) In most of the following cases the chemical's log KOW is varied from 5.0 to 8.0, usually in increments of 1.0. A temperature of 15 Cis Sculpin Mysis Sculpin assumed unless otherwise specified. 5 (TL=3.5) 3 (TL~2.5) (TL~3.5) 5 2.4. Assignment of organism properties

Invertebrates Zooplankton Invertebrates 4 2 (TL=2.5) (TL=2) (TL=2.5) 4 Each organism is assigned a mass and lipid content assuming lipid and n-octanol to have equivalent sorption properties, which

Detritus Phytoplankton Detritus may be a weak assumption for some chemical classes (Seston et al., 0 1 (TL=1.5) (TL=1) (TL=1.5) 0 2014). The phytoplankton compartment is assumed in equilibrium (equi-fugacity) with the water column. Similarly the sediment pore 1C) Linear demersal (benthic) 1D) Combined branched pelagic and demersal food web food chain water and detritus are in equilibrium but have a user-assigned TL such as 1.2 (or 1.5); greater than that of the primary producer. The Fig. 1. Suggested evaluative food webs. A) Linear pelagic food chain, B) Non-linear amphipod with a diet of detritus thus has a TL of 2.2 (or 2.5). For all pelagic omnivorous food web assuming dietary preference is 50% when consuming other species typical ﬁrst order rate constants for respiratory (k ;L/ two different organisms, C) Linear demersal (benthic) food chain, and D) Combined 1 branched pelagic and demersal food web. Numbers in black circles relate to the species kg/d) and dietary uptake (kD; kg/kg/d) and losses by respiration (k2; 1 1 1 listed in Table 1. d ), egestion (kE;d ), biotransformation or metabolism (kM;d ) 102 D. Mackay et al. / Chemosphere 154 (2016) 99e108

Table 1

Relationships between food web organisms and aquatic system properties. The ticks indicate the presence of this compartment or species in the speciﬁed model. fi is a source fugacity.

Compartment or species Name Fugacity source or dependence Pelagic (P) Demersal (D) Pelagic plus demersal (P þ D)

Pelagic water (w) fw user deﬁned ✓✓ ✓ Sediment pore water (pw) fpw deﬁned as multiple of fw ✓✓ 0 Sediment detritus (OC) Equal to fpw ✓✓ 1 Phytoplankton f1 Equal to fw ✓✓ 2 Zooplankton f2 Function of fw and f1 ✓✓ 3 Mysis f3 Function of fw and f2 ✓✓ 4 Benthic invertebrate f4 Function of foc ✓✓ 5 Sculpin f5 Function of fw and f0 ✓✓ 6 Smelt f6 Function of fw and f3 ✓✓ 7 Trout f7 Function of fw,f5 and f6 ✓✓ ✓

1 and growth dilution (kG;d ), are assigned, in most cases using the implications of changes in chemical properties (primarily KOW and correlations for these parameters suggested by Gobas et al. (1988) kM), dietary preferences and effects of sediment-water dis-equi- and in the AQUAWEB model of Arnot and Gobas (2004). It is then librium. Quantifying biotransformation rates to a series of organ- possible to deduce all concentrations and fluxes using the equa- isms is problematic because of obvious dependences on species, life tions described by Mackay et al. (2013) reflecting the assumptions stage, acclimation temperature and internal blood flow limitations. inherent in the model equations. Table 2 lists the rate constant and The conventional application of the MichaeliseMenten equation uptake efficiency correlations. These rate constants are tempera- can result in non-first order kinetics at high chemical concentra- ture dependent and are judged to be typical of conditions at 15 C. tions. In the interests of simplicity, a single biotransformation rate For the present purposes, in the interest of simplicity and to constant is applied to all species for which a mass balance is facilitate interpretation, all lipid contents are 5% by volume, i.e. L is calculated; however, it is preferable to assign species-specific and 0.05. This is clearly an overly simplistic and unrealistic assumption mass-specific rate constants. Similar considerations apply to scaling because lipid contents generally rise from typically 1% for plankton the growth-dilution rate constant. Again in the interests of to as high as 20% for trout. BMFs and TMFs are sensitive to the simplicity we apply the Gewurtz correlation as applied by Gewurtz relative lipid contents of prey and predator, especially if concen- et al. (2006) and Arnot et al. (2008) to address the effect of or- trations are lipid normalized. By assigning a common lipid content ganism size and temperature. The term Q, the ratio of kD to kE is of 5% we circumvent this issue. Clearly for simulations of real sys- assigned a value of 7, but values in the range 3e15 may be possible tems actual lipid contents should be used. Compartment (Gobas et al., 1999), (Kelly et al., 2004), (Arnot and Quinn, 2015). 0 (detritus) and 1 (phytoplankton) are assumed to be in equilibrium Higher values of Q (also may be referred to as BMFmax) indicate a with the sediment and water respectively, thus no diet need be slower rate of egestion relative to the intake from the diet and defined. The pelagic species include species 2 (a zooplankton or possibly more efficient utilization of the diet. This parameter is small fish or invertebrate of mass ~7.85E-5 g), species 3 (a small fish important since it defines the maximum possible BMF. An addi- or crustacean such as a mysis of mass ~6.5E-2 g), species 6 (a me- tional feature is that while it is possible to use a single concentra- dium fish such as smelt of mass ~10 g) and species 7 (a larger fish tion in the water column to calculate respiratory uptake rates, it is such as lake trout of mass ~3000 g). Using the correlations listed in possible to assign different water and diet concentrations to each Table 2 species-specific estimates are made of rate constants for species to simulate the effects of vertical and horizontal hetero- respiratory uptake and loss, dietary uptake, egestion losses (a factor geneity in concentrations caused by stratification or proximity to Q less than the dietary uptake rate constant), biotransformation chemical input sources. It must be emphasised that the TMFs losses and growth dilution. It is emphasised that no attempt is calculated by the model are sensitive to these assumed scaling made to simulate real food webs, rather the aim is to explore the factors. When assessing a specific chemical in a specific ecosystem

Table 2

Summary of uptake and clearance process rate constants for organisms of mass WB (kg). In this table EW is the fractional chemical transfer efﬁciency at the gill, GV is gill ventilation rate, COX is oxygen concentration in the water, L is lipid fraction, ED is the chemical dietary uptake efﬁciency, GD is the feeding rate, Q is the ratio of kD/kE,CW is water concentration and CD is diet concentration. T is temperature.

Process Equation Units

k1 ¼ uptake from water through gills k1 ¼ EWGV/WB k1 ¼ L/kg/d k1,EW, and GV equations from Arnot and Gobas (2004) EW ¼ 1/(1.85 þ 155/KOW) EW ¼ Fractional 0.65 GV ¼ 1400WB /COX GV ¼ L/day (COX ¼ 8 mg/L) WB ¼ kg 1 k2 ¼ loss by respiration k2 ¼ k1/BCFE k2 ¼ d k2 and BCF equations from Gewurtz et al. (2006) BCFE ¼ L*KOW L ¼ Fractional kD ¼ uptake from diet kD ¼ EDGD/WB kD ¼ kg/kg/d 8 kD and GD equations from Arnot and Gobas (2004),ED from Gobas et al. (1988) ED ¼ 1/(2.3 þ 5.3 10 KOW) ED ¼ Fractional 0.85 (0.06T) GD ¼ 0.022WB e GD ¼ kg/day T ¼ C 1 kE ¼ loss by egestion kE ¼ kD/Q kE ¼ d (Q ¼ 7) Q ¼ Unitless ratio [fraction of lipid absorbed ~ 1-(1/Q)] 1 kM ¼ loss by biotransformation Assigned by user kM ¼ d T20 0.2 1 kG ¼ growth dilution rate kG ¼ 0.00586(1.113) (1000WB) kG ¼ d kG equation from Gewurtz et al. (2006) Steady-state mass balance equation from Mackay (2001) CF ¼ (CWk1 þ CDkD)/(k2 þ kE þ kM þ kG)CF ¼ mg/kg D. Mackay et al. / Chemosphere 154 (2016) 99e108 103 parameter values should be selected appropriate to the conditions Q=13 being simulated. 6 Calculations are done in an Excel spreadsheet with separate Q=11 5 sheets for each model, a sample output being included in the Supplementary Information (SI) with the model being available Walters et al. 4 from the web site www.trentu.ca/cprg. Outputs include the per- data TMF Q=9 centage of chemical input by diet for each organism and the TL. In 3 addition to the concentration-based calculations, fugacity-based Q=7 calculations are also done in parallel giving identical results. This 2 requires input of a Z-value for the chemical in water, but this can be Q=5 selected arbitrarily, since it affects only the absolute fugacities and 1 not their relative values. An advantage of the fugacity format is that 5 5.5 6 6.5 7 7.5 Q=3 fugacities in organisms are readily compared with each other and Log Kow with fugacities in water and sediments providing direct informa- Fig. 3. Plots of estimated TMFs as a function of log KOW and Q with growth dilution but tion on their relative equilibrium status for the chemical in the no biotransformation. The temperature is 20 C in this scenario. The slope of the linear deﬁned system. Fugacities are used directly to calculate TMFs as regression obtained by Walters et al. (2011) for PCBs in Lake Hartwell is shown. The described earlier as the slope of the plot of log fugacity vs trophic slopes are as follows: for the Walters regression the slope is 1.60, for Q ¼ 13 is 2.12, for level. Q ¼ 11 is 1.91, for Q ¼ 9 is 1.65, Q ¼ 7 is 1.3, for Q ¼ 5 is 0.97, and for Q ¼ 3 is 0.51.

3. Results 3 No metabolism; 2.5 TMF=4.12 3.1. Effect of hydrophobicity (KOW) on trophic magniﬁcation Half-life=120 2 days;

If TMFs are plotted for a range of K , these plots show the effect ) OW a TMF=2.72 P of hydrophobicity and biomagniﬁcation. Increasing log fugacities ( 1.5 y Half-life=60 t i indicate biomagniﬁcation with each BMF being the antilog of the c days; a

g 1 increase in log fugacity. All simulations employ the pelagic food u TMF=2.12 f g

web except for the sediment/water ratio simulation, which em- o Half-life=30

L 0.5 ploys the demersal food web. Fig. 2 illustrates the results of the days; simple linear pelagic food web illustrated in Fig. 1A with plankton 0 TMF=1.51 (species 1) as TL1 and as the sole dietary source to species 2 in a 12345 homogeneous water column with species 3 consuming only species Trophic Level 2 and similarly species 6 consumes only species 3 and species 7 Fig. 4. Effect of biotransformation half-lives for trophic magnification for a chemical of consumes only species 6. No biotransformation is included in any of log KOW of 7 with a range of half-lives as well as for a chemical that is not metabolized. these modelled simulations, except for in Fig. 4. The TMFs are then obtained from the slope of these lines as outlined earlier. The lines in Fig. 2 are curved indicating non-first order behav- TL 5, giving an average BMF and TMF of 1.64. Because the TL in- iour as is expected because of the structure of the uptake equations creases by 1.0 from a given prey to predator in this scenario, BMFs and especially the contribution of growth dilution that varies non- and TMFs are equal. The average BMF can be calculated as the linearly with species mass and TL. At a log KOW of 5 and below, the average of the individual prey to predator BMFs calculated as in- simulation indicates only slight biomagnification. Both the BCF and creases in fugacity or equivalently using the increase in lipid nor- BAF are approximately equilibrium values with respect to the water malised concentration. At log KOW of 6 as dietary uptake becomes and equal to the product of lipid content and KOW. In this region of more important there is appreciable biomagnification, an average relatively low log KOW, organism fugacities are largely controlled by BMF of 2.32 and with species fugacities increasing to 29 Pa for TL 5. respiratory exchange with the water and the rates of uptake from At log KOW of 7 fugacities increase to 288 Pa for TL5 with a TMF of the diet and loss by egestion are relatively unimportant. For log KOW 4.12. At log KOW of 8 the fugacity increases to 589 Pa and has a TMF of 5.5 the species fugacities increase from 1.0 Pa for TL 1e7.3 Pa for of 4.93. In this region, the BMF approaches 7, namely Q, the assumed ratio of the dietary uptake and egestion rate constants. In general for very hydrophobic and persistent substances with zero 3 Log Kow=8, growth at a TL of n, the species fugacity approaches Q(n 1), i.e. when TMF=4.93 Q ¼ 7, a food web biomagnification factor of 7n 1 or 2401 for an 2.5 Log Kow=7, TMF=4.12 organism with a TL of 5. The transition into a region of high BMFs is 2 Log Kow=6.5, attributable to the increasing importance of dietary uptake and the )

a TMF=3.24 P decreasing rate of respiratory losses. For TL 2, at a log KOW of ( y

t Log Kow=6,

i 1.5 approximately 5 there are equal contributions of chemical uptake c TMF=2.32 a

g from respiration and diet, i.e. when k1CW equals kDCD, with CD u f 1 Log Kow=5.5, $ g being approximately L KOW.

o TMF=1.64 L These trends are consistent with the field data of Kidd et al. 0.5 Log Kow=5, TMF=1.26 (1998), Walters et al. (2011), Fisk et al. (2001), and others. Fig. 2 also shows that the lines are slightly curved thus assuming line- 0 Log Kow=4, 12345TMF=1.03 arity when regressing can introduce a slight error. The non- Trophic Level linearity arises from differences in dietary uptake at higher levels of the food web and in losses by respiration which tends to reduce fi Fig. 2. Effect of KOW on trophic magni cation assuming Q of 7.0 and a linear food web the BMF as discussed by Mackay et al. (2013). Higher in the food with no biotransformation but including growth dilution. 104 D. Mackay et al. / Chemosphere 154 (2016) 99e108 chain, fish concentrations are higher and respiratory losses can 3 become more important, possibly causing a drop in BMF and a S/W=30; 2.5 lower slope. This is usually insignificant when interpreting field TMF=3.08 data for hydrophobic substances, but it can become significant for 2

) S/W=10; a less hydrophobic substances with log KOW values in the range P 1.5 TMF=3.11 (

e y 5.5 6.5. t i

c 1

a S/W=3;

A very comprehensive set of data on TMFs are those of Walters g

u TMF=3.20 et al. (2011) for PCBs in 21 ﬁsh species in Lake Hartwell, South f 0.5 g o

Carolina. Concentrations were measured for 127 congeners and TLs L 0 S/W=1; for fish were determined from stable nitrogen isotope measure- TMF=3.46 ments using conventional techniques, the TLs ranging from 2.5 to -0.5 4.0. TMFs which ranged from about 1.5 to 6.5 were calculated and 12345 2 Trophic Level plotted as a function of log KOW yielding a slope of 1.88 and an r of 0.80, demonstrating the convincing relationship between these Fig. 5. Effect of sediment/water fugacity ratios of 1, 3, 10 and 30 on fugacities in the quantities. Walters et al. (2011) also applied the AQUAWEB model demersal food web as shown in Fig. 1C. giving a generally similar relationship. Fig. 3 gives estimated TMFs for Q values of 3, 5, 7, 9, 11 and 13 with growth included but no biotransformation, enabling comparison with the Walters et al. increasing all the species fugacities and concentrations. All species (2011) data, which has also been plotted in Fig. 3. There is thus encounter the same fugacity in water, thus species higher in the compelling monitoring and model evidence that log KOW as a sur- food web at higher fugacities will consume a diet of higher fugacity rogate for lipidewater partitioning is the key determinant of TMF. A than that of the respired water and now respire water of lower Q of 7 or 9 seems to give a satisfactory fit to the slope. fugacity than the diet. Respiratory exchange between the fish and water can thus result in a net loss of chemical because the fish 3.2. Effect of chemical biotransformation (metabolism) fugacity exceeds that of the water. Their concentrations at high TLs are then reduced as a result of these increased respiratory losses. Fig. 4 illustrates the influence of biotransformation by intro- This is most pronounced for chemicals that are taken up from both ducing varying half-lives (which are due to various rates of diet and water. At high KOW values respiration becomes less biotransformation) for a chemical of log KOW of 7 in the linear important as an uptake and loss process. This demonstrates that pelagic food web (i.e., 1A in Fig. 1) and shows the profound effect of sediment/water fugacity ratio (fS/fW) can be very important as a biotransformation which reduces BMFs and TMFs. The assumed confounding factor in food web biomagnification if there is benthic rate constants correspond to whole body biotransformation half- coupling. At higher KOW values the sediment fugacity dominates lives of 30e120 d and are selected for illustrative purposes. To dietary concentrations in the food web. The implication is that have a significant effect, the biotransformation rate constant (kM) TMFs obtained from monitoring data are sensitive to fS/fW but this must approach or exceed the rate constant for total loss by other sensitivity varies depending on KOW and the food web. mechanisms. For chemical of log KOW 7 the sum of the non- Simple linear regression using all species can be misleading biotransformation rate constants for the representative trout (i.e. because of possible severe non-linearity of the log fugacity vs. TL 1 k2 þ kE þ kG) is 0.0021 d corresponding to a half-life of 324 days. relationships. Because of the uncertainty in the fugacity corre- Clearly, biotransformation profoundly reduces concentrations, fu- sponding to TL1 (especially in detritus and demersal species) it may gacities, BMFs, and TMFs, especially in this case if the biotransfor- be preferable to determine the TMF using only pelagic species. mation half-life is less than 324 days. High biotransformation rates Recently McLeod et al. (2014) used a model to calculate concen- can result in TMFs less than 1.0. For example, if the biotransfor- trations and interpret monitored PCB data from the Detroit River. mation half-life is 10 days the TMF is less than 1.0 in this evaluative They showed that TMFs were highly variable and had a variable model. Growth dilution has a similar algebraic effect as biotrans- dependence on KOW. A possible explanation is that the sediment/ formation because the rate constants kM and kG appear additively water fugacity ratios varied greatly, indeed these ratios were stated in the mass balance equation. Plots such as Fig. 4 can thus apply to to vary by factors as large as 211. growth dilution or to the sum of biotransformation and growth dilution for specified half-lives. The profound effects of growth and 3.4. Effect of omnivorous feeding biotransformation rates on TMFs are likely to be species- and ecosystem-specific because of variability in biotransformation ca- The effect of food web omnivory is illustrated in Fig. 6 and shows pabilities, lipid contents, temperature, feeding, and egestion rates the results for a non-linear pelagic food web depicted in Fig.1B with and rates of growth and respiration. plankton as TL1 in a homogeneous water column with no biotransformation. Species 4 has equal diets of species 1 and 2, and 3.3. Effect of sediment/water fugacity ratio species 5 has a 50% diet of species 4 and 25% diet of each species 2 and 3. There is thus appreciable omnivory of the food web and In Fig. 5, the effect of the sediment to water fugacity ratio is BMFs become less straight-forward. There is a considerable explored using the demersal food web with detritus set at a TL of reduction in the TLs at higher levels in the food web because the 1.5. There is no biotransformation in these simulations. Gobas and predator is consuming dietary items with lower concentrations and MacLean (2003) have described the factors, notably organic matter fugacities from organisms lower in the food web than in the linear mineralisation (diagenesis) that can contribute to high sediment/ case. In general, TLs become non-integers. The result is a lower water fugacity ratios. The sediment to water fugacity ratio is varied fugacity at the top of the food web. This is shown in Fig. 6, for a illustratively from 1 to 3 to 10 and to 30 for a chemical of log KOW of linear food web and a chemical with log KOW of 7. In this scenario 6. The TMFs are calculated only using the fugacities at TL 2 to 5, the fugacity of species 7 is 287 and the TMF is 4.17 (Fig. 6A), but the because as shown, the slopes become highly variable and uncertain fugacity drops to 109 as a result of omnivory and consumption of at low TLs. less contaminated organisms and the TMF increases slightly to 4.64, The higher fugacity in the sediment propagates up the food web (Fig. 6B). The estimated TL of species 7 falls from 5 to 4. The TMFs D. Mackay et al. / Chemosphere 154 (2016) 99e108 105

3 2.5 2.5 2.5 TMF= 4.17 2 TMF= 4.64 2 TMF= 4.29

) 2 a

P 1.5 1.5 ( 1.5 y t i

c 1 1

a 1 g

u 0.5 0.5 f 0.5 g o

L 0 0 0 -0.5 Log fugacity (Pa) -0.5 Log fugacity (Pa) -0.5 12345 12345 12345 Trophic Level Trophic Level Trophic Level ABC

Fig. 6. Effects of food web omnivory on TMF. (A) is an example where no branching occurs (TMF ¼ 4.17), (B) is an example of a food web where extensive omnivory occurs in multiple species in the food chain (TMF ¼ 4.64), (C) is an example where intermediary omnivory occurs (TMF ¼ 4.29). All graphs are for a chemical with log KOW of 7 with no biotransformation. are, however, similar because both the BMFs and TLs are reduced. In Fig. 7 gives the log fugacity vs TL lines for both high and low the intermediate case shown in Fig. 6C the TMF is 4.29. concentration zones. The intercepts differ by a factor F of 10, but the It is at first surprising, and even counter-intuitive, that omnivory slopes and TMFs are identical at 4.17. has such a small effect on TMF, because it implies that TMFs are The model can be used to explore how the concentrations in largely independent of the complexity of the food web. Closer ex- mobile species are affected by partial feeding from each zone. A amination of the uptake equation shows that BMFs are, to a first fractional probability P can be defined for each species describing approximation, dependent on the average diet concentration and the fraction of the time that a zone L fish feeds from zone H. The the individual dietary sources are important only in affecting the average diet concentration increases from the zone L value CL to a average. This constant BMF results in a constant TMF which is higher value for CD because of the consumption of higher concen- essentially an average of the BMFs. An implication is that because tration food from zone H, i.e. CH is FCL. It can be shown that the diet stable N isotope measurements automatically take omnivory into concentration is given by equation (2) and the factor increase in account there is little need for detailed knowledge of the food web. diet concentration (M) is given by equation (3). What matters most is the average number of digestive events that ¼ $ þð Þ$ the chemical has experienced, not the details of the history of these CD P CH 1 P CL (2) digestive events. Omnivory is likely to have a significant effect if BMFs vary between species pairs and if species lipid contents and M ¼ CD=CL ¼ 1 þ P$ðF 1Þ (3) assimilative efficiencies differ. It should be noted that although the Obviously, when F is 1.0 or P is zero M is 1.0 and there is no TMFs increase with more extensive omnivory, the chemical con- increase in concentration. When P is 100%, M equals F and the diet centration is greater in the highest TL species in the non- is exclusively from zone H. omnivorous scenario than in the omnivorous scenario. As omni- In this hypothetical example we select F as 10 and vary P for vory occurs this causes the TLs to be reduced and the overall slope species 6 from 0.10 (M ¼ 1.9) to 0.25 (M ¼ 3.25) and 0.50 (M ¼ 5.5). increases. The log fugacity lines are then intermediate between the L and H lines. The slopes obviously increase as do the TMFs and exceed the 3.5. Effect of spatial variability in diet concentration single diet values. Similar results are obtained if other species are also mobile. The increase in TMF is not attributable to an increase in BMF, it is a result of more mobile fish near the top of the food web To illustrate this effect we consider a chemical of log KOW 7ina linear pelagic food web in a water body that has two concentration having a larger spatial range and being exposed to higher diet zones but identical food webs. In zone H (high) the water and concentrations. This effect can also result in the opposite effect of fi sediment concentrations are high relative to zone L (low) the ratio zone H sh displaying lower TMFs from consumption of less of high/low concentrations being F. The high zone could be the contaminated diets. This issue of spatial variability in concentra- result of local point source contamination such as a waste water tions in water and organisms is addressed in more detail and with treatment plant effluent. greater rigour as in the study by Kim et al. (2016). The authors also undertook a sensitivity and uncertainty analysis on the effects of spatial variability. 3 A simple numerical example and a flow diagram in Fig. 8 also Cw=10; TMF=4.17 demonstrate this effect. Consider a simple linear food web of three 2.5 fish A, B and C residing in zone L of relatively low concentrations. p=50%; ) An adjacent zone H has higher concentrations by a factor of 3. Fish C a 2 TMF=22.96 P

( consumes B, which in turn consumes A. The concentration of ﬁsh A y t i 1.5 in zone L is 10 units and in zone H it is 30. The BMFs are 2.0 so the c p=25%; a

g TMF=13.57 concentrations in ﬁsh B and C in zone L are 20 and 40 respectively. u

f 1

g In zone H the concentrations of B and C are 60 and 120. If B changes

o p=10%; L 0.5 TMF=7.93 its diet to consume 90% of A from zone L and 10% from zone H, its diet concentration (B) is now 0.9 10 þ 0.1 30 or 12, an increase 0 Cw=1; by a factor M of 1.2. Fish B will then have a concentration of 24. 12345TMF=4.17 Similarly, fish C may experience a more concentrated diet by a Trophic Level (TL) factor of 1.2 and, taking into account biomagnification, its concentration increases to 48. Fig. 7. Illustrated effect of spatial concentration variability on TMF with P expressed as In the first case of P of 10% a plot of log fugacity vs TL gives a a percentage. TMFs are calculated between species and TLs 4 and 5 only. P is the 0.301 percentage that a species will feed from the highly concentration zone, CW is the slope of 0.301 and TMF is 10 or 2.0, equal to the BMFs. In the concentration of chemical in the water. second case the slope is 0.38 corresponding to a TMF of 2.4, despite 106 D. Mackay et al. / Chemosphere 154 (2016) 99e108

Immobile Fish Mobile Fish Immobile Fish

40 48 120

10% of diet Uptake=20 90% of diet Uptake=60 Uptake=6 BMF=2 Uptake=18 BMF=2 BMF=2 BMF=2

20 24 60

Uptake=10 90% of diet 10% of diet Uptake=30 BMF=2 Uptake=9 Uptake=3 BMF=2 BMF=2 BMF=2

10 30

ZONE L: Low water C ZONE H: High water C

C=CL C=3*CL

Fig. 8. A mass balance flow diagram for mobile fish that have diets of 90% fish in the low concentration zone and 10% of fish from the high concentration zone. The concentrations in each fish are given as values on each fish diagram and the uptake from the diet is listed beside each arrow. In this scenario P ¼ 0.1 (i.e. 10% feeding from high concentration zone). the BMFs remaining constant at 2.0. This simple numerical example which can be dependent on temperature, on the species present, shows that TMFs can reflect not only biomagnification but also and both the abundance and the nature of the food. It would be changes in dietary location and concentration. TMFs can also be interesting to compare measured TMFs at different temperatures influenced by selection of sampling locations. The TMF in the sec- and latitudes but this is beyond our scope here. The implication is ond case results from an unchanged BMF of 2.0 and a diet con- that TMFs depend on four rate constants for each predator species, centration increase of a factor of 1.2, the product being 2.4. The BMF all of which are potentially important and can be ecosystem- of 2.0 is controlled by the chemical properties and physiological specific. It is thus not surprising that the same chemical may properties of the fish and its diet. The factor M of 1.2 is controlled by exhibit different TMFs in different ecosystems, especially if there either the spatial differences in the diet or sampling locations. If the are spatial differences in feeding patterns from the water column probability P increases to 50%, M equals 2 and the concentrations in and sediments and if gradients exist in water and sediment the mobile fish double. The observed TMF is the product of these concentrations. two quantities. Clearly, spatial variations in concentration and The model used here is steady-state in nature. Although not sampling locations can profoundly affect empirical TMFs, despite addressed here, differences in half-time to steady-state between BMFs being relatively constant. Kim et al. (2016) have discussed this species may also influence the measured TMFs. This is 0.693 issue in more detail. divided by the total loss rate constant and can be very long and could even exceed the lifetime of the fish, in which case a steady- state model is invalid. Food webs may differ in the extent to 4. Discussion which they are approaching steady-state conditions thus TMFs may also differ despite some species having a bioaccumulation “kick- fl 4.1. Factors in uencing TMF start” at birth and possibly faster uptake when young and experi- encing different temperatures. We speculate that this effect may be This study, although hypothetical in nature and applied to especially important for hydrophobic and super-hydrophobic evaluative food webs, employs well-accepted uptake equations and chemicals that may have long half-lives in larger fish and present suggests that BMFs and TMFs are sensitive to KOW and biotrans- both monitoring and modelling challenges. The lipid contents for formation half-life. TMFs can also be dependent on sediment/water each species in the model can be varied by the user, but for these fugacity ratios and spatial variability in dietary sources or sampling. simulations a lipid content of 5% was used for each species. As Variation in lipid contents throughout the food web may also affect previously mentioned this is a simplification made in order to show fi dietary uptake rates and thus BMFs. To a rst approximation, TMFs the effects of each variable studied. A sensitivity analysis on the are independent of the structure of the food web, thus for evalua- effect of lipid content on TMF using a chemical with log KOW of 6 tive purposes simple linear or non-linear food webs may be showed that increasing lipid content increases the estimated TMF, adequate. indicating that accurate lipid contents are an important factor in For a hydrophobic chemical in a linear food web of n species and determining TMFs. (n 1) BMFs with equal species lipid contents In support of the contention that TMFs can be highly variable, X X two chemical-specific analyses of variation in TMFs in ecosystems TMF BMF=ðn 1Þ ðk =ðk þ k þ k þ k ÞÞ=ðn 1Þ D 2 E M G are of note. Franklin (2016) has presented a comprehensive review and analysis of 24 studies of BMFs and TMFs for poly-fluorinated The rate constants apply to the predator species. Large values of substances, showing that BMF values range over orders of magni- fi TMF thus result from large values of kD, or more speci cally kD/kE tude. For TMFs, the factor of variation is about 20. Second are re- (i.e. Q) that characterises the maximum gastrointestinal bio- ports of TMFs of the very hydrophobic, volatile and cyclic methyl fi magni cation factor. For persistent hydrophobic chemicals in slow- siloxane D5 in growing organisms. D5 is known to be subject to fi growing sh, the BMF approaches Q because the other rate con- biotransformation. Relatively high TMFs have been reported by stants are relatively small. Lower TMF values can occur when Borga et al. (2012a,b) in Lake Mjosa (1.22e4.29) and by Jia et al. biotransformation and faster growth rates are applied, both of D. Mackay et al. / Chemosphere 154 (2016) 99e108 107

(2015) in the Northern China marine environment (1.77) compared biomagnification and resulting exposures. Ultimately, these expo- to lower values reported by Powell et al. (2009, 2010) in Lake Pepin sures are major contributors to risk. (0.1e0.2) and Oslo Fjord (0.3e0.4). These inconsistent TMFs are likely due to spatial variability in proximity to effluents to sources 5. Conclusions of water and sediments, water columns, stratification and variation in differences in chemical fate processes (such as volatilization and It is hoped that the concepts outlined above may be useful for sediment deposition), feeding patterns, as well as differences in explaining the variability in reported TMFs, as a justification for species including biotransformation rates, lipid contents, food webs measuring TMFs, for designing TMF measurement programs, and and sediment-water fugacity ratios. Interpretation of these obser- ultimately for including the effects of trophic magnification in vations thus requires a detailed understanding of the ecosystem in chemical hazard and risk assessment. A useful path forward would question and preferably the application of mass balance models be to develop fairly simple evaluative food web models similar in calibrated for the local hydrodynamics and the species in that principle to those developed here and to KABAM that could be ecosystem, and their predatoreprey interactions. validated for one or more specific ecosystems and a range of chemicals. Models could then play a role in evaluation of specific 4.2. Possible regulatory implications for assessing exposure and risk chemicals by demonstrating the extent to which it is likely to experience trophic magnification as a function of its hydropho- This study suggests that two groups of factors control the con- bicity and potential for biotransformation. centrations in upper trophic level organisms and thus the risk to these organisms and those that consume them. Most fundamental Acknowledgements is the level of contamination in the water and sediment. This factor controls the lipid based concentration at trophic level 1, or equiv- The authors gratefully acknowledge the critical comments alently the fugacity at trophic level 1. Second is biomagnification as provided by Dr. Charles Staples. D.M. and A.C. thank Dow Corning controlled by KOW, dietary preferences, rate constants for uptake Corporation and DMER for financial support. J.A. thanks the Euro- and loss in all species and thus the TMF and the length of the food pean Chemical Industry Council (Cefic) (ECO.15) Long-Range web. This results in an increase in fugacity from f1 at trophic level 1 Research Initiative (LRI) program for research funding. Open access (n1) (n1) by a factor TMF to fn or f1 TMF at trophic level n. Alter- publication of this paper was kindly provided by CES e Silicones natively, if the average BMF for pairs of predatoreprey organisms is Europe (CES), a non-profit sector group of Cefic. BMFAV then the ratio of fugacities at the top (TLn) and bottom (TL1) of the food web can be estimated as BMFAV raised to the power Appendix A. Supplementary data (TLneTL1). At the base of the food web f1 can be ambiguous if the sediment and water fugacities differ and it may be preferable to Supplementary data related to this article can be found at http:// determine the TMF using concentrations or fugacities of higher dx.doi.org/10.1016/j.chemosphere.2016.03.048. food web organisms. When comparing exposure or risk from a chemical in two eco- References systems it is the absolute values of fn that control relative risk. It is entirely possible for a high TMF in one system to cause less risk if f1 Arnot, J.A., Gobas, F.A., 2004. A food web bioaccumulation model for organic is lower or if the food web is short i.e. n is low. A large TMF (>1.0) is chemicals in aquatic ecosystems. Environ. Toxicol. Chem. 23, 2343e2355. Arnot, J.A., Gobas, F.A., 2006. A review of bioconcentration factor (BCF) and bio- certainly a cause for concern but it is only one of several factors accumulation factor (BAF) assessments for organic chemicals in aquatic or- influencing risk. When assessing risk, it is thus desirable to ganisms. Environ. Rev. 14, 257e297. compare the values of f1, TMF and n as controlling fn and the cor- Arnot, J.A., Mackay, D., Bonnell, M., 2008. Estimating metabolic biotransformation fi e responding concentrations. When considering a TMF as a ‘line of rates in sh from laboratory data. Environ. Toxicol. Chem. 27, 341 351. Arnot, J.A., Quinn, C.L., 2015. Development and evaluation of a database of dietary evidence’ in chemical assessments it is essential to recognize that bioaccumulation test data for organic chemicals in fish. Environ. Sci. Technol. its value is sensitive to a number of factors but especially the se- 49, 4783e4796. lection of species corresponding to TL1, i.e. f , sediment/water Barber, M.C., 2003. A review and comparison of models for predicting dynamic 1 chemical bioconcentration in fish. Environ. Toxicol. Chem. 22, 1963e1992. fugacity ratios, and spatial variability in dietary sources. Barber, M.C., 2008. Dietary uptake models used for modeling the bioaccumulation In a regulatory context, empirical TMFs can provide invaluable of organic contaminants in fish. Environ. Toxicol. Chem. 27, 755e777. insights into the presence of chemical biomagnification in food Borga, K., Fjeld, E., Kierkegaard, A., McLachlan, M.S., 2012a. Food web accumulation e fl of cyclic siloxanes in Lake Mjosa, Norway. Environ. Sci. Technol. 46, 6347 6354. webs because they re ect actual ecosystem conditions. A reported Borga, K., Kidd, K.A., Muir, D.C.G., Berglund, O., Condor, J.M., Gobas, F.A.P.C., TMF for a chemical in a specific ecosystem should not, however, be Kucklick, J., Malm, O., Powell, D.E., 2012b. Trophic magnification factors: con- regarded as a metric of biomagnification in the same sense as a siderations of ecology, ecosystems, and study design. Integr. Environ. Assess. e fi Manag. 8, 64 84. laboratory-derived BCF, BAF or BMF. Comparisons of eld TMFs Burkhard, L.P., Borga, K., Powell, D.E., Leonards, P., Muir, D.C.G., Parkerton, T.F., with ‘bright line’ or numeric pass/fail criteria as are applied for BCFs Woodburn, K.B., 2013. Improving the quality and scientific understanding of are presently inappropriate and can be misleading. It should, trophic magnification factors (TMFs). Environ. Sci. Technol. 47, 1186e1187. Conder, J.M., et al., 2012. Use of trophic magnification factors and related measures however, be possible to use evaluative food web models to to characterize bioaccumulation potentials of chemicals. Integr. Environ. Assess. demonstrate the potential for high TMFs and explain reported Manag. 8, 85e97. values. TMFs are thus complementary to BMFs that can be obtained Czub, G., McLachlan, M.S., 2004. Bioaccumulation potential of persistent organic e under controlled laboratory feeding conditions. There is a chemicals in humans. Environ. Sci. Technol. 38, 2406 2412. Czub, G., Wania, F., McLachlan, M.S., 2008. Combining long range transport and compelling incentive to develop and apply simple mass balance bioaccumulation considerations to identify potential arctic contaminants. En- models to explain and confirm biomagnification and trophic viron. Sci. Technol. 42, 3704e3709. fl magnification of specific chemicals in evaluative or real food webs Fisk, A.T., Hobson, K.A., Norstrom, R.J., 2001. In uence of chemical and biological factors on trophic transfer of persistent organic pollutants in the northwater provided that estimates are available for organism characteristics polynya marine food web. Environ. Sci. Technol. 35, 732e738. and dietary preferences. Models that are suitably parameterised for Franklin, J., 2016. How reliable are field-derived biomagnification factors and tro- organisms in specific ecosystems can provide complementary in- phic magnification factors as indicators of bioaccumulation potential? Conclu- sions from a case study on per- and polyfluoroalkyl substances. Integr. Environ. sights into chemical behaviour as revealed by TMF measurements Assess. Manag. 12, 6e20. and establish a consistent quantitative picture of the extent of Gerber, K., 2009. User's Guide and Technical Documentation KABAM Version 1.0 108 D. Mackay et al. / Chemosphere 154 (2016) 99e108

(KOW (based) Aquatic Bioaccumulation Model). Report of Environmental Fate McLeod, A., Arnot, J.A., Borga, K., Selck, H., Kashian, D.R., Krause, A., Paterson, G., and Effects Division. Office of pesticide programs. US Environmental Protection Haffnert, G., Drouillard, K.G., 2014. Quantifying uncertainty in the trophic Agency, Washington, DC. magnification factor related to spatial movements of organisms in a food web. Gewurtz, S.B., Laposa, R., Gandhi, N., Christensen, G.N., Evenset, A., Gregor, D., Integr. Environ. Assess. Manag. 11, 306e318. Diamond, M.L., 2006. A comparison of contaminant dynamics in arctic and OECD, 2012. Guideline for the Testing of Chemicals, Section 3. Test No. 305: Bio- temperate fish: a modeling approach. Chemosphere 63, 1328e1341. accumulation in Fish: Aqueous and Dietary Exposure [Online] Available at: Gobas, F.A.P.C., de Wolf, W., Burkhard, L.P., Verbruggen, E., Plotzke, K., 2009. http://www.oecd-ilibrary.org/environment/test-no-305-bioaccumulation-in- Revisiting bioaccumulation criteria for POPs and PBT assessments. Integr. En- fish-aqueous-and-dietary-exposure_9789264185296-en;jsessio- viron. Assess. Manag. 5, 624e637. nid¼5clqwm2q314rn.x-oecd-live-02 (accessed 25.09.15.). Gobas, F.A.P.C., MacLean, L.G., 2003. Sediment-water distribution of organic con- Powell, D.E., Durham, J., Huff, D.W., Bohmer, T., Gerhards, R., Koerner, M., 2010. taminants in aquatic ecosystems: the role of organic carbon mineralization. Bioaccumulation and Trophic Transfer of Cyclic Volatile Methylsiloxanes (cVMS) Environ. Sci. Technol. 37, 735e741. Materials in the Aquatic Marine Food Webs of the Inner and Outer Oslofjord. Gobas, F.A.P.C., Muir, D.C.G., Mackay, D., 1988. Dynamics of dietary bioaccumulation Health and Environmental Sciences, Dow Corning Corporation, Norway, and fecal elimination of hydrophobic organic chemicals in fish. Chemosphere Auburn. 17, 943e962. Powell, D.E., Woodburn, K.B., Drottar, K., Durham, J., Huff, D.W., 2009. Trophic Gobas, F.A.P.C., Wilcockson, J.B., Russell, R.W., Haffner, G.D., 1999. Mechanism of Dilution of Cyclic Volatile Methylsiloxane (cVMS) Materials in a Temperate biomagnification in fish under laboratory and field conditions. Environ. Sci. Freshwater Lake. Health and Environmental Sciences, Dow Corning Corpora- Technol. 33, 133e141. tion, Auburn. Jia, H., Zhang, Z., Wang, C., Hong, W.J., Sun, Y., Li, Y.F., 2015. Trophic transfer of Qiao, P., Gobas, F.A., Farrell, A.P., 2000. Relative contributions of aqueous and dietary methyl siloxanes in the marine food web from coastal areas of Northern China. uptake of hydrophobic chemicals to the body burden in juvenile rainbow trout. Environ. Sci. Technol. 49, 2833e2840. Arch. Environ. Contam. Toxicol. 39, 369e377. Kelly, B.C., Gobas, F.A.P.C., McLachlan, M.S., 2004. Intestinal absorption and bio- Seston, R.M., Powell, D.E., Woodburn, K.B., Kozerski, G.E., Bradley, P.W., magnification or organic contaminants in fish, wildlife, and humans. Environ. Zwiernik, M.J., 2014. Importance of lipid analysis and implications for bio- Toxicol. Chem. 23, 2324e2336. accumulation metrics. Integr. Environ. Assess. Manag. 10, 142e144. Kelly, B.C., Ikonomou, M.G., Blair, J.D., Morin, A.E., Gobas, F.A.P.C., 2007. Food web- Seth, R., Mackay, D., Muncke, J., 1999. Estimating the organic carbon partition co- specific biomagnification of persistent organic pollutants. Science 317, efficient and its variability for hydrophobic chemicals. Environ. Sci. Technol. 33, 236e239. 2390e2394. Kidd, K.A., Hesslein, R.H., Ross, B.J., Koczanski, K., Stephens, G.R., Muir, Derek C.G., Starrfelt, J., Borga, K., Ruus, A., Fjeld, E., 2013. Estimating trophic levels and trophic 1998. Bioaccumulation of organochlorines through a remote freshwater food magnification factors using Bayesian inference. Environ. Sci. Technol. 47, web in the Canadian Arctic. Environ. Pollut. 102, 91e103. 11599e11606. Kim, J., Gobas, F.A.P.C., Arnot, J.A., Powell, D.E., Seston, R.M., Woodburn, K.B., 2016. Thomann, R.V., Connolly, J.P., Parkerton, T.F., 1992. An equilibrium model of organic Evaluating the roles of special concentration gradients, species migration, chemical accumulation in aquatic food webs with sediment interaction. Envi- biotransformation and field sampling design on trophic magnification factors. ron. Toxicol. Chem. 11, 615e629. Sci. Tot. Environ. 551e552, 438e451. Vander Zanden, M.J., Rasmussen, J.B., 1996. A trophic position model of pelagic food Mackay, D., 2001. Multimedia Environmental Models: the Fugacity Approach, sec- webs: impact on contaminant bioaccumulation in lake trout. Ecol. Monogr. 66, ond ed. CRC Press, Florida. 451e477. Mackay, D., Arnot, J.A., Gobas, F.A., Powell, D.E., 2013. Mathematical relationships Walters, D.M., Mills, M.A., Cade, B.S., Burkhard, L.P., 2011. Trophic magnification of between metrics of chemical bioaccumulation in fish. Environ. Toxicol. Chem. PCBs and its relationship to the octanol-water partition coefficient. Environ. Sci. 32, 1459e1466. Technol. 45, 3917e3924.