Statistical analysis

Predicted and measured BCFs were plotted as linear correlations using GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla California USA,

www.graphpad.com). The accuracy and applicability of the models were calculated based

on percentages of BCF data that fit within a factor of 10. Also, the performance of the models was tested by performing the Nash-Sutcliffe efficacy test (NSE) with R in order to see how well the observed values versus predicted fit the 1:1 line. The Nash-Sutcliffe efficiency gives an indication of the goodness of fit of the data. When a value ≥ 0, a good match between the modeled and observed data can be concluded; if a value < 0, the observed data mean are more accurate than the modeled.

Model Equation Species coverage Predictors of the model

(Chiou et al., 1977)

Log BCF = 3.41 × Log S - 0.508 Fish (rainbow trout)

The model predicts the BCF for a wide range of chemicals such as hydrocarbons, aromatic acids, etc only with a

physicochemical property such as water solubility of the chemicals.

Technical Guidance Document for Risk Assessment of the European Union,

1996

Log BCF = 0.85 × LogK_ow - 0.70

Fishes and specifically fathed minnow

The model calculates the BCF of organic chemicals, considering the 𝐋𝐨𝐠𝐊𝐨𝐰, a physicochemical property of the chemicals only.

(Meylan et al., 1999) _{Log BCF = - 1.37 × LogK} ow+

14.4 + ∑fi Fishes and specifically fathed minnow

The hypothesis of the model is to provide a better estimation of BCF based on the physicochemical property of a chemical as

𝐋𝐨𝐠𝐊_𝐨𝐰 and different correction factors to apply to each compound (∑fi).

36 (Meredith-Williams

et al., 2012)

Log BCF = 0.71 × LogDlipw - 0.23

Invertebrates and specifically Gammarus pulex and

Notonecta glauca

The model uses the 𝐋𝐨𝐠𝐃𝐥𝐢𝐩𝐰 which is assumed to be a better input physicochemical parameter for the prediction of uptake

for ionizable compounds.

(Karlsson et al.,

2013) Fw ×

1+10(pHint−pKa)

1+10(pHext−pKa₎ + flip × Dlipwater

Invertebrates and specifically Lumbriculus variegatus

The model proves to better and deeply describe the uptake of ionizable compounds linking together physicochemical

properties such as pKa (constant dissociation of a chemical);

biological traits as 𝐅𝐰, 𝐟𝐥𝐢𝐩 𝐚𝐧𝐝 𝐃𝐥𝐢𝐩𝐰 (the water content of the

organism, the lipid content and the liposome water partition coefficient) and environmental properties such as pH.

(Arnot and Gobas, 2003)

BAF = _CCb

w = (1- Lb) + [( k1 × φ + ( kd × β × τ + φ + Ld × Kow)] / (

k₂ + k_E + k_G + k_M)

The model is applicable on three general trophic levels of fishes (lower, middle and

upper)

The model assumes specific fish traits: the weight of the organism (W); the lipid content of the organism (𝐋𝐛); lipid

content of the lowest trophic level (𝐋𝐝); environmental

conditions such as the concentration of particulate organic

carbon (𝛘𝐏𝐎𝐂); concentration of dissolved organic carbon (𝛘𝐃𝐎𝐂) and physicochemical properties of the chemicals (𝐊𝐨𝐰)

37

Table 2. 1. Summary table for selected BCF models including their predictors and applicability domain.

-LogS (mg/L): water solubility of the chemical; -LogKow: octanol-water partition coefficient;

-Σfi: summation of correction factors to apply to the chemicals. Each chemical contains a specific functional group to which a specific correction factor is applied;

-LogDlipw: logarithmic of the liposome-water partition coefficient;

-LogDow: logarithmic of the pH-corrected octanol-water partition coefficient; -Fw (%): water content of an organism;

- ƒlip (%, wet weight): lipid content of the organism;

-Cb (mg/Kg): concentration of a chemical in the upper trophic level; -Cw (mg/L): is the concentration of the chemical in the unfiltered water;

in order to be more representative of natural uptake of chemicals for fish species in aquatic environments.

(Fu et al., 2009a) Log BCF = 0.85 × LogDow - 0.70

The model has been applied to fish species

The model proposes and assumes that the 𝐋𝐨𝐠𝐃𝐨𝐰, a

physicochemical property (pH-corrected octanol-water

partition coefficient) is a better input parameter to describe the uptake of ionizable compounds.

(Dimitrov et al., 2005)

Log BCF_max = log _(aKKown ow+12n) + Fw

Fish species and in particular for salmonids and cyprinids.

The model predicts base-line BCFs assuming several mitigating

factors as molecular descriptors, the octanol-water partition

coefficient (𝐊𝐨𝐰) and fish biological traits such as water

content

38 -Ld (1 %): lipid content of the lowest trophic level organism;

-Lb (20 %): lipid content of the organism; -W (Kg): the weight of the organism;

-χPOC (5 X 10-7 g/mL): concentration of particulate organic carbon; - χDOC (5 X 10-7 g/mL): concentration of dissolved organic carbon; -T ( ̊C): mean water temperature;

-k1 (

1

[(0.01+ _Kow1 )× W0.4_] ): uptake rate constant;

-φ (_{(1+ χ} 1

POC ×0.35 × Kow + χDOC ×0.1 ×0.35 × Kow ): the fraction of the free bioavailable chemical to be taken up by the organisms in the water; -k_d(0.02 × W−0.15× e0.06 × T

5.1 × 10−8_{× K}

ow+2 ): is the rate of uptake of the chemical via the diet; -β: biomagnification process; it is an empirical value to calibrate the model;

-τ: the maximum level of trophic dilution that occurs for substances that are metabolized at a significant rate in organisms of a food web and by default is 1;

-k2 ( k1

Lb × Kow): is the elimination rate constant; -kE (0.125 × kd): fecal egestion rate constant;

-kG (0.0005 × W−0.2): elimination rate constant through growth dilution; - kM (day−1): metabolic transformation rate constant.

39