Environment International 66 (2014) 182–193

Contents lists available at ScienceDirect

Environment International
journal homepage: www.elsevier.com/locate/envint

Environmental hazard and risk characterisation of petroleum substances:

A guided “walking tour” of petroleum hydrocarbons
Johan Bierkens ⁎, Lieve Geerts
Flemish Institute for Technological Research, Unit of Environmental Risk and Health, Boeretang 200, B-2400 Mol, Belgium

a r t i c l e i n f o a b s t r a c t

Article history: Petroleum substances are used in large quantities, primarily as fuels. They are complex mixtures whose major
Received 3 October 2013 constituents are hydrocarbons derived from crude oil by distillation and fractionation. Determining the complete
Accepted 29 January 2014 molecular composition of petroleum and its refined products is not feasible with current analytical techniques be-
Available online 5 March 2014
cause of the huge number of molecular components. This complex nature of petroleum products, with their varied
number of constituents, all of them exhibiting different fate and effect characteristics, merits a dedicated hazard
Keywords:
Hazard and risk assessment
and risk assessment approach. From a regulatory perspective they pose a great challenge in a number of REACH
Hydrocarbon block method processes, in particular in the context of dossier and substance evaluation but also for priority setting activities. In
Petroleum substances order to facilitate the performance of hazard and risk assessment for petroleum substances the European oil com-
PETRORISK pany association, CONCAWE, has developed the PETROTOX and PETRORISK spreadsheet models.
PETROTOX Since the exact composition of many petroleum products is not known, an underlying assumption of the PETROTOX
UVCBs and PETRORISK tools is that the behaviour and fate of a total petroleum substance can be simulated based on the
physical–chemical properties of representative structures mapped to hydrocarbon blocks (HBs) and on the relative
share of each HB in the total mass of the product. To assess how differing chemical compositions affect the simulated
chemical fate and toxicity of hydrocarbon mixtures, a series of model simulations were run using an artificial petro-
leum substance, containing 386 (PETROTOX) or 160 (PETRORISK) HBs belonging to different chemical classes and
molecular weight ranges, but with equal mass assigned to each of them. To this artificial petroleum substance a
guided series of subsequent modifications in mass allocation to a delineated number of HBs belonging to different
chemical classes and carbon ranges was performed, in what we perceived as a guided “walking tour” through the
chemical space of petroleum substances. We show that the PETROTOX and PETRORISK predictions reflect changes
in mass distribution introduced to selected HBs by affecting hazard and risk estimates in correspondence with what
is expected based on physical-chemical properties of individual constituents in the corresponding HBs.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction to characterise many petroleum substances in terms of exact chemical

Petroleum substances and refinery streams are derived from crude
oil, using one or more refinery processes, but due to their method of
production, different origin and complex composition, it is not possible
Abbreviations: 2D-GC, 2-dimensional gas chromatography, or GC × GC; CLP, classifica-
tion, labelling and packaging of substances and mixtures (Regulation (EC) No. 1272/2008);
C-range, carbon range; CSR, Chemical Safety Report; CONCAWE, Conservation of Clean Air
and Water in Europe; CTLBB, Critical Target Lipid Body Burden; DAE, untreated distillate
aromatic extracts; DNEL, derived no effects level; ECHA, European Chemicals Agency; IF,
intake fractions; FF, fate factors; GES, Generic Exposure Scenario; HB, hydrocarbon block;
HBM, hydrocarbon block method; HFO, heavy fuel oil components; LL50, lethal loading to
50% of the organisms; NOEL, No Observed Effect Level; PEC, Predicted Environmental
Concentration; PNEC, Predicted No Effect Concentration; QSAR, Quantitative Structure
Association Relationships; REACH, Registration, Evaluation, Authorisation and Restriction of
Chemicals (Regulation (EC) No 1907/2006); RCR, risk characterisation ratio; SpERC,
Specific Environmental Release Category; TLM, target lipid model; UVCBs, Substances of
Unknown or Variable composition, Complex reaction products or Biological materials; VP,
vapour pressure; WS, water solubility.
⁎ Tel.: +32 14 33 51 15; fax: +32 14 33 55 99.
E-mail address: johan.bierkens@vito.be (J. Bierkens).
composition, molecular formula or structure. From a regulatory per-
spective they are collectively called UVCBs or Substances of Unknown
or Variable composition, Complex reaction products or Biological mate-
rials (Carrillo et al., 2010). UVCBs pose a great challenge in a number of
REACH processes, in particular in the context of dossier and substance
evaluation but also for priority setting activities of the European
Chemical Agency (ECHA).
Under the CLP1 (Classification, Labelling and Packaging of substances
and mixtures) and REACH regulation it is possible to group substances
together into categories where their physical hazards, human and envi-
ronmental toxicological properties and environmental fate properties
are likely to be similar or follow a regular pattern as a result of structural
similarities. As such, in this category approach petroleum substances
can be grouped together according to the processes by which they are
manufactured and basic physical properties (Carrillo et al., 2010).
Examples of categories are kerosines, untreated distillate aromatic
1
Regulation (EC) No 1272/2008.

0160-4120/$ – see front matter © 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.envint.2014.01.030
 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193
extracts (DAE) and heavy fuel oil components (HFO). In total 27 differ-
ent categories are defined (Carrillo et al., 2010).
In order to facilitate the performance of hazard and risk assessments
of petroleum substances CONCAWE has developed the PETRORISK
(HydroQual, 2011) and PETROTOX (HydroQual, 2009; Redman et al.,
2012) spreadsheet models, both based on the Hydrocarbon Block Meth-
od (HBM). In the HBM approach groups of closely related constituents
of petroleum substances, such as isomers and adjacent members of a
homologous series of hydrocarbons are thought to possess such similar
properties, that it is assumed that they can be grouped in hydrocarbon
blocks and treated as single compounds without introducing serious
errors in their hazard and risk assessments (King et al., 1996). A library
of 1512 individual structures has been constructed (Howard et al.,
2006) to represent all types of structures that are present in petroleum
substances. This range of individual hydrocarbons is subsequently
mapped to the various HBs. Mass is allocated to each structure based
on the relative share of each HB in the total mass of the product, in
order to simulate the complex petroleum substance and derive physi-
cal–chemical properties relevant for environmental exposure and risk
evaluations. In PETRORISK and PETROTOX petroleum substances are
modelled as being composed of 160 and 386 HBs, respectively. Product
compositional information (e.g. weight % of each HB, chemical class,
boiling point intervals reflecting carbon number ranges) can be obtain-
ed by two-dimensional gas chromatography (2D-GC) or more conven-
tional total petroleum hydrocarbon (TPH) analysis (mass distribution
for aliphatic and aromatic classes vs boiling point) (Gustafson et al.,
1997; Weisman, 1998, 1999).
The Total Petroleum Hydrocarbon Criteria Working group (TPHWG)
fractionation approach is one of the most commonly used methods for
characterisation and risk assessment of petroleum products (Gustafson
et al., 1997; Verbruggen et al., 2008; Weisman, 1998, 1999). Similar ap-
proaches grouping petroleum hydrocarbons in pseudo-components
(Lee et al., 1993; Mackay and Paterson, 1980; Reijnhart and Rose,
1982), hydrocarbon blocks (King et al., 1996; EC, 2003), chemical groups
(Van de Weghe et al., 2006; Mao et al., 2008), fractions (Gustafson et al.,
1997; Park and Park, 2010, 2011) or a combination of fractions and indi-
vidual components (Pinedo et al., 2013) have also been shown success-
ful for the assessment of petroleum hydrocarbons. Methods may differ
in the analytical techniques, number of fractions and chemical classes
considered and have evolved over time. The HBM approach at different
stages of its development, the most recent version of which is imple-
mented in the PETRORISK and PETROTOX models, has demonstrated
its utility for exposure and risk assessment of gasoline (Macload et al.,
2004; McGrath et al., 2005; Foster et al., 2005). The PETROTOX model
performance was recently further validated by Redman et al., 2012 by
comparing the predicted toxicity distributions for different petroleum
substances from different petroleum categories with independent, em-
pirical distributions of toxicity data for these same categories. Whale
et al. (2013) successfully applied the PETROTOX model to predict the
toxicity of refinery spot samples based on their hydrocarbon composi-
tion. Comparing hydrocarbon block (HBM) and TPHWG methods
Bamard et al. (2011) conclude that the characterisation factors (CF)
used in life cycle assessment obtained using these different blocking
methods yield similar results and are both relevant for characterising
the potential impact for aquatic ecotoxicity of petroleum substances.
Given the variable composition of petroleum substances lumped to-
gether in a single category – variability due to geographical differences
of origin and different production processes – it is crucial to understand
how well the PETRORISK and PETROTOX modelling tools are able to re-
flect compositional mass distribution over different carbon ranges and
chemical classes as obtained from 2D-GC, i.e. to explore to what extent
the different HBs covering the chemical space of petroleum UVCBs con-
tribute to the overall hazard and risk outcome. Several approaches can
be followed to better understand the effect of variable compositional
input on the hazard and risk estimates of both models. In this paper
we report on the results of model simulations for which an artificial
183
petroleum substance was constructed, containing 160 (PETRORISK) or
386 (PETROTOX) HBs belonging to different chemical classes and mo-
lecular weight ranges, but with equal mass assigned to each block. Sub-
sequently, a controlled or “guided” series of modifications in mass
allocation to a delineated number of HBs belonging to different chemi-
cal classes and carbon ranges was made to this artificial petroleum sub-
stance, in what we perceived as a guided “walking tour” through the
chemical space of an artificial petroleum UVCB. The main purpose of
these model simulations was to investigate the influence of the inherent
variability in product composition on predicted environmental distribu-
tion, hazard and risk of petroleum substances as calculated by the
PETRORISK and PETROTOX models.
2. Methods and materials
2.1. Guided “walking tour” of petroleum hydrocarbons
In order to test the overall performance of both modelling tools the
models were run multiple times and for each run changes were made
to the composition of the hydrocarbon product. Two distinct modifica-
tions were performed: (i) adding 30% mass to HBs in incremental mo-
lecular weight ranges (carbon (C)-ranges) from C3–5 up to C21–23
(larger structures are highly insoluble), each range encompassing 3 car-
bon numbers, in PETRORISK and from C3–5 up to C18–20 (larger struc-
tures are highly non-toxic) in PETROTOX and (ii) adding 30% mass to
different chemical classes representing the product, i.e. Paraffins and
iso-Paraffins (n- & i-P), Cyclo-Pentanes and –Hexanes (n-CC5 & -6),
Naphthenics (i- & Di-N), Mono-Aromatics (Mo- & NM-Ar), Di-
Aromatics (Di- & Ndi-Ar) and Poly-Aromatics (Poly-Ar). When adding
mass to HBs in incremental molecular weight ranges distinction has
been made between a first set of runs including both aliphatics and ar-
omatics and a second set of runs targeting aromatic classes only. As no
aromatic structures bC6 exist the range C3–5 has not been addressed
in the second set of model runs.
During the mass re-allocation process mass added to targeted HBs
was subtracted from mass in all other HBs in order to preserve 100%
mass in the total product. In the original (unmodified) artificial petro-
leum each HB was assigned equal mass, i.e. 1.0490% in case of
PETRORISK and 0.3497% in case of PETROTOX. Differences in the initial
mass allocated to each of the HBs in PETRORISK and PETROTOX result
from the use of 3 carbon versus 1 carbon intervals in the input matrix
in PETRORISK and PETROTOX, respectively. Also, as a result of the un-
equal number of HBs in both models—the weight % after the mass re-
distribution process in both models is not identical because the mass
added to the targeted blocks is subtracted evenly from a different num-
ber of remaining HBs in both model inputs. In our approach the weight
% in the PETRORISK input matrix was obtained by summing the weight
% of 3 successive C-ranges in the PETROTOX 1C matrix (for example: the
sum of the weight % of C6, C7 and C8 of PETROTOX, equals the weight %
of block 2 (C-range 6–8) in PETRORISK). Matrices illustrative of the
targeting process for each model are given in Tables 1 and 2, respective-
ly. For PETROTOX an example is given of targeting the chemical classes
n- & i-P. In case of PETRORISK an example is given of mass addition to
Block 2, i.e. C-range 6 to 8. In subsequent model runs 3C-ranges up to
C23 and different chemical classes specified above were targeted in a
similar way. Notice that as Blocks 1 and 10 only contain structures
with 5 and 30 carbons, respectively, they were assigned only 1/3 of
the weight % of the other blocks. Blocks 8 to 10 which contain highly in-
soluble and non-toxic structures (NC23) were excluded from the mass
re-allocation process targeting incremental C-ranges.
Normal olefins (n-O), branched olefins (i-O) and aliphatic sulphur
compounds (Al–S) were excluded from the targeting experiments
(mass set to zero) as no information on these classes is provided in
the category composition by CONCAWE (Leonards et al., 2010). Because
of their limited quantity in most petroleum products, coupled also with
the difficulty of separating them from the other hydrocarbon groups
 184
Table 1
Input matrix PETROTOX: Targeting n-P and i-P.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

n-P i-P n-CC5 n-CC6 i-N Di-N n-Olefins i-Olefins Poly-N AlS MoAr NMAr DiAr NDiAr PolyAr ArS

J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193

Block C# (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %) (weight %)

1 5 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
2 6 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
7 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
8 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
3 9 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
10 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
11 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
4 12 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
13 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
14 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
5 15 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
16 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
17 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
6 18 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
19 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
20 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
7 21 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
22 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
23 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
8 24 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
25 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
26 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
9 27 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
28 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
29 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
10 30 0.4545 0.4545 0.3263 0.3263 0.3263 0.3263 0.0000 0.0000 0.0000 0.0000 0.3263 0.3263 0.3263 0.3263 0.3263 0.0000
 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193 185

with the current 2D-GC settings (Whale et al., 2013), these chemical

(weight %)
classes were ignored by CONCAWE in recent applications of the tool.

0.0000

0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Moreover, HydroQual (2009) argues that the risk contribution from

ArS
16
sulphur classes are negligible compared with the contribution from

(weight %)
the other hydrocarbon classes predominant in petroleum substances

1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360
PolyAr
and well within the uncertainty of the risk assessment methodology
15 (HydroQual, 2009).
Thirty percent mass re-allocation to test the sensitivity of the models
(weight %) was chosen based on an expert judgement on the observed variation

1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360
NDiAr

seen between similar HBs from both closely related petroleum sub-
14

stances and petroleum substances belonging to different categories

such as gas oils, kerosines, heavy fuel oil components, etc. published
(weight %)

by Leonards et al. (2010). For similar HBs from petroleum substances

1.3636
1.0079
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360
DiAr

belonging to different categories 30% is often an underestimation of

13

the observed variability. However, for more closely related products

(weight %)

within one category, 30% will overestimate the variability seen. In

1.3636
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360

0.3360
general, for testing and verifying the overall capability of the models
NMAr

to accommodate a wide range of petroleum products across different

12

petroleum substance categories, 30% was estimated adequate to test

(weight %)

the sensitivity of the model predictions to variable compositional input.

1.3636
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360

0.3360
MoAr
11

2.2. PETRORISK model simulations

(weight %)

PETRORISK is a spreadsheet tool that performs environmental risk

0.0000

0.0000
0.0000

0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000

assessments for petroleum substances, taking into account the ECHA's

AlS
10

REACH Implementation Guidelines for performing risk assessments of

(weight %)

complex petroleum substances using the hydrocarbon block method—

0.0000

0.0000
0.0000

0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Poly-N

0.0000

HBM (ECHA, 2012). It was developed to evaluate environmental expo-

sure and ecological risks at both local and regional scales for a wide
9

range of petroleum products including hydrocarbon-based solvents, as-

(weight %)

sociated with different stages in the product life cycle, i.e. at the produc-
i-Olefins

0.0000

0.0000
0.0000

0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000

tion, formulation, and distribution stages, as well as for generic uses in

8

industrial, professional and consumer use sectors (HydroQual, 2011).

Essentially three types of model inputs are required to run
(weight %)
n-Olefins

PETRORISK, i.e. (i) identification of the HBs in the product and HB

0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000

0.0000

0.0000
0.0000

mass allocation from 2D-GC analysis, (ii) generic uses and tonnages
7

for each Generic Exposure Scenario (GES) relevant for the chosen petro-
(weight %)

leum product and (iii) substance-specific DNELs (Derived No Effect

1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079

0.3360

Levels) and PNECsoral (Predicted No Effect Concentration) for secondary

Di-N

poisoning. Other inputs (e.g. fate factors (FFs), intake fractions (IFs),
6

physical–chemical fate properties and the PNECs for environmental

(weight %)

endpoints of each of the library structures) are fixed and stored in sep-
1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079

0.3360

arate databases included in the PETRORISK model (HydroQual, 2011).

i-N
5

The model tool makes use of a library of 1512 model structures

(Howard et al., 2006) to represent all types of structures that are present
(weight %)

in petroleum substances. This range of individual hydrocarbons is sub-

1.3636

0.3360
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
n-CC6

sequently mapped to the various HBs. To run risk calculations, the

4

user can choose between three distinct formats, based on three distinct
(weight %)

separations or blocking schemes: high resolution, low resolution and

1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360
1.0079
n-CC5

solvent mode. For the current experiments model simulations were

run in the high resolution format allowing compositional information
3

input for up to 16 structural classes for 10 carbon number intervals of

Input matrix PETRORISK: Targeting Block 2 (C# 6–8).

(weight %)

3 carbons (HydroQual, 2011). In this approach a petroleum substance

1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
0.3360
1.0079

is modelled as being composed of 160 HBs. The sixteen structural clas-

i-P
2

ses are: normal alkanes or paraffins (n-P), branched alkanes or paraffins

(weight %)

(i-P), normal olefins (n-O), branched olefins (i-O), normal alkane

substituted cyclopentane (n-CC5), normal alkane substituted cyclohex-
1.3636
0.3360

1.0079
1.0079
1.0079
1.0079
1.0079
1.0079
1.0079

0.3360
n-P

ane (n-CC6), other mono-naphthenics (i-N), di-naphthenics (Di-N),

1

poly-naphthenics (PolyN), Mono-Aromatics (Mo-Ar), Di-Aromatics

C#f

5
8
11
14
17
20
23
26
29
40

(DiAr), Poly-Aromatics (Poly-Ar), naphthenic mono-aromatics (NMo-

Ar), naphthenic Di-Aromatics (Ndi-Ar), aliphatic sulphur compounds
C#i (N=)

(Al-S), and aromatic sulphur compounds (Ar-S) (HydroQual, 2011).

For regular risk assessment the mass fraction i.e. % weight obtained
3
6
9
12
15
18
21
24
27
30

from 2D-GC is allocated to each of these different HBs.

Block
Table 2

Apart from compositional information PETRORISK also requires the

10
1
2
3
4
5
6
7
8
9

yearly tonnage manufactured within the EU and used during various

186 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193
uses of a given petroleum product. Product composition and volumes
determine the distribution of hydrocarbons emitted to a given compart-
ment. Default emission factors and operational conditions (emission
days, waste water flow, receiving water dilution factors) are used to
calculate the load of hydrocarbons that are released to those environ-
mental compartments at regional and local scales (HydroQual, 2011).
Emission factors and operational conditions are provided for a broad
range of relevant Generic Exposure Scenario's (GES) for the petroleum
industry based on Specific Environmental Release Category (SpERC)
factsheets. The default input values are automatically selected by
the model from a library included in the model once a yearly volume
is entered. The PETRORISK and PETROTOX models can be downloaded
from the CONCAWE website https://www.concawe.eu/Content/
Default.asp?PageID=554. In the current set of model runs, model
simulations were executed for only 1 GES, i.e. manufacture of sub-
stances (industrial), with an arbitrarily chosen production volume of
1.5 × 106 tonnes/yr. For this production volume the risk driving
compartment, characterised by the highest risk characterisation ratio
(RCR) (i.e. sediment), is in the order of 0.5. Because this value is below
1, indicating no potential risk, no further adjustment to production
volumes or emissions for other GESs were deemed necessary. Therefore,
the results could be evaluated solely on the basis of product composi-
tion without having to consider the implementation of additional risk
management measures in the model.
PETRORISK calculates multi-media distribution according to TGD-
2003 guidelines (EC, 2003) implemented and modelled in EUSES.2
Major outputs of PETRORISK are emissions to different compartments
(water, soil, air, effluent), PECs (Predicted Environmental Concentra-
tions) for environmental exposure and indirect human exposure, and
RCRs for all compartments considered at regional and local scale. An
RCR (i.e. the PEC/PNEC ratio) is calculated for each structure in the
assigned blocks, by using PEC and PNEC values of each individual struc-
ture. The RCR values of each structure are summed to obtain the RCR for
the hydrocarbon product. PNEC values for all individual structures used
for the calculations are included in the CONCAWE library.
The results presented in this paper pertain to the local assessment. A
distinction is made between PECs for environmental compartments
(PECprimary) and PECs for human exposure via the environment
(PECintermediate). Whereas the former PEC values relate to concentra-
tions in effluent, water, sediment, soil and air, the latter denote con-
centrations in biota (e.g. fish, plants, milk, meat) and drinking water.
The first step in the PETRORISK model calculations is the estimation
of the product-specific Raoult's scaled vapour pressure (VP) and water
solubility (WS) by mapping structures to HBs according to the composi-
tional information on the petroleum substance. This information is used
in PETRORISK to select default emission factors for generic exposure
scenario's (GES), where default emission factor inputs are specified as
a function of the physical–chemical properties, VP and WS, of the sub-
stance being evaluated (HydroQual, 2011). The VP and WS estimates
are examined in the current study as one aspect of the contribution of
the different HBs, characterised by different C-ranges and chemical
class, to the PETRORISK exposure and risk estimates. The effects of guid-
ed modifications to the composition of the artificial product on the VP
and WS are only shown in the SI added to this paper in order to keep
this paper as concise and transparent as possible.
2.3. PETROTOX model simulations
PETROTOX is a spreadsheet modelling tool to assess the hazard of
complex petroleum products. It predicts toxicity of products to different
aquatic organisms based on the narcosis target lipid model (TLM) devel-
oped by Di Toro and co-workers (Di Toro and McGrath, 2000; Di Toro et
al., 2000). It assesses the impact of composition and test design on the
2
European Union System for the Evaluation of Substances.
toxicity predictions and based on these predictions it estimates PNECs
for all the representative structures that are used in PETRORISK for en-
vironmental risk assessment. The model can accommodate two types
of inputs, low and high resolutions, according to the information avail-
able on the mass distribution of hydrocarbon classes in the petroleum
product (HydroQual, 2009).
PETROTOX is a QSAR based model, accepting the general assumption
that partitioning into lipophilic tissues such as cell membranes (the tar-
get lipids), determines the toxicity of chemicals with a narcotic mode of
toxic action (HydroQual, 2009). Partitioning in target lipids is assumed
to be the same for different species. The difference in sensitivity for dif-
ferent species is related to the concentrations of these target lipids in
different species, i.e. the Critical Target Lipid Body Burden (CTLBB).
Log-transformed acute toxicity data for several species and compounds
are expressed as a function of log Kow with an equal slope but with dif-
ferent intercepts for each species. Additionally, some correction factors
are fitted for a few chemical classes that are slightly more toxic than
baseline toxicants (narcotic chemicals). These correction factors are au-
tomatically applied when running the PETROTOX model. They are in-
cluded in a separate sheet in the PETROTOX model. A three-phase fate
model is used to simulate the distribution of each structure among the
water-, air, and oil-phase liquid in the laboratory test system. Toxicity
is then computed based on the predicted aqueous concentrations and
TLM-predicted aquatic toxicity of each structure (Redman et al., 2012).
PETROTOX requires mass fraction distribution by chemical class and
carbon range. In contrast to PETRORISK the range specified in PETROTOX
in the high resolution format is a 1C-range, starting at C5 and going up to
C30 with 16 chemical classes. In this approach a petroleum substance is
modelled as being composed of 386 HBs. For the current experiments
model simulations were run in the high resolution format.
Results from the PETROTOX model runs are reported as critical prod-
uct loadings (in mg/L water), e.g. LL50 (lethal loading) for acute effects
and NOEL (No Observed Effect Loading) for chronic effects (HydroQual,
2009). Similar to the simulations in PETRORISK, the guided tour in
PETROTOX was performed on an artificial petroleum product composed
of all existing HBs, each containing equal mass. As PETROTOX is designed
to model typical aquatic toxicity test preparation methods where a cer-
tain volume of headspace is allowed in the WAF (Water Accommodated
Fraction) chamber to allow for mixing, different exposure conditions
can be chosen to run simulations. For the current model simulations
exposure conditions were kept constant between model runs, i.e. 1/9
ratio for headspace/aqueous phase. Calculations were performed for
two relatively sensitive and commonly used aquatic test species, i.e.
Selenastrum capricornutum (green algae) and Oncorhynchus mykiss
(rainbow trout; fish).
3. Results
3.1. Targeted mass addition to different HBs belonging to different
chemical classes
3.1.1. PETRORISK
Estimated vapour pressure (VP) and water solubility (WS) calculat-
ed in PETRORISK for subsequent mass re-allocations targeting different
chemical classes are shown in Fig. 1 of the SI to this paper. The highest
estimates for VP are recorded for n- and i-P, while the lowest and
highest WS is calculated for n- and i-P and Mo- and NM-Ar, respectively.
These results are in concordance with expectations as mass assigned to
relatively soluble and volatile chemical classes dominate WP and VP
predictions highly. Correlation between C-number and WS (mg/L) and
C-number and VP (atm) for all structures in the CONCAWE library
with C-number ≤ 20 and grouped according to chemical class are
shown in Figs. 2 and 3 of the Supplementary information, respectively.
The effect on PECprimary, PECintermediate and RCR predictions calculat-
ed in PETRORISK for local assessment when different chemical classes
are targeted is shown in Fig. 1. A distinction is made between primary
 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193
and intermediate PECs. As mentioned before, primary PECs relate to the
different environmental compartments and intermediate PECs refer to
biota and other exposure routes involved in human exposure via the en-
vironment. The results are expressed as the relative deviation Δ (in %) of
PEC and RCR from a model run with targeted HBs as compared to a
model run with equal mass over all HBs. The actual estimates for PEC
and RCR are included in Table 3 in the SI. Relative deviations to the
PECprimary vary (top panel Fig. 1) generally in the order of 5% for
marine and fresh water sediment, soil and groundwater. For all other
PECprimary the deviations are in the order of 0.1–1.0%. Relative changes
to PECfresh water sediment and PECmarine sediment coincide. Relative increases
of PEC in fresh water and marine sediment are highest when Di-& Ndi-
Ar and Poly-Ar are targeted, with a maximum of 11% for Poly-Ar. When
the relative share of one of the other remaining classes is increased, the
PECfresh water sediment and PECmarine sediment decrease as compared to the
petroleum product with uniform mass assigned to all HBs. The PEC in
agricultural soil is most affected when the relative share of n-CC5 &-6
(+ 6.8%) and i- & Di-N (+ 7%) is increased. When mass is added to
(n- & i-P),(Mo- & NM-Ar) and (Di- & Ndi-Ar), PECagricultural soil decrease
with −4.4%, −3.0% and −3.9%, respectively. A similar trend is seen for
PECgroundwater. For biota, the highest PEC increases show no discernible
pattern (middle panel Fig. 1): in fish the highest PEC increases are esti-
mated with higher Di-& Ndi-Ar and Poly-Ar concentrations (+5% and
+11%, respectively), while in meat, milk the highest PEC increases are
estimated when the paraffins are targeted. Whereas PECfish are calculat-
ed based on FF, which in turn are based on BCFwater, the concentrations
in meat and milk are driven by diet through soil and air deposition by
relatively more volatile alkanes. The highest relative deviations for
PECroot are seen when the relative share of n-CC5 & -6 (+8.5%), and i-
& Di-N (+8.7%) is increased.
Although the relative deviations observed for PECprimary and
PECintermediate after targeting different chemical classes vary by environ-
mental compartment and the physical–chemical properties that con-
tribute to the emissions (e.g. solubility and vapour pressure) (top and
middle panel Fig. 1), all RCR coincide with the exception of agricultural
soil (bottom panel Fig. 1). The final RCR outcome for all aqueous com-
partments (sediments included) mimics the PECfresh water sediment and
PECmarine sediment to a large extent. In these aqueous compartments Di-
& Ndi-Ar and Poly-Ar clearly drive risk, increasing RCRs with 11% as
compared to uniform mass allocation input for the initial artificial prod-
uct. For the RCR agricultural soil the main differences are seen for the i-
& Di-N (+4%) and Di- & Ndi-Ar (−3.8%).
3.1.2. PETROTOX
In order to estimate the contribution of different HBs to the overall
toxicity of a petroleum substance a selected number of blocks delineat-
ed according to structurally related chemical classes, i.e. Paraffins (n- &
i-P), Cyclo-Pentanes and -Hexanes (n-CC5 & -6), Naphthenics (i- & Di-
N), Mono-Aromatics (Mo- & NM-Ar), Di-Aromatics (Di- & Ndi-Ar) and
Poly-Aromatics (Poly-Ar), were defined for subsequent targeting and
toxicity calculations. The results for the acute (LL 50 ) and chronic
toxicity (NOEL) predictions are given in the SI added to this
paper (Table 4). Lower LL 50 and NOEL values indicate higher toxic-
ity. The relative change Δ (in %) with respect to acute and chronic
toxicity prediction for S. capricornutum and O. mykiss following
mass addition to different chemical classes, as compared to the
original artificial product, is shown in Fig. 2. Both aquatic species
experience reduced toxicity when 30% mass is added to the ali-
phatic chemical classes examined. Reductions amount between
11 and 14% for n- & i-P, and 7–9% for i- & Di-N. In contrast toxicity
increases for both organisms when the mass fractions of aromatic
chemical classes are increased with 30%. The largest increase in
toxicity predictions is seen for Di- & Ndi-Ar, i.e. 14 and 16% for
S. capricornutum and O. mykiss, respectively. From these results
we can conclude that aromatics in general, and di-aromatics in
particular, drive aquatic toxicity. These results also show that the
187
model is sensitive to changes in composition, modelling reduced
toxicity when mass is assigned to less soluble, and less toxic,
aliphatics.
3.2. Targeted mass addition to different HBs belonging to incremental
carbon ranges
3.2.1. PETRORISK
The effects of 30% mass addition to incremental 3C-ranges on VP
and WS estimates are provided in a separate graph in the SI added to
this paper (Fig. 4). Only minor changes are observed. Adding mass
in the C3–C5 range increases VP and WS estimates with 24 and
17%, respectively. The lowest VP and WS predictions are obtained
when C12–C14 range structures are targeted. The decline, both for
VP and WS, for C12–C14 targeting amounts to − 4%. Targeting
higher MW ranges slightly decreases VP and WS, but the estimates
remain at maximum 2.5% below the prediction for the original arti-
ficial product. Again these results are in concordance with expecta-
tions as mass assigned to relatively soluble and volatile HBs
dominate WP and VP predictions highly (see also Figs. 2 and 3 in-
cluded in the SI).
Estimates for PECprimary, PECintermediate and RCR for different model
runs with 30% mass addition to incremental 3C-ranges across all chem-
ical classes are given in the SI to this paper (Table 4). The relative devi-
ation Δ (in %) of estimated PECprimary, PECintermediate and RCR from those
of the original artificial product is shown in Fig. 3. Similar as for the pre-
vious model runs discussed under Section 3.1.1, the highest effects of
mass re-allocation are predicted for PECfresh water sediment and PECmarine
sediment (top panel Fig. 3; graphs coincide). Increases of the PEC values
are highest when HBs in the range C9–C11 and C12–C14 are targeted.
For biota (middle panel Fig. 3) the highest relative positive deviations
to the PEC are observed when HBs in the ranges C6–C8 up to C12–C14
are targeted. The relative changes to the RCR (bottom panel Fig. 3), in-
troduced by targeted mass re-allocation reflect the relative change of
PECmarine sediment and PECfresh water sediment (all graphs coincide with
the exception of agricultural soil). Adding mass to the C3–C5 range de-
creases risk for all compartments with up to 15%. When targeting higher
C-ranges risk increases with 7 to 8% for C9–C11 and C12–C14 ranges.
When higher MW ranges are increased in mass, RCRs decline and the
relative change approaches zero. In contrast to chemical class targeting,
RCRsoil fluctuates almost in parallel with other RCRs, but does not
coincide.
The effect of 30% mass addition to incremental 3C-ranges for
aromatics-only on VP and WS estimates are provided in Fig. 5 in
the SI. VP decreases up to 2.2% when C9–C11 are targeted, but
steadily increases thereafter. Adding 30% additional mass to the
C21–C23 range results in a relative change of − 1.1%. WS increases
with 5% when mass is added to C6–C8 aromatics. For all other 3C-
ranges WS slightly decreases with 0.6–1.5% when additional mass
is assigned to these blocks.
Relative deviation Δ (in %) of estimated PECprimary, PECintermediate and
RCR for artificial products with targeted mass addition to incremental
aromatic 3C-ranges from those of the original artificial product is
shown in Fig. 4. The actual estimates are provided in Table 6 of the Sup-
plementary information.
Concentrations in soil, groundwater and air are highly affected, with a
relative deviation (decrease) in the order of 80%, no matter which aro-
matic 3C-range is targeted (top panel Fig. 4). The high decrease in PECair
is in line with the decreased VP across all targeted aromatic C-ranges
(data provided in SI). The effect on PECs in the other environmental com-
partments are highest when mass is added to C6–C8 and C9–C11 aro-
matics. As a result of the large effects in PECsoil, PECgroundwater and
PECair, all PECsintermediate (which relate to indirect human exposure), are
significantly decreased across all aromatic C-ranges, with the exception
of PECfish, which relates to the concentration in surface water. This indi-
cates that in the original artificial hydrocarbon, the contribution of the
188 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193

Fig. 1. Relative deviation Δ (in %) of estimated PECprimary (top), PECintermediate (middle) and RCR (bottom) for artificial products with uniform (original) and targeted mass addition to
different aliphatic and aromatic classes. Paraffins (n- & i-P), Cyclo-Pentanes and -Hexanes (n-CC5 & -6), Naphthenics (i- & Di-N), Mono-Aromatics (Mo- & NM-Ar), Di-Aromatics (Di- &
Ndi-Ar) and Poly-Aromatics (Poly-Ar). Top panel: PEC marine and fresh water sediment and the PEC effluent, marine and fresh water coincide. Bottom panel: all RCR curves coincide
with the exception of the RCR agricultural soil. Dotted lines represent trend lines, no gradients.

paraffins to the PECsintermediate is substantial. For the concentration in fish, The largest deviations are observed on soil, groundwater and air con-
the contribution of aromatic substances in the ranges C6–C8 up to C12– centrations, the final overall risk outcome for all compartments with the
C14 seem to be more predominant. exception of soil (bottom panel Fig. 4) is driven by the PECfresh water sediment
 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193
Fig. 2. Relative change Δ (in %) to acute (top; LL50) and chronic (bottom; NOEL) toxicity
predictions for S. capricornutum and O. mykiss introduced by targeting different chemical
aliphatic and aromatic classes. Paraffins (n- & i-P), Cyclo-Pentanes and -Hexanes (n-CC5
& -6), Naphthenics (i- & Di-N), Mono-Aromatics (Mo- & NM-Ar), Di-Aromatics (Di- &
Ndi-Ar) and Poly-Aromatics (Poly-Ar). Dotted lines represent trend lines, no gradients.
and PECfresh water (top panel of Fig. 4; graphs for PECfresh water sediment and
PECfresh water coincide), and it is most affected by targeting aromatics in
ranges C9–C11 and C12–C14. For these C-ranges risk increased by 8%.
3.2.2. PETROTOX
In addition to chemical class, the carbon number has a large impact
on fate and toxicity of petroleum products. In the mass re-allocation ap-
proach this has been addressed by adding 30% mass to HBs containing
structures with incremental 3C-ranges. Two distinct modifications
were performed subsequently. In a first set of runs mass re-allocations
were conducted using 3C-blocks over the entire range studied for both
aliphatics and aromatics. In a second set of runs mass re-allocations
were applied only to the aromatic chemical classes in the same manner.
The results for the acute (LL50) and chronic toxicity (NOEL) predictions
for the first set of runs are given in Table 7 in the SI of this paper. Relative
change Δ (%) introduced to acute and chronic toxicity predictions for
S. capricornutum and O. mykiss following mass addition to incremental
189
C-ranges, as compared to the original artificial product, is shown in
Fig. 5. Acute toxicity increases steadily when additional mass is allocat-
ed to increasing C-ranges up to C12–C14 (algae) and C15–C17 (fish).
From then onwards, acute toxicity predictions decline when targeting
higher C-ranges. Chronic toxicity predictions parallel the acute trends,
with a maximum toxicity at the C15–C17 range for both algae and fish.
In the second set of model runs in which mass reallocations were ap-
plied only to the aromatic chemical classes, the enhancement in toxicity
with increasing C-number is even more pronounced (data provided in
Table 8 in the SI), and the highest increase in toxicity is predicted for
the C12–C14 range. In the C12–C14 range a maximum increase in
acute and chronic toxicity predictions of 20% and 17% is calculated for
O. mykiss and S. capricornutum, respectively. The acute toxicity predic-
tions for the C15–C17 range in this second run declined to the same
level as the predictions for the original artificial product with a uniform
mass allocation over all HBs (Fig. 6).
4. Discussion
In the original HBM, groups of closely related structures are expected
to have similar properties, and it is assumed that they can be grouped in
HBs and that HBs can be treated as single compounds without introduc-
ing serious errors (King et al., 1996). However, because it is currently
recognized that even in relatively narrow HBs, individual components
may behave so differently in the environment that PECs of HBs may
not be estimated adequately by modelling them on the basis of the av-
erages of the physical–chemical and toxicological properties of the indi-
vidual compounds in the blocks, a more pragmatic approach to
estimating fate and effects of petroleum substances has been adopted
for the model calculations in PETRORISK and PETROTOX (van de
Meent, 2008). This approach defines HB more narrowly and uses a li-
brary of representative compounds to provide physical–chemical prop-
erties for these HBs. Fate and effect calculations are then performed for
each of these representative compounds (van de Meent, 2008; Redman
et al., 2012, 2013).
In an earlier version of PETROTOX and the HBM tool, the user was
allowed to freely choose the way he wanted to divide his product into
hydrocarbon blocks. As pointed out by Rorije et al. (2012), this could
lead to the incorrect assumption that the HBM will always give the
same quality of results. However, a higher model resolution, will lead
to a better characterization of the product, which will in turn lead to a
more accurate risk assessment, until the method starts to generate
“empty” hydrocarbon blocks, indicating that the resolution starts to be-
come too high in comparison to the size of the library (Rorije et al.,
2012). In the current version of PETRORISK a distinction is made be-
tween 16 chemical classes over a prescribed C-range of 3. Although
this seems a reasonable limitation, forcing the user to work with sensi-
ble hydrocarbon blocks, Rorije et al. (2012) calculated that for the ali-
phatics, a shift in three carbon atoms comprises a log Kow range of 1.5
units, which is too large to consider this as a homogeneous block for
the lower aliphatic substances. Given the potentially large implications
on the hazard and risk calculations the question arises why PETROTOX
and PETRORISK apply a different aggregation scheme, and how sensitive
the model outputs are with regard to different inputs.
Because of the complex composition of petroleum substances and
the adopted computational approach to model petroleum substances
in terms of individual structures it is difficult to validate/test the accura-
cy of the model results. Nevertheless, through a systematic mass re-
allocation process to the HBs, the behaviour of the model predictions
should reflect what is expected when individual targeted/dominant
HBs are considered separately. In other words, the hazard and risk out-
come from the different model simulations for the entire petroleum
substance should reflect the physical–chemical fate and effect proper-
ties of the constituents of the dominant HBs. From a sensitivity analysis
which was performed on a selected number of individual petroleum
components in EUSES (results not shown) it was shown that log Kow
190 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193

Fig. 3. Relative change Δ (in %) introduced to the estimated PECprimary (top), PECintermediate (middle) and RCR (bottom) as a result of targeting incremental 3C-ranges belonging to both
aliphatic and aromatic classes. Top panel: PEC marine and fresh water sediment and the PEC effluent, marine and fresh water coincide. Bottom panel: all RCR curves coincide with the
exception of the RCR agricultural soil. Dotted lines represent trend lines, no gradients.

and log Koc are the physical–chemical input parameters that contribute
most to the variance of calculated exposure and risk estimates. As log
Koc values are highly correlated to log Kow values, and are often calculat-
ed from log Kow, we focused on log Kow values as a key parameter deter-
mining the distribution of hydrocarbon structures over different
environmental compartments, also knowing that in the TLM toxicity is
directly proportional to the log Kow. Correlations between C-number,
log Kow, C-number and PNEC are provided in the SI (Figs. 6–10). The
correlations were made based on the information provided for the indi-
vidual structures included in the CONCAWE library.
Both water solubility and toxicity are inversely related with log Kow.
An initial screening of petroleum substances for HBs that drive risk will
 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193 191

Fig. 4. Relative change Δ (in %) introduced to the estimated PECprimary (top), PECintermediate (middle) and RCR (bottom) as a result of targeting incremental 3C-ranges belonging to aromatic
classes. Top panel: PEC marine and fresh water sediment and the PEC effluent, marine and fresh water coincide. Bottom panel: all RCR curves coincide with the exception of the RCR
agricultural soil. Dotted lines represent trend lines, no gradients.

focus on HBs containing structures with intermediate log Kow values, i.e.
structures with sufficient water solubility and toxicity to drive potential
risk (e.g. log Kow 2–6). Structures with high log Kow values are too insol-
uble to be toxic to aquatic organisms and structures with low log Kow
have low toxicity or are not found in aquatic environments due to
their high biodegradation or volatilization.
The final screening of petroleum products for risk driving HBs corre-
sponds to identifying HBs with a high mass allocation, intermediate log
Kow values and low PNECs. Inversely, enriching the mass allocation to
certain HBs, should affect hazard and risk predictions in a fashion that
reflects the fate and effect factors of the dominant HBs. In doing so,
and in correspondence with the results of the model simulations, it is
predicted that the HBs that contributed most to risk, situate in the
range C9–C11 to C12–C14, i.e. those HBs that are characterised by inter-
mediate log Kow, keeping in mind that differences exist in the relation
between C-number and log Kow and C-number and PNEC for aromatic
and aliphatic chemical classes due to differences in polarity. Also, envi-
ronmental fate highly depends on log Kow as seen when half-life values
192 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193
Fig. 5. Relative change Δ (in %) to acute (top; LL50) and chronic (bottom; NOEL) toxicity
predictions for S. capricornutum and O. mykiss introduced by adding 30% mass to targeted
incremental 3C-ranges belonging to both aliphatic and aromatic classes. Dotted lines
represent trend lines, no gradients.
for the aqueous compartment of individual structures in the CONCAWE
library are plotted in function of log Kow.
Additionally, it was observed that the chemical classes that contrib-
ute most to risk (with the exception of the risk for soil) and toxicity are
Di-Ar, Ndi-Ar and Poly-Ar. As soil concentrations are driven by
emissions to air more volatile (aliphatic) constituents end up in this
compartment.
C-ranges for aromatics that contribute most to toxicity are slightly
lower (C12–C14) as compared to aliphatics together with aromatics
(C15–C17). Also evident from the mass re-allocation process was a
clear shift in the toxicity profile when HBs of aromatic classes are
targeted, clearly indicating that toxicity is driven by aromatic com-
pounds. Likewise, heavier petroleum substances, simulated by targeting
higher molecular weight ranges, redirect concentrations towards solid
Fig. 6. Relative change Δ (in %) to acute (top; LL50) and chronic (bottom; NOEL) toxicity
predictions for S. capricornutum and O. mykiss introduced by adding 30% mass to targeted
incremental 3C-ranges belonging to aromatic classes. Dotted lines represent trend lines,
no gradients.
phase compartments (e.g., sediment) where they become less bioavail-
able. In contrast, substances containing more low molecular weight
aromatic compounds are mainly directed towards aqueous compart-
ment and exhibit higher toxicity.
5. Conclusion
Since the exact composition of petroleum products is often not
known, the underlying assumption of the PETROTOX and PETRORISK
tools is that the behaviour and fate of a total petroleum substance can
be simulated based on the physical–chemical properties of representa-
tive structures mapped to predefined HBs and on the relative share of
the HBs in the total mass of the product. To assess how differing chem-
ical compositions affect the simulated chemical fate and toxicity of
 J. Bierkens, L. Geerts / Environment International 66 (2014) 182–193
hydrocarbon substances a series of model simulations was run based on
a set of gradually modified mass distribution profiles targeting HBs from
different chemical classes and molecular weight ranges. It is shown that
the PETROTOX and PETRORISK predictions reflect changes in mass dis-
tribution introduced to selected HBs by affecting hazard and risk esti-
mates in correspondence with what is expected based on physical–
chemical properties of individual constituents in the corresponding
HBs.
Acknowledgements
The current results were obtained as part of a larger project, commis-
sioned by the European Chemicals Agency, evaluating the PETRORISK
and PETROTOX tools developed by CONCAWE for performing environ-
mental hazard and risk characterisation for petroleum products. The re-
sults build on previous studies performed by CONCAWE and others and
are published under permission of the Agency. All results or rights there-
on, including copyright and other intellectual or industrial property
rights are owned solely by the European Chemicals Agency. The opinions
expressed in this paper are those of the Authors only and do not repre-
sent the official position of the European Chemicals Agency.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.envint.2014.01.030.
References
Bamard E, Bulle C, Deschênes L. Method development for aquatic ecotoxicological charac-
terization factor calculation for hydrocarbon mixtures in life cycle assessment. Envi-
ron Toxicol Chem 2011;30:2342–52.
Carrillo JC, Djemel D, Hedelin A, Hovius H, Moore N, Riley A, et al. Hazard classification
and labelling of petroleum substances in the European Economic Area—2010.
CONCAWE report no. 11/10. CONCAWEBrussels: CONCAWE; 2010.
Di Toro DM, McGrath JA. Technical basis for narcotic chemicals and polycyclic aromatic
hydrocarbon criteria. II. Mixtures and sediments. Environ Toxicol Chem 2000;19:
1971–82.
Di Toro DM, McGrath JA, Hansen DJ. Technical basis for narcotic chemicals and polycyclic
aromatic hydrocarbon criteria. I. Water and tissue. Environ Toxicol Chem 2000;19:
1951–70.
EC. Technical Guidance Document on risk assessment in support of Commission Directive
93/67/EEC on risk assessment for new notified substances and Commission Regula-
tion (EC) No 1488/94 on risk assessment for existing substances and Directive
98/8/EC of the European Parliament and of the Council concerning the placing of bio-
cidal products on the market. 2nd ed. European CommissionEuropean Commission;
2003 [2003].
ECHA. Guidance on information requirements and chemical safety assessment.
Chapter R.7c: endpoint specific guidance. European Chemical Agency (ECHA)
Helsinki: European Chemical Agency (ECHA); 2012.
Foster KL, Mackay D, Parkerton TF, Webster E, Milford L. Five-stage environmental exposure
assessment for mixtures: gasoline as a case study. Environ Sci Technol 2005;39:2711–8.
Gustafson JB, Tell JG, Orem D. Total Petroleum Hydrocarbon Criteria Working Group Se-
ries Vol. 3. Selection of representative TPH fractions based on fate and transport con-
siderations. Total Petroleum Hydrocarbon Criteria Working Group. Amherst,
Massachusetts, USA: Amherst Scientific Publishers; 1997.
Howard P, Meylan W, Aronson D, Stewart S, Parkerton T, Comber M. Prediction of envi-
ronmental fate and transport properties in support of the hydrocarbon block ap-
proach to risk assessment. Society of Environmental Toxicology & ChemistryThe
193
Hague, The Netherlands: Society of Environmental Toxicology & Chemistry; 2006
[8–11 May 2006].
HydroQual. PETROTOX v3.04 user's guide. CONC.006. HydroQualMahwah: HydroQual;
2009.
HydroQual. PETRORISK version 6.02, public manual. HydroQualMahwah: HydroQual;
2011.
King D, Lyne RL, Girling A, Peterson DR, Stephenson R, Short D. Environmental risk assess-
ment of petroleum substances: the hydrocarbon block method. CONCAWE report no.
96/52. CONCAWEBrussels: CONCAWE; 1996.
Lee C, Eastcott L, Shiu WY, Mackay D. Petroleum contaminated soil: chemistry and model-
ling. In: Calabrese EJ, Kostecki PT, editors. Principles and practices for petroleum con-
taminated soils. Chelsea, MI: Lewis Publishers; 1993. p. 323–39.
Leonards P, Comber M, Forbes S, Whale G, den Haan K. CONCAWE effluent speciation pro-
ject. CONCAWE report no. 3/10. CONCAWEBrussels: CONCAWE; 2010.
Mackay D, Paterson S. The physical properties of fresh weathered crude oils: publication
EE-13: a report to the Artic marine oil spill program. Environment CanadaOttawa,
ON: Environment Canada; 1980.
Macload M, McKone TE, Foster KL, Maddalena RL, Parkerton TF, Mackay D. Applications of
contaminant fate and bioaccumulation models in assessing ecological risks of
chemicals: a case study for gasoline hydrocarbons. Environ Sci Technol 2004;38:
6225–33.
Mao D, Van De Weghe H, Diels L, De Brucker N, Lookman R, Vanermen G.
High-performance liquid chromatography fractionation using a silver-modified col-
umn followed by two-dimensional comprehensive gas chromatography for detailed
group-type characterization of oils and oil pollutions. J Chromatogr A 2008;1179:
33–40.
McGrath JA, Parkerton TF, Hellweger FL, Di Toro DM. Validation of the narcosis target lipid
model for petroleum products: gasoline as a case study. Environ Toxicol Chem
2005;24:2382–94.
Park IS, Park JW. A novel total petroleum hydrocarbon fractionation strategy for human
health risk assessment for petroleum hydrocarbon-contaminated site management.
J Hazard Mater 2010;179:1128–35.
Park IS, Park JW. Determination of a risk management primer at petroleum-contaminant
sites: developing new human health risk assessment strategy. J Hazard Mater
2011;185:1374–80.
Pinedo J, Ibañez R, Lijzen JPA, Irabien A. Assessment of soil pollution based on total petro-
leum hydrocarbons and individual oil substances. J Environ Manage 2013;130:72–9.
Redman AD, Parkerton TF, McGrath JA, Di Toro DM. PETROTOX: an aquatic toxicity model
for petroleum substances. Environ Toxicol Chem 2012;11:2498–506.
Reijnhart R, Rose R. Evaporation of crude oil at sea. Water Res 1982;16:1319–25.
Rorije E, Verbruggen EMJ, de Knecht JA. Service request on a critical review of the environ-
mental and physicochemical methodologies commonly employed in the environmental
risk assessment of petroleum substances in the context of REACH registrations. Total
Petroleum Hydrocarbon Criteria Working Group. Bilthoven: RIVM - National Institute
for Public Health and the Environment; 2012. p. 94. http://echa.europa.eu/documents/
10162/13628/review_environmental_physicochemical_methodol_en.pdf.
van de Meent D. Environmental fate factors and human intake fractions for exposure
and risk calculation of petroleum products with the hydrocarbon block method,
Reports Environmental Science no 322. Netherlands Center for Environmental
Modeling (CEM)Nijmegen: Netherlands Center for Environmental Modeling
(CEM); 2008.
Van De Weghe H, Vanermen G, Gemoets J, Lookman R, Bertels D. Application of compre-
hensive two-dimensional gas chromatography for the assessment of oil contaminat-
ed soils. J Chromatogr A 2006;1137:91–100.
Verbruggen EMJ, Beek M, Pijnenburg J, Traas TP. Ecotoxicological environmental risk
limits for total petroleum hydrocarbons on the basis of internal lipid concentrations.
Environ Toxicol Chem 2008;12:2436–48.
Analysis of petroleum hydrocarbons in environmental media. In: Weisman W, editor.
Volume 1 of the Total Petroleum Hydrocarbon Criteria Working Group series.
Amherst, Massachusetts, USA: Amherst Scientific Publishers; 1998.
Human health risk-based evaluation of petroleum releases sites: implementing the work-
ing group approach. In: Weisman W, editor. Volume 5 of the Total Petroleum Hydro-
carbon Criteria Working Group series. Amherst, Massachusetts, USA: Amherst
Scientific Publishers; 1999.
Whale G, Comber M, Djemel N, Keresztenyi I, Villalobos SA, Seljestokken B, et al.
Predicting refinery effluent toxicity on the basis of hydrocarbon composition
by GC × GC analysis. CONCAWE report no. 2/13. CONCAWEBrussels: CONCAWE;
2013.