Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
Probabilistic ecological risk assessment of polycyclic
hydrocarbons (PAHs) in surface water from Tianjin
aromatic
Yu Yang, Xuan Shi, Fuliu Xu, Wenxin Liu, Shu Tao*
Laboratory for Earth Surface Processes, College of Environmental Sciences, Peking University, Beijing 100871,
China
ABSTRACT
Three approaches namely overlapping area between the exposure concentration and toxic effect
distributions, joint probability curve and distribution of hazard quotient were applied and
compared to evaluate additive toxic effect of eight PAHs to aquatic organisms in rivers in Tianjin
area. Although the toxicity of the studied PAH compounds did not raise significant risk to the
aquatic organisms in Tianjin’s rivers due to relatively low concentrations, it was demonstrated by
the results of all three approaches that the additive effect of the eight PAHs was significantly
stronger than any individual compound acting alone and anthracene was a major contributor to the
overall toxic effect of the mixture. The calculated geometric means of the hazard quotient for the
additive effect varied from 0.00055 to 0.00062, compared to the hazard quotient of individual
PAH ftom 5.110-6 to 0.00053. With slight differences of the calculated results, all three
approaches led to same conclusion, suggesting that the additive effect of 8 PAHs were
significantly stronger than any individual compound acting alone. According to the hazard
quotient distribution approach, the geometric mean of the additive hazard quotient was 0.00058,
while 95% of the quotient fall within a range between 6.6105 and 0.051. The overlapping areas
varied from 0.00015 to 0.02 for individual PAHs and was 0.03 for the additive toxicity.
Keywords: probabilistic risk assessment; equivalent concentration; additive toxicity; PAHs
INTRODUCTION
Tianjin is one of the largest industrial centers in China and is severely polluted with a wide range
of contaminants (Tianjin Environmental Protection Bureau. 2001), among which polycyclic
aromatic hydrocarbons (PAHs) have been detected in all kinds of media including air, dust, water,
soil, sediment, fish, and crops at relatively high levels (Cui, 2003; Shi, 2003; Zhen, 2001). A
recent survey revealed that the concentrations of the 16 USEPA priority PAHs (PAH16) in 30
surface water samples from more than 10 rivers in Tianjin ranged from 45.8 to 1271.6 ng/L with a
geometric mean of 174.1 ng/L (Shi, 2003). The primary sources of PAHs in Tianjin are fossil fuel
combustion, industral discharge, and wastewater irrigation (Zhen, 2001). Both industrial and
domestic users burn a huge amount of coal in the area and the annual consumption of coal was
* To whom correspondence should be addressed: Tel/Fax: 8610-62751938, Email:
taos@urban.pku.edu.cn
159
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
estimated to be approximate 25 million tons (Tianjin Statistical Bureau, 1995). In addition,
because of drastic water shortage, wastewater irrigation and sludge application is a common
agricultural practice in the area and the amount of wastewater used for irrigation is around one
billion tons per year (Tianjin Environmental Protection Bureau. 2001). Enormous amount of PAHs
are brought into environment through the irrigation (Shi, 2003).
PAHs are among potentially hazardous substances in the environment. They may have acute and
chronic toxicity on survival, feeding, reproduction and behavior of organisms (Lotufo, 1997;
Bispo et al., 1999). Using embryos and larvae of the oyster (Crassostrea gigas), Geffard et al.
(2003) demonstrated that PAHs could be accumulated by aquatic organisms inducing
metallothioneins. Ecological risk assessment can be applied to quantitatively analyze the adverse
effects of pollutants to organisms and ecosystem. Specific methods often used range from simple
hazard quotient calculation to probabilistic assessment (USEPA, 2000; Solomon and Sibley, 2002;
Wang et al., 2002). Hazard quotient, defined as the quotient of the measured or estimated
concentration divided by a toxicant reference value, is appropriate as a single-value estimate for
early stage screening-level risk assessment (van Beelen and Doleman, 1997; Solomon et al., 2000;
Staples and Davis, 2002; Newsted et al., 2002; Brooks et al., 2003). Probabilistic risk assessment
provides quantitatively possibility of the ecological risk on either individual or a group of species
based on the distribution of the exposure concentrations and toxic reference data (Duvall and
Barron, 2000; Price et al.,2001; Solomon and Sibley, 2002). Based on the Predicted No Effect
Concentration (PNEC) for all species at a protection level of 95%, Hall and Anderson (1999)
applied the hazard quotient for the probabilistic assessment of the ecological risk of copper in
European saltwater environments. The probability distribution of hazard quotients were also used
to analyze the risk of selenium in Jilin province, China (Ma and Zhang, 2000). The common
approach for probability risk assessment is to calculate the overlapping area between the exposure
and the effect distributions and to derive a joint probability curve, which can fully address the
uncertainty and stochastic properties of the exposure and the effects (Solomon et al., 2000;
Poletika, 2002).
In the reality, there are always more than one toxicants in natural environment and species often
expose to many contaminants simultaneously. With known quantitative concentration-response
relationships of target chemicals, the equivalent concentrations for each individual chemical can
be derived and the overall effect, therefore, can be assessed using the sum of the equivalent
concentrations of all chemicals involved (USEPA, 1989; Petry et al., 1996; Viluksela et al.,1998;
Bosveld et al., 2002). The hazard quotient, therefore, can again be used to analyze the ecological
risk of a mixture, if only the toxic effects of the individual chemicals are linearly addible (Logan
and Wilson, 1995; Cuppen et al., 2002; Soldán, 2003). After exposed to a mixture of zinc,
cadmium, copper, and lead, it was found that the sum of hazard quotients of the metals to Potworm
(Enchytraeus albidus) were positively correlated to the measured toxicity of the mixture (Lock and
Jassen, 2002). For the PAH compounds studied, such additive toxicities are often reported. For
160
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
instance, Knutzen (1995) reviewed relevant literature and found that that the effects mechanism of
PAHs on organisms were similar. (association with lipids and reaction with macromolecules like
DNA, RNA and proteins). Other researches also demonstrated the additivity of the toxicity of
PAHs (Erickson et al., 1999; Landrum et al., 2003). The equivalent concentrations of various
individual PAH compound can be derived based on the concentration-response relationship of
organisms to PAHs (Chen et al., 2001; Bosveld et al., 2002) or the quotient of toxicity data of
PAHs, which were used for hazard quotient calculation (Collins et al., 1998; Tsai et al., 2001).
Sum of hazard quotients was applied for risk assessment when the toxicity of more than one PAHs
compounds were assessed at same time (Swartz et al., 1995; Rogers, 2002). Swartz et al. (1995)
calculated the sum of hazard quotients, which was defined as the quotient of the predicted
concentrations divided by the predicted LC50 using QSAR models, and found that the calculated
sum value was positively correlated with the observed mortality of amphipods exposed to a
mixture of PAHs. When the distribution functions of hazard quotients for single chemical are
defined, the Monte-Carlo simulation can be used to derive the probability distribution of sum of
hazard quotients(Logan and Wilson, 1995). If the concept of the equivalent concentration is
combined with risk probability, it is possible to develop procedures for probabilistic risk
assessment for additive toxicity of a number of chemicals. This approach was commonly
employed in health risk assessment. For example, Price et al. (2001) applied the Monte Carlo
simulation for calculation of the concentrations of pesticides to human exposure and evaluated
the risk of pesticides based on the probability distribution of the hazard quotients. Theoretically,
this approach can also be used in ecotoxicological assessment.
The primary objectives of this study were two folds: 1) to establish procedures for assessment of
probabilistic risk of mixture of PAHs based on the equivalent concentration and the concepts of
overlapping area, joint probability curves and hazard quotient; and 2) to apply the assessment
method for evaluating the toxicity of 8 PAH compounds on aquatic organisms in rivers in Tianjin
area.
METHODOLOGY
Study area
Tianjin lies on Hai river alluvial plain formed by alluvial and marine sedimentation with Baohai
Bay to its east. Bordered by Beijing on the northwest and by Hebei on the north, southwest and
south, most of the plain has an elevation of 3 m to 30 m. With an area of 12,000 km2, the majority
of the land in Tianjin is alluvial/marine deposit plain with hills and mountains occupy a small
portion of the area in north. Tianjin is one of the most important industrial areas in northern China
and the population was around 8 million in 1980’s (Chen and Chen, 1986). The plain is
crisscrossed with a network of rivers and water-ways, most of them man-made, that are largely
used as open sewers (Fu et al., 1990). The main rivers in the study area include Hai River, Ji Canal,
Caobaixin River, North Canal, South Canal, Beijingpaiwu Canal, Beipaiwu Canal, Nanpaiwu
Canal, and Yongdingxin River. All these rivers are severely contaminated with PAHs (Tianjin
Environmental Protection Bureau. 1996., 2001).
161
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
Data collection
Eight PAH compounds were selected for this study based on the availability of toxic data. The
compounds included are naphthalene (Nap), acenaphthylene (Any), fluorine (Fle), phenanthrene
(Phe), anthracene (Ant), pyrene(Pyr), fluoranthene (Fla), and benzo[a]pyrene(Bap). Table 1
summarizes the measured concentrations of these PAHs in river water from Tianjin. The detailed
procedures for sample collection, preparation, and determination, as well as the original data were
presented elsewhere(Shi, 2003).
Table 1 Concentrations of the 8 PAHs in surface water from Tianjin (g/l)
PAHs
Ant
Bap
Fla
Pyr
Fle
Phe
Nap
Any
Arithmetic mean
0.0060
0.0017
0.0069
0.0089
0.023
0.035
0.11
0.0019
Arithmetic S.D.
0.0070
0.0018
0.0069
0.0096
0.059
0.038
0.19
0.0024
Geometric mean
0.0051
0.0013
0.0071
0.0088
0.010
0.029
0.047
0.011
Geometric S.D.
0.0079
0.0021
0.0074
0.011
0.073
0.044
0.21
0.0079
46
46
46
46
46
46
46
46
N
Two toxicity data for PAHs were collected from a database from USEPA (ECOTOX, USEPA,
www.epa.gov/ecotox) and a toxicity handbook (Verschueren, 2001). 24-96 h acute LC50
of the 8 PAHs to a number of aquatic species were collected. The species include
green algae (Selenastrum capricornutum), diatom (Skeletonema costatum),
southernhouse mosquito (Culex quinquefasciatus), yellow fever mosquito (Aedes
aegypti), water flea (Daphnia magna), sheepshead minnow (Cyprinodon variegatus),
channel catfish (ctalurus punctatus), brown trout (Oncorhynchus mykiss), fathead
minnow (Pimephales promelas), scud (Gammarus annulatus), freshwater prawn
(Palaemonetes), and pond snail (Physa heterostropha). The statistics of the toxicity
data are presented in Table 2.
Table 2 Acute toxicity (LC50) of the 8 PAHs to aquatic organism (g/l)
PAHs
Ant
Bap
Fla
Pyr
Fle
Phe
Nap
Any
0.016
2
6.6
4
63
40
110
240
(LC50)median
20
221
109
44
360
534
3700
1030
(LC50)max
360
12000
45000
2000
100000
4240
N
13
11
38
6
8
18
(LC50)min
220000 160500
31
14
Probabilistic risk assessment on individual compound
Three methods adopted in this study for probabilistic risk assessment of toxic effect
of each of the 8 PAH compounds were 1) calculation of the overlapping area between
the exposure and the effect distributions; 2) generation of the joint probability curve, and
3) derivation of probability distribution of hazard quotient. The probability
density functions were derived based on the relevant descriptive statistics. For
the overlapping area approach, the distribution of the exposure concentration of a
PAH in a number of rivers in Tianjin and LC50 of the same compound on various aquatic
species were drawn on the same axes and the overlapping area of the two distributions
was calculated (Solomon et al., 2000). The overlapping area represents a probability
of aquatic ecological risk given the fact that the central tendency of the LC50
162
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
distribution is larger than that of the exposure concentration distribution. The
joint probability curve was derived by plotting the fraction of organisms affected
against the probability of the exposure exceeding the LC50. The joint probability
curve illustrates the relationship between the proportion of species affected and the
likelihood that their response concentration would be exceeded (USEPA, 2000; Solomon and
Sibley, 2002). To calculate the probability distribution of hazard quotient, non-structural
Monte-Carlo simulation was applied based on a procedure described in the literature (USEPA,
2000). The exposure concentration and toxicity data were derived stochastically from the
distribution function to compute the hazard quotient repeatedly for 1000 times to derive the
distribution of hazard quotients. This process took the variability of the environmental parameters
into consideration compared to a single value assessment (USEPA, 2000).
Probabilistic risk assessment of additive effect
The concept of equivalent concentration was adopted for the assessment of additive effect. The
toxic effect of each PAH compound can be expressed as equivalent concentration of an index PAH
that would produce the same response (Price et al.,2001). Therefore, the additive effect can be
expressed as the sum of the equivalent concentrations of all PAHs studied. The approach has been
widely used in human health assessment (Collins et al., 1998; Tsai et al., 2001). To convert the
observed concentration of a given PAH to the equivalent concentration of an index one, Toxicity
Equivalent Factor (TEF) is required, which can be derived from specific concentration-response
relationship (Chen et al., 2001; Bosveld et al., 2002). For health risk analysis, a TEF is usually
based on measures of the relative toxicity including no adverse effect levels (NOAEL), reference
doses (RfD), or Allowable Daily Intakes (ADIs) of the target and the index compounds (Price et
al., 2001). For ecological risk assessment in this study, LC50 was used as the relative toxicity for
TEF computation. Therefore, TEF and the equivalent concentration can be expressed as:
TEF =
LC50i
LC50t
(1)
Ci  Ct
LC50i
LC50t
(2)
where Ct and Ci represent the observed concentration of the target compound (t) and the equivalent
concentration of the index compound (i), respectively. LC50t and LC50i are geometric means of the
LC50 values of the target and the index compound to aquatic organism. When a given PAH was
selected as the index compound, the equivalent concentrations of the PAH were calculated for rest
of 7 PAHs by multiplying the appropriate TEF values to the observed concentrations. The
calculated equivalent concentrations of the 7 PAHs and the original concentration of the index
compound were summed up to provide an estimate representing the total potency of all 8 PAHs.
Theoretically, any of the 8 PAH can be used as the index compound. If all assumptions hold, the
outcomes of the procedure would be identical, or at least similar to one another, no matter which
index compound was chosen. In this study, the 8 PAHs were used as the index compound in
turn and the calculated total equivalent concentrations were compared to address the model
uncertainty. The total equivalent concentrations thus derived were used for calculation of the
163
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
overlapping area, the joint probability function, as well as the probability distribution of hazard
quotients using the same procedures for evaluation of the effect of individual compounds
previously presented.
Matlab v.6.5 (MathWorks, 2002) was used for the calculation of overlapping area, joint probability
function and probability distribution of hazard quotients, while MS Excel and SPSS were
employed for data manipulation and distribution test.
RESULTS
Statistical distribution of the PAH concentration and the LC50
Statistical distributions of the collected data were examined before the risk assessment was
conducted. Since both observed concentrations and toxicity data showed unimodality, the
distributions were tested for their skewness and kurtosis using t-test. Two examples were
presented in Figure 1 as benzo[a]pyrene concentration in river water from Tianjin and LC50 value
of benzo[a]pyrene to aquatic organisms. As can be seen in Figure 1, the original data of PAH
concentration was leptokurtic and right skewed and a normal curve can fit the log-transformed
concentration well, suggesting typical lognormal distribution pattern. For the toxicity data,
however, since there were a few data available (n=11), the distribution pattern was difficult to be
tested. However, at least the right-skewness of the original data was removed by the
log-transformation. A log-normal distribution seems to be better compared with normal
distribution, if not the best, for description of the toxicity data distribution in this study..
Frequency
0.3
9.0
0.2
4.5
0.0
0
0
3
6
0
3
9
1.5
Frequency
Conc. (ng/l)
0
1.5
3
log-Conc. (ng/l)
7.0
5.0
3.5
2.5
0
0
0
5000
10000
15000
LC50 (g/l)
0
2.5
5
7.5
10
log- LC50 (g/l)
Figure 1 The distribution of benzo[a]pyrene concentration in river water from Tianjin (upper)
and LC50 value of benzo[a]pyrene to aquatic organisms (bottom), ether original (left)
or log-transformed (right)
Based on the results of t-test, the hypothesis of both coefficients of skewness and kurtosis do not
differ significantly from zero were rejected for the original data but accepted for most of
log-transformed data, with a few exceptions of LC50 with relatively small sample size. In fact, the
164
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
lognormal distribution pattern of both concentration of PAHs in aquatic environment and toxicity
to aquatic organisms were often reported in the literature (Logan and Wilson,1995; Ma and Zhang,
2000; Solomon et al.,2000; Wang et al., 2002). Accordingly, following lognormal
distribution function was used for both concentration and toxicity data in this study.
f (C ) 
a  b (ln C   ) 2
e
C
(3)
where f(C) is the distribution function of variable C which represent either concentration or LC50,
 is geometric mean of C, while a and b are two constants describing the shape of the lognormal
distribution density functions. Table 3 listed the parameters describing the distribution functions of
the exposure concentration and LC50 values of the 8 PAHs collected in data based on
log-transformed data.
Table 3 Distribution parameters for log-transformed exposure concentrations and
toxicity data of the 8 PAHs in surface water from Tianjin (g/l)
PAHs
Phe
Pyr
Fla
Ant
Nap
Any
Bap
Fle
Concentra Mean
-3.56
-4.74
-4.95
-5.28
-3.05
-5.13
-6.62
-4.84
S.D.
0.80
0.85
0.79
0.95
1.25
1.54
1.11
1.31
a
0.50
0.47
0.51
0.42
0.32
0.26
0.36
0.30
b
-0.78
-0.69
-0.81
-0.55
-0.32
-0.21
-0.40
-0.29
Mean
6.29
4.09
5.20
2.26
8.50
7.08
5.20
6.58
S.D.
1.42
1.89
2.50
2.82
1.19
0.65
3.32
2.33
a
0.28
0.21
0.16
0.14
0.33
0.62
0.12
0.17
b
-0.25
-0.14
-0.08
-0.06
-0.35
-1.20
-0.05
-0.09
-tion
LC50
In general, the concentration of naphthalene in surface water was the greatest, followed by
phenanthrene, while the level of benzo[a]pyrene was much lower than all other PAHs studied.
The mean LC50 varied over 6 orders of magnitude from the most toxic anthracene to the least
toxic naphthalene. The standard deviations of the PAH concentrations were similar to one
another, compared with the LC50, which varied from 0.65 for acenaphthylene to 3.32 for
benzo[a]pyrene, showing a relatively high variability. For the aquatic toxicity of PAH compounds
studied, the sample sizes range from 6 to 38 with a median value of 13.5 (Table 1). The limited
and more or less random coverage of aquatic species in terms of the toxicity data set let to the
wide dispersion of the standard deviation. It seems that the calculated standard deviation,
subsequently the calculated “a” and “b” are sample size and species dependent.
According to the result of a preliminary modeling, the outcome of the assessment are very
sensitive to the standard deviation thus calculated and the difference of the standard deviation and
the parameters ‘a’ and ‘b’ would lead to contradiction when different PAH were used as the index
compound. Therefore, it was assumed the standard deviation of the LC50 of various PAH
compounds should be identical if the sample size increase to infinitive. For the purpose TEF
calculation and risk assessment, the calculated standard deviations for individual PAH compounds
listed in Table 3 were replaced with a mean standard deviation. Subsequentially, the parameters ‘a’
and ‘b’ were replaced as well. For this study with the 8 PAH compounds involved, 2.01, 0.19, and
–0.12 were adopted as standard deviation, ‘a’, and ‘b’ values.
Ecological risk of the individual PAHs
165
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
The risk of individual PAHs studied was evaluated using three approaches as overlapping area of
the exposure concentration and LC50 distributions, joint probability function, and probability
distribution of hazard quotient. The calculated overlapping areas of the 8 PAHs for all aquatic
species are tabulated in Table 4.
Table 4 The calculated overlapping areas of the distribution functions of the
toxicity to aquatic organism and exposure concentrations of individual PAH
compounds in surface water from Tianjin
PAHs
Ant
Overlapping area
0.020
Pyr
Any
Fle
Phe
Nap
Fla
Bap
0.0031 5.8104 5.7104 4.3104 4.2104 2.6104 1.5104
The overlapping areas thus calculated represent to certain extent the probability of risk of the
individual PAHs to the aquatic organisms as a whole. The risks shown in Table 4 were relatively
small with the highest overlapping area around 2% for anthracene, which happens to be the most
toxic compounds to aquatic organisms among the 8 PAHs (Table 3). Although the concentrations
of naphthalene and phenanthrene are the highest in the list (Table 3), the very low toxicity of them
lead to relatively small risk.
The risk of various PAHs can be illustrated in a more explicit way by calculation of joint
probability functions which were derived based on the probability of exposure exceeding the
observed concentration with respect to the affected level of species at the corresponding
concentration. The results are presented in Figure 2. As can be seen in Figbure 2, all joint
probability curves of the 8 PAHs bent towards the left and bottom axis, suggesting a moderate risk
to aquatic organisms. Also like the results of the overlapping area calculation, anthracene again
caused the highest risk among all compounds studied, followed in distance by all the others.
However, the order of the risk demonstrated by the joint probability functions shown in Figure 2 is
not identical to that from the overlapping area calculation. For instance, benzo[a]pyrene with the
smallest overlapping area among the 8 PAHs is not the one showing the smallest risk in Figure 2.
Instead, the area under the joint probability curve of acenaphthylene is the smallest. It appears that
although the results of overlapping areas and joint probability curves show similar results for the
risk of the 8 PAHs in rivers in Tianjin in general, they are not identical in terms of detailed order
of the risk for individual compounds.
166
Probabilty of exceeding
the toxicity data
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
1.0
Ant
Pyr
Fle
Any
0.5
0.0
0
6×10-3
1.2×10-2
3×10-5
6×10-5
6×10-7
1.2×10-6
3×10-6
6×10-6
Probabilty of exceeding
the toxicity data
Fraction of species affected
1.0
Phe
Nap
Fla
Bap
0.5
0.0
0
6×10-6
1.2×10-5
6×10-6
1.2×10-5
3×10-5
6×10-5
3×10-5
6×10-5
Fraction of species affected
Figure 2 Joint probability functions showing individual toxicity of the 8 PAHs compounds to
aquatic organism in surface water of Tianjin
Figure 3 illustrates the distribution of hazard quotient of the 8 PAHs derived from Monte Carlo
simulation. Since all calculated distribution of hazard quotient skewed to right with positive
kurtosis coefficients ranging from 0.012 for naphthalene to 0.281 for pyrene, the data were
log-transformed. The geometric means of the hazard quotients of the 8 PAHs are 0.00053, 0.00016,
5.110-6 1.310-5, 5.410-5, 9.510-6, 4.010-5, 7.010-6 for anthracene, pyrene, acenaphthene,
fluorene, phenanthrene, naphthalene, fluoranthene and benzo(a)pyrene respectively. Again,
according the calculated geometric means of the hazard quotients, the ecological risk is small, in
accordant with the results of the overlapping area and joint probability function calculation in
terms of order of magnitude. Swartz et al. (1995) have calculated the hazard quotients of 13 PAHs
in interstitial waters of many sites (San Diego Bay, Eagle Harbor, Curtis Creek and so on)and also
found that pyrene was the most toxic compounds taking both both toxicity and concentration into
consideration. As can be seen in Figure 3, the calculated hazard quotient distributions present not
only the level of the toxicity of the individual compounds, but also the likelihood of the risk occur
over a large range. If a 95% of the possibility is adopted, the ranges of the hazard quotients
covered are 5, 6, 4, 3, 5, 3, 3, and 3 orders of magnitude for anthracene, benzo(a)pyrene,
fluoranthene, pyrene, fluorene phenanthrene, naphthalene, and acenaphthene, respectively.
167
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
160
120
160
Frequency
Ant
80
0
18
80
0
2
120
14
8
120
Frequency
0
16
12
4
0
-6
80
log-HQ
12
10
Fla
6
2
Bap
40
18
0
18
120
Nap
60
10
50
0
18
2
Phe
60
Fle
Any
60
10
100
Pyr
60
0
12
6
log-HQ
0
0
24
log-HQ
12
0
log-HQ
Figure 3 Probability distribution of hazard quotient (HQ) of the 8 individual PAHs to surface water
aquatic organisms in Tianjin. The calculated hazard quotients were log-transformed
All the results of described above indicate the acceptable ecological risk of individual PAHs with
anthracene is the most hazardous one in surface water from Tianjin area. Both overlapping areas
and joint probability curves suggested that only a small fraction of the aquatic organisms is
affected. The fractions are from 2% for anthracene to negligible for most PAH compounds under
investigation. On the other hand, the hazard quotients indicate the degree that organisms are
affected rather than a fraction of the organism under toxic effect. Again, anthracene imposes the
relatively higher risk than the others do.
The ecological risk of 8 PAHs acting additively
Usually, a representative compound with sufficient toxic data available was selected as the index
compound for TEF calculation and benzo[a]pyrene was often used for this purpose (Petry et al.,
1996). In fact, the implied assumption of using the concept of the index compound and TEF is that
all compounds involved are equivalent as an alternative index one. However, the result of a
preliminary analysis indicated the calculated TEF values were very much index compound
dependent. Therefore, all 8 PAHs studied were used as the index compound in turn to investigate
the possible influence of index compound selection on the calculated results. The derived TEFs
are listed in Table 5.
Table 5 TEFs calculated by using different PAHs as index compound
Index PAH
Ant
Bap
Fla
Pyr
Fle
Phe
Nap
Any average
Ant
1.0
0.053
0.053
0.16
0.013
0.018 0.0019 0.0079
0.16
Bap
19
1.0
1.0
3.0
0.25
0.34
0.037
0.15
3.1
Fla
19
1.0
1.0
3.0
0.25
0.34
0.037
0.15
3.1
Pyr
6.3
0.33
0.33
1.0
0.083
0.11
0.012
0.050
1.0
Fle
77
4.0
4.0
12
1.0
1.4
0.15
0.60
13
Phe
56
2.9
2.9
9.1
0.71
1.0
0.11
0.44
9.1
Nap
530
27
27
83
6.7
9.1
1.0
4.1
86
Any
130
6.7
6.7
20
1.7
2.3
0.24
1.0
21
By using the 8 sets of the TEFs derived from different index compounds as listed in Table 5, 8
equivalent concentrations of each PAHs at each sampling locations and 8 total equivalent
concentrations at each sampling location were computed. According to the results of normality
168
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
tests using the coefficients of skewness and kurtosis, no matter which PAH was used as the index,
the calculated total equivalent concentrations were always log-normally distributed. Means,
standard deviations, as well as the statistics for distribution description are tabulated in Table 6.
Table 6 Log-transformed sum of the equivalent concentration calculated by using
different index compound
Index PAH
Ant
Bap
Fla
Pyr
Fle
Phe
Nap
Any
Mean
-1.16
-3.36
-2.25
-3.91
1.05
-0.38
-2.25
-0.87
Std. deviation
1.08
1.08
1.08
1.08
1.08
1.08
1.08
1.08
a*
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
b*
-0.43
-0.43
-0.43
-0.43
-0.43
-0.43
-0.43
-0.43
* Parameters of the distribution function in Equation (3)
The difference among mean values is expected because they refer to different PAHs as index
compound. It is interesting to see that the standard deviation of the calculated results are identical
(1.08), so are distribution parameters (‘a’ and ‘b’) due to the fact that the exact same
exposure-response assumption was adopted. Based on the total equivalent concentrations listed in
Table 6, the additive effect of the 8 PAHs on the aquatic organisms was evaluated followed the
same procedure used for individual PAHs. The overlapping areas were calculated based on 8
values of the calculated total equivalent concentrations derived when each of 8 PAHs were used as
index compound. It was revealed in a preliminary modeling that the calculated overlapping areas
were differ from one another if the original standard deviations of LC50 of individual PAH
compounds (Table 3). With a replacement of the standard deviation using the averaged one, the
overlapping area thus calculated became identical (0.03), no matter which compound was used as
the index. The result shows a 50% increase from the highest overlapping area (0.02, anthracene)
of the individual PAH, demonstrating a significant additive effect.
The joint probability curves of the equivalent toxicity of the 8 PAHs are shown in
Figure 4 as another way of describing the additive effect of the 8 PAHs. For the
two diagram using anthracene and phenanthrene as the index compound, the joint
probability curves of anthracene and phenanthrene as individual toxicants are also
presented as broken curves for comparison. Of the 8 PAHs in concern, anthracene and
phenanthrene are the most influential compounds to aquatic ecosystem in Tianjin.
Joint probability functions showing additive toxicity of the 8 PAHs compounds to
aquatic organism in surface water of Tianjin. The results were derived using each
of the 8 PAHs as the index compound which are marked on top-right corner of each
diagram, The joint probability curves for toxicity of anthracene is also drawn as
broken curves in Figure 4 for comparison, while the individual curves for other 7
PAHs can not be displayed at such the same scale of the figure. Again, the additive
effect demonstrated by the joint probability curves are much higher than the most
influential compound (anthracene) acting alone, though the toxic effects of all
other 7 PAHs are much weak than anthracene. This result matches well with that
deducted from the overlapping area calculation. Unlike the overlapping areas,
however, the joint probability curves for the additive effect generated by using
different PAHs as the index compound are different from one another, even though
the standard deviation of the toxicity data was averaged for deviation of the
169
Probabilty of exceeding
the toxicity data
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
1.0
Ant
Pyr
Fle
Any
0.5
0.0
6×10-3
0
1.2×10-2
7.5×10-3
1.5×10-2
4.5×10-3
9×10-3
6×10-3
1.2×10-2
Probabilty of exceeding
the toxicity data
Fraction of species affected
1.0
Phe
Nap
Fla
Bap
0.5
0.0
7.5×10-3
0
1.5×10-2
6×10-4
1.2×10-3
4×10-3
4×10-3
2×10-3
4×10-3
Fraction of species affected
Figure 4 Joint probability functions showing additive toxicity of the 8 PAHs compounds to aquatic
organism in surface water of Tianjin. The results were derived using each of the 8 PAHs
as the index compound which are marked on top-right corner of each diagram, The joint
probability curves for toxicity of individual anthracene is also shown as broken curve
for comparison, while the individual curves for other 7 PAHs can not be displayed at
such a scale (Figure 2)
distribution density function.
Finally, the total equivalent concentrations calculated based on different index PAH were used for
derivation of probability distribution of hazard quotient. The resulted distribution all showed
typical log-normal pattern and histograms based on log-transformed data are shown in Figure 5.
The coefficients of skewness and kurtosis before and after log-transformation are tabulated in
Table 7. The coefficients of skewness and kurtosis changed from high positive values to those very
close to zero after log-transformation, which is accordant with theoretical distribution.(Logan and
Wilson, 1995). The result of normality test confirmed the conclusion at a significant level of less
than 0.05.
Table 7 The coefficients of skewness and kurtosis of the raw and the log-transformed
data
Index PAH
Ant
Frequency
150
Pyr
0
60
0
12
6
0
LogHQ
2
0
14
Fle
75
7
3
LogHQ
2
170
0
15
0
15
120
Fla
45
8
Any
150
0
11
90
Nap
60
7
Nap
40
0
12
120
Phe
Phe
Any
50
9
Fle
80
Pyr
75
120
Fla
100
Ant
0
18
Frequency
Bap
7.5
0
8
4
Bap
60
7.5
LogHQ
0
0
20
LogHQ
Figure 5 The distribution of log-transformed total hazard quotient (HQ) of the 8 PAHs to aquatic
organisms in Tianjin based on different index PAH compound (marked on top-left of each histogram) for
the total equivalent concentration calculation.
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
Skewness
Kurtosis
raw
31
26
17
12
15
13
7.8
6.6
log
0.065
0.13
0.099
0.048
0.025
0.055
-0.066
0.049
raw
0.0098
0.0074
0/0033
0.0018
0.0029
0.0023
84
65
log
-0.083
0.16
0.057
-0.16
0.061
0.13
0.14
-0.21
Similar to the overlapping areas, no matter which PAH was used as the index compound, the
calculated distributions of hazard quotient are similar to each other (Figure 5). In fact, the 8
calculated geometric means of the hazard quotient varied within a narrow range from 0.00055 to
0.00062, which was larger than all hazard quotient derived for individual PAH compounds.
Given the distributions of hazard quotients for individual PAHs, the Monte-Carlo simulation could
be applied to derive the distributions of sum of the hazard quotients. without calculation of
equivalent factors (Logan and Wilson, 1995). The sum of the hazard quotients thus calculated was
again log-normally distributed with a geometric mean of 0.0021, which is almost a order of
magnitude higher than that derived from equivalent factors (0.00055-0.00062). In fact, the
approach of the sum of the hazard quotients calculation is rely on an important assumption, e.g.
the toxicity data for various pollutants are recorded for same species. This was just not the case of
this study. The overestimation could be induced from the difference of species in the toxicity data
set. For a situation like this study, the approach based on equivalent factors is more justified.
Moreover, the additive toxic effect shown by using the equivalent factor deduced hazard quotients
are more comparable with the results of the overlapping area and joint probability calculation.
CONCLUSION
For risk assessment of mixed PAHs to aquatic organism, distribution of hazard quotient derived
from the total equivalent concentration are totally independent to index compound selection. The
overlapping area of the exposure concentration and the toxic effect distributions calculated using
different index was constant only the standard deviation of the toxicity was averaged. With slight
differences of the calculated results, all three approaches lead to same conclusion, suggesting that
the additive effect of 8 PAHs are significantly stronger than any individual compound acting alone.
According to the hazard quotient distribution approach, the geometric mean of the additive hazard
quotient are 0.00058, while 95% of the quotient fall within a range between 6.6105 and 0.051.
Acknowledgment
Funding was provided by Junzheng Foundation and The National Scientific Foundation of China
[40031010, 40021101].
Literature Cited
Bispo, A.; Jourdain, M.J.; Jauzein, M. 1999, Toxicity and genotoxicity of industrial soils polluted by polycyclic
aromatic hydrocarbons (PAHs), Org. Geochem., 30, 947-952.
Brooks, B. W.; Foran C. M; Richards S. M; James Weston; Turner P. K; Stanley J. K; Solomon ,K. R; Marc
Slattery ; La Point ,T.W Aquatic ecotoxicology of fluoxetine Toxicology Letters Volume: 142, Issue: 3, May
15, 2003, pp. 169 - 183
Bosveld, ATC.; de Bie, PA.F.; van den Brink, NW.; Jongepier, HK, Anette V., 2002, In vitro EROD induction
171
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
equivalency factors for the 10 PAHs generally monitored in risk assessment studies in the Netherlands,
Chemosphere, 49, 75-83.
Chen, S.P.; Chen, Z.W. 1986, Atlas of Environ. Qual. (in Chinese); Tianjin, Sci. Press: Beijing.
Chen, JJ.; Chen, YJ; Rice, G.; Teuschler, LK; Hamernik, KP, Alberto; RLK, 2001, Using dose addition to estimate
cumulative risks from exposures to multiple chemicals, Regulatory Toxicol. Pharmacol., 34, 35-41.
Collins, JF.; Brown, J. P.; Alexeeff, G. V.; Salmon, A. G.. 1998, Potency equivalency factors for some polycyclic
aromatic hydrocarbons and polycyclic aromatic hydrocarbon derivatives, Regulatory Toxicol. Pharmacol., 28,
45-54.
Cui Y.H. 2003, PAHs in soil-plant system in wastewater irrigated area in Tianjin, Ph.D. thesis (in Chinese), Peking
University, Beijing.
Cuppen,Jan G. M.; Crum,Steven J. H.; Van den Heuvel,Harry H.; Smidt,Rob A.; Van den Brink,Paul J.; 2002,
Effects of a mixture of two insecticides in freshwater microcosms: I. fate of chlorpyrifos and lindane
and responses of macroinvertebrates, Ecotoxicol., 11, 165-180.
Duvall SE, Barron MG., 2000, A screening level probabilistic risk assessment of mercury in Florida everglades
food webs, Ecotoxicol. Environ. Safe.,] 47, 298-305.
Erickson, R. J.; Ankley, G. T.; DeFoe, D. L.; Kosian, P. A.; Makynen, E. A.1999, Additive toxicity of binary
mixtures of phototoxic polycyclic aromatic hydrocarbons to the Oligochaete Lumbriculus variegates, Toxicol.
Appl. Pharmacol. 154, 97-105.
Fu, S.X.; Zhang, C.H.; Cao, G.F. Atlas of Eco-environment of Beijing-Tianjin Area (in Chinese); Sci. Press:
Beijing, 1990; pp74-75.
O Geffard, A Geffard, E His, and H Budzinski
Assessment of the bioavailability and toxicity of
sediment-associated polycyclic aromatic hydrocarbons and heavy metals applied to Crassostrea gigas
embryos and larvae.Mar Pollut Bull, Apr 2003; 46(4): 481-90.
Hall Jr, Lenwood W.; Anderson, Ronald D. A Deterministic Ecological Risk Assessment for Copper in European
Saltwater Environments Marine Pollution Bulletin Volume: 38, Issue: 3, March, 1999, pp. 207-218
Landrum, PF.; Lotufo, GR.; Gossiaux, DC.; Gedeon, ML.; Lee, JH, 2003, Bioaccumulation and critical body
residue of PAHs in the amphipod, Diporeia spp.: additional evidence to support toxicity additivity for PAH
mixtures, Chemosphere Volume: 51, Issue: 6, May, 2003, pp. 481-489
Lock, K.; Janssen, C. R., 2002, Mixture toxicity of zinc, cadmium, copper, and lead to the potworm enchytraeus
albidus, Ecotoxicol. Environ. Safe., 52, 1-7.
Logan, D. T.; Wilson, H. T. 1995, An ecological risk assessment method for species exposed to contaminant
mixtures, Environ. Toxicol. Chem.,. 14, 351-359.
Ma B., Zhang XL, 2000, Regional ecological risk assessment of selenium in Jilin province, China, Sci Total
Environ., 262. 103-110.
MathWorks, 2002, Using MATLAB version 6, The MathWorks, Inc.: Natick, MA.
Newsted, J N, Cousins, I, Werner, K., Giesy JP., 2002, Predicted distribution and ecological risk assessment of a
"segregated" hydrofluorrother in the Japanese environment, Environ. Sci. Technol., 36, 4761-4769.
Petry, T.; Schmid, P.; Schlatter, C. 1996, The use of toxic equivalency factors in assessing occupational and
environmental health risk associated with exposure to airborne mixtures of polycyclic aromatic hydrocarbons
(PAHs), Chemosphere, 32, 639-648.
Poletika NN, Woodburn KB, Henry KS, 2002, An ecological risk assessment for chlorpyrifos in an agriculturally
dominated tributary of the San Joaquin River. Risk Anal, 22, 291-308.
Price, PS.; Young, JS.; Chaisson, CF. 2001, Assessing aggregate and cumulative pesticide risks using a
probabilistic model, Annals Occupational Hygiene, 45, S131-S142.
172
Series of Selected Papers from Chun-Tsung Scholars,Peking University(2003)
Rogers, H., 2002 Assessment of PAH contamination in estuarine sediments using the equilibrium
partitioning–hazard quotient approach
Sci. Total Environ., 290, 139-155.
Shi Z. 2003, PAHs in river system in Tianjin, MS thesis (in m). 2002. Peking University, Beijing.
Soldán, P, 2003, Toxic risk of surface water pollution—six years of experience, Environ. Int., 28, 677-682.
Solomon KR, Giesy J, Jones P. 2000, Probabilistic risk assessment of agrochemicals in the environment, Crop
Protection, 19, 649~655.
Solomon, KR., Sibley, P. 2002, New concepts in ecological risk assessment: where do we go from here ? Mar.
Pollut. Bull., 44, 279-285.
Staples CA, Davis JW. 2002, An examination of the physical properties, fate, ecotoxicity and potential
environmental risks for a series of propylene glycol ethers. Chemosphere, 49: 61~73.
Swartz, RC.; Schults, DW.; Ozretich, RJ.; Lamberson, JO.; Cole, FA.; De Witt, TH; Redmond, MS; Ferraro, SP,
1995, ΣPAH: A model to predict the toxicity of polynuclear aromatic hydrocarbon mixtures in field-collected
sediments, Environ. Toxicol. Chem. 14, 1977-1987.
Tianjin Environmental Protection Bureau. 1996. Environmental quality report of
Tianjin in 1990-1995 (in
Chinese), Tianjin.
Tianjin Environmental Protection Bureau. 2001, Environmental quality report of
Tianjin in 1996-2000 (in
Chinese), Tianjin.
Tianjin Statistical Bureau. 1995, Tianjin Statistics Year Book; Chinese Statistical Press, Beijing.
Tsai, PJ; Shieh, HY; Lee, WJ; Lai, SO, 2001, Health-risk assessment for workers exposed to polycyclic aromatic
hydrocarbons (PAHs) in a carbon black manufacturing industry, Sci. Total Environ. 278, 137-150.
USEPA 1989. Risk Assessment Guidance for Superfund. Vol. I. Human Health Evaluations Manual(Part A). Office
of Emergency and Remedial Response, U.S.EPA/540/1-89/002
USEPA,2000 Environmental implementation plan for probabilistic ecological assessments - terrestrial systems
(www.epa. gov/scipoly/sap/, accessed April 5, 2000).
van Beelen, P.; Doelman, P., 1997, Significance and application of microbial toxicity tests in assessing
ecotoxicological risks of contaminants in soil and sediment, Chemosphere, 34, 455-499.
Verschueren, K., 2001, Handbook of Environmental Data on Organic Chmicals. Van Nostrand-Reinhold, New
York.
Viluksela, M.; Stahl, BU.; Birnbaum, LS.; Rozman, KK., 1998, Subchronic/chronic toxicity of a mixture of four
chlorinated dibenzo-p-dioxins in rats, Toxicol. Appl. Pharmacol., 151, 70-78.
Wang, X.L., Tao, S., Dawson, R.W., Xu, F.L.. 2002, Characterizing and comparing risks of polycyclic aromatic
hydrocarbons in a Tianjin wastewater irrigated area. Environ. Res., 90: 201-206.
Zhen Y. 2002, Contents and spatial characterization of PAHs in surface soil of Tianjin, MS thesis (in Chinese),
Peking University, Beijing.
173