Environmental Pollution 127 (2004) 109–115

www.elsevier.com/locate/envpol

Regression models to predict water–soil heavy metals partition

coefficients in risk assessment studies
C. Carlona,*, M. Dalla Valleb, A. Marcominia
a
Department of Environmental Sciences, University of Venice, Calle Larga Sta Marta, 2137, I-30123 Venice, Italy
b
Department of Environmental Science, Institute of Environmental and Natural Sciences, Lancaster University,
Lancaster LA1 4YQ, UK
Received 23 January 2003; accepted 20 June 2003

‘‘Capsule’’: The adoption of regression models for heavy metals could provide a substantial improvement in risk
assessment practices.

Abstract
Risk assessment studies apply fate and transport models to predict the behaviour of chemicals in the environment. The
definition of physico-chemical properties is crucial to predict the mobility of pollutants and heavy metals in particular within
the environmental compartments. The conservative approach normally adopted at a screening level in attributing a value to
the Kd value, results in an extremely variable mobility in soil. In this paper a regression model to estimate rapidly the Kd for
heavy metals is proposed and applied to Pb, allowing a considerable reduction (3–4 orders of magnitude) of the estimation
uncertainty. The application of a stepwise forward multiple regression to literature data provided a pH-dependent regression
equation of the soil–water distribution coefficient (Kd) for Pb: log Kd=1.99+0.42 pH.
# 2003 Elsevier Ltd. All rights reserved.
Keywords: Risk assessment; Heavy metals; Regression model; Partitioning

1. Introduction with solutions containing varying quantities of the

The risk assessment procedure requires an under-
standing of the transport of contaminants from the
source to the receptor. Fate and transport models are
commonly used to simulate the transport of con-
taminants in environmental media, such as unsaturated
soils, aquifers and ambient air. Uncertainties in model-
ling include the choice of the physico-chemical proper-
ties of contaminants. In particular the solubility and the
soil–water distribution coefficients of heavy metals (KD)
are of paramount importance to predict the behaviour
and mobility of pollutants (especially metals) within the
soil compartment and the extent of their transfer to
groundwater.
The soil–water distribution coefficient (KD) is the ratio
of sorbed contaminant to dissolved contaminant under
equilibrium conditions. Normally sorption studies
regarding inorganic contaminants are performed in
batch or column tests which involve leaving soil to react
* Corresponding author.
0269-7491/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0269-7491(03)00253-7
compound for a specified time period.
The mobility of a contaminant in soil corresponds to
the ability of the contaminant to dissolve into soil pore
water and to be transported by advective and diffusive
processes through the soil medium. Therefore, the
mobility description consists of two steps, the physico-
chemically driven desorption process, also named
‘‘environmental availability’’ (Peijnenburg et al., 1997),
and transport process. Heavy metals can be transported
within soil in different forms, including as free metal
ions (hydrated ions), as inorganic and organic com-
plexes and adsorbed to colloids. Therefore, to estimate
the mobility of the metals all these forms should be
taken into account when a soil–water partition coeffi-
cient is measured.
Since the KD value depends on the metal properties
as well as on the soil characteristics, the values repor-
ted in literature show extremely high variability, span-
ning several orders of magnitude. Although an
experimental measure of KD would be more accurate
and highly recommended in site specific risk assessment
studies, a conservative approach is usually preferred,
110 C. Carlon et al. / Environmental Pollution 127 (2004) 109–115
particularly during the screening level of the analysis,
in order to save money and time. In most cases it
implies the selection of the lowest KD values (i.e.
high mobility) reported in literature. It can be
argued that the KD estimation should take into
account soil properties, wherever known, in order to
derive more likely realistic than conservative values.
However, suitable algorithms to derive the KD from
soil properties in risk assessment studies are not yet
available.
A considerable number of studies have tried to
explain the mobility of heavy metals in soils based on
the multiple regression analysis of extractable metal
concentration versus soil properties, considered as
dependant variables such as total metal content, pH,
organic matter content (OM), cation exchange capacity
(CEC), and the presence of Fe and Al oxides (Feox and
Alox) (e.g., John, 1972; Kuo et al., 1985; Bogacz, 1994,
He and Singh, 1993; Hornburg and Brummer, 1993;
Jopony and Young, 1994; McBride et al., 1997; Sauvé et
al., 1997, 1998). Other authors have used different
dependent variables, such as sorption coefficients (e.g.,
Harter, 1979; Anderson and Christensen, 1988; Kuo
and Jellum, 1991, Janssen et al., 1997) and soil migra-
tion rate (Dragun, 1988). A number of authors have
considered the fraction of extractable metal against the
total soil metal as the dependent variable (Hornberg
and Brummer, 1993; Rieuwerts et al., 1998). Most of
these multivariate regression exercises aim to infer the
relationships between metal solubility and soil proper-
ties, while some studies aim to develop predictive metal
solubility algorithms. Some authors (Sauvé et al., 1997;
Janssen et al., 1997; Jopony and Young, 1994;
McBride et al., 1997) considered a large number of soil
samples covering a broad range of different soil types,
pollution sources and pollutant concentrations, in
order to obtain more general results. However, these
algorithms were rarely compared to each other and
their validity was limited to specific case studies. Dif-
ferences between algorithms may derive from differences
in the experimental procedures used to measure the
metal solubility as well as from different sorption mod-
els and regression methods used to relate solubility to
soil properties.
In this paper, a method to provide suitable KD
values in risk assessment studies based on literature
data is proposed and an example for Pb KD is
provided. Literature datasets were collected and vali-
dated for this purpose. The KD-soil properties corre-
lations were explored for each dataset by a
multivariate regression exercise. Different data sets
were combined to create a master data set, which
was used to develop a multivariate regression predic-
tion model of Pb KD depending on soil properties.
The guiding principle was that although the
combination of different datasets may decrease the
regression, it normally increases the representativity
of the regression model.
2. Methods
2.1. Multivariate regression method
Kd and Pbdis were correlated to the soil character-
istics by a stepwise forward regression analysis based
on the linear least square method. This consisted of
calculating the linear regression so as to minimise the
squared deviations of the observed points from the
predicted points (the squared residuals). The stepwise
forward procedure consisted of including the inde-
pendent variables in the model one by one, setting
the condition so that the F value, calculated as the
regression mean square on residual mean square, was
higher than 1. This method allows the inclusion in
the regression model of only the most significant
variables.
The goodness-of-fit of the multiple regression model
was evaluated by the adjusted R2-value (R2-adj), which
is the approximation of the coefficient of determination
adjusted for the number of added predictor variables,
the F-value, and the Standard Deviation Error of the
Estimate (SDEE), which provides the dispersion of the
observed values around the regression line. The level of
significance of the interpolation, the intercept and the
dependent variable coefficients were evaluated on the
basis of the relative calculated p-values. An analysis of
residuals was carried out to test the goodness of fit and
the homoschedasticity of the equation (Einax et al.,
1997).
The predictive capability of the regression model was
determined using the training/validation splitting
method, which consists of splitting the dataset into a
training set, used for calculating the regression model,
and a validation set, used for validating the model
(Massart et al., 1996).
The predictive performance of the regression model
was estimated using the predicted or cross validated
variance, expressed as Q2, and the Standard Deviation
Error of Prediction (SDEP) (Tosato et al., 1992; Piazza
et al., 1995):

0:5
SDEP ¼ i yobs  ypre 2 =n ;
where yobs are observed values external to the dataset
used for the model construction, and ypre are the model
predictions.
The multiple stepwise forward regression was per-
formed by using the Statistica 4.1 (Stat Soft Inc., 1992)
program, while the calculation of Q2 and SDEP was
performed by using the Excel97 data management pro-
gram (Microsoft, 1997).
 C. Carlon et al. / Environmental Pollution 127 (2004) 109–115 111

3. Results and discussion
3.1. Data selection
An extensive literature survey was undertaken to find
studies which may contain suitable data. Finally, five
data sets were selected, reported by Jopony (1991),
Sauvé et al. (1997), Janssen et al. (1997), Aten and
Gupta (1996), Gerritse and Van Driel (1984).
Mean, median, maximum, minimum and standard
deviation of individual data for each data set are pre-
sented in Table 1.
Since all the parameters of each dataset showed a log-
normal type distribution, the data used in this study
were log-transformed (except pH data) before multiple
regression was carried out, in order to meet the
assumption of normality required by the regression
method. Even when the selected dataset had been
already used in multivariate regressions to relate Pbdis
or Kd to soil properties (i.d., Sauvé et al. for Pbdis, 1997,
and Jansen et al. for Kd, 1997), a new regression was
performed to obtain and compare the same statistical
parameters for each dataset.
The correlation matrix of each dataset showed that
the correlation coefficient of dependent variables was
negligible, i.e. lower than 0.05, so that they could be
included in the multiple regression exercise.
The combined dataset is assumed to be representative
of a wide variety of soil types and experimental condi-
tions. Since the parameters OM, Alox and PO3 4 were
not present in all dataset, only the parameters Kd, Pbtot
and pH were retained in the combined dataset. A plot of
log Pbtot against pH of the Master dataset shows a
homogenous distribution of the predictive variables
covering a pH range from ca. 3.5 to ca. 8.5 and a log
Pbtot range from ca. 0.5 to ca. 5 mg/kg d.w. (Fig. 1).
3.2. Log Kd and log Pbdis vs soil properties
For each dataset, a stepwise forward regression pro-
cedure was applied separately to the log transformed
values of Pbdis and Kd (logKd or log Pbdis=a+b pH+c
log Pbtot+d log Alox+e log PO3 4 +f log OM). The
results are reported in Tables 2 and 3 respectively for
log Pbdis and log Kd with dependent variable coeffi-
cients, intercepts, relative standard errors, adjusted R2,
F value, p-level and standard deviation errors of the
estimates. Janssen et al. (1997) used the multiple regres-
sion method to relate the Kd to the soil characteristics
and reported results very similar to those recalculated in
this study. However, since the data of Fe oxyhydroxides
(log Feox) were not available, the relation between log
Kd and log Feox could not be determined, therefore a
significant correlation between log Kd and log Alox was
found. While Sauvé et al. (1997) applied the multiple
regression technique to the raw Pbdis data, in this study
the data was log-transformed in order to meet the
assumption of normal distribution.
The results showed that pH was a significant pre-
dictive parameter in all the regression equations for log
Pbdis and log Kd. Log Pbtot was also a significant pre-
dictive parameter in all the regression equations for log

Table 1
Mean, median, minimum, maximum and standard deviation of the selected data setsa

Authors and no. of samples Chemical parameters Mean Median Minimum Maximum St. Dev.

Jopony (1991) pH 6.4 6.8 3.9 7.4 0.8

100 samples Pbtot (mg/kg) 8878 4470 330 38,178 9416
Pbdis (mg/L) 177 68 4 3600 496
Sauvé et al. (1997) 84 samples PH 6.6 7.3 3.5 8.1 1.2
Ptot (mg/kg) 857 323 10 14,860 2071
Pbdis (mg/L) 9.7 3.5 0.6 124.1 18.0
Molybdo-PO4 (mmol/L) 18.4 16.7 2.2 47.6 10.8
OM (gC/kg) 29 25 4 107 16
Janssen et al. (1997) 20 samples pH 5.6 6.1 3.0 7.2 1.5
Pbtot (mg/kg) 186 186 74 832 2
Pbdis (mg/L) 15 12 3 1 123
OM (gC/kg) 60 55 20 280 20
Alox (mg/kg) 28 25 4 112 2
Aten and Gupta (1996) 13 samples pH 6.8 6.9 5.6 7.6 0.6
Pbtot (mg/kg) 95.1 32.5 18 281 93.4
Pbdis (mg/L) 5.2 2.04 0.92 33.12 8.6
CEC(meq/100g) 25.67 23.06 12.75 53.92 11.81
Gerritse and Van Driel (1984) pH 6.7 6.8 6.6 4.5 1.8
Pbtot (mg/kg) 514 52 1.89 8500 1562
Pbdis (mg/L) 1.56 0.6 0.04 9.9 2.18
OM (gC/kg) 65 70 11 270 30
a
Alox (mg/kg) is the amount of aluminium extracted by ammonium oxalate/oxalic acid, OM is the organic matter determined after dichromate
oxidation, and molybdo-PO4 (mM of P) is molybdo-reactive phosphate in soil solution.
112 C. Carlon et al. / Environmental Pollution 127 (2004) 109–115
Fig. 1. A log Pbtot vs pH bivariate plot showing the range of variation of the Master data obtained combining the data from six datasets.
Pbdis but only in three out of the five equations for log
Kd. Since Kd is [Pbtot]/[Pbdis], a positive correlation
between log Kd and log Pbtot indicates that Pbdis does
not increase proportionally with the increase in Pbtot.
Amorphous oxyhydroxides (Alox) in soil and dis-
solved phosphate in pore water (molybdo-PO4) were
proved significant predictors of both log Pbdis and log
Kd for the Janssen and Sauvé datasets, respectively.
However, Alox in soil and phosphate in soil solution
are not widely measured parameters included in soil
investigation and hence do not provide a relevant
increase of the explained variance (R2).
On the basis of the Sauvé, Janssen and Aten & Gupta
dataset, log OM and log CEC in soil are not significant
predictors of either log Pbdis or log Kd. However, log
OM was highly positively correlated to log Pbdis and log
Kd in the Gerritse and Van Driel dataset.
This apparent discrepancy can be ascribed to the
difficulty in treating the effect of organic matter and
CEC by a simple correlation analysis because: (i) pH
is a controlling variable on metal complexation by
organic matter, (ii) the low variability of organic
matter content can make the tests of statistical sig-
nificance misleading; (iii) treating organic matter con-
tent as a single variable ignores the heterogeneous
nature of soil organic matter. The relationship between
OM and pH is also evident in the Gerritse and Van
Driel regression exercise, since the log OM significance
decreases after the addition of pH and log Pbtot to the
independent variables.
In conclusion, pH and log Pbtot were found to be the
most suitable parameters for predicting Pbdis and log
Kd. Considering the pH and log Pbtot dependent
regressions for both log Pbdis and log Kd, the regres-
sions of the Jopony dataset (Jo.2 and Jo-k.2 for log
Pbtot and log Kd respectively, Table 2) exhibited the best
fit, i.e. the highest fraction of explained variance, and
the lowest standard deviation errors of estimate, coeffi-
cients and intercept. For both the log Pbdis algorithms
and the log Kd algorithms, it can be seen that the pH
coefficients obtained from different datasets are very
similar. On the basis of these results, log Kd depends on
pH with a regression coefficient ranging from 0.34 to
0.47. The standard errors of the intercepts are generally
higher than those of the dependent variable coefficients.
The plots for the equations of the predicted values ver-
sus the observed values with a linear regression line fit-
ting the data with 95% confidence interval. None of the
scatterplots underscored relevant outliers and ethero-
schedasticity effects.
Table 2
Results of the multiple stepwise forward regression log Pbdis=a+b pH+c log Pbtot+d log Alox+e log PO3-
4 +f logOM, where Pbdis is the total dissolved Pb (mg/L), Pbtot (mg/kg d.w.) is the total Pb
content in soil (after HNO3 digestion), Alox (mg/kg) is the amount of aluminium extracted by ammonium oxalate/oxalic acid, OM is the organic matter content (g/kg) determined after dichromate
oxidation, and PO4 (mM of P) is molybdo-reactive phosphate in soil solution

Dataset Step Equation Adj. R2 F-value Standard Intercept Regression coefficients

code deviation
error of
estimate
pH b log Pbtot c 3th variable d, e, f

Jopony (1991) n=100 Step 1 Eq. Jo.1 0.54 118 Est.  0.36 0.5  0.2 (P=0.019) 0.65 0.06 (P=1E-18)
Variables=Pbdiss, Pbtot, pH Step 2 Eq. Jo.2 0.91 49 Est.  0.16 1.2 0.1 (P=6E-15) 0.47 0.02 (P=7E-36) 1 0.03 (P < 1E-18)
Sauvé et al. (1997) n=84 Step 1 Eq. Sa.1 0.075 8 Est.  0.50 0.06 0.21 (P=0.8) 0.23 0.08 (P=0.007)
Variables=Pbdiss, Pbtot, pH, OM, PO4 Step 2 Eq. Sa.2 0.32 20 Est.  0.43 1.1 0.3 (P=8E-05) 0.26 0.05 (P=5E-07) 0.51 0.09 (P=1E-07)
Step 3 Eq. Sa.3 0.38 18 Est.  0.41 1.6 0.3 (P=8E-07) 0.30 0.05 (P=9E-09) 0.54 0.08 (P=1E-08) PO3-
4 0.01  0.004 (P=0.003)

C. Carlon et al. / Environmental Pollution 127 (2004) 109–115

Janssen et al. (1997) n=19 Step 1 Eq. Ja. 1 0.36 11 Est  0.46 0.4  0.4 (P=0.4) 0.25 0.07 (P=0.004) log Alox 0.3 0.3 (P=0.25)
Variables=Pbdiss, Pbtot, pH, Alox. Step 2 Eq. Ja. 2 0.58 13 Est  0.37 6.2  1.9 (P=0.005) 0.37 0 .07 (P=1E-04) 1.2 0.4 (P=0.006)
Step 3 Eq. Ja. 3 0.59 9 Est.  0.37 5.6  2.0 (P=0.01) 0.33 0.08 (P=0.001) 1.2 0.4 (P=0.01)
Gerritse and Van Driel (1984) n=31 Step 1 Eq. Ge. 1 0.23 10 Est.  0.48 0.8  0.2 (0.001) log OM 0.8  0.3 (P=0.003)
Variables=Pbdiss, Pbtot, pH, OM. Step 2 Eq. Ge. 2 0.24 6 Est.  0.47 0.1 0.8 (P=0.91) 0.1 0.1 (P=0.25) log OM 0.6  0.3 (P=0.04)
Step 3 Eq. Ge. 3 0.31 6 Est.  0.45 0.7 0.8 (P=0.38) 0.2 0.1 (P=0.04) 0.3 0.2 (P=0.06) 0.1  0.4 (P=0.80)
Atens and Gupta (1996) n=13 Step 1 Eq. At. 1 0.72 32 Est.  0.23 4.5 0.7 (P=6E-05) 0.6 0.1 (P=0.0001)
Variables=Pbdiss, Pbtot, PH, CEC. Step 2 Eq. At. 2 0.78 22 Est  0.20 4.7 0.6 (P=3E-05) 0.7 0.1 (P=8E-05) 0.31 0.16 (P=0.08)

Table 3
Results of the multiple stepwise forward regression log Kd=a+b pH+c logPbtot+d logAlox+e log molybdo-PO4 +f log OM, where Kd (L/kg) is the soil-water Pb distribution coefficient, Pbtot
(mg/kg d.w.) is the total Pb content in soil (HNO3 digestion), Alox (mmol/kg) is the amount of aluminium extracted by ammonium oxalate/oxalic acid, OM is the organic matter (% of dry matter)
determined after dichromate oxidation, and PO3- 4 (mM of P) is molybdo-reactive ortophosphate in soil solution

Dataset Step Equation Adj. R2 F-value Standard Intercept Coefficient

code deviation
error of
estimate
pH (b) log Pbtot (c) 3th variable (d), (e), (f)

Jopony (1991) n=100; Step 1 Eq. K-Jo.1 0.85 564 Est.  0.16 1.8  0.1 0.47  0.02
Variab.=Pbdis, Pbtot, pH (P=7E-25) (P=2E-42)
Sauvé et al. (1997) n=84 Step 1 Eq. K-Sa.1 0.50 85 Est.  0.50 2.9  0.2 (P=1E-23) 0.76 0.08 (P=3E-14)
Variab.=Pbdis, Pbtot, pH, OM, molybdo-PO4 Step 2 Eq. K-Sa.2 0.63 72 Est.  0.43 1.9  0.3 (P=1E-10) 0.26  0.05 (P=5E-07) 0.49 0.09 (P=3E-07) Molybdo-PO4 -0.01 0.004 (P=0.003)
Step 3 Eq. K-Sa.3 0.67 56 Est.  0.41 1.4  0.3 (P=4E-06) 0.30  0.05 (P=9E-09) 0.46 0.08 (P=4E-07)
Janssen et al. (1997) Step 1 Eq. K-Ja.1 0.37 35 Est.  0.37 2.1  0.3 (P=1E-05) 0.35  0.06 (P=1E-05) Log Alox 0.4  0.3 (P=0.21)
n=19 Variab.=Pbdis, Pbtot, pH, Alox. Step 2 Eq. K-Ja.2 0.66 19 Est.  0.36 1.8  0.4 (P=0.0005) 0.31  0.06 (P=1E-04)
Gerritse and Van Driel (1984) Step 1 Eq. K-Ge.1 0.51 32 Est.  0.51 3.7  0.3 (5E-14) 0.7 0.1 (P=4E-06)
n=31 Variab.=Pbdis, Pbtot, pH, OM. Step 2 Eq. K-Ge.2 0.64 27 Est.  0.44 2.1  0.5 (P=0.0002) 0.25  0.07 (P=0.002) 0.7 0.1 (P=3E-06)
Atens and Gupta (1996) Step 1 Eq. K-At.1 0.77 41 Est.  0.33 2 1 (P=0.05) 1.0  0.2 (P=5E-05)
n=13 Variab.=Pbdis, Pbtot, pH, CEC. Step 2 Eq. K-At.2 0.91 64 Est.  0.20 1.8  0.7 (P=0.02) 0.7  0.1 (P=8E-05) 0.7 0.2 (P=0.001)

113
114 C. Carlon et al. / Environmental Pollution 127 (2004) 109–115
3.3. A predictive model for estimating the Pb soil–water
coefficient (Kd) in soil
The combined dataset, named Master, was composed
of the values of log Kd, pH, and log Pbtot from each
dataset, whereas other variables, such as OM, Alox and
CEC, were not included because they were not present in all
the original datasets. However, the regression exercise
applied to each original dataset proved that the significance
of these latter variables for predicting log Kd is low.
The combined dataset was divided into two sub-data-
sets, named Master 1 and Master 2, by splitting alternate
cases of the dataset ranked according to a log Kd value
increasing order. Master 1 and Master 2 datasets were
equally representative of the original dataset, as they cov-
ered the whole range of log Kd variation and were homo-
geneously distributed across the range of variation of log
Pbtot and pH, as showed by the log Pbtot vs pH in Fig. 2.
The Master 1 dataset was used as a training dataset
for calculating a regression model, whose prediction
capability was tested using the Master 2 dataset. Then,
the reciprocal validation was performed, using Master 1
as the training set and Master 2 as the validation set, in
order to verify the stability of the model.
The results of the multivariate regression of log Kd
against log Pbtot and pH for the Master, Master 1 and
Master 2 datasets are reported in Table 4. It can be
noticed that log Pbtot gives a very small contribution to
the log Kd prediction on the basis of the combined
dataset, and appears not to be a significant predictor for
Fig. 2. Plot of Master 1 and Mater 2 data according to log Pbtot and
pH.
Table 4
Predictive capability of Master 1 and Master 2 based regression mod-
els calculated by reciprocal external validation: PRESS values of
Master 1 model vs Master 2 data and Master 2 model vs Master 2 data
Q2 SDEP
Master 1 model vs Master 2 data 0.47 0.49
Master 2 model vs Master 2 data 0.37 0.52
the Master 1 and Master 2 datasets. As a result, log Kd
could be explained as a univariate function of pH.
The predictive capability of Master 1 and Master 2
based models was expressed in terms of Q2 and SDEP
(Table 4). The predicted variance Q2 is rather low and
the Standard Deviation of Prediction is quite high.
However, the similarity of Q2 and SDEP values for the
Master 1 and Master 2 models proves the model stabi-
lity. According to the training/validation splitting
method (Massart et al., 1996), the Master 1 model,
which showed the best predictive performance, was
selected as representative of the Master dataset.
A visual presentation of the predictive capability of
the Master 1 model against the Master 2 dataset is given
by a plot of predicted vs external validation data
(Fig. 3), along with the training set values. The dashed
lines around the 1:1 line represent the 95% upper and
lower confidence limits calculated as  Standard
Deviation of Error of Prediction (SDEP).
Since the distribution of residuals was proved to be nor-
mal, 2SDEP can be regarded as the 95% confidence
limits of the data population. The scattered distribution of
data around the 1:1 line indicates the low predictability of
log Kd values. However, an accurate prediction of log Kd
is beyond the scopes of this study and the lower 95% con-
fidence limit of the prediction can be regarded as a con-
servative estimation of log Kd. It is remarkable that this
model allows the reduction of the range of uncertainty
of the Kd estimation from four to one order of magni-
tude simply on the basis of the soil pH measurement.
Fig. 3. Plot of both the Pb log Kd values predicted by using the
Master 1 regression model and the Pb log Kd training set values
against the observed data of the validation Master 2 dataset.
 C. Carlon et al. / Environmental Pollution 127 (2004) 109–115
4. Conclusions
The application of a stepwise forward multiple
regression to literature data provided a pH-dependent
regression equation of the soil-water distribution coeffi-
cient (Kd) for Pb: log Kd=1.99+0.42 pH.
In a risk screening exercise, to guarantee a con-
servative approach, the Pb log Kd can be calculated
using this equation and subtracting two times the
SDEP (0.48) which provides, with 95% confidence, an
assessment that the true value is higher than the esti-
mated log Kd.
On the basis of our regression results, the inclusion of
additional variables into the linear model is not expected
to provide significant model improvement. However, the
application of different chemometric methods (Lee et al.,
1996) to include in the model the organic matter effects on
metal retention/dissolution seems promising.
The use of algorithm-calculated KD values allows a
considerable reduction of the estimation uncertainty. In
particular, the application of the regression model to
estimate KD instead of using conservative assumptions,
i.e. the lowest KD value in variability range for the spe-
cific element as reported in literature, for fate and
transport modelling risk assessment procedures, may
reduce the risk of estimation by three to four orders of
magnitude. Such a reduction is a result of a more accu-
rate prediction and does not imply a decrease in the
protection level for the risk receptor (human health or
environmental endpoints). The use of a combined data-
set has been showed to decrease the regression goodness
of fit, but in practice it provided a regression model that
can be applied widely. The example with Pb shows that
in spite of low predictability and conservative assump-
tions, the use of a regression model can provide a sub-
stantial increase in risk estimation accuracy. The
construction of similar regression models for most
common heavy metals, to be updated and adjusted
according to new literature data, could provide a sub-
stantial improvement in risk assessment practices. It can
be also expected that the need for data to be added in a
unique master dataset would enforce a general harmo-
nisation of experimental partitioning methods.
Acknowledgements
I would like to thank Dr Scott Young, University of
Nottingham, for the kind help in the acquisition of
useful data.
References
Anderson, P.R., Christensen, T.H., 1988. Distribution coefficients of
Cd, Co, Ni, and Zn in soils. J. Soil Sci. 39, 15–22.
Aten, C.F., Gupta, S.K., 1996. On heavy metals in soil; rationalisation
115
of extractions by dilute salt solutions, comparison of the extracted
concentrations with uptake by ryegrass and lettuce, and the possible
influence of pyrophosphate on plant uptake. The Sci. Total
Environ. 178, 45–53.
Bogacz, W., 1994. Influence of soil properties on the distribution of Zn
in Polish soils. Geomicrobiology J. 11, 299–308.
Dragun, I., 1988. Element adsorption and mobility in soil. In The Soil
Chemistry of Hazardous Materials, Hazardous Materials Control
Research Institute. Silver Spring, Maryland.
Einax, J.W., Zwanzinger, H.W., Geib, S., 1997. Chemometrics in
Environmental Analysis. VCH A Wiley company, Weinheim.
Gerritse, R.G., Van Driel, W., 1984. The relationship between
adsorption of trace metals, organic matter, and pH in temperate
soils. J. Environ. Qual. 13, 197–204.
Harter, R.D., 1979. Adsorption of copper and lead by Ap B2 horizons
of several Northeastern United States soils. Soil Sci. Soc. America J.
43, 679–683.
He, Q.B., Singh, B.R., 1993. Plant availability of cadmium in soils.
Acta. Agric. Scand. 43, 142–150.
Hornburg, V., Brummer, G.W., 1993. Behaviour of heavy metals in
soils. 1. Heavy metal mobility. Zeitschrift fur Planzenernahrung und
Bodenkunde 156, 466–477.
Janssen, R.P.T., Peijnenburg, W.J.G.M., Posthuma, L., Van Den
Hoop, M.A.G.T., 1997. Equilibrium partitioning of heavy metals in
Dutch fields, soils. I. Relationship between metal partition coeffi-
cients and soil characteristics. Environ. Toxicol. Chem. 16, 2470–
2478.
John, M.K., 1972. Lead availability related to soil properties and
extractable lead. J. Environ. Qual. 1, 295–298.
Jopony, M., 1991. The solubility of lead and cadmium in con-
taminated soils. PhD thesis, University of Nottingham.
Jopony, M., Young, S.D., 1994. The solid-solution equilibria of lead
and cadmium in polluted soils. Eur. J. Soil Sci. 45, 59–70.
Kuo, S., Jellum, E.J., 1991. Affinity constants and behaviour of Cd
sorption in some acid soils. Water, Air, Soil Pollut. 57-58, 369–376.
Kuo, S., Jellum, E.J., Baker, A.S., 1985. Effects of soil type liming and
sludge application on zinc and cadmium availability to Swiss chard.
Soil Sci. 139, 122–130.
Lee, S.Z., Allen, H.E., Huang, C.P., Sparks, D.L., Sanders, P.F.,
Peijnenburg, W.J.G.M., 1996. Predicting soil-water partition coeffi-
cients for cadmium. Environ. Sci. Technol. 30, 3418–3424.
Massart, D.L., et al., 1996. Chemiometrics: A Textbook. Data Hand-
ling in Science and Technology. Elsevier, Amsterdam.
McBride, M., Sauvé, S., Hendershot, W., 1997. Solubility control of
Cu, Zn, Cd and Pb in contaminated soils. Eur. J. Soil Sci. 48, 337–
346.
Peijnenburg, W.J.G.M., Posthuma, L., Eijsackers, H.J.P., Allen, H.F.,
1997. A conceptual framework for implementation of bioavail-
ability of metals for environmental management purposes. Ecotox.
Environ. Safe. 37, 163–172.
Piazza, R., Pino, A., Marchini, S., Passerini, L., Chiorboli, C., Tosato,
M., 1995. Modelling physico-chemical properties of halogenated
benzines: QSAR optimisation through variables selection. SAR and
QSAR in Environ. Res. 4, 59–71.
Rieuwerts, J.S., Thornton, M.E., Farago, M.E., Ashmore, M.R.,
1998. Factors influencing metal bioavailability in soils: preliminary
investigations for the development of a critical loads approach for
metals. Chem. Spec. Bioavail. 10, 61–75.
Sauvé, S., McBride, M., Hendershot, W., 1998. Soil solution specia-
tion of lead(II): effects of organic matter and pH. Soil Sci. Soc.
America J. 62, 618–621.
Sauvé, S., McBride, M.B., Hendershot, W.H., 1997. Speciation of lead
in contamination soils. Environ. Pollut. 98, 149–155.
Tosato, M.L., Piazza, R., Chiorboli, C., Passerini, L., Pino, A., Cru-
ciani, G., Clementi, S., 1992. Application of chemiometrics to the
screening of hazardous chemicals. A case study. Chemiomet. Intell.
Lab. Syst 16, 155–167.