the solid–solution partitioning of heavy metals (cu, zn, cd, pb) in

8/3/2019 The solidsolution partitioning of heavy metals (Cu, Zn, Cd, Pb) in

1/13

The solidsolution partitioning of heavy metals (Cu, Zn, Cd, Pb) inupland soils of England and Wales

E. Tippinga,*, J. Rieuwertsb,1, G. Panb,c,2, M.R. Ashmorec, S. Loftsa, M.T.R. Hillc,3,M.E. Faragob, I. Thorntonb

aCentre for Ecology and Hydrology (Windermere), Ambleside, Cumbria LA22 0LP, UKbDepartment of Environmental Science and Technology, Imperial College of Science, Technology and Medicine,

Royal School of Mines, London SW7 2BP, UKcDepartment of Environmental Science, University of Bradford, Bradford BD7 1DP, UK

Received 3 July 2002; accepted 31 January 2003

Capsule: Solidsolution distributions of heavy metals can be described quantitively by multiple regression and

mechanistic modelling.

Abstract

Ninety-eight surface soils were sampled from the uplands of England and Wales, and analysed for loss-on-ignition (LOI), and

total and dissolved base cations, Al, Fe, and trace heavy metals (Cu, Zn, Cd, Pb). The samples covered wide ranges of pH (3.48.3)

and LOI (998%). Soil metal contents measured by extraction with 0.43 mol l1 HNO3 and 0.1 mol l1 EDTA were very similar,

and generally lower than values obtained by extraction with a mixture of concentrated nitric and perchloric acids. Total heavy

metal concentrations in soil solution depend positively upon soil metal content and [DOC], and negatively upon pH and LOI,

values ofr2 ranging from 0.39 (Cu) to 0.81 (Pb). Stronger correlations (r2=0.760.95) were obtained by multiple regression analysis

involving free metal ion (Cu2+, Zn2+, Cd2+, Pb2+) concentrations calculated with the equilibrium speciation model WHAM/

Model VI. The free metal ion concentrations depend positively upon MHNO3 and negatively upon pH and LOI. The data were also

analysed by using WHAM/Model VI to describe solidsolution interactions as well as solution speciation; this involved calibrating

each soil sample by adjusting the content of active humic matter to match the observed soil pH. The calibrated model provided

fair predictions of total heavy metal concentrations in soil solution, and predicted free metal ion concentrations were in reasonable

agreement with the values obtained from solution-only speciation calculations.

# 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Chemical speciation; Heavy metals; Modelling; Soils; Uplands

1. Introduction

Solidsolution partitioning exerts a major control onthe transport and retention of heavy metals in soil

water systems. Furthermore, for many soil biota, the

bioavailability of soil heavy metals depends upon con-

centrations and chemical forms in the soil solution (see

e.g. Allen, 1993). In particular, free metal ion con-centrations (Cu2+, Zn2+, etc.) provide the best guide to

both partitioning and bioavailability. Therefore, quan-

titative, predictive, descriptions of the solidsolution

interactions are needed. Two approaches to the problem

can be identified. Firstly, empirical relationships have

been derived, by multiple regression analysis, relating

concentrations of either total dissolved metal or free

metal ions to key soil variables such as total soil metal,

pH and organic matter content (Jopony and Young,

1994; McBride et al., 1997; Sauve et al., 1997, 1998a,b,

2000). This approach has provided good descriptions of

0269-7491/03/$ - see front matter # 2003 Elsevier Science Ltd. All rights reserved.

doi:10.1016/S0269-7491(03)00058-7

Environmental Pollution 125 (2003) 213225

www.elsevier.com/locate/envpol

* Corresponding author. Tel.: +44-15394-42468; fax: +44-15394-

46914.

E-mail address: [email protected] (E. Tipping).1 Present address: Faculty of Environment, Trinity College, College

Road, Carmarthen SA31 3EP, UK.2 Present address: State Key Laboratory of Environmental Aquatic

Chemistry, Chinese Academy of Sciences, Beijing 100085, China.3 Present address: Maslen Environmental Ltd, Albion House, Vicar

Lane, Bradford BD1 5AH, UK.
http://-/?-http://-/?-http://-/?-http://www.elsevier.com/locate/envpol/a4.3dmailto:[email protected]:[email protected]://www.elsevier.com/locate/envpol/a4.3dhttp://www.sciencedirect.com/http://www.sciencedirect.com/http://www.sciencedirect.com/http://-/?-http://-/?-http://-/?-


2/13

metal chemistry, principally in metal-contaminated soils

with relatively low levels of organic matter, i.e. with

loss-on-ignition (LOI) values of 10% or less. However,

there do not appear to have been applications to more

organic-rich upland soils.

A second approach is to use mechanistic models, that

describe the interactions more comprehensively, takinginto account competitive binding to both the solid soil

materials and to solution ligands, including dissolved

organic matter. For example, Benedetti et al. (1996)

applied the NICADonnan model, which describes

proton and metal binding to organic matter, to the

upper, organic-rich horizons of a podzol and were able

to account for solid-solution distributions of Cu and

Cd. Another model that describes ion binding by

organic matter is WHAM (Tipping, 1994). Tipping et

al. (1995a) showed that WHAM could account for pro-

ton, aluminium, and base cation binding by organic-rich

upland soils, while Tipping et al. (1995b) reported the

successful prediction of solidsolution distributions ofmetallic radionuclides in the same soils. Tipping (2002)

reported the application of WHAM to two soils in

which Cd partitioning had been determined by Lee et al.

(1996). For the more organic-rich soil, the solidsolu-

tion partitioning of Cd could be accounted for, but the

involvement of mineral sorbents appeared necessary to

explain the results with the other soil.

One policy context in which such soilsolution parti-

tioning is important is in the estimation of critical loads,

i.e. the metal deposition at which metal accumulation in

soils threatens soil microbes, plants, and higher organ-

isms (de Vries and Bakker, 1998). Models of soilsolu-tion partitioning are central to the critical load

approach, both in modelling solution concentrations

and free-ion activities, which may be more closely rela-

ted to adverse biological effects, and in modelling the

dynamics of metal inputs and outputs for different soils.

However, the multiple regression partitioning models

currently proposed for the calculation of critical loads

are based primarily on data from mineral soils (e.g. de

Vries and Bakker, 1998; Paces, 1998), and there is a

need to establish effective models of soilsolution parti-

tioning for the organic-rich soils which dominate large

areas of the northern hemisphere.

The work described here focuses on organic-rich soilsfrom the uplands of England and Wales. These soils have

accumulated heavy metals from atmospheric deposition,

both long-distance and localised, and in some cases also

from the weathering of soil mineral matter. Measure-

ments were made of key soil parameters, including heavy

metal contents, together with soil solution compositions.

The resulting data set (98 samples) covers wide ranges of

conditions, enabling speciation and regression analyses

to be performed to identify the factors responsible for

solidsolution partitioning, and to attempt to predict

free metal ion concentrations.

2. Methods

2.1. Sampling

Samples of surface soil (05 cm) were collected from

upland moorland sites at Dartmoor, the English Lake

District, North Wales, the Peak District, and the York-shire Dales (Fig. 1). They were from the following soil

types: brown earth, humic brown podzol, humic ranker,

peat, peaty gley, podzol, stagnohumic gley, stagno-

podzol. In each case, a block of intact soil, of approx-

imate area 14 cm2, was encased in an air-tight container

and placed immediately into cold storage in preparation

for extraction of soil porewater. A separate quantity

(100200 g) of soil was collected in preparation for

analysis of total and extractable metals and soil prop-

erties.

2.2. Soil analyses

Determinations were made of soil water content, by

oven drying, and of LOI by ashing at 450 C. Total

Fig. 1. Location map. DM=Dartmoor, LD=Lake District,

NW=North Wales, PD=Peak District, YD=Yorkshire Dales.

214 E. Tipping et al. / Environmental Pollution 125 (2003) 213225


3/13

soil metal contents were determined by digestion with a

mixture of concentrated nitric and perchloric acids, fol-

lowed by leaching of the residues with 5 mol l1 HCl,

and analysis by ICPAES. Extractions with 0.43 mol l1

HNO3 were performed at a ratio of 1 g air dried soil (2

mm sieved) to 10 cm3 of extractant. After extraction for

2 h, the samples were centrifuged, 5 cm

3

of the super-natant were removed to separate tubes and 0.5 cm3 of 5

mol l1 HCl was added prior to analysis by ICPAES.

Extractions with 0.1 mol l1 Na2EDTA (ethylenedini-

trolotetraacetate) were also made at a ratio of 1 g to 10

cm3. The supernatants after centrifugation were diges-

ted with concentrated nitric and perchloric acids, as for

the total soil metal determinations, prior to ICPAES

analysis, in order to remove the EDTA.

2.3. Soil solution extraction and analysis

Blocks of field moist soil, stored in air tight plastic

boxes at 4 C since sampling, were brought into thelaboratory. Drainage holes were punched into the base

of each box. The lid of each box was removed in turn and

high purity water was added gradually, in small aliquots,

to the surface of the soil block. On the first appearance of

water drops in the saucer under the box, addition of

water was stopped and further drainage allowed. When

drainage had ceased, the contents of the saucer were

added to the surface of the soil block. If no more drai-

nage occurred, small amounts of water were again added

to the surface. This process was repeated until the soil

block could hold no more water, i.e. when the soil was at

field capacity. The lids of the boxes were then replacedand the boxes were wrapped in plastic wrapping film and

returned to storage at 4 C for a period of 1 week to

allow equilibration of the soils with the added water.

Before extraction of soil solution, the boxes of soil

were brought to the laboratory and left for one or two

days at room temperature. The wrapping film was

removed (but not the lid) and two soil solution Rhi-

zonTM samplers were inserted diagonally into the soil

block through two diagonally opposite corner drainage

holes at the base of the box, such that they extracted

solution from the entire depth of the soil block. Once

the samplers were in place, a 10-cm3 syringe was

attached, with the syringe plunger fully inserted. Theplunger was slowly withdrawn and held open (to coun-

teract the resulting initial vacuum). The boxes were

again wrapped in cling film to prevent evaporation and

left overnight for extraction to take place. Extracted

solution, collected in the syringes overnight, was

removed to centrifuge tubes for analysis.

The soil solutions were analysed for pH with a com-

bination glass/calomel electrode, and for DOC with a

Dohrmann DC-190 TOC analyser. Metals were ana-

lysed by ICPMS after acidification with HNO3 (to 2%)

and passage through a 0.2-mm filter.

2.4. Preliminary data manipulations

Of the total of 116 samples collected, 17 of the data

lines were incomplete. Except for EDTA extracts, if

any data were missing for a sample the whole sample

was rejected. In the case of EDTA extracts this was

also done except for Cd, for which a relatively largeproportion of samples gave values at the detection limit

of the analytical method. In cases where duplicate

determinations had been made, means were used. This

led to a data set comprising 99 samples. However, one

of these soils had a very high DOC concentration (>

700 mg l1), which was considered suspect. This sample

was therefore rejected, giving a final data set of 98

samples.

The data are analysed in terms of relationships

between dissolved metal concentrations (either total

dissolved or free ions) and metal contents of the soil

solids. Also involved are solid-phase contents of organic

matter and concentrations of dissolved organic matter.The available measurements are of dissolved concentra-

tions and total concentrations. Therefore it was neces-

sary to modify the total concentrations in order to

obtain values of solid-phase contents. The following

assumptions were made;

(a) Under the conditions of the experiments that

yielded pore water, all the soil pore space was

occupied by soil solution.

(b) The soil samples in the experiments had the same

bulk density (BD, g cm3) as in the field, for

which the following relationship connected BDto organic carbon (OC) content (% C);

BD 1:38 0:29 ln OC n 19; r2 0:96

1

This equation was derived by A. Colgan and J.R.

Hall (personal commununication) from data

reported by McGrath and Loveland (1992).

(c) Organic carbon was 58% of the LOI (Rowell,

1994).

Then, the LOI values determined for the present

samples were used to obtain BD, from which solid:so-lution ratios in the experiments were obtained. The dis-

solved metal and organic carbon concentrations were

combined with the solid:solution ratios to calculate the

fractions of the total soil contents in solution, and the

amounts associated with the soil solids in the experi-

ments were obtained by difference. The modifications

were generally very small, average amounts in solution

being


4/13


5/13

assume that the HNO3 and EDTA extractions provide

the best measures of geochemically active metal, i.e.

metal that enters into interactions with the soil solids

that control solution concentrations, while metal

extracted with the concentrated acid mixture includesmore recalcitrant forms. In the present work, we based

our analysis on MHNO3, principally because we have

used dilute HNO3 as an extractant in related work

(Lawlor and Tipping, 2002; Tipping et al., 2003).

The data for heavy metals (HNO3-extractable and

dissolved concentrations) were examined for corre-

lations with soil properties, by testing logarithmic rela-

tionships (except for pH). Only low values of r2 were

obtained (data not shown). Metal extractable with

HNO3 was weakly positively correlated (r240.13) with

LOI for all four metals. Dissolved metal concentrations

were weakly positively correlated with DOC in all four

cases (r240.23), and weakly negatively correlated with

pH (r240.21). For Zn, Cd and Pb there were weak

positive correlations with log%LOI (r240.11). Dis-

solved metals were positively correlated with HNO3-extractable metal, giving r2 values of 0.13 (Cu), 0.28

(Zn), 0.30 (Cd) and 0.62 (Pb).

3.2. Multiple regression without speciation of the soil

solution

The most likely major variables explaining total solu-

tion metal concentrations (Msol) are loss-on-ignition,

DOC concentration, pH, and the soil content of geo-

chemically active total soil metal (MHNO3). The follow-

ing regression equation was used:

Fig. 2. Distributions of soil chemical variables. The ordering is from lowest to highest, and is different for each variable.

E. Tipping et al. / Environmental Pollution 125 (2003) 213225 217


6/13

log Msol a log %LOI bpH c log MHNO3

d log DOC e 2

and Table 2 shows the results obtained. The geochemi-

cally active soil metal content was either the most or thesecond most important variable. For Cu and Pb, both

of which are strongly bound by organic matter, DOC

was important, whereas pH was more important for

both Zn and Cd. Overall, LOI was the least important

variable. Inclusion of HNO3-extractable Al and Fe in

the multiple regressions had negligible effect on the

regression coefficients or r2 values. The regression model

does not provide a sufficiently good prediction of dis-

solved metal concentrations for practical use, and so

speciation of the soil solution was tried (Section 3.3).

3.3. Speciation of the soil solution

A schematic picture of the assumed distribution of

ions in soil is given in Fig. 3. The inorganic master

species of the soil solution are protons, base cations

(Na+, Mg2+, K+, Ca2+), anions (Cl, NO3, SO4

2,

CO32), Al3+, Fe3+, Cu2+, Zn2+, Cd2+ and Pb2+.

(Since all the samples were from presumably oxic sur-

face soil horizons, it was assumed that the iron was

present exclusively in the ferric form.) The metals may

form complexes with the anions, and they may undergo

hydrolysis reactions, which are especially important for

Al and Fe(III). In addition, they may interact with dis-

solved organic matter, represented by FA. The metalsalso undergo solidsolution partitioning. All of them

can react with solid-phase organic matter, while Al may

equilibrate with Al(OH)3 and Fe(III) with Fe(OH)3.

Although the other metals are known to adsorb to oxide

surfaces, such reactions are ignored here, because the

soils are high in organic matter, which is assumed to

dominate partitioning. Part of the soil solids is assumed

to be inert with respect to metal binding; this fraction

includes unreactive organic matter and mineral compo-

nents.

Two approaches were taken to describe the partition-

ing reactions. Firstly the chemical speciation only of the

soil solution was modelled, and multiple regression

analyses were performed to try to establish relationships

between solid and solution metal concentrations, and

other soil variables (Section 3.4). This is basically the

same approach that has been used by Sauve , McBrideand co-workers (see references in Section 1), except that

those workers made direct measurements of either the

free metal aquo ion concentrations, or the concentra-

tions of labile (assumed to represent inorganic) forms of

the metal. The second approach was to include the soil

solids in the speciation calculation, and to predict metal

distributions in the whole system (Section 3.5).

Applications of WHAM and WHAM/Model VI to

speciate the samples of soil solution require as input

data total concentrations of the significant reactants.

Ideally, these are H+, base cations, strong acid anions,

Al, Fe(III) and humic substances, together with the

heavy metals (Cu, Zn, Cd, Pb), and the partial pressureof CO2, or the total carbonate concentration. Con-

centrations of strong acid anions were not determined in

the present work. Their concentrations were estimated

by forcing a charge balance, assuming the anions to

comprise Cl and SO4 at a ratio of 0.75:0.25 in terms of

charge equivalents, typical for runoff in the UK uplands

(Patrick et al., 1995); however, the calculated free metal

ion concentrations were insensitive to the Cl:SO4 ratio.

This approximation does not introduce serious uncer-

tainty, because the strong acid anions do not enter into

significant complexation reactions with the metals of

interest, except that Cl complexation of Cd can accountfor up to 16% of the inorganic solution forms. Con-

centrations of humic substances were estimated from

[DOC] by assuming that 65% of the DOC was due to

FA, to describe ion-binding, while the remaining 35%

was assumed inert. This division of DOC is based on

modelling interactions with organic matter in surface

Table 2

Multiple regression parameters (Eq. 2) for total solution concentra-

tions (Msol) of heavy metals

a b c d e r2

Cu 0.30 0.03 0.38 0.51 4.50 0.39

Zn 0.54 0.18 0.60 0.39 1.11 0.57

Cd 0.61 0.20 0.78 0.28 0.31 0.55

Pb 0.47 0.20 0.89 0.79 1.23 0.81

All values of r2 are significant (P


7/13

and soil waters involving Al and H+ (Tipping et al.,

1991, 2002), Cu (Dwane and Tipping, 1998; Vulkan et

al., 2000; Bryan et al., 2002), and Cd (Tipping, 2002).

Equilibrium with atmospheric CO2 was assumed, but

very similar results were obtained if the pCO2 was set to

10 times the atmospheric value.

Both Al and Fe(III) have to be taken into account,because they can compete significantly with trace metals

for binding by humic substances (Tipping et al., 2002).

These two components were investigated by first

assuming that the measured concentrations in soil solu-

tion represented truly dissolved metal. This yielded ion

activity products (IAP: aAl3+/aH+3 and aFe3+/aH+

3 ), as

shown for WHAM/Model VI in Fig. 4. The values of

log IAP for Al range from ca. 5 at low pH to ca. 12 at

pH $ 7, when calculated with either WHAM or

WHAM/Model VI. The low-pH value is appreciably

smaller than the solubility products (log Kso) for various

forms of Al(OH)3, the lowest of which at the tempera-

ture of the experiments is ca. 8.0, for gibbsite ( Palmerand Wesolowski, 1992). The value of 12 is greater than

those in the range 89 reported for soil waters at pH>5

(LaZerte, 1989). We interpret these results to indicate

that, above pH $ 5, the soil solutions are oversaturated

with respect to Al(OH)3, i.e. some of the Al in soil

solution is present as colloidal suspended particulate

matter, including Al(OH)3. Thus, activities of Al3+ are

controlled either by solidsolution partitioning, com-

bined with solution interactions, principally hydrolysis

and binding by organic matter, or by equilibrium with

Al(OH)3. We adopt a mid-range value of log Kso,25 of

8.5 for the Al(OH)3 solubility control. The IAP valuesfor Fe(III) range from ca. 2 to 8 (Fig. 4). Values of log

Kso,25 for Fe(OH)3 as high as 5 have been reported, for

fresh precipitates, but aged material is appreciably less

soluble, with log Kso,25 $ 2.5 (e.g. Baes & Mesmer,

1976). Therefore, it seems likely that virtually all the soil

solutions are oversaturated with respect to Fe(OH)3,

and that much of the dissolved Fe(III) is due to col-

loidal Fe(OH)3. The low-pH results obtained with

WHAM/Model VI (Fig. 4) suggest a solubility product

of ca. 3 at 20 C, which corresponds to log Kso,25=2.7,

if an enthalpy of reaction of 102 kJ mol1 (Liu and

Millero, 1999) is applied.

Fig. 5 compares free metal ion concentrations calcu-lated with the two models. The ranges of variation are

quite large, 6 log units for [Cu2+], 4 for [Zn2+], 2 for

[Cd2+] and 8 for [Pb2+]. For nearly all the samples, and

for each metal, agreement is within one order of mag-

nitude. WHAM/Model VI tends to give higher values of

[Zn2+] than WHAM, whereas the reverse is true for

[Cu2+]. Results for [Cd2+] and [Pb2+] are in closer

agreement. The differences between the models in their

predictions of free metal concentrations arise partly

from their different assumptions, in particular the

inclusion of low-abundance, high-affinity binding sites

in WHAM/Model VI, and partly because their para-

meter values are derived from different data sets (moredata were available to parameterise Model VI). Fig. 6

compares free metal ion concentrations (WHAM/

Model VI) with total dissolved concentrations. It is seen

that Cu and Pb are extensively complexed, Zn and Cd

less so. These differences reflect the much stronger

interactions of Cu and Pb with dissolved organic mat-

ter. (Hereinafter, the results refer only to speciation

calculations performed with WHAM/Model VI.)

3.4. Multiple regression analysis based on speciated soil

solution

Multiple regression analysis was applied in order to

derive relationships between solid-phase metal contents

and the free metal ion concentrations, calculated as

described earlier. Since the complexing effects of dis-

solved organic matter are accounted for by WHAM/

Model VI, the most likely determinants of solidsolu-

Fig. 4. Ion activity products for Al(OH)3 (filled circles) and Fe(OH)3(open circles) as functions of pH, calculated for the soil solutions using

WHAM/Model VI (see Section 3.3).

Fig. 5. Comparison of free metal ion concentrations calculated with

WHAM and WHAM/Model VI. Dotted lines indicate one order of

magnitude.



8/13

tion distributions were taken to be MHNO3, LOI and

pH, suggesting the following regression equation:

log M2

a log %LOI b pH c log MHNO3 d 3

The regressions using these variables explained much of

the variation in the free metal ion concentrations, with

values ofr2 in the range 0.760.95 (Table 3). For all four

metals the values of r2 are appreciably greater than

those for the prediction of total dissolved concentra-

tions (Table 2). In all cases pH was the most important

variable, followed by MHNO3 and LOI. The relative

importance of the soil variables is due in part to their

degrees of variation. Thus, the range of pH is ca. 5 (log)units, that of log MHNO3 23 units, and that of log%

LOI only 1 unit. Inclusion of soil Al and Fe contents in

the regressions had hardly any effect, increases in r2

being


9/13

3.5. Modelling solidsolution distributions with WHAM/

Model VI

The purpose here is to represent the ion-binding

components of the soil solid phase by humic substances,

and to use WHAM/Model VI to predict solidsolution

distributions of metal ions. The modelling requires all

significant ions to be taken into account. Thus we are

attempting to describe simultaneously the distributions

in the soilwater systems of H+, base cations, Al,

Fe(III), and trace heavy metals (see Section 3.3 and

Fig. 3). As input data, we have total soil contents of

metals, DOC concentrations, and estimated solution

concentrations of acid anions (Section 3.3). Solubilitycontrol by Al(OH)3 and Fe(OH)3 was assumed, as

described in Section 3.3. The soil version of WHAM/

Model VI operates by computing pH, forcing the sys-

tem to be in charge balance. It was therefore applied by

adjusting the soil content of active humic substances in

order to make the calculated pH the same as the

observed value, for each soil sample. The ratio of HA to

FA was taken to be 84:16, as determined in previous

work (Tipping et al., 1995a). Note that this procedure

does not involve the optimisation of metal soilsolution

distributions, and therefore the models are being used

purely predictively with regard to those distributions.

Model success or failure can be judged by comparing

the predictions either with total dissolved metal con-centrations or with free metal ion concentrations com-

puted just from the solution compositions.

Application of WHAM/Model VI to the data was

successful for 93 of the 98 data sets; in the remaining

five, the required total concentration of humic sub-

stances exceeded the total organic content of the soil,

suggesting either model failure or analytical error.

Ignoring those five results, the average ratio, RHS (g

g1), of active humic substances to LOI ranged from

0.09 to 0.86. Essentially, the fitting exercise has deter-

mined the content of active HS in each soil. If these

values are plotted against %LOI (Fig. 9) it is seen that,

at low values of LOI the active HS content increasesapproximately linearly with %LOI, whereas at %LOI

values greater than about 40% the active soil content is

approximately constant, with an average value of 0.17 g

g1. This value is in agreement with values found in

previous studies, where WHAM was applied to the

results of batch titration experiments with organic soils

(Tipping et al., 1995a; Lofts et al., 2001).

Predicted total solution concentrations of heavy

metals are compared with observed values in Fig. 10.

The results show considerable scatter, but the predicted

values are generally within an order of magnitude of the

observations. The average ratios of predicted toobserved total metal for Cu and Zn are close to unity

(1.4 and 0.7, respectively), whereas for Cd and Pb the

model predictions are generally too high (average ratios

of 2.8 and 8.9, respectively). In the case of Cu, the pre-

dictions are poorly correlated with the observations

(r2=0.34, for log data). Better correlations are obtained

Fig. 8. Comparison ofKD values predicted by multiple regression with

values calculated from WHAM/Model VI speciation of the soil solu-

tions. Dotted lines indicate one order of magnitude.

Fig. 9. Soil contents of active humic substances obtained in the

calibration of WHAM/Model VI (Section 3.5).

Fig. 10. Comparison of total solution metal concentrations, predicted

by whole-soil modelling using WHAM/Model VI, with observations.

Dotted lines indicate one order of magnitude.



10/13

for the other three metals, values ofr2 being 0.63 for Zn,

0.54 for Cd and 0.83 for Pb. This trend in r2 follows the

range of total solution concentrations, which is only

about a factor of 30 for Cu, a factor of ca. 100 for Zn

and Cd, and a factor of ca. 1000 for Pb.

The plots in Fig. 11 show that the whole-soil model

applications provide estimates of free metal ion con-centrations that correlate well with the estimates made

by speciating only the soil solution. The values of r2 for

log-transformed data are 0.89, 0.81, 0.71 and 0.94 for

Cu, Zn, Cd and Pb, respectively. We find that con-

centrations of Cu2+ tend to be underestimated, by an

average factor of 2.2, and those of Zn2+ by a factor of

3.7, while concentrations of Cd2+ and Pb2+ are over-

estimated by factors of 1.1 and 3.7.

3.6. Comparison with the results of Sauve and

colleagues

As mentioned in Section 1, Sauve and colleagues haveanalysed metal chemistry data for contaminated mineral

soils by multiple regression, using measured free metal

ion concentration as the key variable to be explained. It

is of interest to explore how well their equations account

for the data presented in this paper, and also to examine

whether our equations can account for their observa-

tions. For copper, the following equation was obtained

by McBride et al. (1997):

pCu 1:28 1:37 pH 1:95 log CuT 1:95 log OM

n 68; r2 0:80

6

Here, pCu is the negative logarithm of the Cu2+ activ-

ity, CuT is the total soil copper, in mg kg1, and OM is

the organic matter content of the soil in gC kg1. An

additional equation for copper was derived by Sauve et

al. (1998a):

pCu 3:42 1:4 pH 1:7 log CuT

n 66; r2 0:857

Table 5 shows that Eq. (6) predicts our data poorly, allthe values of Cu2+ activity (aCu2+) being two-to-three

orders of magnitude too small, although the correlation

between predicted and observed values was good. Eq.

(7) is much more successful, predicting aCu2+ values on

average only five times too small, and with a high

correlation. Sauve et al. (2000) reported the following

equation for Cd:

pCd 5:14 0:61 pH 0:79 log CdT

n 64; r2 0:708

and this gave predictions that were, on average, within a

factor of two of the values of aCd2+ obtained in thepresent study, although the correlation was relatively

low (r2=0.60). Finally, Sauve et al. (1998a) produced

this regression for Pb:

pPb 6:78 0:62 pH 0:84 log PbT

n 84; r2 0:649

which predicted aPb2+ values for our data that were, on

average, 2.5 times the observations, with a high corre-

lation (r2=0.92). When our regression Eq. (3), with the

coefficients given in Table 3, was applied to the data of

Sauve et al. (1997) for Cu, and Sauve et al. (2000) forCd, the metal activities were of the right magnitude, on

average, but the correlations were low, with r2 $ 0.5

(Table 5). In both cases, the predicted values tended to

be too high when metal activities were low. The pub-

lished data sets of Sauve and colleagues for Cu and Cd

Fig. 11. Comparison of free metal ion concentrations, predicted by

whole-soil modelling using WHAM/Model VI, with values obtained

from speciation calculations for the soil solution only. Dotted lines

indicate one order of magnitude.

Table 5

Comparisons of multiple regression equation predictions. Eqs. (6)(9)

were used to predict the results of the present study. Eq. (3), derived in

the present study, was used to predict the results of Sauve et al. (1997,

2000)

Data set n Eq. Mean r2

Present Cu 98 6 0.003a 0.83

Present Cu 98 7 0.2a 0.87

Present Cd 98 8 1.6a 0.60

Present Pb 98 9 2.5a 0.92

Cu, Sauve et al. (1997) 68 3 0.7b 0.50

Cd, Sauve et al. (2000) 61 3 4.0b 0.54

n=no. of data points.a apred/aWHAMVI is the ratio of the activity predicted with the spe-

cified equation (see text) to the activity estimated by speciation of the

soil solution with WHAM/Model VI.b apred/aobs is the ratio of the activity predicted with Eq. (3) to the

activity determined experimentally.



11/13

can be combined with the data reported in the present

study to generate the following regression equations:

log aCu 5:35 1:17 pH 1:09 log MCu

0:52 log %LOI

n 165; r2 0:87;

10

log aCd 0:22 0:45 pH 0:77 log MCd

0:21 log %LOI

n 158; r2 0:69:

11

4. Discussion

The present data set covers wide ranges of soil condi-

tions in terms of pH, base cations and organic matter,

although it does not include LOI levels lower than 10%.

The geographical spread of sampling sites means that

most situations in the uplands of England and Walesare represented. Although acid soils are the most com-

mon, there are appreciable numbers of samples with

high pH, the metal chemistries of which appear to be

governed by the same processes that operate in the acid

samples. The main inadequacy of the present data is

that they do not include measurements of either free

metal ions or other labile forms of the metals. For

logistical reasons, and with the available resources, this

would have been difficult to achieve for four metals in

so many samples. Therefore, we have used speciation

modelling, and the analysis depends very much on the

ability of the models to provide accurate estimates offree metal ion concentrations. As mentioned in Section

3.3, there is published evidence that the models have

such capabilities, but model-testing is far from com-

plete, and the calculated values should therefore be

regarded with appropriate caution. The fact that the

calculated free metal ion concentrations lead to a

coherent analysis of the partitioning data adds indirect

support to the speciation modelling approach.

The advantage of having a large, representative data

set is that it provides a good test of the ability of mod-

elling approaches to provide general predictions, that

could be used to map actual and potential toxic effects

(see below). Logistical constraints and resource limi-tations mean, however, that only one solidsolution

partitioning state is observed for each metal in each soil.

This means that the effect on partitioning of, for exam-

ple, pH is seen only by comparing soil samples of dif-

ferent pH, rather than arranging for a single sample to

experience a range of pH. Therefore, useful com-

plementary information could be obtained from more

detailed study of a sub-set of samples, for example by

acid-base titrations (Tipping et al., 1995a,b; Lofts et al.,

2001). Work on this aspect of soil metal chemistry is

currently in progress in our laboratories.

The measured concentrations of DOC include some

high values (>100 mg l1), which far exceed the con-

centrations (130 mg l1) that are observed either in the

porewaters of upland organic soils, collected by suction

lysimeters (Hughes et al., 1994; Adamson et al., 2001) or

in upland UK surface waters (see e.g. Tipping et al.,

1988; Scott et al., 1998; Monteith and Evans, 2000).Therefore it seems unlikely that the soil solutions sam-

pled in the present work are representative of soil water

in situ. For example, it may be that the disturbance of

soil structure caused by sampling leads to the solubili-

sation of entrapped dissolved organic matter, or to

peptisation of colloidal forms. This uncertainty is

important if soil solution composition is used to deter-

mine soil metal budgets, since metal associated with

apparently soluble, but practically immobile, DOC

would cause the overestimation of metal leaching losses.

With regard to metal bioavailability, the uncertainty

would be significant if total solution concentrations

were considered, but inconsequential if free metal ionconcentrations were used.

It can be concluded from the results of multiple

regression analysis involving total solution metal con-

centrations that metal distributions are determined

mainly by sites on organic matter, at which metal ions

and protons compete for binding (cf. Fig. 3). Although

significant correlations can be obtained from such a

regression analysis, considerably higher values of r2 are

achieved when the soil solution is speciated (by model

calculation). The improvements arise because metal

binding by organic matter cannot be represented satis-

factorily by combinations of linear terms, as is requiredby multiple regression analysis for unspeciated soil

solutions. The results provide strong evidence for the

key role of the free metal ion concentration in deter-

mining solidsolution partitioning. The comparisons of

our regression equations and data with those of Sauve

and colleagues (Section 3.6), and the modelling of com-

bined data sets, attest to the robustness of the multiple

regression approach. The regression equations are easy

to use, and require relatively little input data, but should

not be used outside the ranges of input data used to

derive them. They are well suited to the estimation of

metal partitioning from basic soil characteristics, with

limited information about soil solution composition,and for large numbers of examples, as might be required

in mapping exercises.

The prediction of metal solidsolution partitioning

with WHAM/Model VI can be considered reasonably

successful (Section 3.5, Figs. 10 and 11), and given the

consistency of under- or over-prediction of free metal

ion concentrations (Fig. 11), the model could be opti-

mised straightforwardly by making adjustments (which

would be modest) of equilibrium constants for metal

binding. Calibration of the model by estimation of the

active content of humic matter is also required, and the



12/13

results in Fig. 9 could be used to derive average humic

contents, dependent upon LOI. WHAM/Model VI as

presently applied to soils requires the system to be

charge balanced, and therefore needs information, or

assumptions, about total soil composition, if it is to be

used to estimate metal speciation in a given soil. On the

other hand, the ability of the model to speciate theentire soil means that it can be used to estimate how soil

solution composition, notably pH and free metal ion

concentrations, would depend upon, for example,

changing inputs of acidifying pollutants. Thus,

WHAM/Model VI is potentially a powerful tool in

describing and predicting temporal changes in soil

chemistry, within a consistent framework.

One reason for carrying out the present work was to

contribute to the evaluation of critical loads approaches

in the environmental risk assessment of heavy metals

transported by long-range (transboundary) atmospheric

processes. The critical load is derived from a critical

limit, above which the metals are judged to have dele-terious effects. In the case of soils, the critical limit for a

given metal may be expressed, for example, by the total

soil metal content, the total solution metal concen-

tration, or the free metal ion concentration. Calculation

of the critical load requires an understanding of metal

behaviour, including solidsolution partitioning, fol-

lowing atmospheric deposition, in order to calculate the

deposition (critical load) at which the critical limit

would be exceeded if the system were at steady-state (de

Vries and Bakker, 1998). The multiple regression mod-

els, combined with the application of WHAM/Model

VI to compute solution speciation, would be suitablefor use in the calculation of steady state metal con-

centrations, given information about metal inputs

(including deposition and weathering), removal pro-

cesses (e.g. the harvesting of plants), soil properties, soil

pH, and the DOC concentration in drainage water. In

this way, calculations could be performed to determine

the critical load per se, and also, combined with soil

inventory data, to assess whether current solution con-

centrations and/or free ion activities exceed critical

limits.

An objection to steady-state modelling is that soil

systems can take long periods to reach steady state (see

e.g. Paces, 1998). Dynamic modelling may therefore bemore appropriate, at least to explore the time scales

involved, and again the models described here are

potentially applicable. Information, or assumptions,

about temporal changes in metal inputs to the system

would be required, and account should also be taken of

changes in soil conditions, notably in acidification sta-

tus. The use of the multiple regression equations would

require input data describing variations in soil pH,

whereas WHAM/Model VI could be used to simulate

acidification processes and metal behaviour simulta-

neously.

Acknowledgements

We are grateful to B.M. Simon (CEH Windermere)

for performing DOC determinations, to Barry Cole

(Imperial College) for assistance with metal analyses,

and to L. Baldwin (CEH Windermere) for secretarial

assistance. This work was supported by the Environ-mental Diagnostics Programme of the UK Natural

Environment Research Council, and by the Department

of the Environment, Transport and the Regions (now

the Department for Environment, Food and Rural

Affairs), the Scottish Executive, the National Assembly

of Wales and the Department of the Environment (in

Northern Ireland) under contract EPG 1/3/144.

References

Adamson, J.K., Scott, W.A., Rowland, A.P., Beard, G.R., 2001. Ionic

concentrations in a blanket peat bog in northern England and

correlations with deposition and climate variables. Eur. J. Soil Sci.

52, 6979.

Allen, H.E., 1993. The significance of trace metal speciation for water,

sediment and soil quality criteria and standards. Sci. Tot. Environ.

Supplement (Part 1), 2345.

Baes, C.F., Mesmer, R.E., 1976. Hydrolysis of Cations. Wiley, New

York.

Benedetti, M.F., van Riemsdijk, W.H., Koopal, L.K., Kinniburgh,

D.G., Gooddy, D.C., Milne, C.J., 1996. Metal ion binding by nat-

ural organic matter: from the model to the field. Geochim. Cosmo-

chim. Acta 60, 25032513.

Bryan, S.E., Tipping, E. Hamilton-Taylor, J., 2002. Comparison of

measured and modelled copper binding by natural organic matter in

freshwaters. Comp. Biochem. Physiol. C, 133, 3749.

de Vries, W., Bakker, D.J., 1998. Manual for Calculating Critical

Loads of Heavy Metals for Terrestrial Ecosystems (Report 166).

DLO Winand Staring Centre (SC-DLO), Wageningen, The Nether-

lands.

Dwane, G., Tipping, E., 1998. Testing a humic speciation model by

titration of copper-amended natural waters. Environ. Int. 24, 609

616.

Hughes, S., Norris, D.A., Stevens, P.A., Reynolds, B., Williams, T.G.,

1994. Effects of forest age on surface drainage water and soil solu-

tion aluminium chemistry in stagnopodzols in Wales. Water, Air

and Soil Pollution 77, 115139.

Jopony, M., Young, S.D., 1994. The solidsolution equilibria of lead

and cadmium in polluted soils. Eur. J. Soil Sci. 45, 5970.

Langford, C.H., Khan, T.R., 1975. Kinetics and equilibrium bindingof Fe3+ by a fulvic acid: a study by stopped flow methods. Can. J.

Chem. 53, 29792984.

Lawlor, A.J., Tipping, E., 2002. Metals in bulk deposition and surface

waters at two upland locations in northern England. Environ. Pol-

lut. 121, 153167.

LaZerte, B.D., 1989. Aluminium speciation and organic carbon in

waters of central Ontario. In: Lewis, T.E. (Ed.), Environmental

Chemistry and Toxicity of Aluminium. Lewis, Chelsea, MI, pp.

195207.

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, 34183424.

Liu, X., Millero, F.J., 1999. The solubility of iron hydroxide in sodium

chloride solutions. Geochim. Cosmochim. Acta 63, 34873497.



13/13

Lofts, S., Woof, C., Tipping, E., Clarke, N., Mulder, J., 2001. Mod-

elling pH buffering and aluminium solubility in European forest

soils. Eur. J. Soil Sci. 52, 116.

McBride, M., Sauve , S., Hendershot, W., 1997. Solubility control of

Cu, Zn, Cd and Pb in contaminated soils. Eur. J. Soil Sci. 48, 337

346.

McGrath, S.P., Loveland, P.J., 1992. The Soil Geochemical Atlas of

England and Wales. Blackie, London.

Monteith, D.T., Evans, C.D., 2000. UK Acid Waters Monitoring

Network: 10 Year Report. ENSIS, London.

Paces, T., 1998. Critical loads of trace metals in soils: a method of

calculation. Water Air Soil Pollut. 105, 451458.

Palmer, D., Wesolowski, D.J., 1992. Aluminium speciation and equi-

libria in aqueous solution: II. The solubility of gibbsite in acidic

sodium chloride solutions from 30 to 70 C. Geochim. Cosmochim.

Acta 56, 10931111.

Patrick, S., Monteith, D.T., Jenkins, A., 1995. UK Acid Waters

Monitoring Network: The First Five Years. ENSIS Publishing,

London.

Rowell, D.L., 1994. Soil Science: Methods and Applications. Long-

man, Harlow.

Sauve , S., Dumestre, A., McBride, M., Hendershot, W., 1998a. Deri-

vation of soil quality criteria using predicted chemical speciation of

Pb2+ and Cu2+. Environ. Toxicol. Chem. 17, 14811489.

Sauve , S., McBride, M., Hendershot, W., 1998b. Soil solution specia-

tion of lead(II): effects of organic matter and pH. Soil Sci. Soc. Am.

J. 62, 618621.

Sauve , S., McBride, M., Norvell, W.A., Hendershot, W.H., 1997.

Copper solubility and speciation of in situ contaminated soils:

effects of copper level, pH and organic matter. Wat. Air Soil Pollut.

100, 133149.

Sauve , S., Norvell, W.A., McBride, M., Norvell, W.A., Hendershot,

W.H., 2000. Speciation and complexation of cadmium in extracted

soil solutions. Environ. Sci. Technol. 34, 291296.

Scott, M.J., Woof, C., Jones, M.N., Tipping, E., 1998. Concentrations

of dissolved organic carbon in drainage water from an upland peat

system. Environ. Internat. 24, 537546.

Snedecor, G.W., Cochran, W.G., 1967. Statistical Methods, sixth ed.

Iowa State University Press, Ames.

Tipping, E., 1994. WHAMa chemical equilibrium model and com-

puter code for waters, sediments and soils incorporating a discrete

site/electrostatic model of ion-binding by humic substances. Comp.

Geosci. 20, 9731023.

Tipping, E., 1998. Humic Ion-Binding Model VI: an improved

description of the interactions of protons and metal ions with humic

substances. Aq. Geochem. 4, 348.

Tipping, E., 2002. Cation Binding by Humic Substances. Cambridge

University Press, Cambridge.

Tipping, E., Berggren, D., Mulder, J., Woof, C., 1995a. Modelling the

solid-solution distributions of protons, aluminium, base cations and

humic substances in acid soils. Eur. J. Soil Sci. 46, 7794.

Tipping, E., Hilton, J., James, J.B., 1988. Dissolved organic matter in

Cumbrian lakes and streams. Freshwat. Biol. 19, 371378.

Tipping, E., Hurley, M.A., 1992. A unifying model of cation binding

by humic substances. Geochim. Cosmochim. Acta 56, 36273641.

Tipping, E., Rey-Castro, C., Bryan, S.E., Hamilton-Taylor, J., 2002.

Al(III) and Fe(III) binding by humic substances in freshwaters, and

implications for trace metal speciation. Geochim. Cosmochim. Acta

66, 32113224.

Tipping, E., Smith, E.J., Lawlor, A.J., Hughes, S., Stevens, P.A., 2003.

Predicting the release of metals from ombrotrophic peat due to

drought-induced acidification. Environ. Pollution 123, 239253.

Tipping, E., Woof, C., Kelly, M., Bradshaw, K., Rowe, J.E., 1995b.

Solid-solution distributions of radionuclides in acid soils: applica-

tion of the WHAM chemical speciation model. Environ. Sci. Tech-

nol. 29, 13651372.

Vulkan, R., Zhao, F.-J., Barbosa-Jefferson, V., Preston, S., Paton,

G.I., Tipping, E., McGrath, S.P., 2000. Copper speciation and

impacts on bacterial biosensors in the pore water of copper-con-

taminated soils. Environ. Sci. Technol. 34, 51155121.


the solid–solution partitioning of heavy metals (cu, zn, cd, pb) in

Documents