application of a geochemical transport model to predict
TRANSCRIPT
-
8/8/2019 Application of a Geochemical Transport Model to Predict
1/8
Application of a geochemical transport model to predict
heavy metal retention (Pb) by clay liners
Ray E. Ferrell a,*, Per Aagaardb, Johan Forsman a, Lea Greenwood a, Zuoping Zhengb
aDepartment of Geology and Geophysics, Louisiana State University, Baton Rouge, LA 70803, USAbDepartment of Geology, University of Oslo, Oslo, Norway
Received 25 October 2000; received in revised form 2 February 2001; accepted 16 July 2001
Abstract
PHREEQC, a geochemical transport model, is used to simulate diffusive transport of Pb through a 10-cm-thick clay liner.
The results are compared to those of Roehl and Czurda [Applied Clay Science 12 (1998) 387] who studied Pb migration by
diffusion in a carefully monitored laboratory experiment. The computer simulation accounts for effects due to adsorption by ion
exchange, changes in CEC, variable ion selectivity, and porosity or compacted density. It facilitates evaluation of changes in the
diffusion coefficient and solution input parameters. The effective Pb diffusion coefficient determined for the simulation is
3 10 10 m2 s 1 and for the 520-day experiment of Roehl and Czurda it is 2.3 10 10 m2 s 1. Differences in theretardation factors (23.6 and 503, respectively) indicate that the model does not account for all of the adsorption mechanisms
suggested by the experimental investigation. Thus, less Pb is retained and the liner is predicted to fail more rapidly than theactual results indicate. Models have great flexibility, but need to be verified by field data before they can be applied to specific
waste site conditions. D 2002 Elsevier Science B.V. All rights reserved.
Keywords: Transport; Lead; Clay liner; Diffusion
1. Introduction
Geochemical transport models have great utility in
assessing the retention of potential heavy metal con-taminants by clay liners. Natural and compacted clay-
rich materials and geosynthetic textiles containing
clay minerals are among the most commonly em-
ployed methods to isolate landfill wastes from the
environment. Clay liners are particularly important
with respect to the role that they may play in the
isolation of high-level nuclear wastes (Madsen, 1998).
Computer models incorporating the results of modern
studies of reaction mechanisms provide the opportu-
nity to examine factors affecting waste retention andoffer the potential to identify those factors that are
most likely to influence the integrity of natural and
engineered clay barriers. Several models (Selim, 1992)
based on the assumption of local equilibrium are
available to assess processes such as ion exchange,
complexation, and competitive adsorption. Others
have developed multireaction multisite models incor-
porating varying kinetic effects and irreversible reac-
tions (Selim, 1992; Hinz et al., 1992). One of the most
thorough laboratory and field studies of Cd, Pb, Zn
0169-1317/02/$ - see front matterD 2002 Elsevier Science B.V. All rights reserved.P I I : S 0 1 6 9 - 1 3 1 7 ( 0 1 ) 0 0 0 9 2 - 8
* Corresponding author. Fax: +1-225-578-2302.
E-mail address: [email protected] (R.E. Ferrell).
www.elsevier.com/locate/clay
Applied Clay Science 21 (2002) 5966
-
8/8/2019 Application of a Geochemical Transport Model to Predict
2/8
and Cr diffusion and sorption by clay minerals is that
reported by Wagner (1992).
PHREEQC (Parkhurst, 1995) is one of the com-
puter programs for geochemical reaction modelling incommon use. Based on an ion-association aqueous
model, it can perform speciation and saturation index
calculations; simulate reaction-paths and 1D-transport
involving reversible reactions, and inverse modelling
of reactions to account for differences in composition
between waters. Reactions considered include those
involving aqueous, mineral, and gaseous phases;
solid-solution, surface complexation and ion-ex-
change equilibria; and irreversible reactions involving
specified mole transfers of reactants, kinetically con-
trolled reactions, mixing of solutions, and temperature
changes. The latest version of PHREEQC (Parkhurst
and Appelo, 1999) includes capabilities to simulate
dispersion or diffusion in 1D-transport calculations,
with or without advective transport. A WINDOWS
version of PHREEQC prepared by Vincent Post may
be obtained free of charge from the website, http://
www.geo.vu.nl/users/posv/phreeqc.html.
The purpose of this paper is to illustrate the use of
PHREEQC to simulate changes in clay content, cation
exchange capacity, metal concentration, ion exchange
and other factors on the diffusive transport of Pb
through a 10-cm barrier. Diffusion may be the mostimportant transport processes when the permeability is
about 110 9 m s 1 (Shackelford, 1991). Allcalculations were performed without advective trans-
port, thus focusing on reactions within the liner. The
results represent the best-case situation when com-
pared to actual waste repositories, as some advective
transport is often observed (Johnson et al., 1989).
Computed results are compared with the experimental
results of Roehl and Czurda (1998) for Pb in order to
assess the validity of the model assumptions.
2. Methods
2.1. Simulation conditions
PHREEQC commands for simulation of Pb trans-
port by diffusion are presented in Table 1. The major
command blocks are identified with all upper case
letters. SOLUTION is used to define the temperature
and composition of up to 20 initial solutions. In this
example, a low ionic strength calciumsodium bicar-
bonatechloride solution has been used to represent
the initial composition of the pore water in the barrier.
The solution contained 0.9 mmol/kgw Ca, 0.1 mmol/
kgw Na, 0.2 mmol/kgw Cl and alkalinity of 1.7 mmol/
kgw expressed as HCO3. The amounts and types of
exchangers are defined in the EXCHANGE block. X
equals the CEC (equivalents per l i ter) of the
Table 1
Typical input for simulation of diffusive flux through a clay barrier
(Pb) with ion exchange
SOLUTION 1 20
-units mmol/kgw
Ca 0.9
Na 0.1
Alkalinity 1.7 as HCO3
Cl 0.2
EXCHANGE 1 20
X 1.0
-equil 120
EXCHANGE_SPECIES
P b + 2 + 2 X = PbX2log_k 1.05
END
SOLUTION 0
-units mmol/kgwpH 6
Pb 1
Cl 2
END
TRANSPORT
-cells 10 #number of cells
-length 0.01 #length (m) of cells
-flow_direction diffusion_only
-shifts 1 #value times time_st is total time for
diffusion calculation
-diffc 3.0e 10 # diffusion coefficient = 3.0e 10 m2/s-bcon constant flux
-time_st 3.15e7 #1 year
USER_GRAPH-heading Distance/m Cl Cd CdX2 CaX2 NaX
-axis_scale x_axis 0 1
-axis_scale y_axis 0 1
-axis_titles Distance/m mol/l
-chart_title
-initial_solutions false
-start
10 graph_x Dist
20 graph_y tot(Cl) * 1e6, tot(Pb) * 1e6,
mol(PbX2) * 1e6, mol(CaX2) * 1e6, mol(NaX) * 1e6
-end
END
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 596660
-
8/8/2019 Application of a Geochemical Transport Model to Predict
3/8
exchanger utilised in these simulations and CaExchinitially occupies more than 99% of the exchange
sites. Half reactions and relative log k values for
exchange reactions are defined next. This block(EXCHANGE_SPECIES) is ordinarily left blank
and values in the database file are used automatically.
Alternatively, the user can identify new half reactions
or modify the relative log k. SOLUTION 0 establishes
the composition of the input fluid. In this case, it is a
pH 6.0 PbCl2 solution containing 1 mmol/l of the
metal. Conditions for the TRANSPORT calculations
are established next. The cell was divided into 10
segments 0.01 m in length, the effective diffusion
coefficient (Deff) was 3 10 10 m2 s 1 and the
experiment was conducted for 3.15 108 s, or 10years. Deff is often used to represent the diffusion
coefficient derived from laboratory or field determi-
nations of element profiles. It is equal to the product
(Roehl and Czurda, 1998) of the free solution diffu-
sion coefficient (D0) and the tortuosity factor (s)
divided by the retardation factor (R). Graphing and
output conditions are established in the last block.
END statements throughout signify the end of each
input block.
2.2. Variable clay properties
The influence of changes in clay type on the metal
diffusion profile can be simulated in three ways.
Variations in the total quantity of clay and the type
of clay can be used to modify the total sample CEC.
Ten weight percent clay with a CEC of 100 meq/100 g
is equivalent to 50 wt.% clay with a CEC of 20 meq/
100 g. Porosity (e) and bulk density (qb) also influ-
ence the available cation exchange capacity. CEC in
meq/l can be obtained by multiplying the value
measured as meq/100 g by 10qb/e (Appelo and
Postma, 1994). A sample with a CEC of 18.5 meq/100g with a bulk density (qb) of 1.68 g/cm3 and a porosity
(e) of 35% will have a CEC of 885 meq/l. Increasing
this value increases the quantity of an element
adsorbed by the solids and thus increases the effective
retardation. Converting CEC to a constant volume
facilitates comparison with the quantities of elements
in the pore solution and mass transfer associated with
reaction and transport through the clay. In most of the
simulations described below, the exchange capacity
was either 1.0 or 1.1 mol of a univalent exchanger/l.
A third effect attributable to changes in the char-
acter of the clay or clay minerals can be simulated by
changing the apparent exchange constant for the
metal/exchanger half-cell with regard to the Na/exchanger half-cell used as the reference reaction
(log k= 0.0). For example, changing the log k for Pb
exchange from 1.05 to 0.0 will eliminate the specific-
ity of the exchanger for the Pb ion and Na and Pb
exchange will proceed as a linear reaction.
2.3. Initial concentration of metal
The effects of initial metal concentration in the
input solution on values associated with a given depth
in a penetration profile can be determined from Ficks
second law of diffusion. The partial differential equa-
tion can be integrated to obtain a solution (Appelo and
Postma, 1994) for the concentration of a mobile
component at a given point and time that can be
expressed in terms of the initial or boundary condi-
tions. Increasing the input concentration increases the
concentration or apparent rate of penetration at a
given point.
3. Results
3.1. Ion exchange capacity
Increasing clay content or the CEC of the clay
produces the changes in the apparent penetration
profile of a metal illustrated in Fig. 1. In this diagram,
results for 10 years of diffusion are compared for
initial exchange capacities varying from 0.5 to 2.0
equivalents/l. The highest value is equivalent to a
sample with a CEC of 41 meq/100 g with 35%
porosity and a density of 1.7 g/cm3. This represents
a solid with 100 wt.% of a high CEC illite orvermiculite, or 41 wt.% smectite with a CEC of 100
meq/100 g mixed with a material with no ion
exchange capacity. Profiles shown depict changes in
exchangeable Ca and Pb plus Cl and Pb in the pore
fluid. Exchangeable Na is not illustrated because the
values are smaller than 0.0001 mol/l. After 10 years,
Pb occupies 50% of the exchange sites (0.5 mol PbX2)
at a depth of 1.0 cm when the CEC is 2 equivalents/l
(Fig. 1A). The depth to 50% exchangeable Pb
increases to 1.4 and 2.0 cm, respectively, when the
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 5966 61
-
8/8/2019 Application of a Geochemical Transport Model to Predict
4/8
exchange reservoir is reduced to 1.0 and 0.5 equiv-
alents/l. Cl in solution (Fig. 1B) has reached a steadystate equal to the input value in all cells because it is a
conservative element and is not involved in the
simulated reactions. The diamond-shaped data points
for ExcPb in the 0.5 equivalents/l simulation illustrate
how much more Pb is held in exchange positions than
in pore fluids. Depth of penetration for Pb in the pore
fluid (SolPB) varies inversely with the CEC of the
liner material. The depth to C/Co of 0.1 is 4.3, 2.7, and
2.0 cm when the CEC is 0.5, 1.0, and 2.0 equivalents/
l, respectively.
3.2. Ion exchange selectivity
Ion exchange selectivity can be modelled by mod-
ifying the log k of the element with regard to thereference half-cell. Changes in the relative ion
exchange constant for Pb produce the effects shown
in Fig. 2. After 10 years of diffusion, there is less ion
exchange when the relative ion exchange constant for
the Pb half cell (log k= 0.1) is close to that for Na. The
system retains a strong selectivity for Ca. The depth to
50% Pb exchanged decreases from 1.4 to 0.9 and 0.3
cm as the relative log k changes from 1.05 to 0.53 to
0.1 (Fig. 2A). The change in adsorption is accompa-
nied by a change in the depth to the Pb C/Co of 0.1
Fig. 2. Changes in Pb penetration profiles associated with variable
log k for Pb exchange in the simulated liner. (a) Exchangeable Ca
and Pb by clays with relative log kof 0.1, 0.53 and 1.05 with respect
to Na. (b) Pb and Cl in solution with one of the exchangeable Pb
values reproduced for comparison with a.
Fig. 1. Changes in Pb penetration profiles associated with variable
total cation exchange capacity (CEC) of the simulated liner. (a)
Exchangeable Ca and Pb at total exchange capacities of 0.5, 1.0 and
2.0 mol/l. (b) Pb and Cl in solution with one of the exchangeable Pb
values reproduced for comparison with a.
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 596662
-
8/8/2019 Application of a Geochemical Transport Model to Predict
5/8
(Fig. 2B). Decreasing the relative log k decreases the
adsorbed Pb, so the apparent migration takes place
more rapidly. Depths to C/Co of 0.1 are 2.7, 4.5, and
7.5 cm, respectively. Exchangeable Pb is still signifi-cantly higher than the quantity in solution. When the
barrier material has a stronger preference for the
naturally occurring elements in the pore fluid, the
metal contaminant will migrate more rapidly than
when there is a strong ion exchange preference for
the metal.
3.3. Effective diffusion coefficient
Changes in the effective diffusion coefficient uti-
lised in the simulation produce the changes in the
penetration profiles illustrated in Fig. 3 after 10 years
of transport. As the coefficient increases from 1.5 to
3.0, and 6.0 10 10 m2 s 1, the depth to 50%exchangeable Pb increases to 1.5, 1.9, and 2.4 cm,
respectively (Fig. 3A). The depth to C/Co of 0.1 for Pb
in the pore fluid also increases from 1.8, 2.4, and 3.8
cm (Fig. 3B) A 4 increase in the diffusion coef-ficient increases the rate of Pb penetration about 2 .
3.4. Input fluid composition
Increasing the concentration of Pb in the input fluidincreases the depth of Pb penetration as shown in Fig.
4. When Pb is 0.1 mmol/l and the ion exchange
capacity is 1.0 equivalent/l, exchangeable Pb is only
detectable in the first centimeter of the cell. When Pb
in the input solution is 10.0 mmol/l, the depth to the
50% exchangeable level increases to 3.7 cm (Fig. 4A).
At this high concentration level, Pb is no longer
confined to the simulated liner and C/Co in the last
cell of the barrier is 0.3 (Fig. 4B). Increases in the Pb
concentration in the input solution satisfy the demands
for ion exchange more rapidly. Subsequently, the build up and transport through the pore network is
more rapid.
3.5. Ionic strength of pore fluid
The chief consequence of increasing the ionic
strength of the pore fluid is a decrease in the apparent
selectivity of the ion exchanger for the ions that are
more strongly attracted to the exchanger at low con-
centrations. The exchange relative to Na becomes
more linear and effects similar to those illustrated in
Fig. 2 are observed. A 10 increase in millimoles ofNa and Ca in the pore fluid increases the mole fraction
of Na in the exchanger from < 0.01 to almost 0.04, sothe magnitude of the expected change in Pb retention
is small.
3.6. Summary of modelling results
The results of the transport simulations described
above illustrate several of the basic principles govern-
ing the retention of metal wastes by clay liners. The
magnitude and selectivity of the ion exchanger has a
major effect on the retardation of diffusing wastes.
Fig. 3. Changes in Pb penetration profiles associated with variable
effective diffusion coefficients (m2 s 1). (a) Exchangeable Ca and
Pb with coefficients of 1.5e 10, 3.0e 10, and 6.0e 10 m2 s 1.(b) Pb and Cl in solution with one of the exchangeable Pb values
reproduced for comparison with a.
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 5966 63
-
8/8/2019 Application of a Geochemical Transport Model to Predict
6/8
Important relationships controlling diffusive transport
include:
Increased CEC of the barrier materials retards
metal movement. Increased clay content increases the effective
CEC and thus promotes retardation. When the clay has a strong exchange preference
for an ion in the waste fluid compared to those
naturally present in the pore fluids of the barrier,
the input ion movement will be strongly retarded. Increasing the effective diffusion coefficient
increases the rate of penetration.
Rate of metal penetration is directly related to
increases in its concentration in the input
solution.
Large changes in the composition of the porefluid produce minor changes in the mole
fraction of weakly adsorbed ions, thus creating
the potential for minor increases in metal
penetration rate.
3.7. Comparison with experimental results
The experimental results obtained by Roehl and
Czurda (1998) for Pb provide an excellent basis for
comparison with PHREEQC simulations. The authors
assessed the retardation and solid speciation of Pb(also Cd) migrating into an illitic/smectitic loess loam
during experiments lasting from 43 to 520 days. The
loam had a CEC of 18.5 meq/100 g and experiments
were performed at a Proctor density of 1.68 g/cm3 at
moisture content of 17.5 wt.%. Calculated apparent
diffusion coefficients for input solutions containing
0.001 M Pb or Cd chloride were in the range of 3.6
7.4 10 9 cm2 s 1 for Pb and 4.9 6.7 10 8
cm2 s 1 for Cd, with a slight tendency to decrease
with the length of the experiment. The bulk of the
adsorbed metal was in the fraction defined as ex-
changeable (6080%) by selective sequential extrac-
tion procedures. The operationally defined carbonate
fraction (representing specific fixation on pH-depend-
ent sites) was next most abundant.
Fig. 5. Comparison of simulated Pb penetration profiles with data
(open circles and triangles) from the 520-day experiment of Roehl
and Czurda (1998).
Fig. 4. Changes in Pb penetration profiles associated with variable
Pb content and ionic strength of the simulated waste. (a)
Exchangeable Ca and Pb at total Pb concentrations of 0.1, 1.0,
and 10.0 mol/l. (b) Pb and Cl in solution with one of the
exchangeable Pb values reproduced for comparison with a.
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 596664
-
8/8/2019 Application of a Geochemical Transport Model to Predict
7/8
A comparison of the Roehl and Czurda (1998) data
for the 520-day experiment with PHREEQC-simu-
lated Pb pore fluid profiles is presented in Fig. 5. The
simulations were conducted with a CEC of 18.5 meq/
100 g, bulk density of 1.68, porosity of 36% and the
other input parameters identified in Table 1. Pb input
was constant and the bottom of the column was open.
Pb concentrations were modified by ion exchange
with a relative log k of 1.05. Effective diffusion
coefficients (Def f) were 1.0, 3.0, 4.0, 5.0 and
6.0 10 10. The best visual fit to the experimentaldata was obtained with Deff=3.0 10
10 m2 s 1.
Deff reported by Roehl and Czurda (1998) from other
laboratory experiments was 2.3 10 10 m2 s 1.The two values are essentially equal.
Retardation (R) in the simulation was evaluated at
several values of C/Co by dividing the depth of Cl
penetration by the corresponding depth for Pb. Results
are presented in Table 2. The simulated R is 23.6
when C/Co is equal to 0.5 and 22.3 for C/Co of 0.1.Simulated retardation decreases to about one-half
these values when C/Co is 0.01, presumably due to
dispersion introduced by the choice of cell size for the
model and other factors increasing the apparent dis-
persion of the advancing pore fluid front. Simulated
retardation values near 23 are 22 smaller than thevalue (503) reported by Roehl and Czurda (1998).
Differences in R are reflected in differences in the
apparent diffusion coefficients (Dapp), 4.7 10 13
and 1.3 10 11 m2 s 1, respectively. These coef-
ficients suggest slower rates of Pb transport in theexperimental situation because more Pb is retained by
reaction with the soil.
4. Discussion
The experiment and the simulation produced
nearly identical results for Deff, the diffusion coeffi-
cient that accounts for the influence of tortuosity on
the free-solution aqueous diffusion coefficient. The
differences between the simulated and experimental
results are most obvious with respect to R and Dapp. R
is a dimensionless retardation factor reflecting the
relative rate of transport of a non-reactive solute to areactive one. The apparent diffusion coefficient (Dapp)
is the parameter that most closely approximates the
movement of a reactive solute through a particular
soil. Dapp is the value that is often used to compare
element behaviour in different samples, but is mean-
ingless without an estimate of the retardation factor.
The factors are related by:
Dapp Deff=Retardation R
Differences in the estimate of the retardation factor
account for the large differences in the apparent dif-
fusion coefficient derived from the experiment and the
simulation. A good estimate of this R is critical when
attempting to evaluate the retention of Pb by clay liners.
R is different in the two instances described above
because it is calculated differently. Roehl and Czurda
(1998) used actual determinations of Pb in the soil and
an independent experimental determination of Deff to
calculate retardation. The PHREEQC simulation used
input values of Deff to match the Pb in the pore fluid
of the 520-day experiment and only precipitated the
amount of Pb required to satisfy a simple linear ionexchange process. Retardation was calculated by com-
parison with the hypothetical penetration profile of a
non-reactive solute. Actual measurements indicate
that the clay materials are retaining more Pb than
the simulation predicts.
The differences in retained Pb are apparent in Fig.
6, which compares the total quantity of Pb adsorbed by
ion exchange under the simulated transport conditions
with the actual measurements of Roehl and Czurda
(1998) after 121 and 520 days. The simulated values
are always lower than the measured ones but are veryclose to the quantities attributed to the exchangeable
fraction by Roehl and Czurda (1998). Comparing the
solid circles that represent measured values to the
curve drawn through the open circles (Fig. 6) illus-
trates that the simulated values are within 7080% of
the measured total after 520 days of reaction. The data
for 121 days provide a similar comparison. The
simulations fail to account for (up to 30%) the Pb
contained in the carbonatic fraction. Other sorption
mechanisms should be used to account for all the
Table 2
Calculated retardation factors (R)
C/Co Cl distance (cm) Pb distance (cm) R (Pb)
0.01 50.0 3.65 13.70.1 31.2 1.4 22.3
0.5 13.7 0.58 23.6
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 5966 65
-
8/8/2019 Application of a Geochemical Transport Model to Predict
8/8
reactions retarding the transport of Pb. In essence, the
PHREEQC model is 7080% effective in simulating
the adsorption observed in laboratory studies, and thus
predicts a more rapid failure of the clay liner.
The demonstration of the relative importance of
factors related to specific liner characteristics is one of
the most useful products of transport simulation.
PHREEQC provides a way to compare clays with
differing CECs, mineral content, ion exchange selec-tivity, and compacted density or porosity. Apparent
diffusion coefficients and the effects of different reac-
tion mechanisms can also be considered. The model
may also utilise variable input such as the composition
of the initial pore fluid or the metal-containing solution.
The chief limitations of the model are ones that
affect all models; they are only as good as the input
data and their ability to reproduce natural conditions.
It is best to use a model to extend interpretations based
on solid laboratory and field investigations. The model
can be used as a sensitivity tool to determine howchanges in field conditions such as porosity might
affect the integrity of clay liners or chemical conditions
that might accelerate transport by diffusion.
5. Conclusions
Simulated transport modelling provides an excellent
means to evaluate the relative influence of the variables
affecting Pb migration through a clay liner. The uncer-
tainties associated with the natural environment can
be modelled to evaluate their relative importance.
PHREEQC is a flexible routine that provides ways to
assess the roles of changing clay content, particularlyCEC and exchange ion specificity, on the calculated
element profiles produced by diffusive transport. Spe-
cific adsorption reactions can be included in the model
to account for the retardation of reactive elements.
Modelling is a useful approach for the study of metal
retention by clay liners. However, it must not be used
as a complete substitute for field and laboratory inves-
tigations. PHREEQC results are in generally good
agreement with Deffdetermined in the laboratory study
of Roehl and Czurda (1998) but differences in the
observed and predicted retardation factors result in the
more rapid transport of Pb in the model clay liner.
References
Appelo, C.A.J., Postma, D., 1994. . Geochemistry, Groundwater,
and Pollution. A.A. Balkema, Rotterdam, p. 536.
Hinz, C., Buchter, B., Selim, H.M., 1992. Heavy metal retention in
soils: applicationof multisite modelsto Zinc sorption. In: Islander,
I.K., Selim, H.M. (Eds.), Engineering Aspects of Metal-Waste
Management. Lewis Publishers, Boca Raton, FL, pp. 141 170.
Johnson, R.L., Cherry, J.A., Pankow, J.F., 1989. Diffusive contam-
inant transport in natural clay: a field example and implicationsfor clay-lined waste disposal sites. Environmental Science and
Technology 23, 340349.
Madsen, F.T., 1998. Clay mineralogical investigations related to
nuclear waste disposal. Clay Minerals 33, 109129.
Parkhurst, D.L., 1995. Users guide to PHREEQCa computer
program for speciation, reaction-path, advective-transport, and
inverse geochemical calculations, U.S. Geological Survey
Water-Resources Investigations Report, 95-4227, 143 pp.
Parkhurst, D.L., Appelo, C.A.J., 1999. Users guide to PHREEQC
(Version 2)a computer program for speciation, reaction-path,
advective-transport, and inverse geochemical calculations, U.S.
Geological Survey Water-Resources Investigations Report, 99-
4259, 312 pp.
Roehl, K.E., Czurda, K., 1998. Diffusion and solid speciation of Cd
and Pb in clay liners. Applied Clay Science 12, 387402.
Selim, H.M., 1992. Transport and retention of solutes in soils; mul-
tireaction and multicomponent models. In: Islander, I.K., Selim,
H.M. (Eds.), Engineering Aspects of Metal-Waste Management.
Lewis Publishers, Boca Raton, FL, pp. 117140.
Shackelford, C.D., 1991. Laboratory diffusion testing for waste dis-
posal: a review. Journal of Contaminant Hydrology 7, 177 217.
Wagner, J.-F., 1992. Verlagerung und Festlegung von Schwerme-
tallen in tonigen Deponieabdichtungen. Ein Vergleich von La-
bor- und Gelandestudien. Schriftenreihe Angewandte Geologie
Karlsruhe 22 246 pp.
Fig. 6. Total quantity of adsorbed Pb predicted by simulations based
on ion exchange (curves) compared with the data from 121- (open
squares) and 520-day (open circles) experiments of Roehl andCzurda (1998).
R.E. Ferrell et al. / Applied Clay Science 21 (2002) 596666