Dissolution and mass transfer of multiple chlorinated hydrocarbons under field conditions

Abstract

The time taken for a residual source of chlorinated hydrocarbons in an aquifer to dissolve without active intervention is controlled by the process that transfers the dissolved phase from the source to the flowing groundwater. Although various equilibrium and kinetic models have been proposed, the nature of this mass transfer process in the field is not yet well understood. To investigate mass transfer under field conditions, a three-dimensional multicomponent kinetic mass transfer model is calibrated against detailed field measurements of dissolved-phase concentrations at an emplaced DNAPL source at the Canadian Forces Base Borden, Ontario. The results show that, at least for the three year observation period, the mass transfer process is clearly equilibrium, and that observed early tailing is due to the declining mole fraction rather than to mass transfer kinetics.

Full text

Groundwater Quality: Remediation and Protection (Proceedings of the GQ'98 Conference held at
Tubingen, Germany, September 1998). IAHS Publ. no. 250, 1998. 173
Dissolution and mass transfer of multiple
chlorinated hydrocarbons under field conditions
EMIL O. FRIND, JOHN W. MOLSON & MARIO SCHIRMER
Department of Earth Sciences,
University
of
Waterloo,
Waterloo, Ontario, Canada N2L3G1
Abstract The time taken for a residual source of chlorinated hydrocarbons
in an aquifer to dissolve without active intervention is controlled by the
process that transfers the dissolved phase from the source to the flowing
groundwater. Although various equilibrium and kinetic models have been
proposed, the nature of this mass transfer process in the field is not yet well
understood. To investigate mass transfer under field conditions, a three-
dimensional multicomponent kinetic mass transfer model is calibrated against
detailed field measurements of dissolved-phase concentrations at an
emplaced DNAPL source at the Canadian Forces Base Borden, Ontario. The
results show that, at least for the three year observation period, the mass
transfer process is clearly equilibrium, and that observed early tailing is due
to the declining mole fraction rather than to mass transfer kinetics.
INTRODUCTION
Chlorinated hydrocarbons can be persistent sources of serious groundwater
contamination. Various methods for the remediation of such sources have been
proposed, but costs can be high and results uncertain (Pankow & Cherry, 1996). As
a result, there is an increasing interest in passive remediation. This option requires a
sound understanding of the controlling processes involved in the dissolution of a
hydrocarbon source.
Our focus here is on the processes that transfer a dissolved hydrocarbon from a
residual source in an aquifer to the flowing groundwater. Because the organic liquid
will occupy the larger pores in the medium, reducing the hydraulic conductivity of
the residual zone, flow will be partly redirected around the source and diffusion
could play a role in transferring the dissolved phase to the zone of flowing
groundwater. The time required to reduce the effluent below drinking water levels
will depend to a large extent on whether the mass transfer process from the source to
the flowing groundwater is an equilibrium process (i.e. the concentration of the
dissolved phase in the flowing groundwater is at the solubility limit of the
hydrocarbon), or a kinetic process driven by diffusion where the concentration in the
groundwater is below the solubility limit. Under a kinetic process, the effluent
concentration is subject to tailing which will extend the time taken for passive
remediation.
A large body of recent literature is devoted to the study of this mass transfer
process (see for example Mayer & Miller, 1992; Geller & Hunt, 1993; Powers et
al., 1994; Imhoff et al., 1994). Most of this work is based on the interpretation of
laboratory column experiments using some form of diffusion layer model (Sherwood
et al., 1975), and various kinetic or equilibrium mass transfer relationships have

174 Emil O. Frind et al.
been proposed. For example, Powers et al. (1994) found that in their experiments,
the mass transfer process is equilibrium at early times and kinetic at later times.
Mayer & Miller (1996) found that a homogeneous source will tend toward equilibrium
mass transfer while a heterogeneous source will tend toward kinetic mass transfer.
However, none of these models have as yet been tested under actual field conditions.
Under field conditions, DNAPL equilibrium concentrations have rarely been
observed (Feenstra & Cherry, 1988). As a result, there is some reason to believe that
mass transfer under field conditions might be kinetic. On the other hand, recent work
by McWhorter and Sale (1998) shows that under homogeneous conditions in the
field, mass transfer kinetics are relatively unimportant compared to the process of
carrying the dissolved NAPL away in the aquifer. Similarly Grafhwohl (1997) calcu-
lates that equilibrium concentrations should be reached in the groundwater relatively
quickly. The objective of the present work is to test these theories by simulating the
mass transfer process at a well-documented field site, using a kinetic model.
THE BORDEN EMPLACED SOURCE
This source (Fig. 1) consists of a mixture of three chlorinated hydrocarbons (TCM,
TCE, and PCE) that was emplaced into a sandy aquifer at the Canadian Forces Base
Borden, Ontario, in the late 1980s (Rivett et al., 1992). For placement, the source
area was isolated by sheetpiles, the sand removed and mixed above ground with the
DNAPL and a quantity of gypsum powder. The characteristics of the source are
summarized in Table 1, and the hydraulic parameters are provided in Table 2. The
plume of dissolved DNAPL was monitored for about three years by a three-
dimensional monitoring network of 2300 multilevel sampling points arranged in
several fences located at various distances along the axis of the plume. Of these
samplers, 173 are located at the 1-m fence (Fig. 1). Analyses of cores taken from the
source in 1994 and 1996 showed a negligible amount of TCM residual in the source,
but high amounts of TCE and PCE.
The peak concentrations, observed at the 1-m fence up to 1000 days for the three
organics, are shown in Fig. 2. In the transverse section (not shown), the observed
plumes are very narrow and sharply peaked, concentrating most of the mass flux
over an area of only one-quarter the width of the source area. The figure also shows
the solubility limit of each DNAPL calculated on the basis of the initial mole fractions
^^"
i
i 2m
2m{
7m 1
source 3.5 In
! •
1
qround surface
^
w watertable
^s*
1
"
1
"
1
>l
•
¥
flow
tâ*m MMB^mmMMM &>
Fig. 1 Site layout of the Borden emplaced source showing 1-m fence.

Dissolution and mass transfer of multiple
chlorinated hydrocarbons
under field conditions 175
Table 1 Emplaced source DNAPL properties.
Initial volume (1)
Density (kg l"1)
Initial mass (kg)
Molecular weight (kg mol"')
Moles
Initial mole fraction, X'"
Pure phase solubility, C0 (mg
1"
Initial effective solubility (mg 1
Diffusion coefficient, D'" (m2s"
Initial NAPL saturation, Sn0
Retardation coefficient, R
')
')
')
TCM
0.97
1.49
1.44
0.119
12.10
0.078
8670
676.3
9.0 x 10"10
1.1
TCE
6.11
1.46
8.92
0.131
68.09
0.439
1270
557.5
7.0 x 10"10
1.5
PCE
7.63
1.63
12.44
0.166
74.94
0.483
242
116.9
6.0 x 10"10
1.1
Total
14.71
22.8
155.13
1.000
0.05
Table 2 Physical properties.
Aquifer hydraulic conductivity, K (m s"1)
Porosity, 0
Hydraulic gradient
Median grain size, di0 (mm)
Residual water saturation, Snv
Specific storage, 5,
0.5 x 10"4- 1.0 x 10"4
0.33
0.0065
(spring)-0.0034 (summer)
0.15
0.07
0.001
present in the source. Most of the measured peak concentrations for TCE and PCE
are at or just below the solubility limit, while those for TCM decline below this
initial solubility limit after 400 days. The scatter in the measurements is most likely
due to the plume peak passing between the sampler points, which are spaced at 25-
50 cm intervals, as the flow gradient changed seasonally. Thus it is likely that TCE
and PCE dissolve at equilibrium rates during the period of measurement. A kinetic
process is possible for TCM, or for TCE and PCE at later times. To obtain insight
into the mass transfer process at this site, we will calibrate a three-dimensional (3-D)
kinetic mass transfer model to the concentration data at the 1-m fence.
MASS TRANSFER MODEL
The model is designed to handle the spatially and temporally variable mass transfer
of multiple DNAPL components in 3-D saturated groundwater systems. It is based
on the well-known diffusion layer concept where a mass flux of dissolved component
jio2
£101
o
°10°
LD
0-
x<
t
< 1 —1 —1
--^o =
9
_l
-.=s-
e
i
-=a~-
0
TXK
0
:
--B--
.«
observed CjC (y
• TCM
• TCE
x PCE
200 800 1000
400 600
Time (days)
Fig. 2 Peak concentrations observed at the 1-m fence, showing initial solubility limits.

176 Emil O. Frind et al.
m from the residual zone to the zone of flowing groundwater at point / is expressed
by:
j»'
=4'"(c;'-c;;;) (i)
where C" —
CQX"
is the effective solubility of component m as calculated by Raoult's
law (Mackay et al., 1991), with C0 being the pure-phase solubility and X'" being the
mole fraction, and C", is the concentration of component m in the flowing water at the
source. The mass transfer rate coefficient of component m at point / is expressed as:
D"' ( f-'SnY
0*50
) V S"0 J
where Sh is the Sherwood number, D'" is the aqueous diffusion coefficient of
component m, «SW, is the NAPL degree of saturation at point /, with Sn0 being the
initial value, and //" is the local volume fraction of NAPL m. Sh and p'" are
empirical parameters whose values are to be determined by calibration. Sh represents
the effects of the porous medium within the source area, including the surface area of
the NAPL blobs and the thickness of the conceptual diffusion layer; these geometric
factors are the same for each component in the source. The component volume
fraction
f"'Sn/Sn0
serves to represent a possible time dependence of the mass
transfer process. The exponent
(3'"
can be component-specific.
This mass transfer mechanism is built into a 3-D finite element model that
handles variably saturated flow coupled with multicomponent transport. The solution
is iterative and time-marching. The discretization varies from 0.05 m within the
source area to 0.1 m elsewhere in each coordinate direction, and the time step is 2.0
days.
The model was validated by applying it to the column experiments conducted
by Powers et
al.
(1994).
MODEL CALIBRATION
Because most measurements of dissolved TCE and PCE at the 1-m fence are at or
near equilibrium concentrations (Fig. 2), the source concentrations must be essentially
at equilibrium at least part of the time. Furthermore, the observed behaviour suggests
that dispersion between the source and the fence must be small. In order to reduce
transverse and upstream dispersion, we use dispersivity values of aL = 0.025 m and
aTH = aTV = 0.001 m, which are well below the range of dispersivity values used
previously in a large-scale transport simulation of the Borden aquifer (Frind &
Hokkanen, 1987). These values are reasonable because of the small spatial scale (1 m)
of the transport process observed here. The DNAPL properties are listed in Table 1
and the physical properties of the medium and the flow system are listed in Table 2.
The source is assumed to be isotropic with respect to hydraulic conductivity.
The mass transfer parameters that can be adjusted in the calibration are Sh
(constant for the three components) and P"\ In addition, the source hydraulic
conductivity Ks""m and the source porosity
Bsmrce
are unknown due to the presence of
gypsum. To minimize the number of fitting parameters, we assume a constant

Dissolution and mass transfer of multiple
chlorinated hydrocarbons
under field
conditions
111
porosity of 0.33, and take
Kso"rce
as
the additional calibration parameter. The source
is assumed to be isotropic with respect to hydraulic conductivity. The calibration was
carried out by fitting the calculated effluent concentrations to the observed peak
concentrations at the 1-m fence. Because some of the peak values might have been
missed in the sampling, the high data values are considered to have the greatest
credibility. As a check on the calibration, we use the total mass remaining in the
source, obtained from the average mass flux passing through the fence.
We first obtain an initial estimate of Sh by utilizing a plot of the Sherwood
number versus the Reynolds number obtained for a number of ID column
experiments (Imhoff et al., 1994). With a value of the Reynolds number
characteristic of the source area (Re = 1.46 x 10"5) we obtain a value of Sh = 0.002
from this plot. The model was first calibrated with respect to the source hydraulic
conductivity, and the best fit for all three components was obtained with a value of
rm = 4x lO-'ms-1.
The exponent affects the tail of
the
breakthrough curve, and Powers et
al.
(1994)
found values of
(3
ranging between 0.75 and 0.96, depending on the porous material,
in simulations of a column experiment with a single component. In the Borden case,
only the TCM effluent curve shows enough tailing during the 1000 day observation
period to allow p to be fitted. Figure 3 shows the peak concentrations at the 1-m
fence obtained with p = 0.5, 0.5, 0.5, as well as the effective solubility limit for
each DNAPL, valid for the downstream edge of the source closest to the fence.
Favouring the higher-valued observations, the fit is good for all three DNAPLs. The
aqueous concentrations lie about 3 mg l"1 below the corresponding solubility limits
throughout the observation period, accounting for the diffusion gradient across the
stagnant layer and for transverse dispersion between the source and the 1-m fence.
The slight undulation in the TCM curve is due to the seasonal variation of the
hydraulic gradient. The mass remaining in the source is also fitted well. The results
suggest that dissolution within the source takes place essentially at equilibrium
Peak concentrations at the
1 -m
fence
|io2
c 10
o
P*°^®^^$
fxx
"
X^X><W7^
=48:
@ ^ .
X
t
,w~~~^
observed simulated
•
TCM
Cw
QTCE c"xm
x PCE
200 400 600 800 1000
Mass remaining in the source
200 c—1#^-
800 1000
400
Time (days)
Fig. 3 Peak concentrations at the 1-m fence and mass remaining in the source to
1000 days, fitted with
K!°"rce
= 4 x 10"6 m s"1, Sh = 0.002, p= 0.5, 0.5, 0.5; also
showing effective solubility limits
C0X'".

178 Emil O. Frind et al.
concentrations throughout the observation period, and that the drop in the TCM
values is due to the changing mole fraction and source depletion, and not due to mass
transfer kinetics. This agrees with the results of McWhorter & Sale (1998) to the
effect that for homogeneous sources, kinetics is not a controlling process.
A value of Sh = 0.000 05 (40 times lower than the initial value) was found to
give,
after a suitable adjustment of
Kso"m,
a fit almost as good as with the original Sh
value. Evidently, the model is quite insensitive with respect to Sh. We tested other
possible Sh values and found that a range of approximately \Q's<Sh<
10"2
will
produce acceptable fits, provided Ks""m is adjusted accordingly. The model,
therefore, converges to an equilibrium or near-equilibrium mass transfer process for
any value of the mass transfer rate coefficient within the above range.
EVOLUTION OF THE DISSOLUTION PROCESS
The evolution of source dissolution is illustrated in Fig. 4 for the case of TCM,
showing the flownet, the mole fraction remaining in the source, and the aqueous
concentration at 20, 100, and 400 days. Much of the flow is diverted around the
source, while some passes through the source and is focused in a narrow streamtube
downstream of the source. At 100 days, TCM dissolution has progressed almost
halfway through the source, while at 400 days, most of the TCM has dissolved.
However, flow continues to be diverted around the source zone because the low
source conductivity is maintained on account of the TCE, PCE and gypsum that are
still present. The aqueous concentrations show a plume developing which has
reached the 1-m fence at 20 days. At 100 days, the plume is narrowly focused along
its centre, where it reaches close to the effective solubility limit, while at 400 days
TCM Mole Fraction
20 days
TCM Aqueous Phase Concentration
0 1 2 3 0 1 2 3
Distance (m) Distance (m)
Fig. 4 Flow system, source dissolution, and plume evolution for TCM.

Dissolution and mass transfer of multiple chlorinated
hydrocarbons
under field conditions 179
the plume has dispersed to very low values. These results show that although
solubility limit concentrations are approached throughout the source, they are
observed downstream only at the very centre of the plume, consistent with field
observations. When monitoring the plume, these high concentrations could be easily
missed.
Acknowledgements We thank Mike Rivett for generously providing the data for the
emplaced source experiment, John Cherry for setting the stage and providing
continual encouragement, and Dave McWhorter, Peter Grathwohl, and Georg
Teutsch for insightful comments on mass transfer processes. Funding for this work
was provided through a research grant to the first author by the Natural Sciences and
Engineering Research Council of Canada.
REFERENCES
Feenstra, S. & Cherry, J. A. (1988) Subsurface contamination by dense non-aqueous phase liquid (DNAPL) chemicals.
Proc. IAHInt.
Groundwater
Symp.
(Halifax, Nova Scotia, Canada).
Frind, E. O. & Hokkanen, G. E. (1987) Simulation of the Borden plume using the Alternating Direction Galerkin
Technique. Wat.
Resour.
Res. 23(5), 918-930.
Geller, J. T & Hunt, J. R. (1993) Mass transfer from nonaqueous phase organic liquids in water-saturated porous media.
Wat. Resour. Res. 29(4), 833-845.
Grathwohl, P. (1997) Gefahrdung des Grundwassers durch Freisetzung organischer Schadstoffe: 1. Methoden zur
Berechnung der in-situ Schadstoffkonzentrationen. Submitted to
Grundwasser.
Imhoff,
P. T., Jaffé, P. R. & Pinder, G. F. (1994) An experimental study of complete dissolution of a nonaqueous phase
liquid in saturated porous media. Wat.
Resour.
Res. 30(2), 307-320.
Mackay, D. M., Shiu, W. Y., Maijanen, A. & Feenstra, S. (1991) Dissolution of non-aqueous phase liquids in
groundwater. J.
Contain.
Hydrol. 8, 23-42.
Mayer, A. S. & Miller, C. T. (1992) Simulating non-aqueous phase liquid dissolution in heterogeneous porous media.
Proc.
9th
Int.
Conf.
on
Computational Methods
in
Water
Resources.
Comput. Mech. Publ., Southampton, UK.
Mayer, A. S. & Miller, C. E. (1996) The influence of mass transfer characteristics and porous media heterogeneity on
nonaqueous phase dissolution.
Wat.
Resour.
Res. 32(6), 1551-1567.
McWhorter, D. B. & Sale, T. C. (1998) Steady-state mass transfer from single-component NAPLs. Submitted to Wat.
Resour. Res.
Pankow, J. F. & Cherry, J. A. (1996) Dense
Chlorinated Solvents
and
other DNAPLs
in
Groundwater.
Waterloo Press,
Portland, Oregon, USA.
Powers, S. E., Abriola, L. M. & Weber, W. J. Jr (1994) An experimental investigation of nonaqueous phase liquid
dissolution in saturated subsurface systems: transient mass transfer rates.
Wat.
Resour. Res. 30(2), 321-332.
Rivett, M. O., Feenstra. S. & Cherry, J. A. (1992) Groundwater zone transport of chlorinated solvents: a field
experiment. In:
Modern Trends
in
Hydrogeology
(Proc.
Conf.
held at Hamilton, Ontario, Canada).
Sherwood, T. K., Pigford, R. L. & Wilke, C. L. (1975) Mass
Transfer.
McGraw-Hill, New York, USA.