Development and validation of a model of bio-barriers for remediation of
Cr(VI) contaminated aquifers using laboratory column experiments
Authors:
T. Shashidhar, S. Murty Bhallamudi, Ligy Philip
Abstract
Bench scale transport and biotransformation
experiments and mathematical model simulations were carried
out to study the effectiveness of bio-barriers for the containment
of hexavalent chromium in contaminated confined aquifers.
Experimental results showed that a 10 cm thick biobarrier
with an initial biomass concentration of 0.205 mg/g of soil
was able to completely contain a Cr(VI) plume of 25 mg/L
concentration. It was also observed that pore water velocity
and initial biomass concentration are the most significant
parameters in the containment of Cr(VI). The mathematical
model developed is based on one-dimensional advection-dispersion
reaction equations for Cr(VI) and molasses in saturated,
homogeneous porous medium. The transport of Cr(VI) and molasses
is coupled with adsorption and Monod’s inhibition
kinetics for immobile bacteria. Itwas found that, in general,
the modelwas able to simulate the experimental results satisfactorily.However,
therewas disparity between the numerically simulated and
experimental breakthrough curves for Cr(VI) and molasses
in cases where therewas high clay content and high microbial
activity. The mathematical model could contribute towards
improved designs of future bio-barriers for the remediation
of Cr(VI) contaminated aquifers.
© 2006 Published by Elsevier B.V.
Keywords: Bio-barrier;
Biotransformation; Hexavalent chromium; Column experiments;
Contaminant transport model
1. Introduction
As reported from several parts of the
world, both anthropogenic and natural processes may lead
to hexavalent chromium contamination in soils and aquifers.
Industrial activities such as electroplating, leather tanning,
and wood preservation, etc. release large quantities of
liquid and solid waste containing Cr(VI) [1]. Hexavalent
Chromium is highly toxic, highly soluble in water, and is
likely to be transported over long distances in the subsurface.
Two alternatives are available for treating the Cr(VI) contaminated
groundwater: (i) pump and treat, and (ii) in situ. These
can be achieved either by physico-chemical or biological
processes. Latter method appears to be more economical and
environmentally friendly. Several researchers have reported
that many micro-organisms, under various environmental conditions,
can reduce highly toxic and mobile Cr(VI) to less toxic
and less mobile Cr(III) [2–4].
Studies have been conducted in the past
to evaluate the potential of biotransformation for the remediation
of Cr(VI) contaminated soil and wastewater. Most of these
studies pertain to ex situ treatment option [5–7].
Experiments have also been conducted to understand the combined
transport and geo-chemical processes pertaining to Cr(VI)
in different soils through batch and continuous column studies
[8–12]. Several batch studies on biotransformation
of Cr(VI) to Cr(III) under various environmental conditions
[13–15] have been reported. Recently, combined transport
and biotransformation studies have been reported by Guha
[16] and Shashidhar et al. [17]. Guha [16] focused on the
transport of Cr(VI) through saturated column with manganese
coated sand, under the influence of adsorption, competitive
redox and biotransformation.However, he used small laboratory
scale columns (10–30 cm long) for this purpose. Experimental
data was available only for the inlet and outlet, which
may not be sufficient to completely understand the interplay
between geo-hydrology and chromium containment. Shashidhar
et al. [17] conducted bench-scale column experiments to
evaluate the effectiveness of Cr(VI) containment in aquifers
using bioremediation. Effects of ground water velocity,
initial microbial concentration, and aquifer soil characteristics
on Cr(VI) containment were studied. Experiments were also
conducted to study only the transport and adsorption of
Cr(VI) in order to assess the role of bioremediation in
Cr(VI) containment. Although there have been many studies
which considered the effectiveness of Cr(VI) containment
in confined aquifers using bio-barriers, not many studies
on validation of mathematical models for such systems using
systematic laboratory experimental data are reported.
The objective of the present study was
to conduct, in continuation of authors’ earlier study
[17], further bench-scale column studies for evaluating
the performance of bio-barriers for the remediation of Cr(VI)
contaminated confined aquifers. Also, an attempt was made
to develop and validate a mathematical model for predicting
the containment of Cr(VI) in contaminated confined aquifers
by in situ bioremediation.
2. Materials and methods
2.1. Soil
Soils used in this study were collected
from the I.I.T. Madras campus, Chennai, India. The portion
which passed through 4.75mm sieve opening was used for the
experiment. River sand which passed through 0.6mm and retained
in 0.425mm sieve opening was washed thoroughly with distilled
water and oven dried at 103 ?C over night before being used
in sand column experiments. The soil and sand characteristics
were analyzed as per the standard methods [18] and are presented
in Table 1.
2.2. Chemicals
All the chemicals used in this study were
of AR grade and were supplied by Ranbaxy Chemicals Ltd.,
Chennai, India. Glassware used for analysis were equilibrated
with Cr(VI) and washed with acid solution followed by distilled
water.
2.3. Nutrient medium
The medium (M2) for Cr(VI) reduction experiments
consisted of K2HPO4 (0.03 g/L), KH2PO4
(0.05 g/L), MgSO4·7H2O (0.01
g/L), NH4Cl (0.03 g/L), NaCl (0.01 g/L), carbon
source (2 g/L), and 1mL of trace element solution. The carbon
source used in the present study was molasses (measured
as COD), which was prepared synthetically from crude sugar,
known as “jaggery”. Trace element solution consisted
of FeCl2·4H2O (12.2 g/L), MnCl2·4H2O
(4.09 g/L), CoCl2·6H2O (0.927
g/L), ZnCl2 (0.37 g/L), CuCl2 (0.61
g/L), NaMoO4·2H2O (0.579 g/L),
H3BO3 (0.16 g/L), KI (0.148 g/L),
NiCl·6H2O (0.067 g/L), and EDTA Na2·4H2O
(6.5 g/L). The pH was maintained at 7±0.2 by using
HCl or NaOH. Molasses was used as a carbon source. Sterilized
medium was used for all the studies.
2.4. Enrichment of Cr(VI) reducing
bacterial strains
Bacterial strains were isolated from the
soil samples collected from the chromium contaminated site
located at Ranipet, Tamilnadu, India. Detailed methodology
for the enrichment, cultivation and harvesting of Cr(VI)
reducing bacterial strains is presented elsewhere [6,17].
2.5. Experimental setup
Fig. 1a shows the schematic of the bench
scale experimental setup used in this study. The length
of the column was 1m and it had a 10 cm2 cross
section, with an effective radius of 10 cm. This was fabricated
in the laboratory using PVC (Perspex) sheets of 6mm thickness.
The column was filled with thoroughly washed pure river
sand from 0 to 50 cm, and again from 60 to 100 cm. To simulate
the bio-barrier conditions, the portion between 50 and 60
cm was filled with Soil-A augmented with Cr(VI) reducing
microbes. An overhead tank with provision for an adjustable
head served as the inlet to the column. A uniform entry
of water into the soil was ensured by providing a porous
plate at the inlet. The effluent was collected in a tank
provided at the outlet. Saturated conditions in the column
were achieved by maintaining the water level in the outlet
reservoir above the top of the soil column. Sampling ports
were provided on the sides at five different cross sections
at distances 20, 40, 49, 60, and 80 cm from the inlet, respectively.
As shown in Fig. 1b, four ports were provided at each cross-section.
Sterilized disposal syringes (Dispovan, India) were used
to collect liquid samples from these ports at regular time
intervals.
2.6. Batch studies
Sorption equilibrium studies for Cr(VI),
alone and in presence of molasses and lithium were conducted
using Soils A, C, and sand, to estimate the adsorption coefficients.
The details of these studies are presented elsewhere [17].
Similarly, biokinetic studies were conducted as per standard
procedure to estimate the kinetic parameters of Cr(VI) reducing
microbial consortia [19].
2.7. Transport and biotransformation
studies
2.7.1. Transport studies without biotransformation
The column was filled with prepared soil
in 33 layers. In order to get a more or less uniform compaction,
each layer was compacted with 25 blows of a 1.2 kg hammer
falling from a 30 cm height. Bulk density was determined
by measuring the dry weight of the soil before adding the
required moisture and filling the column. The porosity was
determined using the formula relating the bulk density,
and dry weight. Initially, steady flow rate through the
soil column was obtained by maintaining constant heads in
the head and tail tanks for a long time. The flow rate was
monitored with respect to time by collecting the water at
the outlet. Cr(VI) contaminated water was introduced into
the system through the head tank after the steady state
was attained. Liquid samples were taken from all the sampling
ports as well as the head tank at regular time intervals,
and were analyzed for Cr(VI) concentration. Experiment was
continued until the breakthrough occurred at the last sampling
cross section. These experiments were conducted for Soil-A
with a porosity of 0.375, for three different velocities
of 22.4, 11.2, and 5.6 cm/h, respectively.
2.7.2. Transport studies with biotransformation
Transport studies with biotransformation
were conducted in a similar manner. However, in these tests,
the soil was mixed with bacterial cells and mineral medium.
The column was fed with mineral medium until a steady state
velocity was attained. The column was then fed with medium
containing 25 mg/L of Cr(VI), molasses (approximately 2000
mg/L of COD), and lithium (50 mg/L) along with minerals.
Samples were withdrawn from various sampling ports at regular
intervals using syringes and were analyzed. Other details
of the procedure are given elsewhere [17]. These studies
were conducted with two different soils namely Soils A,
and C and with sand as given in Table 1.
2.7.3. Transport studies with bio-barriers
Transport studies with bio-barriers were
conducted in a similar manner to those without bio-barriers,
except that a bio-barrier of 10 cm thickness was provided
at a distance of 50 cm from the inlet. The 10 cm bio-barrier
was filled with Soil A, mixed with microbes, whereas the
rest of the column was filled with sand without any microbes.
Also, four more sampling ports were provided at 49 cm from
the inlet, i.e. just at the starting of the bio-barrier.
Two vents were provided for release of gasses generated
in the bio-barrier. As in the column studies without bio-barriers,
concentrations of microbes, substrate Cr(VI), and lithium
in liquid phase were monitored continuously at all the ports.
Flow rate and microbial concentration were also monitored
at the outlet. After the completion of column run, soil
samples were taken from different ports and were analyzed
for adsorbed Cr(VI), total chromium, COD, lithium, and microbial
concentration. These column studies with bio-barriers were
conducted for two different initial microbial concentrations
in the barrier. The experimental details for all the transport
studies with biotransformation are presented in Table 2.
2.8. Analytical procedures
2.8.1. Liquid phase chromium analysis
Diphenyl carbazide methodwas used to determine
the Cr(VI) concentration [20]. Total chromium concentration
was analyzed using atomic absorption spectrometer (Perkin
Elmer, USA).
2.8.2. Extraction and analysis of Cr(VI)
and total chromium in soil
Alkaline digestion method and nitric acid/sulfuric
acid digestion method as per Standard Methods were used
for the extraction of Cr(VI) and total chromium from soil,
respectively. Diphenyl carbazide method was used to determine
the Cr(VI) concentration [20]. Potassium permanganatewas
used to oxidize Cr(III) to Cr(VI) in the case of Cr(III).
2.8.3. Measurement of cell density
in liquid phase
Cells were grown overnight, centrifuged,
washed with physiological saline water thrice, re-suspended
in saline water, homogenized, and used as stock solution.
Different dilutions were made from the above stock solution.
Dry weights of cells were measured by filtering a known
volume of these solutions through 0.45 m filter paper (Millipore,
USA). Corresponding absorbance was measured at 440 nm using
a spectrophotometer. This information was used to prepare
a calibration curve between dry weight and absorbance. For
unknown samples, the absorbance was measured at 440 nm and
was converted to dry weight using absorbance versus dry
weight calibration curve.
2.8.4. Microbial quantification
The bacterial cell count (colony forming
units) was carried out as per standard procedure [20]. The
total protein of intact cells was determined according to
the method of Herbert et al. [21]. The cell suspension (0.5
mL)was mixed with 2mL of 1.0N NaOH andwas kept in boilingwater
bath for 5 min. The contents were then cooled in cold water.
To this, 5mL of freshly prepared alkaline copper reagentwas
added and allowed to stand for 10 min. 0.5mL of Folin-Ciocalteu
reagent was then added and allowed to stand for 30 min for
the color development. Reagent blank containing 0.5mL distilled
water instead of bacterial suspension was treated in a similar
way. The optical density was measured at 750 nm using a
spectrometer against the reagent blank. Known bacterial
concentrations were used for preparing the calibration curve.
2.8.5. Chemical oxygen demand
COD of liquid and soil samples were estimated
as per standard methods [20]. Closed reflex method was followed.
2.8.6. Lithium
Lithium was analyzed using flame photometer
(Elico, India) method as described in standard methods [20].
3. Mathematical model
Mathematical model for the transport accounts
for the advective-dispersive-reactive transport of three
aqueous species: Li, Cr(VI), and substrate (molasses). Lithium,
a conservative pollutant, was used as a tracer. Therefore,
Li transport data was used to determine the dispersion coefficient.
The column experiments showed that Cr(III) formed due to
biotransformation of Cr(VI) did not remain in the liquid
phase and it was either precipitated and retained or adsorbed
onto the soil matrix almost immediately. Liquid samples,
collected from all the 16 ports and the outlet tank, did
not contain any Cr(III). Therefore, Cr(III) transport is
not included in the present mathematical model.
In the column experiments, there was washout
of microbes during the initial stabilization. During this
period, water with only mineral medium was allowed to pass
through the column until steady state flow conditions were
achieved. The amount of washout depended upon the soil.
There were more washouts of microbes in columns with sand
as compared to those in columns with soil. It may be noted
that although therewaswashout during the stabilization period,
the liquid samples taken from 16 ports as well as the outlet
tank did not contain significant amount of microbes once
the stabilization was achieved. Therefore, it was assumed
that the microbes were immobile and attached to the soil
matrix. Only the microbial growth equation was considered
and the transport of microbes in the liquid phase was neglected.
It may be noted that in an earlier study
on bio-geochemical transport of Cr(VI) through sand columns,
Guha [16] used a similar approach to modeling the transport
and transformation of Cr(VI). However, in that model, it
was assumed that some bacteria were mobile and some were
immobile. Also, a double Monod’s kinetic equation was used
for microbial growth, in which both Lactate (electron donor)
and Cr(VI) (electron acceptor) were treated as substrate.
In this study, the Monod’s equation with inhibition was
used to model the microbial growth. Batch studies indicated
that molasses (substrate) concentration was not limiting,
while Cr(VI) concentration used was much above the inhibition
concentration (3.05 mg/L). Also, the batch studies indicated
that not all the molasses present was available for microbial
utilization. In the present study, the carbon source used
(referred to as molasses) was crude sugar, known as ‘jaggery’.
It contains a mixture of sucrose, cellulose, etc. In this,
sucrose is the only easily biodegradable substrate available.
After the complete utilization of sucrose, the microbes
start utilizing cellulose slowly, as it is not an easily
biodegradable substrate. In the present study, the time
available was limited for degradation of cellulose. Therefore,
the concept of utilizable substrate [22] was adapted in
the model for microbial growth rate. Another modification
made to the usual microbial growth rate equation was the
introduction of a parameter, 
which is similar to the microbial metabolic potential factor
used by other researchers [16,23].
The governing advection-dispersion reaction
equations for one-dimensional transport of lithium, hexavalent
chromium, substrate, and microbes can be written as follows:
where Li is the lithium concentration in
the liquid medium (mg/L), Cr6 the hexavalent chromium concentration
in the liquid medium (mg/L), S the molasses concentration
in the liquid medium (mg/L), M the bacterial concentration
expressed as mg/L of liquid in the column, Su the utilizable
concentration of molasses (mg/L), ST the total
inlet molasses concentration (mg/L), u the pore water velocity
(cm/h), D the coefficient of dispersion (cm2/h),
RCr6 the retardation coefficient for hexavalent
chromium, Rs the retardation coefficient for
substrate, Rsink Cr6 the sink term for hexavalent
chromium due to biotransformation (mg/L/h), RsinkS
the sink term for substrate due to microbial utilization
(mg/L/h), µ the specific growth rate (h-1),
µmax the maximum specific growth rate (h-1),
kd the decay constant (h-1), Y the
observed yield coefficient, 
the efficiency factor for chromium reduction with respect
to substrate utilization, 
the proportionality constant which accounts for the differences
in the microbial activity in a suspended batch system and
attached continuous system. It also implicitly accounts
for metabolic retardation due to starving in the stabilization
and acclimatization periods. In the present model, the pore
velocity, u is obtained by dividing the Darcy velocity,
U by the porosity of the soil column, 
.
The coefficient of dispersion, D is obtained by multiplying
the pore velocity, u with the dispersivity, l.
Computation of the retardation coefficients for hexavalent
chromium and substrate is based on the equilibrium adsorption
studies. Constant 0.63 in Eq. (7) was obtained from batch
studies.
The basic assumptions made in deriving
the model can be summarized as follows:
-
The flow in the column is one-dimensional.
-
The porous medium is homogeneous,
and the porosity remains constant through out the study
period.
-
Adsorption is assumed to occur under
equilibrium conditions.
-
The model is based on the “macroscopic
modeling” of microbiological reactions. This is a single
phase model where all the microorganisms present in
a given control volume are equally exposed to the substrate
concentration prevailing in the bulk liquid volume [24].
-
The microbes are immobile.
-
The contaminant is toxic and has
inhibitory effect on microbial growth rate.
-
The Monod’s equation with inhibition
describes the microbial growth.
-
Only a fraction of substrate is available
for Cr(VI) reduction.
-
Cr(III) generated due to biotransformation
is either adsorbed or precipitated and retained on the
soil matrix.
-
The temperature is constant.
A direct numerical substitution approach
with Picard iteration was used to solve the non-linear partial
differential equations [25–27]. Advection-dispersion part
was discretized using the implicit–explicit approach for
time discretization because of its better numerical stability
and accuracy. The spatial discretization for advection term
was based on an Essentially Non-Oscillating scheme in which
MINMOD limiter was employed for suppressing numerical oscillations.
A central difference scheme was used for spatial discretization
of the dispersive term.
4. Results and discussion
4.1. Batch studies
Batch studies were conducted to determine
the equilibrium adsorption constants, and the biokinetic
parameters for bacterial growth. Adsorption equilibrium
studies were conducted for Cr(VI), Cr(III), Li, and COD
(molasses) for all the three soils. Adsorption studies were
also conducted for Cr(VI) and Cr(III) in presence of COD
and Li to understand the interference of these components
on adsorption. Freundlich isotherm was used for fitting
the experimental data. Table 3 shows the Freundlich coef-
ficient (Kf), exponent (1/n) and the corresponding correlation
coefficient for all the isotherms [17]. It can be seen that
adsorption of Cr(III) is much higher than Cr(VI). Adsorption
of Li is almost negligible, indicating that it is a conservative
pollutant, which serves as a tracer to determine dispersion
characteristics.
4.2. Studies to estimate biokinetic
parameters
Biotransformation studies were conducted
with initial chromium concentrations of 0, 1, 5, 10, 20,
50, 100, 200, 300, and 500 mg/L. Bacterial cells (36 mg/L)
were added to the media and hexavalent chromium, COD and
bacterial concentrations were measured at various time intervals.
µmax and Ks were determined
using the data from experiments conducted without chromium,
and with an initial molasses concentration of 5000 mg/L
as COD. The bacterial growth in the exponential phase was
fitted to the equation M = M0eµmaxt,
where M0 is the initial biomass concentration.
Using this µmax value, the Ks
value was determined such that the simulated growth curve
matched with the experimental growth curve. Ki
was then determined using the experiments for microbial
growth rate in the presence of chromium, using Monod’s equation
with inhibition. The biokinetic parameters obtained are:
µmax = 0.3 h.1, Ks = 40.0 mg/L
(as COD), Ki = 3.05 mg/L of Cr(VI), and Y = 0.263.
The efficiency factor, ,
i.e. ratio between amount of Cr(VI) biotransformed to the
amount of molasses consumed, as well as were
determined by back-fitting the Cr(VI) breakthrough curve
at port 20 cm using Genetic Algorithms. The same values
were used in the mathematical simulations for breakthrough
curves of Cr(VI), and molasses at all other ports located
at 40, 60, and 80 cm. -values
were 0.1, 0.065, and 0.1 for Soils A, sand, and soil C,
respectively, whereas estimated value of =
0.3, which is almost the same as reported in the literature
[16].
Table 4 Modified coefficients
of efficiency (E) for the transport and biotransformation
studies
4.3. Transport studies with no biotransformation
It is essential to understand the transport
of Cr(VI) without any biotransformation in order to study
the role of biotransformation in the containment of Cr(VI)
in aquifers, considering only adsorption. These studies
would help in validating the numerical solution of the advection-dispersion-adsorption
part of the mathematical model. In this study, the model
performance was statistically evaluated using the dimensionless
modified coefficient of efficiency, E [28,29].
where E(ti) is the numerically
simulated value of a variable at time ti, O(ti)
the observed value of the same variable at time ti,
and 
is
the mean value of the observed variable. E varies between
-
and 1.0, the higher values indicating better model prediction.
As suggested by Kohne et al. [29] a positive value of E
represents an “acceptable” simulation whereas E > 0.5 represents
a “good” simulation. E equal to one indicates a “perfect”
simulation. Values of E for all the simulations carried
out in this study are presented in Table 4.
Numerically simulated results along with
the experimental data for the breakthrough of Cr(VI) at
20, 40, 60, and 80 cm ports, for three pore velocities in
column with Soil A are presented in Fig. 2a–d. The measured
breakthrough data at 20 cm port was used to back fit the
dispersivity, áL, and the same was used to simulate the
breakthrough curves at other ports. The dispersivity in
these studies was equal to 4.46 cm, and the dispersion coefficient
varied linearly with pore velocity. Dimensionless modified
coefficient of efficiency, E for these simulations (Table
4) varied from 0.54 to 0.94 indicating that the mathematical
model simulates the experiments well, which is also evident
from Fig. 2a–d. However, there was disparity between the
model and experimental results for the case of v = 5.6 cm/h
at x=60 cm (E = 0.54), although the matching was good at
x=80 cm(E = 0.94). Onewould expect that the disparity between
the simulated and observed data to increase in the direction
of transport due to retardation. Here, it may be noted that
a single dispersion coefficient was used for the entire
column. On the other hand, the dispersion coefficient may
be varying spatially due to non-homogeneity in compaction.
This is especially true in case of soil with high clay content
(Soil A). It is also obvious from these results that adsorption
alone was not able to contain Cr(VI) in the aquifer. The
maximum Cr(VI) concentration at 80 cm port was almost equal
to the inlet concentration irrespective of pore velocity.
Fig. 2. Experimental and
numerical Cr(VI) breakthrough curves of Soil A column for
different pore velocities; no biotransformation (pH 6.7–7,
inlet Cr(VI) concentration 25 mg/L): (a) x = 20 cm, (b)
x = 40 cm, (c) x = 60 cm), and (d) x=80 cm.
Fig. 3. Experimental and
numerical lithium breakthrough curves of Soil C column;
with biotransformation (pH 6.2–7.2, inlet Lithium concentration
46 mg/L): (a) x = 20cm, (b) x = 40 cm, (c) x = 60 cm, and
(d) x=80 cm.
Fig. 4. Experimental and
numerical substrate breakthrough curves of Soil C column;
with biotransformation (pH 6.2–7.2, inlet COD concentration
2000 mg/L): (a) x = 20cm, (b) x = 40 cm, (c) x = 60 cm,
and (d) x=80 cm.
Fig. 5. Experimental and
numerical Cr(VI) breakthrough curves of soil C column; with
biotransformation (pH 6.2–7.2, inlet Cr(VI) concentration
25mg/L): (a) x = 20 cm, (b) x = 40 cm, (c) x = 60 cm, (d)
x=80 cm.
Fig. 6. Experimental and
numerical lithium breakthrough curves of soil A column;
with biotransformation (pH 6.2–7.2, inlet Li concentration
36 mg/L): (a) x=20 cm, (b) x = 40 cm, (c) x = 60 cm, (d)
x=80 cm.
Fig. 7. Experimental and
numerical substrate breakthrough curves of Soil A column;
with biotransformation (pH 6.2–7.2, inlet COD concentration
2000 mg/L): (a) x = 20cm, (b) x = 40 cm, (c) x = 60 cm,
and (d) x=80 cm.
Fig. 8. Experimental and
numerical Cr(VI) breakthrough curves of soil A column; with
biotransformation (pH 6.2–7.2, inlet Cr(VI) concentration
25mg/L): (a) x = 20 cm, (b) x = 40 cm, (c) x = 60 cm, (d)
x=80 cm.
Fig. 9. Experimental and
numerical Cr(VI) breakthrough curves of sand column; with
biotransformation (pH 6.2–7.2, inlet Cr(VI) concentration
25 mg/L, pore velocity 6.67 cm/h): (a) x = 20cm, (b) x =
40 cm, (c) x = 60 cm, and (d) x=80 cm.
Fig. 10. Experimental and
numerical Cr(VI) breakthrough curves of sand column; with
biotransformation (pH 6.2–7.2, inlet Cr(VI) concentration
25mg/L, pore velocity 1.16 cm/h): (a) x = 20 cm, (b) x =
40 cm, (c) x = 60 cm, and (d) x=80 cm.
4.4. Transport studies with biotransformation
Bench scale experiments were conducted
for transport along with biotransformation in saturated,
confined aquifer systems. As given in Table 2, experiments
were conducted for two different soils (Soils A and C),
and for sand. For sand, experiments were conducted for two
different pore velocities.
Fig. 3a–d shows the comparison between
the numerically simulated and experimentally measured breakthrough
for lithium tracer at x = 20, 40, 60, and 80 cm, for transport
in column with Soil C. This figure also shows the pore velocity
variation with time. Pore velocity was obtained from the
measured Darcy velocity by dividing it with porosity. It
was assumed that the porosity remained constant, though
it might have changed. It can be seen from Fig. 3 that the
pore velocity was decreasing drastically as the time progressed.
Periodically the head in the upstream tankwas adjusted to
increase the velocity to reasonable levels. Similar trend
in the reduction of pore velocity with time was observed
in all the other experiments also. The pore velocities reported
here were used as velocity input for the solution of transport
equations in all simulations. Fig. 4a–d shows the comparison
between the numerically simulated and experimentally measured
breakthrough for substrate (molasses) at x = 20, 40, 60,
and 80 cm, for the same experiment. Fig. 5a–d shows the
same for Cr(VI). In the numerical simulations for the transport
of Cr(VI) and molasses, the biokinetic parameters as determined
from the batch experiments were used. The dispersivity value
was equal to 3.5 cm. This valuewas obtained by fitting the
breakthrough curves for lithium tracer at x = 20, 40, 60,
and 80 cm, since the lithium transport gives the hydraulic
characterization. E values for these simulations varied
from 0.75 to 0.96, indicating that the parameter estimation
for dispersivity using the lithium breakthrough data was
good. As mentioned earlier, a parameter ë has been introduced
in the model to account for the differences in the microbial
growth in a suspended batch system and attached continuous
system. It also implicitly accounts for metabolic retardation
due to starving in the stabilization and acclimatization
periods. It was assumed that this parameter
was constant through out the column. Therefore, the chromium
breakthrough curve at x = 20 cm was used to back fit the
value of this parameter (E = 0.4), and the same
was used for simulating the substrate and chromium breakthrough
curves at the remaining ports. This value was equal to 0.1.
Fig. 4a–d shows that simulation of molasses transportwas
satisfactory (E value varied from 0.24 to 0.56).
It can also be seen that numerically simulated breakthrough
curves for hexavalent chromium at x = 40, 60, and 80 cm
match well with the experimental data (E = 0.65,
0.70, and 0.83). It can be inferred from these results that
the proposed model is able to explain the transport and
biotransformation of hexavalent chromium in the confined
aquifer. One calibrating parameter,
was able to implicitly include most of the uncertainties
associated with biotransformation in a confined silty aquifer.
Fig. 11. Experimental and
numerical lithium breakthrough curves of BB1; (pH 6.2–7.2,
inlet Li concentration 46 mg/L, pore velocity 1.6 cm/h):
(a) x = 20 cm, (b) x = 49 cm, (c) x = 60 cm, and (d) x=80
cm.
Experimental results for Cr(VI) breakthrough
indicate that there was a high concentration of Cr(VI) initially
at all the ports, which was due to dominance of advection
as compared to the biotransformation. Subsequently, the
microbial activity increased due to an increase in the microbial
population, which resulted in significant Cr(VI) reduction.
This trend was well simulated by the mathematical model.
Figs. 6–8 show the numerically simulated
and experimental breakthrough curves for lithium, molasses,
and hexavalent chromium, respectively, for SoilAwith 6.19%
clay content. The same value of ë as obtained for Soil C
was used in this case also. Dispersivity in this experiment,
as obtained from the lithium breakthrough data, was equal
to 4.46 cm. E values for these simulations varied
from -0.9 to -0.1, indicating that the parameter estimation
for dispersivity using the lithium breakthrough data was
not satisfactory. It is clear from these figures that as
the clay content increases, it becomes difficult to simulate
even the transport of lithium, which is a conservative pollutant.
The E values for Cr(VI) simulations varied from
-10.1 to -0.37, again indicating an unsatisfactory performance
by the mathematical model. However, the simulation of molasses
transport was satisfactory (E values ranged from
0.44 to 0.62), may be because of high concentrations. It
may be noted that more than 80% of molasses was left even
after the complete biotransformation of Cr(VI). The gas
released due to microbial metabolic activity might have
been trapped unevenly in the column and introduced non-homogeneities.
These non-homogeneities affected the dispersivity with respect
to time, which was not accounted in the present model. It
may be noted here that, for the same soil without biotransformation,
transport of both lithium and Cr(VI) were simulated well
by the proposed model. Thus, it may be concluded that for
modeling the transport and biotransformation of Cr(VI) in
aquifers with high clay content, non-homogeneities introduced
by biotransformation process and the consequent changes
in the hydro-geological conditions should be considered
for a better simulation.
Figs. 9a–d and 10a–d present the breakthrough
for Cr(VI) at 20, 40, 60, and 80 cm ports for the case of
transport and biotransformation experiments in sand, for
two different velocities of 6.67 and 1.16 cm/h, respectively.
Dispersivities in these experiments (as obtained using the
lithium tracer data, results not presented) were 0.1 and
0.3 cm, respectively. The E values for lithium
transport in the case of velocity of 6.67 cm/h varied from
0.75 to 0.93. They varied from 0.67 to 0.93 when the velocity
was equal to 1.16 cm/h. The
value (as obtained from the Cr(VI) breakthrough data at
x = 20 cm) was 0.065. It is very clear that the effect of
biotransformation on Cr(VI) containment is very significant
in the case of lowpore velocity. In case of high pore velocity,
breakthrough of Cr(VI) occurred much earlier and also the
maximum concentration was almost equal to the inlet concentration
even after 150 h. These effects are well simulated by the
mathematical model as evident from the figures (E
values ranged from 0.71 to 0.95). Pore velocity had a significant
effect on bacterial retention on the soil matrix. High pore
velocity resulted in significant bacterial cell washout
from 0.021 to 0.005 mg/g, while the bacterial concentration
reduced from 0.04 to 0.027 mg/g in the case of column with
low pore velocity. It can be seen from Fig. 10a–d that the
rate of Cr(VI) containment increased with respect to time
because of corresponding increase in biomass concentration
in the system. Increased microbial activity in the case
of low pore velocity might have introduced some non-homogeneity
and had affected the hydrogeology. As a result, the performance
of the mathematical model was not as good as in the case
of high pore velocity (E values ranged from 0.78
to .0.32).
To summarize, comparison of the mathematical
model and experimental data for Soils A, C, and sand shows
that in case of sand, the prediction of breakthrough curves
was more accurate. This may be due to more homogeneity in
the case of sand, compared to that of Soils A and C, which
contained various levels of clay and silt. Uneven accumulation
of gas generated due to biotransformation might have also
introduced considerable non-homogeneity in case of clayey
soils. This effect was not considered in the simple mathematical
model. The mathematical model did not consider the change
in the dispersivity due to microbial activity. It may be
also noted that we used the same value of porosity to determine
the pore velocity from the measured Darcy velocity. Non-uniform
entrapment of gas might have resulted in non-uniformity
in pore velocity. This effect was also not considered in
the present model.
Fig. 12. Experimental and
numerical substrate breakthrough curves of BB1; (pH 6.2–7.2,
inlet COD concentration 1000 mg/L, pore velocity 1.6 cm/h):
(a) x = 20cm, (b) x = 49 cm, and (c) x=60 cm.
4.5. Transport and biotransformation
studies with bio-barriers
As given in Table 2, two experiments were
conducted with a bio-barrier in place from x=50 to x = 60
cm. In these experiments, Soil A was used for the bio-barrier,
while sand was used in the rest of the column. Initial bacterial
concentration in Bio-barrier one (BB1) was 0.0205 mg/g of
soil while it was 0.205 mg/g in BB2. Breakthrough curves
for lithium for BB1 at 20, 49, 60, and 80 cm are presented
in Fig. 11a–d. The breakthrough, curves for molasses and
Cr(VI) for the same ports are given in Figs. 12 and 13,
respectively. Breakthrough curves for lithium, molasses
and Cr(VI) at 20, 49, and 60 cm for BB2 are presented in
Figs. 14–16. It can be seen from these figures (Figs. 13
and 16) that the 10 cm bio-barrier was able to contain the
hexavalent chromium, even when the inlet chromium concentration
was as high as 25 mg/L. Also, the pore velocity (1.6 cm/h)
in these experiments was quite high compared to pore velocities
normally encountered in the field. The chromium containment
was almost complete in BB2, in which the initial microbial
concentration was relatively high. From this, it can be
concluded that bio-barrier using enriched microbes is a
viable method for the remediation of chromium contaminated
aquifers.
Table 5 shows the input values for various
parameters used in the mathematical simulations for the
bio-barrier experiments. In these experiments, measurement
of bacterial concentration in the outlet reservoir showed
that there was considerable bacterial washout during the
stabilization period. Assuming that washout had occurred
uniformly from the bio-barrier and there was no retention
of washed out biomass in the down stream column, the initial
biomass concentration for biotransformation of Cr(VI) in
the bio-barrier was estimated as 0.0205 mg/g in BB1 and
0.1805 mg/g in BB2. The column experiments also indicated
some biotransformation on the upstream side of the bio-barrier.
This may be due to the presence of small amount of biomass
in the sand portion, which might have entered and accumulated
through the feed or due to themovement of bacteria from
the biobarrier section of the column upstream into the sand
sections. In the mathematical simulations, this concentration
was assumed as 10 mg/L, which was almost negligible compared
to the concentration of biomass in the barrier. As earlier,
the dispersivity values for sand and barrier (Soil A) portions
were estimated using the lithium breakthrough curves (0.01,
1.0). These values matched closely with the values estimated
earlier for sand and Soil A. Values of
for sand and Soil A as determined from the earlier experiments
were used here also.
Table 5 Input values for mathematical
simulations of bio-barrier experiments
Figs. 11–13 show the comparison between
numerically simulated and experimental breakthrough curves
for lithium, molasses, and Cr(VI) for BB1. It can be seen
from Fig. 11a–c that the parameter estimation for dispersivity
for BB1 is good, E value varied from 0.77 to 0.96.
However, the matching between the numerically simulated
and experimental breakthrough curves was not good for molasses
(E value varied from 0.08 to 0.65). Similarly,
the matching between the numerically simulated and experimental
breakthrough curves for Cr(VI) was not as satisfactory (E
varied from -0.26 to +0.21). Although the proposed mathematical
model simulated the overall trend of increasing biotransformation
with distance and time, the quantification by the mathematical
model was not satisfactory. Comparison between the numerical
model and experimental results for BB2 are presented in
Figs. 14–16, for lithium, molasses, and Cr(VI), respectively.
It is evident from these figures that the matching between
the numerical model and experimental results for Cr(VI)
breakthrough was not as good, especially at the 60 cm port
(downstream of barrier, E =-0.46). These results
are in conformity with the earlier results that biotransformation
affects the hydrogeology of the aquifer and increases the
non-homogeneity. This affect was more significant in cases
where the biomass concentration was high, as expected. Temporal
variation of head loss from x=0 to x = 49 cm (sand portion),
and from x=49 cm to x = 60 cm (bio-barrier) showed that
there was a considerable increase in head loss (0–6 cm in
80 h) with respect to time in the bio-barrier portion, whereas
the change in the head losswas almost negligible in the
sand portion. This clearly indicates that the biotransformation
process has an effect on the hydrogeology. As discussed
earlier, present mathematical model does not consider the
effect of biotransformation on hydrogeology.
Breakthrough curves for molasses at x =
60 cm clearly show that significant amount of molasses was
unutilized even after the complete biotransformation of
Cr(VI). This means more molasses was introduced, than what
was truly required. In order to reduce the cost as well
as associated contamination problems in field applications,
optimization is required to determine the best combination
of initial biomass concentration in the barrier, substrate
concentration, and the bio-barrier thickness.
Fig. 13.
Experimental and numerical Cr(VI) breakthrough curves
of BB1; (pH 6.2–7.2, inlet Cr(VI) concentration 25
mg/L, pore velocity 1.6 cm/h): (a) x = 20 cm, (b)
x = 49 cm, (c) x=60 cm. |
|
Fig. 14.
Experimental and numerical lithium breakthrough curves
of BB2; (pH 6.2–7.2, inlet Li concentration 46 mg/L,
pore velocity 1.6 cm/h): (a) x=20 cm, (b) x = 49 cm,
(c) x=60 cm. |
| |
|
|
Fig. 15.
Experimental and numerical substrate breakthrough
curves of BB2; (pH 6.2–7.2, inlet COD concentration
1000 mg/L, pore velocity 1.6 cm/h): (a) x = 20 cm,
(b) x = 49 cm, and (c) x=60 cm. |
|
Fig. 16.
Experimental and numerical Cr(VI) breakthrough curves
of BB2; (pH 6.2–7.2, inlet Cr(VI) concentration 25
mg/L, pore velocity 1.6 cm/h): (a) x = 20 cm, (b)
x = 49 cm, (c) x=60 cm. |
5. Conclusions
Bench scale transport and biotransformation
studies showed that bio-barriers are an effective way of
chromium containment in contaminated aquifers. Most significant
parameters in the containment of Cr(VI) are pore water velocity
and the initial biomass concentration. A simple mathematical
model for the transport of Cr(VI) and molasses, coupled
with adsorption and Monod’s inhibition kinetics for immobile
bacteria, was able to simulate the experimental results
satisfactorily when the clay content was less and the microbial
activity was not very high. Clay content and increased heterogeneity
in the system due to high bacterial activity altered the
hydro-geological conditions. In such cases, there was disparity
between the numerically simulated and experimental breakthrough
curves for Cr(VI) and molasses.
