&EPA
United States
Environmental Protection
Agency
Robert S. Kerr Environmental
Research Laboratory
Ada, OK 74820
Research and Development
EPA/600/S-93/006 August 1993
ENVIRONMENTAL
RESEARCH BRIEF
Spatial Heterogeneity of Geochemical and Hydrologic
Parameters Affecting Metal Transport in Ground Water
J.A. Davis1, C.C. Fuller1, J.A. Coston1, K.M. Hess2, and E. Dixon1
INTRODUCTION
Reliable assessment of the hazards or risks arising from ground-
water contamination and the design of effective rehabilitation
procedures require the capability to predict the movement and
fate of dissolved solutes in ground water. The modeling of metal
transport in ground water requires adsorption coefficients to
describe ion adsorption to soils, sediments, and rock surfaces.
The term adsorption coefficient, as used here, means a parameter
that quantifies the distribution of an ion between aqueous and
mineral phases, after accounting for speciation in the aqueous
phase. Ideally, these adsorption coefficients would be determined
from surface compiexation models, which have been highly
successful in simulating adsorption data in laboratory studies of
reference mineral phases (see review by Davis and Kent, 1990).
However, there are difficulties in determining the appropriate type
and surface density of reactive functional groups for the mixtures
of mineral phases in natural systems. Metal-ion sorption in
natural systems is commonly thought to be controlled by surface
reactions with iron (Fe) and aluminum (Al) oxyhydroxides and
organic coatings on particles. The importance of surface coatings
makes it difficult to relate the bulk mineralogical composition of a
sample to its adsorptive reactivity.
1 U.S. Geological Survey, MS 465, 345 Middlefield Ftd.
Menlo Park, CA 94025
2 U. S. Geological Survey, 28 Lord Rd., Suite 280
Marlborough, MA 01752
Application of hydrogeochemical transport models to real-world
problems involving toxic metals is limited by a lack of kinetic and
thermodynamic data for describing appropriate reactions.
Adsorption-desorption reactions of metal ions with the porous
medium should be included within the transport model. Current
approaches to the problem usually involve laboratory
measurements of adsorption coefficients in batch or column tests.
The relevance of adsorption coefficients determined in laboratory
tests to the field scale is not well known, and important questions
in this context concern the spatial variability of adsorption
coefficients and the representativeness of small samples of
subsurface material used in laboratory tests.
Determining the predominant adsorbing surface in a mineral
assemblage can be a useful approach to modeling adsorption
with a surface compiexation model. One objective of this study
was to search for a geochemical indicator of the mineral surface(s)
controlling lead (Pb) and zinc (Zn) sorption on the aquifer sand.
To determine the important mineral phases (or surfaces) for Pb2+
and Zn2* adsorption, batch adsorption experiments were done
with several grain-size and mineral fractions in which the surface
area to volume ratio was constant. A pure SiO2 sample was also
studied for comparison. The effect of removing surface coatings
on metal adsorption was investigated with commonly used selective
extraction techniques (Tessier et al., 1979; Chao, 1984).
The spatial variability of parameters used in models for solute
transport in ground water are of significance in the simulations of
solute movement. Variability of hydraulic conductivity has been
shown to control longitudinal macroscale dispersion in sand and
gravel aquifers (Garabedian etal., 1991; Hess eta!., 1992). The
random, but spatially correlated, variationsinhydraulicconductivity
U.S. Environnr
i on Recycled Paper
-------
cause small-scale variations in fluid velocity within an aquifer. It
is commonly believed that the velocity variations cause the scale-
dependent observations of macrodispersion.
Like hydraulic conductivity, adsorption coefficents and other
geochemical properties of the porous medium are expected to
vary spatially, because of variations in surface area, grain coatings,
and sedimentation patterns. A number of theoretical studies have
shown the potential importance of correlations between hydraulic
conductivity and adsorption or ion exchange coefficients
(Garabedian, 1987; Valocchi, 1989; Chrysikopoulos et al., 1990;
Kabala and Sposito, 1991). In particular, it has been shown that
an adsorbing tracer will exhibit "enhanced mixing" relative to a
nonreactive tracer if the hydraulic conductivity and adsorption
coefficients are negatively correlated. Robin et al. (1991) found
a weak, negative correlation between measured values of hydraulic
conductivity and distribution coefficients for strontium. No other
detailed measurements of the spatial variability of adsorption
coefficients in aquifers have been reported, and the importance of
this type of variability to solute transport modeling is still unknown.
In this study, 14 continuous cores of subsurface material were
collected from a shallow, sand and gravel aquifer in Falmouth,
Massachusetts, where the spatial variability of hydraulic
conductivity has been previously studied (Hess et al., 1992).
Eight of these cores were located along a transect (16 m in length)
oriented parallel to the direction of ground-water flow in order to
estimate correlation scales of physical and chemical properties.
Six other cores were collected in various locations in the field, with
distances on the order of several tens of meters separating each
core. Each core was sectioned into subsamples 10 cm in length,
yielding approximately 500 subsamples in total. For about 400 of
these subsamples, measurements were made of: 1) grain-size
distribution and 2) lead (Pb2*) and zinc (Zn2+) adsorption for a
constant set of experimental conditions (in an artificial ground-
water solution at pH 5.3). Grain-size measurements were used
to estimate hydraulic conductivity; it has been shown that these
two properties are highly correlated at this field site (Wolf et al.,
1991).
Partial chemical extractions of aquatic sediments and aquifer
materials are commonly used to estimate the partitioning of trace
metals among secondary phases such as amorphous iron and
manganese oxide (Tessier et al, 1979; Chao, 1984; Robinson,
1984/85; Tessier et al, 1989). For a smaller set of subsamples,
we have used several chemical extraction methods to determine
if Pb2* and Zn2* adsorption correlate with the amount of iron (Fe),
aluminum (Al), and manganese (Mn) extracted from the aquifer
material. If the variability in metal ion adsorption is caused by a
greater reactivity with the surfaces of a secondary mineral phase
or coating, the extraction results could reveal the cause of the
variability. If strong positive correlations exist, it may justify the
use of such selective extraction techniques as geochemical
indicators of variability in the adsorptive reactivity of subsurface
sediments.
Surface area determined by gas adsorption can be used to
estimate the total, non-selective adsorption capacity of mineral
surfaces (Davis and Kent, 1990). We measured the BET surface
area of a set of subsamples to examine the variability in reactive
surface area and the usefulness of this parameter as an indicator
of metal ion adsorption and its variability. A strong positive
correlation between surface area and metal ion adsorption might
only be expected if the sediment surfaces were relatively
homogeneous with respect to adsorptive reactivity. In this case,
the spatial variability in metal ion adsorption might only be related
to the variability in sediment surface area, which in turn might be
inversely correlated with the average grain-size.
FIELD SITE DESCRIPTION AND CORING
PROCEDURES
The study was conducted with sediments collected at the U.S.
Geological Survey Toxic Substances Hydrology Research site in
Falmouth, Massachusetts (Fig. 1). The aquifer at the site is
shallow, unconfined, and composed of glacial outwash sediments
(LeBlanc et al., 1991). The unconsolidated sediments consist
predominantly of medium to coarse sand, with localized zones of
gravel, fine sand, and silt. The aquifer has an average porosity of
0.39 and an average ground-water velocity of 0.4 meters/day.
The amount of sedimentary organic carbon in the sand and silt
fractions is between 0.01% and 0.05% of the dry weight (Barber
etal. 1992).
Ground water at the site is contaminated by the discharge of
treated sewage effluent onto rapid infiltration sand beds, located
about 300 meters upgradient (LeBlanc, 1984). The effluent
percolates into ground water below the infiltration beds, creating
a plume of contaminated water which extends over 4 km
downgradient. The contaminant plume contains elevated
concentrations of major cations (Ca, Mg, K, Na), anions (sulfate,
phosphate), dissolved metals (e.g., Zn), and various detergent
compounds (LeBlanc, 1984; Davis et al., 1991; Barber et al.,
1988).
Cores were collected using a wireline-piston core barrel (Zapico
et al., 1987) and plastic core liners 5 cm in diameter and 1.5
meters in length and were frozen until used. Three successive 1.5
meter sections were collected from each bore hole. At the
inception of the study, four cores were collected from the suboxic
zone of the aquifer (see Davis et al., 1991 for a description of
aquifer chemistry) to provide material for a large sample of
homogenized aquifer sand. This sand (see Composite Sand
below) was used to measure adsorptive properties, to test
experimental methods, and for detailed characterization of bulk
geochemical properties of the sand.
Aquifer material from 14 cores was used to evaluate the spatial
variability of metal adsorption. Eight closely-spaced cores (1 to 4
meters apart) were collected in a transect parallel to the ground-
water flow direction (see map, Fig. 1). An additional six cores
used to assess geochemical and hydrologic variability were
collected at more widely-spaced intervals (approximately 30
meters apart on average). In the vertical direction, the transect
and widely-spaced cores spanned from the pristine, recharge
zone near the water table down into the suboxic, sewage-
contaminated zone of the aquifer (see Davis et al., 1991).
EXPERIMENTAL MATERIALS
The Composite Sand. Subsections of the four cores collected
from the suboxic zone were air-dried and sieved to remove
material greater than 1000 jim diameter (>1000 n.m = 20.7% by
weight of the total). The >1000 n.m fraction was removed because
it was difficult to sample representatively in small samples (10 g)
used for batch adsorption experiments. The remaining sand,
called the <1000 jam or composite sand, was mixed together and
-------
73° 00'
r
42° 30'-/
70° 00'
TRANSECT CORES
F415C38
F415C40,
SF415C39
F415C41 _JJ
F415C42*
F415C43
F415C53*
F415C66 *
30 FEET
10 METERS
0 100 FEET
0 30 METERS
F415C54
EXPLANATION
LOCATION OF CORES AND IDENTIFIER
Figure 1. Map showing the U. S. Geological Survey Toxic Substances Hydrology Research site in Falmouth, Massachusetts (USA) and the
locations of cores taken in the study.
-------
TABLE 1. DISTRIBUTION OF GRAIN-SIZE FRACTIONS AND MINERALOGY FOR THE CAPE COD COMPOSITE SAND
Size Fractions, j
Weight % 48 hours. The
composition of the artifical ground-water solution (Table 2) was
designed to closely match that of the sewage-contaminated
ground water in the suboxic zone of the aquifer (Davis et al.,
1991). The suboxic zone of the aquifer, however, contains highly
variable concentrations of bicarbonate and nitrate anions, and
these ions were not included in the composition of the artifical
solution.
Mineral abundances in the composite sand were determined by
density separations using bromoform diluted with acetone (Table
1). Diamagnetic sand fractions of the composite sand were
prepared using a Frantz electromagnetic barrier separator.
The composite sand was predominately quartz (90-95% by
weight) with 5-10% heavy minerals, feldspars, and lithic fragments.
The abundance of heavy minerals as well as the amount of
magnetic/paramagnetic minerals increased as grain-size
decreased (Barber et al., 1992). The 250-500 |im fraction makes
up approximately 50% (by weight) of the <1000 u.m fraction and
accounts for 27% of the BET surface area (Table 1). The 500-
1000 and 64-250 |o.m fractions comprise 36% and 16% by weight,
respectively (Coston et al., 1993). The fine fraction material
makes up <1 % by weight of the composite, which is typical for this
aquifer (Barber et al., 1992). However, the specific surface area
of the <64 |im fraction is an order of magnitude greater than other
size fractions; therefore, the fine fraction accounts for nearly 10%
of the composite sand surface area.
The grain surfaces typically had a heavily weathered appearance,
and well-rounded quartz grains were common. Orange-red
coatings of variable thickness covered portions of grain surfaces.
Magnetite, hematite, and glauconite were identified in the heavy
mineral fraction. No carbonate minerals have been reported in
this aquifer (Barber, 1990).
Transect and widely-spaced core samples. Cores were subdivided
into subsamples using vertical intervals approximately 10 cm in
length and dried in a laminar flow hood, yielding 476 subsamples
in total. After drying, each subsample was sieved to remove
grains greater than 1000 urn, and the weights were recorded. The
<1000 urn fraction was then divided into two parts with a riffle
splitter, taking care to minimize the loss of fines. One part of each
sample was used for batch adsorption experiments and
geochemical characterization methods; the other part was used
for grain-size analysis of the <1000 u/n fraction. Grain-size
distributions were determined on each subsample using a dry
sieve analysis (ASTM, 1986) in the USGS Sediment Laboratory
in Harrisburg, Pennsylvania. Grain-size sieves of 710,500, 355,
250, 180, 125, 90, and 63 microns were employed for each
sample.
Pure quartz analog. Min-U-Sil. A cleaned and sized fraction of
Min-U-Sil 30 (Pennsylvania Glass and Sand Co., Pittsburgh,
Pennsylvania) was used to simulate the reactivity of Pb2+ and Zn2+
with a purified quartz surface. Min-U-Sil is a crushed and sieved
quartz powder sized for commercial applications. The powder
-------
TABLE 2. ARTIFICIAL GROUND-WATER SOLUTION USED IN
ADSORPTION MEASUREMENTS
Ion
Na
Ca
Mg
K
Cl
SO,
Actual aquifer
concentrations, jjN
1350
274
180
130
1380
305
Artificial solution
I concentrations, nM
1300
250
150
100
1400
400
was treated to remove organic material, leach out contaminant
metal oxides, and dissolve amorphous silica from the surface (D.
Kent, personal communication, U.S. Geological Survey, Menlo
Park, California, 1991). The grain-size distribution after the
treatment was 8-25 urn and the BET surface area was 0.32 m2/g.
EXTRACTION METHODS
Hydroxylamine hydrochloride sand extraction. Sand samples
were extracted with 0.25M NH2OH-HCI in 0.25M HCI at 50°C for
0.5, 72, or 96 hours (see Table 3). Hydroxylamine-HCI (HA)
extracts iron by reductive dissolution at low pH (Chao and Zhou,
1983). The HA extraction is not selective for Fe; for example, Al,
P, Ca, and Mn were also extracted from the sand. Al dissolved by
HA may be produced by partial dissolution of feldspars and other
aluminosilicate phases as well as aluminum oxide coatings on
grain surfaces. The HA extraction for 0.5 hours was used because
it has been reported to be the most effective method for selective
dissolution of amorphous iron oxyhydroxides with minimal
dissolution of crystalline iron oxyhydroxides (Chao and Zhou,
1983). Dissolved Fe increased with time in HA extracts of the 250-
500 urn diamagnetic fraction of the sand and reached a constant
concentration after 48 hours of extraction. The 96-hour HA
extractions were used to examine the effects of partial dissolution
of crystalline iron oxyhydroxide coatings on Pb2* and Zn2*
adsorption by diamagnetic fractions of the composite sand. The
72-hour HA extractions were used to study the spatial variability
of Fe and Al dissolution from a set of subsamples taken from
transect cores.
TABLE 3.
YIELDS OF DISSOLVED Fe AND Al FOR SEVERAL EXTRACTION TECHNIQUES APPLIED TO FRACTIONS OF THE CAPE
COD SAND.
Fe
Al
Mn
P
Fe
Al
Mn
P
<1000
0.87
1.8
0.02
0.29
17
20
0.18
nm
500-
1000
0.41
0.90
0.01
nm
Size Fractions, tim
250-500 250-500 250-500 64-250 64-250
diamagnetic Quartz diamagnetic
Hydroxylamine-HCI, 0.5 hours
0.62 0.45 0.35
1.3 0.86 0.82
0.01 0.00 0.00
nm 0.19 0.12
Hydroxylamine-HCI, 96 hours
— 2.7 —
— 3 —
— 0.04 —
— nm —
1.1
2.5
0.02
nm
<64
23
36
0.31
nm
Fe
Al
Mn
P
Fe
Al
Mn
P
Dithionite-Citrate
9.5 5.8 8.1 4.0
1.7 0.99 1.3 0.83
0.04 0.04 0.03 0.02
nm 0.6 0.8 0.5
HCI, 8 hours
8.1
17
0.1
nm
20
2.4
0.08
1.5
124
69
0.82
nm
5.9
1.4
0.03
nm
12
31
0.09
nm
22
8.3
0.19
5.0
nm = not measured
— = not extracted
-------
Dithionite-citrate (DC) sand extraction. Sand samples were
extracted at 25°C for 24 hours with 0.08M Na2S2O4 in 0.2M
ammonium citrate solution buffered at pH 8.5. The low-temperature
DC extraction at pH 8.5 was used to extract diamagnetic sand
because it is effective in dissolving crystalline iron coatings with
minimal dissolution of aluminum (Table 3) (S. Short, ANSTO,
Australia, personal communication).
4M HCI sand extraction. The 64-250 u,m diamagnetic sand
fraction was extracted with 4M HCI at 100°C for 1 hour in Teflon
centrifuge tubes, as a preparative procedure for batch adsorption
experiments. This sand fraction and a density-separated quartz
fraction (250-500 urn) were also extracted for 8 hours under the
same conditions to estimate the total amount of Fe and Al present
on the surfaces of grains (Table 3).
SURFACE SENSITIVE SPECTROSCOPIC TECHNIQUES
Microscale analysis of the surface coatings was accomplished
using three techniques:
1) scanning electron microscopy-energy dispersive
spectroscopy (SEM-EDS),
2) time of flight-secondary ion massspectroscopy(TOF-SIMS),
and
3) auger electron spectroscopy (AES).
Handpicked sand grains from the 250-500 H.ITI fraction of the
composite sand were examined before and after treatment with
the extractants discussed above. Compositional data for the
surface coatings were collected by TOF-SIMS and AES on grains
mounted in conducting foils. The sensitivity of these two techniques
to the outer few monolayers of the surface is orders of magnitude
betterthan SEM-EDS, which collects compositional dataaveraged
over depths of 1 micron thickness.
ADSORPTION EXPERIMENTS
Kinetics of adsorption. The rates of Pb2* and Zn2+ adsorption by
the composite sand were determined in experiments conducted
with a constant partial pressure of CO2 (1%), which resulted in a
constant pH of 5.3. These experiments demonstrated that Pb2>
and Zn2+ adsorption appeared to reach adsorptive equilibrium
within 48 hours (Costonetal., 1993). The approach to equilibrium
was probably limited by diffusion to adsorption sites located within
intragranular pores (Wood etal., 1990). In accordance with these
results, all batch experiments were conducted for a 48-hour
reaction period.
Batch adsorption experiments with composite sand. All adsorption
experiments were conducted in the AGW solution (Table 2). The
dried sand was rinsed with AGW before initiating an adsorption
experiment. Adsorption experiments with different size or mineral
fractions used a constant surface area to water volume ratio, the
same value as used for experiments with the entire <1000 \im
fraction. The surface area to volume ratio used for Zn2+ experiments
was 176 m2/L; for Pb2* experiments the ratio was 22 m2/L The
total concentration of Pb2+ or Zn2+ was 10 nM- To begin an
adsorption experiment, the sand-AGW slurry was spiked with an
acidified metal stock solution, and then acid or base (HNO^
NaOH) was added such that the desired pH was reached at the
end of the reaction period (48 hours, see below). For Zn2+
experiments, 0.73 milliequivalents (mequiv) of acid per liter of
artificial ground-water solution were added with the Zn2+ stock
solution; for Pb2+ experiments, 0.66 mequiv per liter of solution
were added. Subsequent experiments (Coston et al., 1993)
showed that the Zn2+ adsorption results were slightly affected by
the amount of initial acid added to the batch experiments (possibly
by dissolution and reprecipitation of aluminum), but that Pb2*
adsorption was not affected. Metal adsorption by the sand was
calculated from the difference between the amount of metal
added to experiment blanks (AGW and metal solution only) and
the amount remaining in solution at the end of the adsorption
experiment. In some cases, dissolved Zn concentrations were
estimated from the activity of a 65Zn radiotracer. Otherwise,
experimental solutions were analyzed by inductively-coupled
plasma atomic emission spectrometry (ICP-AES).
Batch experiments to determine the spatial variability of metal
adsorption. For each of 374 subsamples (all transect samples
plus half of the widely-spaced core samples), Pb2* and Zn2+
adsorption was measured in three batch experiments, using the
same method outlined above, but using a sand to water ratio of 50
g/L. Pb2+ and Zn2* were added simultaneously from a combined
stock solution to yield total concentrations of 20 and 5 nM,
respectively. The metal stock solutions had a lower acid
concentration than that used in the composite sand batch
experiments; 0.065 mequiv of acid were added per liter of artificial
ground-water solution. Each of the three batch experiments for a
subsample was designed to attain a different pH value after 48
hours of reaction by addition of predetermined aliquots of 0.04M
NaOH or 0.04M HNO3. The target pH values of the three
experiments were 5.1, 5.3, and 5.5. The metal adsorption
measurements for each subsample were then interpolated to a
pH value of 5.3, as described in the Results section, and averaged.
In cases where the agreement among the three experiments was
poor, or all three pH values were greater or less than 5.3, the
experiments were rerun.
RESULTS AND DISCUSSION
The Composite Sand
Metal adsorption by size and mineral fractions. Metal adsorption
experiments were conducted on four size fractions (<64,64-250,
250-500, 500-1000 jim) of the composite sand and on a
diamagnetic fraction of the 250-500 urn size fraction. Pb2+ was
preferentially adsorbed by the <64 p.m and 250-500 ^m fractions
(Fig. 2); both fractions were more reactive at a given pH than the
<1000 urn (the composite) sand on a surface area per volume
basis. The diamagnetic sand from the 250-500 JJ.ITI fraction
adsorbed less Pb2+ than the entire 250-500 urn fraction (Fig. 3).
The decreased Pb2+ adsorption by the diamagnetic fraction
suggests that iron oxide minerals and/or mineral grains with
coatings removed by magnetic separation had more reactive
sorption sites. The greater reactivity of Pb2+ with the <64 urn
fraction (Fig. 2) could be due in partto the greater amount of heavy
minerals in that size fraction (Table 1). Zn2+ adsorption varied less
among the different size fractions (Fig. 4). In contrast to Pb2*,
removing magnetic material had noeffecton Zn2+ adsorption (Fig.
5).
The composite sand is dominated by quartz in all the size fractions
examined (Table 1). Density separations were used to determine
the distribution of adsorbed Pb2+ and Zn2+ among mineral
components of the sand. Although the heavy minerals and
-------
TJ
0)
.Q
O
(0
TJ
(0
.0
Q.
v$
IUU
90
80
70
60
50
40
30
20
10
0
1 | 1 i,| M '
D 4% U g
++,-£ °A
: *& . '
A rt°
A O
1 *
•f j»
- QD* ^^
- ^oo
CD
-A • <1000^m
- o ° 500-1000 \im
- ° D 250-500 urn -
A 64-250 |4.m
- •+ <64 jim
i i i i i
U.1-O
0.36
CM
0.27 |
a.
0.18 |
^
0.09
0.00
4567
Final pH
Figure 2. Pb2* adsorption on various size fractions of the composite sand. The surface area of sand used in each experiment
was held constant at 22 m2/L; total Pb2* concentration was 10 micromolar in each experiment. Both the 250-500 |im and <64 |am
fractions adsorbed more Pb2* than the total <1000 nm fraction.
TJ
a>
o
^M
"D
0}
n
a
c£
IUU
90
80
70
60
50
40
30
20
10
n
1 ' ' ' n ' *i»A' + '
D 250-500 ^m fraction +
• diamagnetics Q D
• • 96hrHA • ^
- A 0.5 hr HA n ^ ~
- + 24 hr DC <-"
- O Si09 + A
• *
A
"
on m
<
+ A * O
"^ A
-+ + o
- +
- o -
L i* i O i O Oi i i
0.45
0.36
CM
0.27 -i
JD
Q.
(I)
0.18
0.09
_
o
E
3.5
4.5
5.5
Final pH
6.5
7.5
0.00
Figure 3. Pb2* adsorption by diamagnetic and extracted diamagnetic sand from the 250-500 urn fraction. The surface area of
sand used in each experiment was held constant at 22 m2/L; total Pb2* concentration was 10 micromolar in each experiment. Pb2*
adsorption after three different extraction procedures is shown (DC refers to the dithionite-citrate-ammonium extraction; HA
refers to the hydroyxlamlne-hydrochloric acid extraction). Pb2* adsorption on a pure SiO2 surface (22 m2/L; Min-U-Sil) is plotted
for comparison with the extracted surfaces.
-------
0)
o
W
•o
(0
c
N
IVV
90
80
70
60
50
40
30
20
10
0
, , ^F"ns v — -
-t-
A
I -^
• dtf-0
^k
-
• -f*
• A B
A
— rr
^\
AtS1
o
•o
(0
N
o^
IUU
90
80
70
60
50
40
30
20
10
0
' *, ' + '- —
D 250-500 jim fraction **
• diamagnetics B
+ 24hrDC o -
" • 96hrHA +
" O SiO5 *
+
'- + O
B
-
+
1 ++ •
& •
• o
_
o
1 1 XV 1 Ol 1 1 1
u.uo/
0.038 CM
E
N
O
0.019 ZL
0.000
3.5 4.5 5.5 6.5 7.5
Final pH
Figure 5. Zn2* adsorption by diamagnetic and extracted diamagnetic sand from the 250-500 urn fraction. The surface area of
sand used in each experiment was held constant at 176 m2/L; total Zn2* concentration was 10 micromolar in each experiment. Zn2*
adsorption after two different extraction procedures is shown (DC refers to thedithionite-citrate-ammonium extraction; HA refers
to the hydroyxlamine-hydrochloric acid extraction). Zn2* adsorption on a pure SiO2 surface (176 m2/L; Min-U-Sil) is plotted for
comparison.
-------
feldspars were more reactive than quartz grains on a weight-
normalized basis, greater than 80% of Zn2* was adsorbed by the
quartz fraction and more than 90% of Pb2* by the quartz and
feldspar fractions (Coston et al., 1993). The quartz accounts for
such a large weight percentage of the sand that it dilutes the effect
of preferential adsorption by the heavy minerals.
Metal adsorption after extraction of the composite sand. The
diamagnetic fraction of the composite sand had a much greater
average reactivity than that observed for the pure quartz surface
(Min-U-Sil)(Figs. 3 and 5). Even though quartz represents a
large fraction by weight of the sand, adsorption on a clean
quartz surface (i.e., complexation with surface silanol groups)
of equal surface area to water volume does not account for the
observed adsorption of either metal ion. A coating must be
associated with the quartz grain surfaces that enhances metal
adsorption. Chemical extractions (hydroxylamine-HCI (HA),
dithionite-citrate (DC), and HCI) were used to: 1) dissolve and
quantify the extent of grain coatings, and 2) prepare diamagnetic
sand fractions for further metal adsorption experiments (Coston
et al., 1993). The visible orange-red hue of the sand was
lightened or completely removed by each of these extraction
procedures. Amounts of Al and Fe removed by each extraction
technique are shown in Table 3. In addition, sand grains were
examined by surface analytical techniques to determine the
composition and distribution of surface coatings before and
after extraction.
A 0.5 hr HA extraction, intended to dissolve noncrystalline hydrous
iron oxides (Chao and Zhou, 1983), did not affect Pb2+ or Zn2+
adsorption (e.g., Fig. 3). Pb2+ adsorption was reduced after HA
extraction of Fe from the 250-500 urn diamagnetic fraction was
complete (96 hours, Fig. 3). This extraction, however, had no
significant effect on Zn2* adsorption (Fig. 5). Neither the 0.5 or 96
hr HA extractions affected the specific surface area measured for
these samples. The 96 hr HA extraction, although intended to
selectively dissolve crystalline iron oxides, also dissolves a
significant amount of aluminum. The experimental and modeling
results (Coston etal., 1993; Davis etal., 1993) suggest that Pb2+
adsorption is influenced more by crystalline iron oxides in the
sand, while Zn2+ adsorption is likely influenced more by aluminum-
bearing coatings or minerals.
SEM-EDS backscatter images showed Fe enrichment at
irregularities in the surface topography such as cracks, cleavage
planes, and pits. The coatings appear to fill in irregularities on the
grain surfaces. After extraction by HA for 96 hours, detectable Fe
remained along the base of cleavage fractures on feldspars and
in conchoidal fractures on quartz grains. However, the surfaces
of extracted grains examined by the TOF-SIMS technique
appeared to retain coatings rich in Fe and Al at least several
monolayers thick. The differences between the SEM-EDS and
TOF-SIMS results can be attributed to the different depths over
which the surface composition is analyzed.
For all size fractions, the ratio of Fe to Al dissolved was much
greater in the DC extraction than for HA extractions (Table 3).
Therefore, the DC-extracted surface was enriched in Al and
depleted in Fe relative to the HA-extracted sand surface. There
was 10 to 30 times more Al than Fe detectable by AES at the
surface (10-100 A deep) of unextracted quartz grains. Fe was not
detected by AES on the surface of either the HA or DC extracted
quartz grains. However, significant Al remained on the surface of
all extracted quartz grains examined.
DC extraction of the 250-500 (xm diamagnetic fraction dissolved
more Fe than the 96-hour HA extraction (Table 3), had a small
effect on Pb2* adsorption (Fig. 3), and did not affect the specific
surface area measured for the sample. The Pb2+ adsorption
results suggest the surface coating structure may be more open
after the extraction, and that aluminol surface sites may be
important after Fe is dissolved. The DC extraction appeared to
cause a slight increase in Zn2+ adsorption (Fig. 5), suggesting that
the Al-enriched surface is more reactive with Zn2*. It was also
found that excess acid added with metal stock solutions caused
an increase in Zn2+ adsorption in batch adsorption experiments (where
pH is allowed to drift), as compared to experiments exposed to a
constant partial pressure of CO2. Pb2* adsorption, however, was not
affected. It was hypothesized that the addition of excess acid caused
a dissolution and reprecipitation of Al at the surface, resulting in
increased Zn2+ adsorption (Coston et al., 1993).
Adsorption of Zn2+ by 64-250 |im diamagnetic sand was decreased
by a 4N HCI extraction (1-hour) at 100°C (Fig. 6). This harsh
extraction nearly doubled the BET surface area (from 0.25 to 0.48
m2/g) of the sample. Al was easily detected on the surface of the
HCI-extracted sand by TOF-SIMS; the amount of Fe at the
surface, however, was reduced to just above the limit of detection
under the same beam conditions. These results, the apparent
redistribution of Al on the sand surface when excess acid was
added to batch experiments, and the effects of the DC treatment
on the diamagnetic sand each suggest that aluminol (AIOH)
groups may be important in complexing Zn2+ at the sand surface
(Coston etal., 1993).
SPATIAL VARIABILITY OF SUBSURFACE SEDIMENT
PROPERTIES
Spatial variability of metal adsorption. Adsorption results for each
of the three batch experiments with each vertical subsample were
interpolated to pH 5.3 using the Kurbatov equation with a
macroscopic proton coefficient of one (Kurbatov et al., 1951;
Honeyman and Leckie, 1986). Figure 7 shows the variability of
Pb2+ and Zn2* adsorption data for core F415-C38 (core 38), a
transect core (see Fig. 1). The results from this core were typical
of those from other cores in the transect. In most cores, Pb2*
adsorption varied by a factor of two over the length of the core.
Zn2+ adsorption was less than that observed for Pb2+, but the
variability of Zn2* adsorption was generally greater (usually about
a factor of three or four) than that observed for Pb2* adsorption.
Table 4 summarizes the results of Pb2* and Zn2* adsorption for all
cores. The mean values of metal adsorption for the transect cores
were not significantly different from the mean values observed for
widely-spaced cores, but a larger range of vertical variability in
Pb2+ adsorption was observed in some widely-spaced cores. The
range of Pb2+ adsorption observed for all cores was 0.091-0.371
nmoles/g sand; the Zn2+ range was 0.00-0.028 ixmoles/g. Pb2+
and Zn2* adsorption were significantly correlated (p<0.01; r=0.53,
n=374). In this report, a linear correlation between two variables
is considered significant if the absolute value of the correlation
coefficient (r) is greater than the critical value of r for the 1 % level
of significance (p<0.01) for n-2 degrees of freedom (Sokal and
Rohlf, 1973).
The greatest adsorption measurements reported here do not
represent adsorption maxima for those subsamples; they are
simply the greatest measurements observed under the constant
conditions imposed in the batch experiments (pH 5.3; total Zn of
-------
0)
1
to
c
N
^
IUU
90
80
70
60
50
40
30
20
10
n
A x
— A /A
X
&
-
-
A X
A
* * x X
_
A
Ax A
- A 64-250 H.ITI fraction
A diamagnetics
X 1 hr HCI extraction
i i i i i
0.057
0.038
0.019
N
CO
jj>
O
4
7
0.000
Final pH
Figure 6. Effect of 4N HCI-extraction (100°C) on Zn2* adsorption by the 64-250 \nn diamagnetic fraction. The surface area of
sand used in each experiment was held constant at 176 m2/L; total Zn2* concentration was 10 micromolar in each experiment
TABLE 4. RANGE, MEAN, AND STANDARD DEVIATION OF Pb2* AND Zn2* ADSORPTION MEASUREMENTS (nmoles/g) AT pH 5.3 IN
TRANSECT AND WIDELY-SPACED CORES
Core
F415C38
F415C39
F415C40
F415C41
F415C42
F415C43
F415 C53
F415C66
All transect cores
F415C44
F415C51
F415C54
F509C2
F510C3
F511 C2
All wide-spaced cores
Average Pb
adsorbed
(Std. Dev)
0.22 (0.02)
0.22 (0.02)
0.23 (0.02)
0.22 (0.03)
0.23 (0.02)
0.23 (0.02)
0.24 (0.02)
0.22 (0.02)
0.23 (0.02)
0.22 (0.02)
0.22 (0.05)
0.23 (0.06)
0.22 (0.06)
0.22 (0.02)
0.23 (0.03)
0.22 (0.04)
RANGE
High
0.28
0.27
0.28
0.34
0.26
0.29
0.27
0.27
0.34
0.26
0.32
0.37
0.28
0.27
0.29
0.37
Low
0.18
0.19
0.19
0.20
0.18
0.18
0.20
0.20
0.18
0.19
0.14
0.15
0.09
0.19
0.19
0.09
Average Zn
adsorbed
(Std. Dev.)
0.012 (0.005)
0.010(0.005)
0.014 (0.002)
0.012 (0.004)
0.012(0.003)
0.01 1 (0.003)
0.014 (0.003)
0.011 (0.002)
0.012 (0.004)
0.008 (0.005)
0.01 1 (0.003)
0.014 (0.004)
0.007 (0.005)
0.012 (0.001)
0.01 1 (0.004)
0.010(0.005)
RANGE
High
0.028
0.022
0.019
0.026
0.018
0.018
0.022
0.015
0.028
0.017
0.018
0.021
0.014
0.015
0.021
0.021
Low
0.003
0.000
0.011
0.006
0.006
0.006
0.006
0.004
0.000
0.003
0.006
0.008
0.000
0.010
0.004
0.000
All cores
0.22 (0.03)
0.37
0.09
0.012 (0.004)
0.028
0.000
10
-------
13
12
(1)
§ 11
a>
-o
1 10
<
% Pb Adsorbed at pH 5.3
0 20 40 60 80 100
% Zn Adsorbed at pH 5.3
0 10 20 30 40 50
i i i i i i i i i i i i i
I i i i i I
®
A:
_ Odata
0 interpolated
13
12
11
10
9 -
I
B
© interpolated .
i , I . , , , I , , .
0 0.1 0.2 03 0.4
Pb Adsorbed at pH 5.3 (u.moles/g)
0 0.01 0.02 0.03 004 0.05
Zn Adsorbed at pH 5.3 (umoles/g)
Figure 7. Pb2* and Zn2* adsorption measured in batch experiments with artificial ground water and subsamples of subsurface
material (Core 38). Diamonds show actual measurements made near pH 5.3. Circles are interpolated to pH 5.3 using data of 3 batch
experiments in the pH range 5.1-5.5. Horizontal error bars represent the standard deviation of the measurement. MSL = mean
sea level.
5 uM; total Pb of 20 ^M; 50 grams sand per liter of AGW solution).
An adsorption maxima of 1.7 u.moles/g can be estimated from the
surface area (Table 1) of the composite sand (<1000 u.m) and the
recommended adsorption site density of 3.84 nmoles/m2, from
Davis and Kent (1990).
Spatial variability of grain-size distribution and estimates of
hydraulic conductivity. K. The range of grain-size distributions in
the 376 subsamples is shown in Figure 8. The results were similar
to those observed by Wolf et al. (1991), for cores collected
approximately 6 meters west of our transect cores (Hess et al.,
1992). Table 5 shows the maximum, minimum, average, and
standard deviation values of d10 for subsamples in the transect
and widely-spaced cores. d1p is the grain diameter at which 10%
of thesubsample mass was of smaller size. Hydraulic conductivity
was estimated for each subsample from the empirical relation of
Hazen(1893):
K=A-(d10)2
(D
where A is 1.157 for Kin cm/sec (at 10°C) and d,0 is in mm.
Figure 9 illustrates the variability of /(estimates for subsamples of
core 38. The range of values for K for all samples (Table 5) is
similar to that reported by Wolf et al. (1991) for Kestimated from
grain-size distributions and permeameter measurements. The
average Kestimated from a tracer test (LeBlanc et al., 1991) and
from borehole flowmeter measurements (Hess et al., 1992) is
larger by about 30 percent. The small-scale variability in K
estimated from the flowmeter method is similar to that estimated
from permeameter measurements (Hess et al., 1992) and from
grain-size distributions (Wolf et al., 1991, and this study).
Variability of surface area and sand extraction yields. The BET
surface area of subsamples from core 38 varied by a factor of two,
with no apparent trend with depth (Figure 10). The d10 values in
this core ranged from 0.174 to 0.362 mm. An inverse relationship
between surface area and d, 0 for core 38 subsamples is observed
(Fig. 11). Excluding the obvious outlier, a linear regression
through the data gave a significant correlation (r= -0.62). The
relationship does not appear to be linear, however; and a parabolic
function would probably yield a better fit.
The range in extracted Fe, Al and Mn from this core is shown in
Table 6. No apparent trend with depth was observed for any of the
extracted metals. In units of u.moles/g, Al was significantly
correlated with Fe in the three extractions, despite the greater
selectivity of DC for Fe over Al. The dissolution rates of Fe and
Al in the extracting solutions is likely proportional to the exposed
surface areas of minerals containing these elements. The fact
that Fe and Al were highly correlated in all three extractions
suggests that the source is largely from the dissolution of mixed
oxide coatings on the grain surfaces (Coston et al., 1993). If the
source was from dissolution of separate minerals, such as a
mixture of magnetite and feldspar, it is unlikely that the abundances
of these minerals would covary throughout the core. Extracted
Mn was strongly correlated with Fe in both the 0.5 and 72-hour HA
extractions, but not in the DC extraction. Extracted Fe, Al, or Mn
by the three methods did not correlate with BET surface area. The
lack of correlation may be due to two factors:
11
-------
100 I i i i i
Particle-Size Diameter (mm)
Figure 8. Range of grain-size distributions measured for the 376 subsamples from transect and widely-spaced cores. The right curve
shows the grain-size distribution for the subsample with the largest value of d10. The left curve shows the grain-size distribution
for the subsample with the smallest value of d10. The average value for d,0 (and one standard deviation) are shown; d10 is the
grain diameter at which 10% of the sample (by weight) is of smaller size.
1 w i i I i i r i i i I T~l i f i i (~"~1
12
u>
Q>
1 "
CO
£
CD
3
-------
13
12
W
o>
S 11
E,
Q>
-------
TABLE 5. RANGE, MEAN, AND STANDARD DEVIATION OF d,0 FOR TRANSECT AND WIDELY-SPACED CORES
Transect
Widely Spaced
Mean, mm 0.252
(Standard deviation) 0.046
Maximum, mm 0.475
Minimum, mm 0.076
K
Mean, cm/s 0.076
(Standard deviation) 0.029
Maximum, cm/s 0.261
Minimum, cm/s 0.007
0.264
0.046
0.387
0.170
0.083
0.029
0.173
0.033
All Cores
0.255
0.046
0.475
0.076
0.078
0.029
0.261
0.007
1) the BET surface area is indicative of the entire surface area
of samples, including surfaces other than those containing
the extracted elements, and
2) variable thickness of coatings on grain surfaces could break
down any expected relationship between elements dissolved
and surface area.
Characterization techniques as geochernjcal indicators of
adsorption pptentiql. Linear regressions of the Pb2+ adsorption
data in units of jimoles per gram of sand did not yield significant
correlations with grain-size (d10) or surface area. This result
suggests that Pb2* adsorption occurs at specific areas on the sand
surface that are more reactive, whereas the BET surface area is a
measurement of the entire surface area. Pb2* adsorption was also not
correlated well with 0.5 hour HA extractable Fe, Al, or the sum of
extractable iron and aluminum (Fe+AI) in units of nmoles/g. These
elements are expected to dissolve from the the reactive portion of the
surface area (Davis and Kent, 1990). However, significant correlations
were observed for linear regressions of Pb2* adsorption with 72-hour
HA Al and Fe+AI and with the DC Fe, Al, and Fe+AI (table 7). Because
the molar ratio of Pb2* adsorbed to Mn extracted was greater than one
for all extractions, Mn was not used as a variable for correlation analysis
with Pb2*. No significant correlations were found between Zn2+
adsorption (in units of nmoles/g) and BET surface area, grain-size, or
any of the dissolved elements determined in the three extraction
methods.
Because the batch adsorption experiments (to determine spatial
variability) were conducted using conditions of equal sand mass
per volume of AGW, the adsorption and extraction data were
normalized per unit surface area to eliminate surface area as a
variable. Significant, positive correlations between Pb2+ and Zn2*
adsorption and 0.5 hour HA, 72-hour HA, and DC-extracted Fe,
Al, Fe+AI, Fe+AI+Mn were obtained after normalizing to surface
area (for example, see Figure 12); the correlation coefficients
were greater than those found without normalization to surface
area. Both Pb2+ and Zn2+ adsorption were more highly correlated
with the 72-hour HA- and DC-extracted Fe and Al than with the
0.5-hour HA extraction (Table 7). Because Fe and Al are highly
correlated with each other in all extractions, Pb2t or Zn2+ adsorption
each correlate well with extractable Fe, Al, or the sum of Fe and
Al, when normalized to surface area. Thus, the extraction data
does not allow one to distinguish between metal adsorption onto
Al or Fe phases.
The significant correlation of Zn2+ adsorption with Pb2+ adsorption
suggests that the abundance of adsorption sites for these two
metals covaries within a core, or that Pb2+ and Zn2+ adsorb to the
same type of mineral surface sites. Assuming that the 0.5 hr HA
TABLE 6. RANGE OF EXTRACTED IRON, ALUMINUM, AND MANGANESE FOR SUBSURFACE SAMPLES OF CORE 38
Extraction
Element 0.5-hour HA
nmoles/g \imoles/m2
72-hour HA
\imoles/g \imoles/m2
DC
nmo/es/g (imotes/m2
Fe
Al
Mn
2.4 - 0.4
2.4 - 0.7
0.16-0.02
10-1.2
9.9-2.6
0.79 - 0.05
19- 12
1 -11
0.20-0.11
86-33
81 -33
1.5-0.3
11-5.6
1.9-1.0
0.12-0.03
52-22
7.9 - 3.7
0.60-0.12
14
-------
TABLE 7. SUMMARY OF REGRESSION ANALYSIS FOR PB*' ADSORPTION DATA AT pH 5.3 FOR SUBSURFACE SAMPLES OF CORE 38
Dependent
variable
Independent
variable
n-2
slope intercept
nmoles/g Pb, Fe, Al
Pb
Pb
Pb
Pb
Pb
Znc
(imoles/m2 Pb, Fe, Al
72 hr Ala
72 hr Fe+Ala
DCFe"
DCAI"
DC Fe+AI"
Pbc
19
19
21
21
21
266
0.60
0.52
0.63
0.62
0.64
0.58
0.55
0.55
0.53
0.53
0.53
0.16
0.006
0.003
0.012
0.082
0.011
0.099
0.144
0.145
0.121
0.112
0.115
0.011
Pb
Pb
Pb
Pb
Pb
Pb
Pb
Pb
Pb
0.5 hr Fed
0.5 hr Ald
0.5 hr Fe+Ald
72 hr Fea
72 hr Ala
72 hr Fe+Ala
DCFe»
DCAI"
DC Fe+Alb
24
24
24
19
19
19
21
21
21
0.60
0.76
0.73
0.86
0.88
0.89
0.80
0.87
0.82
0.50
0.50
0.50
0.55
0.55
0.55
0.53
0.53
0.53
0.045
0.060
0.030
0.009
0.010
0.005
0.017
0.111
0.015
0.718
0.533
0.595
0.365
0.336
0.326
0.308
0.279
0.293
• 72 hour hydroxylamine hydrochloride extraction at 50°C
b 24 hour dithionite-ammonium citrate extraction
c All transect cores
d 0.5 hour hydroxylamine hydrochloride extraction at 50°C
8 Critical value of r for 1% level of significance
"o
E
JD
CL
20
40 60
Fe (jimoles/m2)
80
100
Figure 12. Pb2* adsorption (in units of micromoles per square meter of surface) as a function of iron extracted from sand grain coatings by
hydroxylamine hydrochloride (micromoles per square meter) for subsamples of subsurface material (Core 38).
15
-------
extraction accurately estimated the ferrihydrite content (Chao
and Zhou, 1983), the adsorptive reactivity of the composite sand
for Zn2+ was calculated using a two-site, diffuse double layer
surface complexation model (Dzombak and Morel, 1990). The
model simulations (Davis et al., 1993) suggest that insufficient
ferrihydrite was present to appreciably affect Zn2+ adsorption by
the sand, which is consistent with the experimental results (Fig.
5). The experiments showed that a HA extraction (0.5 hr) did not
significantly change Pb2+ adsorption either (Fig. 3).
The results forZn2+ and Pb2+ adsorption by diamagnetic fractions
of the sand showed that very strong extractions were required to
alter the adsorptive interactions, i.e., extractions capable of
dissolving crystalline iron oxides, such as goethite or hematite.
Using literature data, we estimated apparent stability constants
for the adsorption reactions with goethite and hematite, and the
surface site density of the crystalline iron oxides on the sand was
estimated from the DC extractions. The modeling (Davis et al.,
1993) suggests that goethite surfaces are not present in sufficient
abundance to explain the reactivity of the sand. A similar
conclusion was reached assuming that all iron dissolved was
present as hematite.
As mentioned above, normalization of the extracted Fe and Al to
surface area significantly improved the correlation with metal
adsorption. Thesignificantcorrelationsbetween metal adsorption
and the extracted Fe and Al (DC or HA-72 hr) suggest that these
extractions, combined with surface area measurements, could be
used as an effective geochemical indicator of the potential variability
of metal adsorption, at least in this aquifer. Because of the many
potential artifacts and time-intensive nature of adsorption
experiments conducted with natural materials (Davis and Kent,
1990; Coston et al., 1993), simple laboratory characterization
methods are needed to provide a first approximation of the
adsorptive reactivity of site-specific materials at contaminated
field sites. Such measurements will obviously never provide a
more accurate measurement than carefully-controlled adsorption
measurements, but the measurement of simple geochemical
indicators may be preferable in site assessment or other non-
research applications. However, the usefulness of the indicator
discovered in this study needs to be tested at several other sites
before it can be considered an effective characterization tool.
Correlation between metal ion adsorption and hydraulic
conductivity. When all the data are considered, a statistically
significant linear correlation between K and Pb2+ adsorption is
found (r=-0.29,n=374). In contrast, Zn2+adsorption was statistically
independent of K(r=-0.089, n=374). Figure 13 shows the scatter
diagram for K and Pb2* and Zn2* adsorption (data from the
transect cores only). The transect data exhibited larger correlation
coefficients for both Pb2+ and Zn2+ adsorption with /(than did the
data from the widely-spaced cores. For example, the correlation
of Pb2+ adsorption with K (transect data only) had a greater
absolute value for the correlation coefficient (r=-0.41, n=268).
The overall correlations can be compared to that determined for
Sr sorption by a calcareous, sandy aquifer material at the Borden
experimental site (Robin et al., 1991). These authors also found
a very weak, negative correlation between the natural log (In) of
the distribution coefficient for Sr (Kd) and In K. The r of this
correlation was -0.13, thus having an absolute value for the
correlation coefficient larger than found in this study for Zn but
much weaker than that found for Pb.
0.4
CM
0.3
CO
_CD
o
e
Q)
.a
o
CO
T5
03
JD
0.
0.2
0.1
0
1 1 1 1 1
-
-
x x
~ X
xX *$jft **
xx^^y
-x 4- ^^^^X*5
+ 4- *
+
1 1 1 1 1 1 1 1 I 1 1 1 1 | 1 1 1 1 | 1 1 1 1
xpb
x
+ Zn
5< x x x x
^M^ip"^ ^S< «> ^ *
v5$^0 Jv, v* x x
^TVX% x X x x
+ x
+++ ++ 4-H-H- -H- + +
+ +H- +
_ 1 1 n 1 1 1 m -f-
i i iiiinitii
-Hf + +++++ +
i nun + + +
ii t\r nil ll J t
-+ + -H- 111 MI mnm-H- ++ ++ + + +
i it III!'"" ' ' '
T T r 1 ITT
~~ "H" ~tr "ii T~
+ *
- ^- 4.
T IIBT T nr i r i T
"HHt" 1 HHil I "IT ~~
H4+ -H- + +
-m- + + + + + -
-H-+ +
+ -H-+ + + + -H- +
1 1 1 1 1
1 t 1 I I I I 1 I I 1 1 1 1 1 1 1 1 1 .] _J _L
0.04
0.03
^O
"o
E
0.02
o
CO
T3
CO
0.01 ^
0
0 0.05 0.10 0.15 0.20 0.25 0.30
K (cm/s)
Figure 13. Pb2* and Zn2* adsorption as a function of hydraulic conductivity (estimated from grain-size distribution) for the transect cores.
16
-------
STATISTICAL EVALUATION OF PARAMETER
VARIABILITY
Hydraulic-conductivity data sets typically exhibit a skewed
statistical distribution; Kcan vary by several orders of magnitude,
but all values are positive (Freeze et al., 1990). Most workers
have found that a log normal distribution provides a good fit to the
data (e.g., Woodbury and Sudicky, 1991). Thus if,
Y, = In K, (2)
the parameter Y is normally distributed. However, the variability
of /(within the aquifer is not completely random; commonly, the
observed values of Y, are spatially correlated with one another.
Values separated by short distances are generally highly
correlated, and those separated by long distances may be only
weakly correlated or not correlated at all. The function that
displays the drop in correlation with distance is called an
autocorrelation function (Freeze et al., 1990).
The autocorrelation function may take a number of forms; one of
the most commonly used functions is the exponential model,
pY(x) = exp [-|
(3)
where x is the separation distance, py(x) is the autocorrelation
function, and Xy is an exponential decay parameter, known as the
correlation length or scale. The latter parameter is a measure of
the distance over which the Y-value is correlated; specifically, it is
the distance over which py(x) decays to a value of e~1. It is
reasonable to anticipate a direct relationship between the
magnitude of the correlation length of the hydraulic conductivity
and the average dimension of bedding structures within a
sedimentary deposit (Freeze et al., 1990).
Var'iQgram analysis. A spatial autocorrelation function requires
continuous data. In practice, we typically have discrete samples
that are continuous in the vertical direction and discontinuous in
horizontal directions. A geostatistical analysis allows quantification
of the spatial structure of discrete data (Journel and Huijbregts,
1978). The basic tool of the analysis is the experimental
semivariogram (referred to below as the variogram) which is
calculated as the mean-squared differences between sample
values at specified separation distances,
where y is the variogram statistic, h is the separation distance
between observations, n(h) is the number of data pairs separated
by h, and the summation is made from i=1 to \=n(h). In practice
the variogram is constucted so that h is the mean distance
between points within a range of distances known as a lag class
(Hess et al., 1992). Data exhibiting spatial correlation will give
variogram statistics that are small at small separation distances
and that increase to the data set variance at distances beyond the
correlation scale. Isotropic and directional analyses can be
performed.
To characterize the spatial correlation structure of the hydraulic-
conductivity and adsorption data sets, directional variogram
analyses were conducted on a subset of each data set. Data from
the 8 cores located in the transect (Fig. 1) were included in the
analyses. This transect is in the vicinity of and is approximately
parallel to the transect used by Hess et al. (1992) to characterize
the correlation scales of the variability in hydraulic conductivity in
this aquifer. The transect is also approximately parallel to the
mean direction of ground-water transport at the site and to the
hypothesized mean direction of the glacial outflow which deposited
the sediments at the site (Hess et al., 1992).
The hydraulic-conductivity and adsorption data sets display
truncated and skewed natures which suggest lognormal
distributions. The lognormal nature of the adsorption data is
speculated here; however, Robin et al. (1991) found little evidence
that the log transformation affected the spatial behavior of the
variables, only the magnitude of the spectral estimates. Because
of the correlated nature of our data sets, a rigorous test of
normality is difficult. The variogram analyses were conducted on
In-transformed data sets. The In-transformed data sets—hydraulic
conductivity (In K), lead adsorption (In Pb ads), and zinc adsorption
(In Zn ads)— contained 268 values from the 8 boreholes.
In the variogram analyses, the lag classes were based on the
sample spacing. In the vertical, samples were on average 0.1 m
in length, and for the most part, sampling was continuous within
boreholes. Therefore, the minumum lag in the vertical was set to
0.1 m, which represented the average vertical spacing between
the centroids of adjacent samples. Each lag class calculation
included at least 100 data pair comparisons in the vertical.
In the horizontal direction, sampling was not continuous; samples
were limited to discrete boreholes. The minimum spacing between
boreholes was approximately 1 m; a mean distance of 1.4 m was
used for the smallest lag in the horizontal variograms to ensure a
minimum of 50 data pairs in each lag class. Partly because of the
continuous nature of the vertical sampling and the discontinuous
nature of the horizontal sampling, vertical variograms were better
defined than were horizontal variograms; this is the case with the
variograms reported below.
Negative-exponential models were chosen to represent the
variogram trends:
y(h) = s2[1 -
(5)
where y(/i) is the variogram statistic (Equation 4), s2 is the
variogram sill, h is the average separation distance between data
pairs compared in the lag class, and A. is the correlation length. To
fit the model to the experimental variograms, both the sill and the
correlation length can be varied. The sill sets the value which the
model approaches asymptotically as the separation distance
increases. The sill is typically set to (or near to) the sample
variance, as was done here. The correlation length is, therefore,
the primary model parameter that is fit to the data; this length
controls the rate at which the model approaches the asymptote
and quantifies the spatial correlation within the data.
Correlation scales for In /Cand In (metal adsorbed). The vertical
variograms (Fig. 14) show greater correlations (smaller variogram
statistics) at smaller separation distances than at larger distances.
As the separation distance increases, the variogram statistic
approaches the sample variance. The vertical correlation scales
identified by the fitted models were 0.1,0.15, and 0.26 m for In K,
In Pb ads, and In Zn ads, respectively. These values are similar
to the vertical correlation scale for In K(0.18-0.38 m) identified by
Hessetal. (1992). The fact that the vertical correlation scales for
In K, In Pb ads, and In Zn ads is of similar magnitude is interesting
and warrants further investigation; however, the similarity in
17
-------
0.015
0.01
0.005
0.2
0.15
0.1
0.05
0.5 1
Separation Distance, in meters
1.5
Figure 14. Vertical variograms of A) In (lead adsorption), B) In (zinc adsorption), and C) In (K) (hydraulic conductivity). Exponential model of
the type y = s2[1 - e<'xrt>] is fit to the data, where 7 is the variogram statistic, s* is the variogram sill, x is the average separation
distance between data pairs compared in the lag class, and K is a characteristic length (correlation scale).
18
-------
correlation scales does not imply causality. The negative
exponential model is only one of several models (Journel and
Huijbregts, 1978) that might adequately fit these variograms.
However, the alternative correlation scales are not expected to
vary in their orders of magnitude from those presented here.
The horizontal spacing between the eight boreholes along the
transect ranged between 1 and 16 meters. At the maximum
spacing of 16m, data pairs from only two boreholes are available
and the number of pairs is less than 50. Therefore, the resulting
variogram statistics at the largest separation distances are not
reported. The resulting horizontal variograms (Fig. 15) provide
correlation estimates between separation distances of 1.4 and 12
m; the variograms show a lack of correlation at these horizontal
separation distances. The horizontal sampling density at small
separation distances was not adequate to resolve a small-scale
correlation structure. Therefore, no models were fit to these
horizontal variograms. Additional data would be necessary to
resolve the horizontal correlation structures of these variables.
CONCLUSIONS
1. Adsorption of Pb2+ and Zn2* on the highly-weathered, coarse
aquifer sand was dominated by quartz grains (95% by
weight) in the sand. Heavy minerals and feldspars were
more reactive than quartz grains on a weight-normalized
basis, but greater than 80% of Zn2+ was adsorbed by the
quartz fraction and more than 90% of Pb2+ by the quartz and
feldspar fractions because of the greater abundance of this
mineral fraction. The reactivity of the metal ions with the
quartz grains in the sand was much greaterthan that observed
for purified, commerically-prepared quartz grains. The greater
reactivity of the natural quartz surfaces can be attributed to
thick (100-300 nm) coatings on the sand grains. The coatings
are composed of surface precipitates of complex mixed
oxides and silicates, containing high concentrations of Fe
and Al. The coatings are derived from the weathering of
feldspars and other accessory minerals in the aquifer.
2. The spatial variability of Pb2+ and Zn2+ adsorption under
constant chemical conditions was moderate in this aquifer at
a scale of 100 meters. Pb2+ adsorption ranged from 0.091-
0.371 nmoles/g sand; Zn2+ adsorption ranged from 0.00-
0.028 nmoles/g sand. The mean values of metal-ion
adsorption for all samples (n=268) of a 16 meter transect
were not significantly different from the mean values observed
for samples (n=106) taken from more widely-spaced cores.
3. The variability in Pb2+ and Zn2+ adsorption was attributed to
both variability in surface area and in the composition of
surface coatings of grains, as indicated by extracted Fe and
Al. Strong correlations between metal-ion adsorption and Fe
and Al extraction yields were observed when both quantities
were normalized per unit surface area. A non-linear, inverse
relationship was observed between surface area and grain-
size distribution.
4. Pb2* adsorption showed a significant, but weak, negative
correlation with hydraulic conductivity (K), estimated from
the grain-size distribution; Zn2+ adsorption exhibited no
correlation with K. The correlation between Pb2+ adsorption
and K were stronger when only the data from the transect
cores were considered.
5. Vertical correlation scales identified by a simple exponential
model were 0.1, 0.15, and 0.26 m for In K, In Pbads, and In
Zn ads, respectively. These values are similar to the vertical
correlation scale identified by Hess et al. (1992) for In K.
Variograms for the three parameters showed a lack of
correlation at the horizontal separation distances examined
(between 1.4 and 12m).
ACKNOWLEDGEMENTS
This work was funded by the U.S. Environmental Protection
Agency under Interagency Agreement DW14934639 with the
Robert S. Kerr Environmental Research Laboratory, Ada,
Oklahoma. It has been subjected to Agency review and approved
for publication. Mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
Technical advice from R. Pulsof the R. S. Kerr Lab throughout the
study is gratefully acknowledged. B. Rea performed the Pb2*
adsorption experiments with different grain-size fractions of the
composite sand. D. Kent (USGS-Menlo Park, California), D.
LeBlanc, C. Mclntosh, and R. Quadri (USGS, Marlborough,
Massachusetts) assisted in obtaining cores of subsurface material
from the field site. K. Bussey (USGS, Marlborough, Massachusetts)
assisted in the preparation of Figure 1.
QUALITY ASSURANCE STATEMENT
All research projects making conclusions or recommendations
based on environmentally related measurements and funded by
the Environmental Protection Agency are required to participate
in the Agency Quality Assurance Program. This project was
conducted under the approved Quality Assurance Program Plan.
The procedures specified in the plan were used without exception.
Information on the plan and documentation of the quality assurance
activities and results are available from the Principal Investigator.
19
-------
U.UI3
0.01
0.005
0
0.25
0.2
.a
Crt
•a
n
on
a 0.15
U
i o.i
'i
0.05
0
0.2
0.15
0.1
0.05
n
T" 1 1 -i 1 1
A
-
•
•
* «
~
Bfi
~
C
-
• •
-
468
Separation Distance, in meters
10
12
Figure 15. Horizontal variograms of A) In (lead adsorption), B) In (zinc adsorption), and C) In (K) (hydraulic conductivity).
20
-------
REFERENCES
ASTM (1986) Annual Book of ASTM Standards: Soil, Rock and
Building Stones. American Society of Testing Materials,
Philadelphia, Pennsylvania.
Barber, L. B. II (1990) Geochemical heterogeneity in glacial
outwash aquifer: Effect of Particle Size and Mineralogy on
Sorption of Nonionic Organic Solutes, Ph.D. Dissertation.
University of Colorado, Boulder, CO.
Barber, L. B. II, Thurman, E. M, and Runnells, D. D. (1992)
Geochemical heterogeneity in a sand and gravel aquifer: Effect of
sediment mineralogy and particle size on the sorption of
chlorobenzenes, Journal of Contaminant Hydrology 9. 35-54.
Barber, L B. II,Thurman, E. M, and Schroeder, M.P. (1988) Long-
term fate of organic micropollutants in sewage-contaminated
groundwater, Environ. Sci. Technol.. 22, 205-211.
Chao, T. T. (1984) Use of partial dissolution techniques in
geochemical exploration, Journal of Geochemical Exploration.
20, 101-135.
Chao, T. T. and Zhou, L. (1983) Extraction techniques for selective
dissolution of amorphous iron oxides from soils and sediments.
Soil. Sci. Soc. Am. J. 47, 225-232.
Chrysikopoulos, C. V., Kitanidis, P. K., and Roberts, P. (1990)
Analysis of one-dimensional solute transport through porous
media with spatially variable retardation factor. Water Resources
Research. 26, 437-446.
Coston, J. A., Fuller, C. C., and Davis, J. A. (1993) Importance of
aluminum- and iron-bearing surface coatings on Pb2* and Zn2+
adsorption by an aquifer sand. 1. Experimental studies. Geochimica
et Cosmochimica Acta. submitted.
Davis, J. A. and Kent, D. B. (1990) Surface complexation modeling
in aqueous geochemistry In: Mineral-Water Interface
Geochemistry. Hochella. M. F., and White A. F. (Eds) Mineralogical
Society of America, Washington, DC, 177-260.
Davis, J. A., Coston, J. A., and Fuller, C. C. (1993) Importance of
aluminum- and iron-bearing surface coatings on Pb2+ and Zn2*
adsorption by an aquifer sand. 2. Surface complexation modeling,
Geochimica et Cosmochimica Acta. submitted.
Davis, J. A., Kent, D. B., Rea, B. A., Garabedian, S. P., and
Anderson, L. D. (1991) Effect of the geochemical environment on
heavy-metal transport in ground water, In: U.S. Geological
Survey Toxic Substances Hydrology Program-Proceedings.
G. E. Mallard and D. A. Aronson (Eds.) WRIR 91-4034, 53-62.
Dzombak, D. A. and Morel, F. M. M. (1990) Surface Complexation
Modeling: Hydrous Ferric Oxide. John Wiley, New York.
Freeze, R. A, Massmann, J., Smith, L., Sperling, T., James, B.
(1990) Hydrogeological decision analysis; 1. A framework, Ground
Water 28, 739-776.
Garabedian, S. P. (1987) Large-Scale Dispersive Transport in
Aquifers: Field Experiments and Reactive Transport Theory,
Ph.D. Thesis, Dept. Civil Eng., Mass. Inst of Tech. Cambridge, MA.
Garabedian, S. P., LeBlanc, D. R., Gelhar, L W., and Celia, M. A.
(1991) Large-scale natural gradient tracer test in sand and
gravel, Cape Cod, Massachusetts. 2. Analysis of spatial moments
for a nonreactive tracer, Water Resources Research. 27(5), 911-
924.
Hazen, A. (1893) Some physical properties of sands and gravels,
Mass. State Board of Health. 24th Annual Report.
Hess, K. M., Wolf, S. H., and Celia, M. A. (1992) Large-scale
natural gradient tracer test in sand and gravel, Cape Cod,
Massachusetts. 3. Hydraulic conductivity variability and calculated
macro-dispersivities, Water Resources Research. 28(8), 2011-
2027.
Honeyman.B.D., and Leckie.J.O. (1986) Macroscopic partitioning
coefficients for metal ion adsorption: Proton stoichiometry at
variable pH and adsorption density, in Geochemical Processes at
Mineral Surfaces: J. A. Davis and K. F. Hayes, Eds. ACS
Symposium Ser. 323, American Chemical Society, Washington
DC, p. 162-190.
Journel, A. G. and Huijbregts, C. J. (1978) Mining Geostatistics.
Academic Press, San Diego, CA.
Kabala, Z. J., and Sposito, G. (1991) A stochastic model of
reactive solute transport with a time-varying velocity in a hetero-
geneous aquifer, Water Resources Research. 27, 341-350.
Kurbatov, M. H., Wood, G. B., and Kurbatov, J. D. (1951)
Isothermal adsorption of cobalt from dilute solutions, Journal
Phys.Chem. 55,1170-1182.
LeBlanc, D. R. (1984) Sewage Plume in a Sand and Gravel
Aquifer, Cape Cod, Massachusetts, U. S. Geological Survey
Water-Supply Paper 2218, 28 p.
LeBlanc, D. R., Garabedian, S. P., Hess, K. M., Gelhar, L. W.,
Quadri,R.D.,Stollenwerk,K.G.,andWood,W.W. (1991) Large-
scale natural gradient tracer test in sand and gravel, Cape Cod,
Massachusetts. 1. Experimental design and observed tracer
movement, Water Resources Research. 27, 895-910.
Rea, B. A., Kent, D. B., LeBlanc, D. R., and Davis, J. A. (1991)
Mobility of zinc in a sewage-contaminated aquifer, Cape Cod,
Massachusetts, In: U.S. Geological Survey Toxic Substances
Hydrology Program-Proceedings. G. E. Mallard and D. A.
Aronson (Eds.) WRIR 91-4034, 88-95.
Robin, J. J. L., Sudicky, E. A., Gillham, R. W., and Kachanoski, R.
G. (1991) Spatial variability of strontium distribution coefficients
and their correlation with hydraulic conductivity in the Canadian
Forces Base Borden Aquifer, Water Resources Research. 27,
2619-2632.
Robinson, G. D. (1984/85) Sequential chemical extractions and
metal partitioning in hydrous Mn-Fe-oxide coatings: reagent
choice and substrate composition affect results, Chemical Geology
47,97-112.
Sokal, R. R., and Rohlf, F. J. (1973) Introduction to Biostatistics.
W. H. Freeman and Co., San Francisco, CA. 368 p.
21
-------
Tessier, A., Campbell, P. G. C., and Bisson, M. (1979) Sequential
extraction procedure for the speciation of particulate trace metals,
Analytical Chemistry. 51, 844-851.
Tessier, A., Carignan, R., Dubreuil, B., and Rapin, F. (1989)
Partitioning of zinc between the water column and the oxic
sediments in lakes. Geochimica et Cosmochimica Acta. 53,
1511-1522.
Valocchi, A. J. (1989) Spatial moment analysis of the transport of
kinetically absorbing solutes through stratified aquifers, Water
Resources Research. 25, 273-279.
Wolf, S. H., Celia, M. A., and Hess, K. M. (1991) Evaluation of
hydraulic conductivities calculated from multiport-permeameter
measurements, Ground Water. 29, 516-525.
Wood, W. W., Kraemer, T. F., and Hearn, P. P., Jr. (1990)
Intragranular diffusion, an important mechanism influencing solute
transport in clastic aquifers? Science. 247,1569-1572.
Woodbury, A. D., and Sudicky, E. A. (1991) The geostatistical
characteristics of the Borden aquifer, Water Resources Research.
27, 533-546.
Zapico, M. M., Vales, S., and Cherry, J. A. (1987) A wireline piston
core barrel for sampling cohesionless sand and gravel below the
water table. Ground Water Monit. Rev. 74-82.
22 ttV.S. GOVERNMENT PRINTING OFFICE: N*3 - 750-07I/80M7
------- |