You are on page 1of 18

Geochemistry at the sulfate reductionmethanogenesis

transition zone in an anoxic aquiferA partial


equilibrium interpretation using 2D reactive transport modeling
Rasmus Jakobsen
a,
*
, Lise Cold
b
a
Institute of Environment and Resources, Technical University of Denmark, Bygningstorvet, Bygning 115, DK-2800 Kongens Lyngby, Denmark
b
Municipality of Herlev, Department of Environment, Herlev Bygade 90, 2730 Herlev, Denmark
Received 12 May 2006; accepted in revised form 15 January 2007; available online 20 January 2007
Abstract
The study addresses a 10 m deep phreatic postglacial sandy aquifer of vertically varying lithology and horizontally varying
inltration water chemistry, displaying calcite dissolution, ion-exchange, and anaerobic redox processes. The simple varia-
tions in lithology and inltration combine into a complex groundwater chemistry, showing ongoing Fe-oxide reduction,
sulfate reduction and methanogenesis. Rates of sulfate reduction, methanogenesis and methane oxidation were measured
directly using radiotracers. Maximum rates were 1.5 mM/yr for sulfate reduction, 0.3 mM/yr for methanogenesis, and only
4.5 lM/yr for methane oxidation. The overlap of sulfate reduction and methanogenesis was very small. The important inter-
mediates formed during the degradation of the organic matter in the sediment, formate and acetate, had concentrations
around 2 lM in the sulfate reducing zone, increasing to 10 and 25 lM in the methanogenic part. The concentration of H
2
was around 0.25 nM in the Fe-reducing zone, 0.4 nM in the sulfate reducing zone, and increased to 6 nM in the methanogenic
zone. Using in situ concentrations of products and reactants the available energies for a range of dierent reactions could be
calculated. The results of the calculations are in accordance with the observed distribution of the ongoing redox processes,
implying that the system is well described using a partial equilibrium approach. A 2D numerical PHAST model of the system
based on the partial equilibrium approach, extended by implementing specic energy yields for the microbial redox processes,
could explain most of the observed groundwater geochemistry as an expression of a closely coupled system of mineral equi-
libria and redox processes occurring at partial equilibrium.
2007 Elsevier Ltd. All rights reserved.
1. INTRODUCTION
Groundwater is an important source of drinking water in
many places of the world. However due to pollution from
industry and agriculture, near-surface aquifers are in many
cases abandoned, and water is obtained from deeper often
anoxic aquifers. Anoxic redox processes are important in
determining natural water quality in these deep-seated aqui-
fers (Park et al., 2006), however, detailed knowledge of these
systems is scarceto some degree due to the large depth
which limits the available data. Several studies have focused
on the microbiology of deep-seated aquifers (e.g. Pedersen,
2000; Amend and Teske, 2005), but the main focus here is
on the geochemistry of Fe-oxide reduction, sulfate reduction
and methanogenesis mediated by microorganisms in the
aquifer and the relations between the processes and the sedi-
ment. Earlier studies of the Danish anoxic Rm aquifer
(Jakobsen and Postma, 1999; Hansen et al., 2001; Larsen
et al., 2006) have shown it to be characterized by the same
geochemical processes found in other anaerobic systems,
e.g. marine, lacustrine and soil sediments, but due to the
importance of the advective ow in aquifers these processes
present themselves dierently. The Rm study indicated a
coupling of inorganic geochemical processes with the
microbiologically mediated redox processes through the
0016-7037/$ - see front matter 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.gca.2007.01.013
*
Corresponding author.
E-mail addresses: raj@er.dtu.dk (R. Jakobsen),
lisecold@ or.dk (L. Cold).
www.elsevier.com/locate/gca
Geochimica et Cosmochimica Acta 71 (2007) 19491966
thermodynamics of these (Jakobsen and Postma, 1994, 1999;
Hansen et al., 2001) and lead to the partial equilibrium ap-
proach to describing redox processes (Postma and Jakobsen,
1996). The study presented here, based on a detailed data set
on water and sediment chemistry from the Asserbo aquifer
on Northern Zealand, Denmark, describes results from a
similar system, in terms of age and reactivity of the organic
matter. However, the Asserbo aquifer consists of shallow
marine sands rather than the dune sands on Rm, has a
higher ow rate, a dierent inltration composition, and is
lithologically less homogeneous. This enables us to further
address howaquifer owand sediment characteristics as well
as microbial processes control the groundwater chemistry of
an anaerobic groundwater system, increasing the potential
for transferring the insights fromthese detailed studies to less
accessible aquifers. The main features of the Rm aquifer
could be modeled using a 1D PHREEQC model (Jakobsen
and Postma, 1999) applying the partial equilibrium
approach. This approach is currently only one of several
approaches used in modeling of redox processes in the sub-
surface as there is also the kinetic approach used by Hunter
et al. (1998), and various combinations (Jin and Bethke,
2005; Brun and Engesgaard, 2002). The latter as well as
Curtis (2003) provide reviews and comparisons of some of
these approaches. More elaborate schemes involving a closer
modeling of the microbial dynamics are described by
Thullner et al. (2005) and Watson et al. (2005). The Asserbo
aquifer presented here appears to be in a quasi-steady-state,
implying that the system will change as the reactivity of the
minerals and the organic matter in the systemchange, but be-
cause reaction rates are low, changes will be slow compared
to the owrate of the system. Furthermore, with no available
information on, e.g. seasonal dynamics, modeling the system
using the simple partial equilibriumapproachused by Jakob-
sen and Postma (1999) seems appropriate. Due to the
non-uniform chemical composition of the inltration at the
Asserbo site, the system is modeled in 2D using PHAST
(Parkhurst et al., 2005). Furthermore, in order to approach
a modeling of the levels of the intermediates, H
2
, acetate
and formate, formed during the decomposition of organic
matter in the system, the simple partial equilibriumapproach
used by Jakobsen and Postma (1999) has been modied to in-
clude the non-zero energy yield that the microbes mediating
the redox processes need. The use of the numerical model
makes it possible to quantitatively address the interactions
between inorganic geochemistry and the microbiologically
mediated redox processes occurring in this type of system
that develops into a very closely coupled system of several
mineral equilibria and terminal electron accepting processes
controlling the groundwater chemistry. This quantitative insight
should be useful when interpreting other anaerobic aquifer
systems, where the dataset is normally much less comprehensive.
2. METHODS AND SETTING
2.1. Groundwater sampling and analysis
In general sampling was carried out as described in Jakobsen
and Postma (1999) and Hansen et al. (2001), the following is a
short summary of this. Groundwater samples were taken using
stainless steel drive point piezometers. Steel pipe (1
//
) was driven
into the aquifer using a pneumatic hammer. The 6 cm long 50 lm
stainless steel mesh screen lter tip was equipped with a check valve
to allow sampling by nitrogen gas displacement, taking care to
sample the middle part of the water in a pump cycle minimizing
eects of degassing. Three sites spaced 10 m apart on a line
intended to coincide with the ow direction were sampled.
For hydrogen sampling, a bundle of 10 mm diameter PVC
tubes equipped with a 20 mm disc-shaped 20 lm nylon screen were
installed in a hand-drilled well using plastic casing and shoe
(Jakobsen and Postma, 1999). To obtain meaningful results, metal
parts were avoided and the well was left to rest for two months
before hydrogen sampling commenced (Jakobsen and Postma,
1999). Hydrogen was measured in the eld using the bubble
stripping method of Chapelle and McMahon (1991), modied,
using lower ow, to account for the small intake area as described
in Jakobsen and Postma (1999) and Hansen et al. (2001).
Methane water samples were collected in a syringe, avoiding
any air contact, and ~3 ml were injected in a pre-weighed 13 ml
evacuated blood sample vial, and then frozen upside down at
18 C. For all other components, water samples were ltered
anaerobically through 0.2 lm lters in the eld. Samples for inor-
ganic anions, acetate and formate were collected in 5 mL poly-
propylene vials and always frozen immediately to below 18 C in
the eld. Also in the eld, pH, O
2
and electrical conductivity were
measured in a ow cell, alkalinity determined by Gran titration and
Fe
2+
(Stookey, 1970) and H
2
S (Cline, 1967) were determined
spectrophotometrically. The pH measurements from the two last
sampling positions appear awed, presumably due to a malfunc-
tioning pH meter. The data from the rst data set show that calcite
equilibrium is obtained at depth. Therefore the pH in the last two
data sets have been adjusted by shifting the data by 0.7 U, in both
cases resulting in an asymptotic approach to calcite equilibrium
matching the development in the rst data set rather closely, sup-
porting this x.
In the laboratory, cations were measured by AAS, the con-
centrations of anions determined by ion chromatography, on
subsamples amended with 4& 0.05 M NaEDTA to avoid precipi-
tation of oxidized Fe, and methane by gas chromatography with
FID detection on the headspace of the vial after thawing, and the
aqueous concentration was calculated from the volume of water
and headspace in the vial. The concentrations of acetate and for-
mate were determined using ion exclusion chromatography with a
Dionex AS-10 column and suppressed EC detection. Samples for
acetate and formate were thawed less than 12 h before analysis and
both standards and samples were amended to hold 0.2% chloro-
form to minimize both loss by microbial oxidation in standards and
an increase in concentrations in samples, presumably due to
microbial decomposition of DOC (dissolved organic carbon). The
detection limit for acetate was 0.2 lM but often unidentied
interfering peaks were observed, leading to a higher eective
detection limit. For other analytes, detection limits were always,
apart from sulde, well below measured concentrations.
2.2. Radiotracer rate measurements
Sediment cores for sediment analyses and radiotracer rate
measurements were taken in 50 mm ID, 1.5 mm thick, stainless
steel tubing, with a barrel free corer (Starr and Ingleton, 1992).
Core depths were adjusted for compaction (if necessary) by
comparing concentrations in core pore water with corresponding
concentrations in well water samples. Displacements were generally
less than 0.4 m. A separate core was taken for each of the radio-
tracers used. After retrieval, the cores were cut in 50 cm sections
and the ends immediately sealed with plastic stoppers and wrapped
1950 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
with aluminum-foil-tape. The core sections were then lowered in a
well for 13 h to bring them back to in situ temperature.
After retrieval of the core sections fromthe well, 1 mmholes were
drilledinthe tubingwall and12.525.0 lLof radiotracer was injected
along a line across the core at intervals of 1012 cm. The holes were
resealed with aluminum-foil-tape. Then the core was relowered into
the well for incubation. For CO
2
reduction100150 kBq (12.5 lL) of
H
14
CO
3

was injected, which changed the in situ total inorganic


carbon concentration by 1%. For acetate 1.3 kBq of
14
CH
3
COO-
Na was injected, which changed the in situ acetate concentration by
<1%. For sulfate reduction approximately 75 kBq of dilute H
2
35
SO
4
was injected, again changing the in situ sulfate concentration by
1%and not aecting the pHsignicantly. The incubation time for
acetate was 14 h, for CO
2
reduction 22 h and sulfate reduction 18 h.
The incubations were ended by freezing the cores to 50 C in
dry-ice, they were stored at 20 C until analysis.
The rate of sulfate reduction was determined using the method
of Howarth and Merkel (1984) modied for use in aquifer material
as described in Jakobsen and Postma (1994, 1999) separating re-
duced sulfur in AVS (acid volatile sulfur) and CRS (chromium
reduced sulfur). The rate was calculated separately for AVS and
CRS from:
SRR =
(SO
4
2
) a
Red-S
a
t a
Sulfate
(1)
where SRR is the rate of sulfate reduction, (SO
4
2
) the sulfate con-
centration, t the incubation time, a a fractionation factor (=1.06)
taking into account the dierence in the SRR of
35
S and natural
sulfur, a
Red-S
and a
Sulfate
are the measured radioactivities in the
reduced (AVS or CRS) and unreduced fractions after the
incubation.
The rate of CO
2
reduction and acetate fermentation was mea-
sured by a modied version of the method of Crill and Martens
(1986) as described in Hansen et al. (2001). The rate of CO
2
reduction was calculated from:
CRR =
(TIC) a
CH
4
a
t a
TIC
(2)
where CRR is the rate of CO
2
reduction, TIC the concentration of
total inorganic carbon calculated from alkalinity and pH, a
CH
4
and
a
TIC
the radioactivities of CH
4
and TIC after incubation, t the incu-
bation time and a the fractionation factor, here amounting to 1.08
(Hansen, 1998).
In contrast to TIC, a signicant amount (2550%) of the acetate
tracer is consumed during the incubation. The concentration of
14
CH
3
COO is accordingly not a constant but is calculated from a
rate expression assuming a rst order decrease. The acetate turn-
over rate (ATR) is then calculated from the relation:
ATR =
(CH
3
COO

) a
t
ln
a
CH
3
COO
a
TIC
a
CH
4
a
CH3COO

(3)
where a is the fractionation factor of 1.08, (CH
3
COO

) the acetate
concentration, t incubation time and a
CH
3
COO
, a
CH4
and a
TIC
the
radioactivity in the subscripted fractions. The acetate fermentation
rate (AFR) and the acetate oxidation rate (AOR) are then calculat-
ed as:
AFR = ATR
a
CH4
a
CH4
a
TIC
(4)
and:
AOR = ATR
a
TIC
a
CH4
a
TIC
(5)
The concentrations of unlabeled TIC and acetate needed for the
rate calculations were interpolated from groundwater concentra-
tion proles at the same site. Methane oxidation rates were calcu-
lated from the expression:
MOR =
(CH
4
) a
TIC
a
t a
CH4
(6)
where (CH
4
) represents the methane concentration measured in the
sediment sample after equilibration in a headspace, a
TIC
and a
CH
4
the activities of the oxidized and unoxidized tracer injected, t the
time of incubation and a the fractionation factor, which for meth-
ane oxidation was set to 1 derived from the very little reported frac-
tionation (Whiticar et al., 1986). The measured rates are extremely
low, perhaps approaching the limit of detection. It was not feasible
to determine this limit statistically, but given the data distribution
(Fig. 6) we estimate it to be around 1.3 lM/yr.
2.3. Sediment parameters
For parts of the cored sections the sediment content of sulde,
Fe, organic and inorganic carbon was determined as described in
Jakobsen and Postma (1999). Sulde as AVS (acid volatile) and
CRS (chromium reducible) was determined from the concentra-
tion of sulde in the traps used for the rate measurements, Fe was
simply measured as total Fe released by ~6 M HCl during the 1 h
sulde extraction. The Jakobsen and Postma (1999) study indi-
cated that little extra information on the Fe-oxide content was
obtained from an anaerobic oxalate extraction, but it did indicate
the presence of siderite. A simple separate measurement of FeII
on the anaerobic ~6 M HCl extract could perhaps have given an
indication of the siderite content. Sediment bound organic carbon
was measured as two fractions: Non acid desorbable sedimentary
organic carbon (NADSOC) measured on a LECO CS-225 after
removing inorganic carbonate using 2 M HNO
3
releasing ADSOC
which was determined on a Dohrman DC-180. Inorganic carbon
was determined by subtracting the sum of NADSOC and
ADSOC from a TC (total carbon) measurement using the LECO
CS-225.
2.4. Site description
The site (Fig. 1) is situated on the north coast of northern
Zealand, Denmark. The aquifer consists of approximately 10 m of
sand underlain by a clayey till. The lower 78 m of the sand are
shallow marine sandy deposits with occasional layers of gravely
sand containing pebbles. The upper 23 m are homogeneous eolian
sand deposits with occasional paleosols formed on the dunes.
Porosities are 2530%. Hydraulic conductivities determined from
sieve analyses on samples from the cores taken at site 2 are around
1.3 10
4
m/s with an interval of lower hydraulic conductivity,
5 10
6
m/s, 45 mbs. Contours of the groundwater table (Fig. 1),
were derived from piezometers installed by hand auger ~50 cm
below the groundwater table. The groundwater table is about
1.2 mbs (meters below surface) giving an aquifer depth of
approximately 9 meters. The groundwater divide is (based on
topography) approximately 600 m upstream from the studied site.
The non-parallel contours indicate heterogeneities in the aquifer.
During the installation of the piezometers an apparently continu-
ous and rather large paleosol was encountered (Fig. 1), but more
local paleosols were seen in single piezometer wells. Paleosols may
exist at various depths in the eolian sediment package, implying
that the measured groundwater head may relate to dierent more
or less interconnected subunits. Annual inltration is estimated at
150 mm/yr, from a regional average. The vertical 2D model (see
section 4.5.1) estimates a horizontal ow velocity of about 40 m/yr
at the site and vertical velocities in the upper studied part of the
aquifer of approximately 0.6 m/yr. The model used for modeling
ow and transport was PHAST (Parkhurst et al., 2005), a reactive
transport model based on HST3D (Kipp, 1997) and PHREEQC
(Parkhurst and Appelo, 1999).
Geochemistry at the sulfate reductionmethanogenesis transition zone 1951
The piezometric heads, and sampling positions are shown in
Fig. 1 together with the ow vectors estimated by the PHAST
model for the 2D cross-section.
3. RESULTS
Three sites 10 m apart (Fig. 1) along a ow line were
sampled for water chemistry, and at the central site cores
were taken and geochemical sediment parameters and
direct radiotracer rate measurements were made.
3.1. Groundwater chemistry
The general groundwater chemistry at the three sites, for
the major ions that are not, or only indirectly, aected by
redox processes, is shown in Fig. 2 together with the pH
and the total alkalinity. The Na and Cl proles show a
distinct maximum around a depth of 57 mbs. At least two
explanations are possible. It could be a result of a higher
level of dry deposition upstream due to, e.g., higher trees
or a release of dissolved ions from high salinity water from
earlier inundation events retained in a layer of low perme-
ability. The similar peak in sulfate (Fig. 3) suggest accumu-
lation of sulfate as dry deposition or that sulfate from pyrite
oxidation in the proposed low permeable layer is released
with the high salinity water. Analysis of water sampled
from the piezometer wells indicate the occasional presence
of high salinity water in the upper groundwater. In the
paleosol indicated in Fig. 1, Cl

reached 3.8 mM, almost


twice the highest concentration in the proles, and sulfate
Fig. 1. (a) Site with sampling positions piezometers, heads, near-surface geology and an indication of (b) the 2D vertical PHAST model cross-
section with the calculated ow vectors shown as a base point with a line indicating velocity and direction (right to left). Note the dierent
scales on the map and in the cross-section.
1952 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
reached 0.74 mM, similar to what is seen in the proles.
Also in the area upstream of the sampling site, high concen-
trations were found in the piezometer, though not as high
as those found in the groundwater. This implies that the
high salinity groundwater, is either found deeper, which
would imply an inundation event, or in a location not sam-
pled. The owpattern indicated by the contours of the
groundwater head makes it dicult to pinpoint the area
by backtracking a ow line, but according to the ow
model the area would be outside the area covered by the
piezometers. Though there could be other explanations
for the anomaly in the water chemistry the modeling de-
scribed later assumes the proposed inundation event. The
increase in calcium at 3.5 mbs presumably corresponds to
the level from where inorganic carbonate, mainly as shell
fragments, is present in the sediment. The dissolution of
these causes an increase in alkalinity along with Ca and
Mg, present in the calcite shells. This does not cause the
pH to increase, because the pH already increased from
around 5 in the top samples to 7.2 at 4 mbs, presumably
related to the reduction of Fe-oxides described below.
The patterns of variation seen in the cations, most distinct
for Mg, indicate that the distribution of these are aected
by ion-exchange processes, presumably induced by the
introduction of the high-salinity water.
Data for components directly aected by redox process-
es are shown in Fig. 3. The Fe
2+
concentration shows that
the aquifer becomes anaerobic just below the water table.
The proles show most of the classical anaerobic redox
sequence; an increase in Fe
2+
due to Fe-oxide reduction,
followed by depletion of sulfate due to sulfate reduction
and an increase in methane from methanogenesis when
the sulfate is depleted. The upper part of the peak in the sul-
fate prole coincides with the Cl prole. The decrease in
sulfate from 5.5 to 6.5 mbs is accompanied by an increase
in sulde from 5.5 to 7 mbs. Sulde disappears or decreases
at depth presumably being xed in an Fe-sulde phase. The
distinct increase in dissolved manganese which should oc-
cur rst in a classical redox sequence appears below the
zone of increasing Fe
2+
concentrations indicating either
reduction of very stable Mn-oxides or perhaps more prob-
able, when noting the coincidence with the pattern observed
for Ca, a release of Mn contained in some of the carbonate
going into dissolution between ~3.5 and 6 mbs.
Concentrations of the intermediates acetate, formate
and H
2
are shown in Fig. 4. Concentrations of formate
are close to constant at site 2 and 3, while at site 1 values
show more scatter and an increasing tendency as the system
becomes increasingly reduced with depth. A similar though
less erratic pattern is seen for acetate. The low concentra-
tions, compared to the measured rates (Fig. 6), imply that
these intermediates have a low residence time and are pro-
duced close to the sampled site. The source is presumably
organic macromolecules present in solution and carbon
compounds bound to the sediment. The H
2
prole from site
2 shows low values around 0.5 nM over most of the prole
with a slight increase over depth to 6 mbs, and below this,
the level increases sharply to 7 nM at 6.5 mbs.
Fig. 5. shows the DOC, the d(TIC-Ca) (the increase in
TIC from one sample depth to the next corrected by the
Fig. 2. Concentrations of major components in the Asserbo
aquifer as symbols and the model results as lines. Note the ow
direction is from left to right.
Geochemistry at the sulfate reductionmethanogenesis transition zone 1953
increase in Ca removing carbonate released from calcite
dissolution) and the ammonium concentration. Only a very
minor part of DOC consists of acetate and formate (Fig. 4)
and though the porewaters presumably also contain other
small organic molecules such as lactate, propionate and
butyrate, these are probably not quantitatively important,
so that the DOC is dominated by macromolecules. DOC
values are high in the upper part and the decrease down
to 5.5 mbs correlates with the increase in Ca from the cal-
cite dissolution. This indicates that the higher Ca
2+
concen-
trations aects the mobility of the DOC, perhaps by
changing the adsorption characteristics of the organic mol-
ecules. The generally positive values of d(TIC-Ca) indicate
oxidation of organic matter over most of the prole, the
few negative values are presumably eects of ion
exchange aecting the Ca concentrations. The steady in-
crease in ammonium over depth also indicates the oxidation
of organic matter as it presumably reects the release of
ammonium to the porewater as organic matter is degraded
as described from marine sediments by Caneld et al.
(1993).
3.2. Measured rates
Rates of sulfate reduction measured by injecting
35
S,
rates of methanogenesis from the reduction of injected
14
CH
3
COOH labeled acetate, and H
14
CO
3

as well as the
rates of oxidation from the labeled
14
CH
3
COOH, and rates
of methane oxidation measured using
14
CH
4
are plotted in
Fig. 6. Oxidation of organic matter appears to be focused in
the 56 mbs interval with sulfate reduction as the most
dominant redox process. The zone of sulfate reduction
shows a slight overlap with the zone of methanogenesis.
Though rates of methanogenesis are not as high, they are
Fig. 4. Measured concentration of intermediates used in terminal
electron acceptor processes with model results shown as lines. Note
the ow direction is from left to right.
Fig. 3. Concentrations of redox-reactive components in the
Asserbo aquifer as symbols and the model results as lines. Note
the ow direction is from left to right.
1954 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
within the same range. Both CO
2
reduction, acetate fermen-
tation and oxidation show a distinct peak just around
7 mbs indicating that rates are related to the reactivity of
the organic matter in the sediment. The acetate oxidation
rate is very similar to the CO
2
reduction rate, while the ace-
tate fermentation rate is considerably smaller. Methane oxi-
dation rates are even lower, showing that though the
process occurs the extent is very limited as only about 1%
of the methane appears to be reoxidized.
3.3. Sediment geochemistry
Sediment parameters are shown in Fig. 7. The Fe
extractable by 6 N HCl acid is 1015 mmol/kg in the upper
interval 57 mbs, and then increases to 25 mmol/kg at
7.5 mbs, part of this could be siderite. Suldes are present
in the entire 57.5 mbs interval, but below 6 mbs only
CRS (mostly FeS
2
) are found while above about one third
of the suldes are found as AVS (mostly FeS), showing that
the conversion of FeS to FeS
2
runs to completion in this
system. For sedimentary organic matter the amounts of
ADSOC and NADSOC are about the same, except for
the peaks around 5.7 and 7 mbs where the extra organic
matter appears to be NADSOC. In the examined interval
the sediment inorganic carbon content (SIC) is relatively
constant around 250 mmol C/kg, with a single outlier at
7 mbs with only 60 mmol/kg, possibly related to the high
rate of methanogenesis observed close to this level
(Fig. 6), producing extra CO
2
, causing or having caused
extra carbonate dissolution.
4. DISCUSSION
4.1. Mineral equilibrium control
Fig. 8 shows activity plots for the most important min-
eral equilibria in the system. For calcite, the plot shows that
a very large part of the data plot very close to the line
representing calcite equilibrium. These data plot in a dense
cluster because the pH, the alkalinity and the Ca concentra-
tion all become close to constant at depth. The points below
the line are from the upper part of the aquifer from where
calcite has either been leached out and is present in such
small amounts that equilibrium cannot be obtained within
the time available for the groundwater owing through.
That calcite equilibrium is obtained at depth ts well with
the presence of SIC in the cored interval (Fig. 7). For sider-
ite many data points plot along a line representing an
SI
siderite
= 0.75. Equilibrium control at supersaturations of
this magnitude, reecting a kinetic inhibition of the ongoing
precipitation that is compensated by the supersaturation,
have been described for other systems, e.g. Postma (1981),
Jakobsen and Postma (1999) and Jensen et al. (2002). Also
the nearly constant Ca
2+
/Fe
2+
activity ratio seen at depth
indicates siderite equilibrium control. Again the points that
are o the line are from the upper part where carbonate
activities are low. For FeS, data points are few due to
the detection limit for sulde, but the plot, where the
log(IAP
FeS
) falls into a rather narrow range, indicates that
FeS precipitation is controlling the observed concentration
of sulde. This implies that there may be intervals where
Fe
2+
is controlled by equilibrium with both Fe-sulde
and Fe-carbonate while at the same time carbonate is
controlled by equilibrium with both Ca-carbonate and
Fe-carbonate, leading to an interlocked system where, e.g.
the production of CO
2
from methanogenesis may aect
the observed sulde concentration.
4.2. Exchange
Comparing the Na and Cl proles (Fig. 2) indicates that
Na is slightly delayed in terms of the vertical transport,
indicating ion exchange processes aect the cations. This
also appears to aect the Ca, Mg, and K, resulting in what
appears to be a chromatographic sequence where Mg is dis-
placing Ca which again is displacing K from the exchanger.
The Ca is also aected by the dissolution of calcite, and the
calcite equilibrium in the lower part implies that if ion
exchange releases Ca it may lead to precipitation of calcite
potentially lowering the pH. The ammonium (Fig. 5) does
not seem to be aected by the ion exchange processes, but
Fig. 5. (Left) The d(TIC-Ca), the change from one observation
point to the one below, derived from observations, representing the
increase in TIC not related to calcite dissolution. (Center) The
concentration of ammonium, released from oxidizing sedimentary
organic matter. Derived or measured values are plotted as symbols,
the model results are shown as lines. (Right) The measured total
dissolved organic carbon. Note the ow direction in this plot is
from top to bottom.
Geochemistry at the sulfate reductionmethanogenesis transition zone 1955
is probably, to a higher degree, controlled by the release
from organic matter being oxidized, implying that it may
be a cumulative indicator for the oxidation of organic mat-
ter in the system. As mentioned dTIC-Ca (Fig. 5) is slightly
negative in three intervals also indicating that ion exchange
processes are aecting the Ca concentration.
4.3. Rates and sediment composition
There is a good coherence between where measured sul-
fate reduction rates are high and where the sulde minerals
are found in the sediment. The AVS minerals are only
found in the interval from 5 to 6 mbs where the rates,
and especially the AVS related rates are high showing that
the kinetics of the transformation from AVS to CRS are
more sluggish than the sulfate reduction. In the lower part
of the prole the suldes are almost exclusively found as
CRS. The rates of sulfate reduction are extremely small
here which must imply that sulfate input has been higher
so that sulfate reduction took place in a larger volume of
the sediment. This indicates that the system is in fact only
in a quasi-steady-state. The total amount of sulde accumu-
lated in the 56 mbs interval is around 7 mmol/kg of sedi-
ment, which using a wet bulk density of 2 and a porosity
Fig. 6. Directly measured radiotracer rates of sulfate reduction, methanogenesis from acetate oxidation and CO
2
reduction, and methane
oxidation. The AVS part of the sulfate reduction rate is not shown directly, but is the dierence between the total rate and the CRS rate. The
rates for dierent redox processes derived from the total organic carbon addition entered in the model, assuming at most two concommitant
processes to enable the calculation, are shown as lines as indicated.
Fig. 7. Sediment content of Fe extractable by 6 N HCl acid, suldes CRS (chromium reducable suldemostly pyrite), AVS (acid volatile
suldesmostly monosuldes) and organic C as TSOC (total sedimentary organic carbon) and ADSOC (acid desorbable organic carbon).
NADSOC is the dierence between the two. The plot also shows SIC (sedimentary inorganic carbon).
1956 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
of 0.33 amounts to 42 mmol per liter of porewater. Given
the rate of around 1 mmol/L/year observed in this interval
implies that the suldes there could have formed in just 42
years. This indicates that the zone where high rates of sul-
fate reduction occurs has moved upwards within the sedi-
ment. What controls this movement is not known, but it
could be that the most bioavailable Fe-oxides have been
used allowing sulfate reduction to occur at a higher level
in the aquifer. It is probably more complicated than that,
because the amount of organic matter in the interval is also
rather low, around 30 mmol/kg or around 180 mmol/L of
porewater, implying that if the high rate was maintained
at this level the organic matter would have disappeared in
90 years. With an age of the sediment around 1000 years,
it seems there is too little organic matter and too little sul-
de to match the measured rates. This implies that zones of
high rates are temporary. It also indicates that the decrease
in DOC seen over depth reects transport of organic matter
from the surface into the aquifer. Considering the low
amount of Fe-oxide in the system compared to the signi-
cant release of FeII to solution 34 mbs (Fig. 3) could indi-
cate that also Fe-oxides are supplied from the soil above
presumably as colloids. Another apparently odd thing relat-
ed to the sulfate reduction is that the measured sulde con-
centration is highest, below the level where the rate of
sulfate reduction peaks. However since the sulde concen-
trations according to Fig. 8 appears to be controlled by
an FeS phase the high measured sulde concentrations
are a function of the low Fe
2+
concentrations rather than
the sulde production. This must imply that FeS can dis-
solve, perhaps explaining the low amount accumulated,
compared to the rate of sulfate reduction.
Acetate oxidation rates are relatively high in the lower
part of the prole where there is no sulfate reduction, indi-
cating another electron acceptor process of which the most
probable would be Fe-oxide reduction, which is at least not
contradicted by the increase over depth in the amount of
6 M HCl extractable Fe (Fig. 7). The acetate oxidation rate
is approximately the same as the rate of methanogenesis by
CO
2
reduction. This could indicate that the acetate is
oxidized to CO
2
and H
2
but as discussed in the section
below this is probably not the case.
The high measured rates of sulfate reduction coincide
with the increase in Ca from 56 mbs. This was also
seen in the similar Rm system (Jakobsen and Postma,
1999) indicating a relation. Several explanations are pos-
sible; the dissolving calcite itself contains reactive organ-
ic matter which is released during dissolution, the
increase in Ca causes changes in the availability of the
organic carbon, or the increased ionic strength as such
changes it.
4.4. Bioenergetics
The Gibbs energy for a range of probable microbiolog-
ically mediated processes have been calculated using the
thermodynamic data shown in Table 1. The values for
the central sampling site where H
2
was measured are plot-
ted in Fig. 9. The values for sulfate reduction are some of
the rst values reported from a groundwater system based
on actual measured values for sulde, which was always
below the detection limit in the Rm aquifer (Jakobsen
and Postma, 1999; Hansen et al., 2001). The values are
close to constant around 4.5 kJ/mol e

from 4.56 mbs


and decrease to 6 kJ/mol e

at 6.8 mbs, and very similar


to the values reported from larger Middendorf aquifer by
Park et al. (2006). The calculated Gibbs energy is, howev-
er, around twice as low as the 2.4 reported by Hoehler
et al. (2001), which was argued to be adequate for ATP
synthesis. The lower Gibbs energies, and the decrease seen
in the lower part of the prole, could be an indication of
a higher energy requirement when sulfate concentrations
are low. The value for methanogenesis by CO
2
reduction
by H
2
is around 3 kJ/mol of e

, in the interval where


methanogenesis was detected by the direct rate measure-
ments, similar to the values found by Park et al. (2006),
but again about twice as low as the value of 1.3 kJ/
mol of e

from Hoehler et al. (2001). Again it could be


related to a dierence in, e.g. the P
CO
2
, but it could also
be related to a dierence in the measurement method,
since the H
2
data from Hoehler et al. (2001) and the even
higher values of 0.9 from Schulz and Conrad (1997) are
from incubations of the sediment. In an overall sense the
observed values are similar to the values found for the
Rm setting (Hansen et al., 2001). Here in the Asserbo
system, 2.56 mbs, there is a large interval where the ener-
gy available for CO
2
reduction is very close to zero, which
together with the low, but measurable, concentration of
Fig. 8. Mineral equilibria. Log plots of the activities of the ions in calcite, siderite and FeS, indicating equilibrium control for calcite and FeS
in a large part of the system. The SI values (saturation indexes), given in the siderite and FeS plots are the values imposed in the model.
Geochemistry at the sulfate reductionmethanogenesis transition zone 1957
methane, 15 lM in this interval, points to the presence of
micro-niches in which methane is produced. Darling and
Goody (2006) also found methane in low concentrations
in a range of not highly reduced aquifers, indicating that
this is a general feature. The methane leaving the micro-
niches will only be oxidized as long as there is energy
available for the microorganisms. This is not the case
when Gibbs energies in the sampled water are close to
zero as neither methane production nor oxidation can oc-
cur. However, a principal dierence is that at this site
methanogenesis from CO
2
reduction by H
2
does become
thermodynamically feasible below 6 mbs, where it is also
observed in the directly measured rates (Fig. 6), so that
stagnant zones or micro-niches with higher H
2
content
need not be implied here. It is also worth noting that sul-
fate reduction via acetate does not seem to be feasible
with a Gibbs energy above 1 kJ/mol e

, and with only


15 lM of methane above 6 mbs, where it would be ther-
modynamically feasible to produce it by, e.g. acetate fer-
mentation, it appears that H
2
as substrate is extremely
important in this system. This is also indicated by the
Gibbs energy of H
2
production from acetate (Fig. 9)
which is feasible above 6 mbs where the H
2
concentration
is low (Fig. 5). Still acetate could be important as a sub-
strate for Fe-oxide reduction, but due to the diculties in
assigning a value for the Gibbs energy of formation for
the Fe-oxide being reduced no attempt has been made
in calculating the free energy for Fe-reduction reactions.
However, the fact that Fe
2+
is present over the entire
depth interval (Fig. 3) indicates that Fe-oxide reduction
is not limited to the upper part. This is supported by
the acetate oxidation rates measured below 6 mbs of
0.050.3 mM/yr (Fig. 6), where the transformation of ace-
tate to H
2
and CO
2
is not feasible according to the Gibbs
energies in Fig. 9, due to the high H
2
concentration.
Fig. 9. The Gibbs energy for a range of possible microbial
processes in the aquifer system calculated from the measured
concentrations of reactants and products using the thermodynamic
data in Table 1.
Table 1
Thermodynamic data used for calculations of in situ energy yields, and the energy yields at the temperature T = 281.15 (8 C) used in the
PHAST simulation obtained by adjusting the K-value for the reaction (see Table 2)
Reaction
a
DG
o
ra
(kJ/mol)
a
DH
o
r
(kJ/mol)
b
DG
oT
r
(kJ/mol)
c
DG
osh
r
(kJ/mol)
d
DG
oTsh
r
(kJ/mol)
e
Energy yield
(kJ/mol)
0
f
2H
+
+ 2e

MH
2
17.6 4.18 16.3 17.6 16.3 0
1 Fe
3+
+ 0.5H
2
MFe
2+
+ H
+
83.1 38.5 80.5 70.6 68.7 11.8
2 4H
2
SO
4
2
H

HS

4H
2
O 262.4 235.0 260.8 220.7 221.5 4.9
3
g
4H
2
HCO
3

CH
4
3H
2
O 229.4 237.9 229.8 206.8 208.6 2.7
4 CH
3
COOH4H
2
O 4H
2
2HCO
3

2H

241.8 228.4 241.0 234.0 233.7 0.91


5 HCOOHH
2
O H
2
HCO
3

40.2 15.1 38.8 46.8 44.9 3.1


6
h
CH
3
COOHSO
4
2
HS

2HCO
3

20.6 6.6 19.8 13.3 12.2 5.81


7
i
CH
3
COOHH
2
O CH
4
HCO
3

12.4 9.5 11.2 27.2 25.1 3.61


8
h
4HCOOHSO
4
2
HS

4HCO
3

3H

101.6 174.6 105.6 33.5 41.9 8.0


9
i
4HCOOHH
2
O CH
4
3HCO
3

3H

68.6 177.5 74.6 19.6 29 5.8


a
All reactions calculated from values in Stumm and Morgan (1996), except the value for acetic acid taken from Atkins and de Paula (2002).
b
Gibbs energy of reaction at 8 C and standard conditions.
c
The shifted value of the Gibbs energy of reaction entered into the model converted to a logK value by dividing by 5.708.
d
The temperature corrected shifted value of the Gibbs energy of reaction calculated using DH
o
r
.
e
The energy that has to be available from the reaction for the microorganisms at the model temperature of 8 C.
f
This reactioncanbe seenas the reactiontransferringexcess electronactivity introducedby adding the organic carbontodissolvedaqueous H
2
.
g
Methane oxidation would be the same with the two sides of the equation swapped.
h
Reactions 6 and 8 and the thermodynamic values are derived from the sum of reactions 2 + 4 and 2 + 4*(5).
i
Reactions 7 and 9 and the thermodynamic values are derived from the sum of reactions 3 + 4 and 3 + 4*(5).
1958 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
4.5. Modeling
4.5.1. An extended partial equilibrium approach
The approach taken is a two-step PEA model using the
typology of Brun and Engesgaard (2002). It is an extension
of the partial equilibrium approach used by Jakobsen and
Postma (1999). In order to better represent the actual sys-
tem, and be able to model the observed H
2
values the ener-
gy needed for the microorganisms to carry out the
microbiologically mediated processes needs to be included
in the thermodynamic description of the system. This corre-
sponds closely to assigning an observed saturation index,
SI, to a mineral in a purely geochemical model. Just as
the SI of a mineral showing a constant value above 0.0
may reect that extra energy is needed for precipitation to
take place at a rate corresponding to the production of a
reactants, the negative Gibbs energy for the microbiologi-
cally mediated reactions reect that energy is needed, in this
case in order to sustain the life functions of the bacteria at a
rate controlled by the rate of electron donor supply. In or-
der to facilitate comparison with observed values, the redox
reactions in the PHREEQC database were rewritten so that
the reductant was H
2
instead of electrons (Table 2). In a
way this is along the lines of Hoehler (1998), who proposes
that the observed H
2
level due to the hydrogenase enzyme
being present in many microorganisms, may actually reect
the internal redox state of many dierent types of anaerobic
microorganisms, regardless of the main electron donor
used. The logK value for the new reaction is calculated
from the DG of the rewritten reaction and then in order
to implement the energy surplus the logK value in the data-
base used by PHREEQC is shifted to lower or higher val-
ues, depending on the direction of the reaction in the
database. An example, using sulfate reduction, is shown
in Table 2. The same approach was used to attempt the
modeling of the observed acetate and formate values. This
was done by adding two reactions to the database where
C(0) as in acetate and C(2) as in formate, are transformed
into H
2
and HCO
3

and assigning logKs for the two reac-


tions that are shifted from the thermodynamic equilibrium
values, again to allow an energy gain for the microorgan-
isms carrying out these processes. The shifting of the logK
values for the redox reactions holds the same general prob-
lem as the setting of a xed supersaturation of a mineral
does, namely that if a saturated solution enters a cell with
the mineral, the mineral will dissolve until the solution
reaches the specied supersaturation, which should not
happen. However, as long as the system is maintained in
a supersaturated state, which for the redox processes imply
that organic matter is always being added to the system,
this is not a problem.
4.5.2. PHAST, Model setup
4.5.2.1. Flow model parameters. The ow model domain has
a no-ow boundary corresponding to the groundwater di-
vide, a bottom no-ow boundary corresponding to the bot-
tom of the aquifer, a constant head boundary at the
downstream end of the model and a ux boundary on
top corresponding to the inltration set to 150 mm/yr. A
uniform hydraulic conductivity of 8.5 10
5
in the hori-
zontal direction was the result of tting the observed
hydraulic gradient (Fig. 1), in the vertical direction a small-
er value of 1.3 10
5
was used assuming that the horizontal
bedding leads to lower values. The horizontal conductivity
is slightly lower than the value of 1.3 10
4
m/s determined
by sieve analysis, but considering the uncertainties it comes
close. As shown in Fig. 1 a section of the upper ux bound-
ary has a high salinity solution associated with it. Upstream
and downstream of this section, two dierent solutions with
lower salinities inltrate. The inltrating solutions used are
listed in Table 4. The variation in the inltrating solution
compositions, represent a horizontal 1D variation needed
to model the features observed in the sampled proles.
The width and the position of the high salinity inltration
was chosen to match the peak in chloride at site 2
(Fig. 2). The grid cells in the 750 m long and 10 m high
model domain were 5 m long and 0.2 m high, the time-step
used was 0.02 yr. The weighting parameters used by
PHAST for the space and time dierencing were set to
0.32 and 0.60, respectively. This set of values is a compro-
mise between calculation time, numerical oscillations and
numerical dispersion. Numerical dispersion is particularly
dicult to avoid when the ow direction is not aligned with
the grid. The free water table and the recharge boundary
along the top makes it inevitable. The modeled ow
directions and velocities are shown as velocity vectors in
Fig. 1. Because of the numerical dispersion, no additional
dispersivity was entered. If the actual zone of high salinity
Table 2
An example of how redox reactions were rewritten in the PHREEQC database so that a minimum energy yield of the reaction is available
The reaction for sulfate reduction in the PHREEQC database is
SO
4
2
9H

8e

HS

4H
2
O with : log K = 33:65=DG
o
r
= 192:07 kJ=mol
To make the value for the available energy comparable to observed values based on H
2
it is rewritten into
4H
2
SO
4
2
H

HS

4H
2
O with : log K = 45:95=DG
o
r
= 262:3 kJ=mol
The logK value is then shifted (similar to imposing supersaturation for a mineral) so that the reaction only occurs when there is energy
available for the microorganisms:
4H
2
SO
4
2
H

HS

4H
2
O shifted log K : 38:5=DG
o
r
= 219:8 kJ=mol
Implying that the left side components need to be higher, e.g. more H
2
. The available energy is 262.3 219.8 = 42.5 kJ/mol reaction or
5.3 kJ/mol of electrons at 25 C. As the enthalpy of reaction is non-zero the actual energy available will change with the temperature (see Table 1)
Geochemistry at the sulfate reductionmethanogenesis transition zone 1959
inltration does have sharp boundaries, the modeled Cl
matches the measured Cl (Fig. 2) with the amount of
numerical dispersion present in the used model setup.
4.5.2.2. Geochemical model parameters. The model setup is
in terms of sediment geochemistry essentially a 1D system.
This 1D system is very similar to the 1D PHREEQC
application by Jakobsen and Postma (1999) for the Rm
aquifer, but in this 2D PHAST model of the Asserbo site,
horizontal zones, essentially layers, are associated with
an initial solution given in Table 4, a number of solid
phases and a rate of organic matter addition summarized
in Table 3.
For a given solid phase the imposed SI and initial
amount of mineral, calcite, siderite, FeS etc., is entered.
In this model, the layers have all been assigned a xed cal-
cite saturation simply matching the calculated SI in the
aquifer, in a way a geochemical boundary condition. It
was chosen not to do a more elaborate matching of the re-
sult of the calcite dissolution using a kinetic expression as
there are no data available on particle distribution, surface
area etc. Because Mn peaks appears to coincide with the
dissolution of calcite, and the increase in Mg also appears
related, the calcite used in the model were dissolution is
dominant is not pure CaCO
3
but Ca
0.849
Mg
0.15
Mn
0.001
CO
3
.
Substitution of Ca by Mn, though much more substantial,
in the form of Ca-rhodochrosites are known from the Baltic
Sea (e.g. Jakobsen and Postma, 1989), and marine calcite
generally contains high amounts of Mg (Mucci, 1987).
The substitution inuences the solubility of the calcite. A
stoichiometric solubility constant, with a logK value of
8.38 was used for the MgMnCalcite (Busenberg and
Plummer, 1989), neglecting the minor amount of Mn. At
depth where PHREEQC calculations show that the pore-
water is close to equilibrium with calcite, pure calcite was
used as the CaCO
3
phase as this is more realistic in case
of reprecipitation. The specied SI of amorphous ferrihy-
drite was used to adjust the stability of the Fe-oxide, and
with that, the tendency for FeIII to go in to solution as
Fe
3+
and further be reduced by H
2
according to Eq. (2)
in Table 1. The Fe
2+
concentration is thereby constrained
indirectly. The coupling implies that the specic value of
the SI is related to the value of 25 kJ/mol chosen for the
energy available for Fe-reductiona value for which there
is currently no observations. The t was obtained by
decreasing the SI over depth. This could reect that the
more unstable Fe-oxides are no longer present at depth,
either due to previous reduction or due to a catalytic trans-
formation as suggested by Pedersen et al. (2005). These
uncertainties related to modeling the Fe-oxides reect the
unresolved issues regarding Fe-oxides in natural systems.
As it was discussed for Fig. 8, it seems that siderite is
controlling the Fe
2+
concentration when an SI ~ 0.75 for
siderite is reached. Based on Fig. 8, an SI of 0.0 for the
FeS precipitate dened in PHREEQC by the reaction:
FeS + H
+
= Fe
2+
+ HS

, has been imposed (using values


from Davison (1991); FeS: logK = 2.95/Fe(HS)
2
:
logK = 6.45), again with no initial FeS. It was chosen not
to model the precipitation of FeS by a kinetic expressions,
as the observations indicate that precipitation takes place at
a constant supersaturation, and data on what might control
the kinetics in this system are not available. Below the level
where sulde was detectable an SI = 1.0 was imposed
implying that the suldes eventually change to more stable
phases.
The redox reactions in the model are driven by the input
of organic carbon, in the form of CH
2
O, to each cell in the
system at a specied rate (Table 3). The organic carbon is
oxidized to inorganic carbon and the increased electron
activity arising from this enters into the system of redox
equilibrium reactions (Table 1), which together with the im-
posed mineral equilibria determine a new stable solution
chemistry for the given cell. The organic matter added to
each layer is put into the cells using the kinetic feature in
PHREEQC, however, in this model the rate is constant
over time, for a given layer. Several of the processes could
have been tted using kinetics rather than using mineral
equilibria, but the obtained parameters could not be com-
pared to actual data on mineralogy, surface area etc. as
these are not available. Kinetic descriptions could poten-
tially still be useful for long term modeling of the system,
but not for a steady-state model. In the chosen setup, the
rate of organic matter oxidation together with the stability
of the Fe-oxide are the main tting parameters. The tted
energy yields (Table 1) for sulfate reduction and methano-
genesis, and the saturation index for amorphous FeS were
all chosen very close to the observations. To produce the
Table 3
Organic matter oxidation rates assumed for the model simulation
Depth
(m)
C(0)rate mM/
yr
SI Mg,Mn-calcite/
calcite
SI
Fe(OH)
3
0.01.8 0.015 1.5
1.82.1 0.003 1.5
2.12.4 0.003 4.80 1.5
2.42.7 0.006 4.40 1.5
2.73.0 0.009 4.20 1.5
3.03.3 0.024 3.20 1.5
3.33.6 0.024 2.80 1.5
3.63.9 0.024 1.80 1.6
3.94.2 0.036 1.30 1.9
4.24.5 0.036 0.92 2.3
4.54.8 0.036 0.89 2.3
4.85.1 0.036 0.74 2.3
5.15.4 0.042 0.56 2.3
5.45.7 0.66 0.30 2.4
5.76.0 0.90 0.22 2.8
6.06.3 0.30 0.20 2.8
6.36.6 0.045 0.12 2.8
6.66.9 0.045 0.08 2.8
6.97.2 0.09 0.00 2.8
7.27.5 0.015 0.00 2.8
7.57.8 0.015 0.00 2.8
7.810.0 0.024 0.00 2.9
Saturation indexes (SI) imposed on the system for Mg,Mn-calcite
(down to 6.6 mbs), calcite below 6.6 mbs, and for Ferrihydrite,
obtained by tting. The initial amount for these two was 10.0
mol/L. The SI for amorphous FeS was set to 0.0 down to 6.9 mbs
and 1.0 below, and for siderite 0.75 was used. Both values were
adjusted based on the plots in Fig. 8. the initial amount was set to
0.0 for both.
1960 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
observed increase in the NH
4

concentration the organic


matter added needed to be given a C:N ratio of 10
((CH
2
O)
10
NH
3
). This is above the Redeld ratio for marine
algae of 6.6 expected due to the marine origin of the sedi-
ment. Well cultured soils have a C:N ratio of 10, indicating
a soil origin for the organic matter. However, the observed
ratio reects the organic matter being consumed, the bulk
organic matter may be quite dierent. The rate of addition
of organic matter given in Table 3 is the total rate, in Fig. 6
it is split up into the dierent redox processes calculated
assuming either Fe-oxide reduction with sulfate reduction
or methanogenesis and plotted with the most relevant
directly measured rates. Assuming a 1D vertical distribu-
tion is a simplifying assumption implying that within the
model area the formation can be viewed as consisting of
homogeneous layers in terms of the controlling geochemical
parameters. The tting of the model was done by compar-
ing the model output to the central prole, site 2.
As the distribution of cations show that ion exchange
reactions are occurring, an exchanger with a capacity of
30 mmol(+)/L was specied for the entire system. This
was initialized by specifying the distribution of cations on
the exchanger to be in equilibrium with a solution based
on the standard exchange constants in the PHREEQC
database. The solution used was a constructed solution
(see Table 4) made by mixing seawater with a water compo-
sition from the upper part of the system and letting this
mixture come into equilibrium with calcite. The solution
composition could be seen as the result of inundating sea-
water mixing with water in the system and equilibrating
with the calcite in the system.
The model has been run for 100 years, the time it takes
to almost exchange all of the water in the model with the
proposed inltrating solutions, implying that what is being
modeled is an assumed quasi-steady state situation. This
seems to be a reasonable approximation for the redox pro-
cesses where the sediment characteristics of a given layer
presumably change slowly, however, for the ion-exchange
processes this is a simplication.
4.5.3. Model results and discussion
Model results for individual parameters are shown in the
Figs. 26 with the measured data, and will be discussed be-
low. To give an impression of the 2D patterns Fig. 10ad
shows the simulation after 100 years of, a) a 2D cross-sec-
tion of chloride, as a tracer of the ow pattern, b) sulfate, as
an example of a solute being removed, c) a plot of the dis-
solved FeII an example of a solute being produced and d)
the amount of precipitated Fe-sulde after 100 years, an
example of a solid phase forming. Fig. 10a and b shows that
there is some numerical oscillation occurring at the bound-
ary between the inltration areas of the high and low
salinity solutions, these oscillations are, however, smeared
out upstream of the sampled area due to the numerical dis-
persion. The sulfate distribution shows the relative com-
plexity that arises when a very simple 1D variation in
inltrating water chemistry, resulting in the pattern shown
for chloride in Fig. 10a is combined with a 1D distribution
of sediment characteristics (organic matter oxidation rates
and mineralogy). The ow pattern seen from Fig. 10a is
presumably a reasonable representation of the Asserbo
aquifer, but the complexity issue is worth keeping in mind
Table 4
Inltrating solutions and exchanger solution used in the model, all values are in mmol/L
Parameter
(mM) (8 C)
a
Solution inltrating in
upstream end of model
(medium TDS)
b
Solution inltrating
in central part of
model (high TDS)
c
Solution inltrating
downstream and upper
initial solution (low TDS)
d
Initial solution
for the lower part
of the aquifer
e
Solution used
for initializing the
exchanger
pH 4.8 4.8 4.9 7.48 8.3
Alkalinity 0.04 0.02 0.02 4.04 0.57
Na 1.2 1.952 0.257 1.02 2.085
K 0.16 0.13 0.03 0.18 0.041
Mg 0.248 0.30 0.04 0.39 0.113
Ca 0.186 0.756 0.07 1.70
g
0.487
Mn 0.002 0.002 0.002 0.002 0.002
FeII 0.04 0.04 0.04 0.04 0.04
FeIII
f
0.00055
f
0.00056
f
0.00041
f
0.000002
O(0) 0.015 0.015 0.015 0.0094
Cl 1.35 2.00 0.254 1.1 2.20
Sulfate 0.443 1.00 0.135 0.008 0.3
Ammonium 0.002 0.002 0.002 0.261 0.001
C(-4) 0.337
Br (tracer) 0.001
a
Solution from site 2, 5.9 mbs modied to resemble the solution from site 1 2.5 mbs used for the downstream part in terms of its oxic state
and pH, and diluted to obtain an adequate ionic strength.
b
Solution from site 2, 5.9 mbs modied to resemble the solution used for the upstream part in terms of its oxic state and further modied by
increasing ionic strength.
c
Solution from site 1, 2.5 mbs, also used as the initial solution in the model from 0 to 5.1 mbs.
d
Initial solution below 5.1 mbs, taken as the solution from site 2, 8.3 mbs.
e
Dilute seawater with extra Ca, alkalinity and sulfate, perhaps from pyrite oxidation and calcite dissolution.
f
Calculated by PHREEQC from equilibrium with amorphous Fe(OH)
3
.
g
Calculated by PHREEQC from charge balance.
Geochemistry at the sulfate reductionmethanogenesis transition zone 1961
when other, very probably, less simple ow systems with a
less simple distribution of sediment characteristics are sam-
pledgenerally at much lower sample resolution. Fig. 10d,
shows, as expected, that most of the FeS precipitation takes
place where the high sulfate water intersects with the layers
with high rates of organic matter degradation.
Columns (1D) of modeled values for the three observa-
tion points are plotted as lines with the observed data in
Figs. 26. As described above, the tting of the model
was done by comparing the model output to the central
prole, site 2. This means that the dierences in the chloride
t for the 3 boreholes indicate either variations in the salin-
ity of the high salinity water over time or space, or probable
variations in the actual ow paths in the system in response
to inltration variations and heterogeneities in the geology.
The dierences in the chloride t are rather small, indicat-
ing a relatively homogeneous geology on this scale, where
the sampling sites are only 10 m apart. The small dierences
will also show up in other parameters controlled by the in-
put solution: sulfate Na, K, Mg, while other parameters,
e.g. pH, Ca, FeII, alkalinity and sulde are mainly con-
trolled by the imposed mineral equilibria and the coupling
to oxidation of the organic matter added to the system. This
also implies that the numerical dispersion aects these
parameters to a lesser degree.
The model t for the cations (Fig. 2) is not perfect in any
way, partly because the system is being modeled as if in a
quasi-steady-state, while any eects of ion exchange pro-
cesses imply that the system is not in a steady state. To rea-
sonably model the cations, information on the actual timing
of events aecting the exchanger composition, and the
water composition involved would be required, The peaks
could reect changes in inltration chemistry on a time
scale of years.
Comparing the shape of the sulfate and the chloride
peak shows that though the peak in the sulfate concentra-
tion is maintained (due to sulfate reduction rates being
low compared to the ux of sulfate into the system) the
presence of a layer with highly reactive organic matter cre-
ates the illusion that sulfate is moving slower than chloride,
while in fact it is the tip of the peak which is being removed
by sulfate reduction, the good t obtained for the sulfate is
directly related to the rate of organic matter oxidation in
the model.
Fig. 10. Model output concentrations in mM for (a) chloride, (b) sulfate, (c) FeII
aq
and (d) solid FeS after 100 years when the model has
reached a quasi-steady-state. Note that the ow in this gure is from left to right. The dashed lined boxes on each plot indicates the positions
and depth intervals of the sampling sites as indicated on the FeS plot (d).
1962 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
The obtained t for the other redox related components
is also pretty close. This is to some extent a result of adjust-
ing the saturation indexes for calcite and FeS, to values
close to the observed values. However if the tted rate of
organic matter oxidation was slightly dierent then, be-
cause of the close coupling with the mineral phases through
the water chemistry implied in the partial equilibrium ap-
proach, the t would be worse. The good t in all three pro-
les indicates that, since all model sediment parameters
only vary in the vertical direction, the aquifer material, in
terms of the reactivity of the sedimentary organic matter
and the mineral phases controlling the water chemistry,
must be rather homogeneous in the horizontal plane, at
least within the scale of sampling. No attempt has been
made to test the uniqueness of the calibrated set of param-
eters, but tting the observations made it quite clear that
minor changes in the parameters have very large eects
on the simulated output. The ammonium proles are
matched using a single C:N ratio, but this leads to an
underestimation of the rate of organic matter oxidation be-
low 6.5 mbs indicating that the C:N ratio of the organic
matter being oxidized in the lower part is higher, possibly
reecting a stripping of more N-rich parts over time.
In order to maintain a reasonable pH while the Fe(II)
concentration builds up, it turned out to be necessary to
have some buering of the pH of the system. In this case
it was simply obtained by adjusting the TIC, by lowering
the pH in the inltration water, while maintaining the same
low alkalinity. The resulting TIC in the modeled water at 2
mbs is around 1.3 mM, compared to the value of the upper-
most groundwater, derived from measured alkalinity and
pH, ranging from 1.8 mM at site 1 to only 0.3 at sites 2
and 3. In the real system some of the pH buering could
be related to Ca
2+
H
+
exchange on the carboxylic acid
groups in the DOC which as mentioned (Fig. 4), shows a
notable decrease at the top of the prole coinciding with
the increase in Ca. It could also be due to buering by pro-
tons being released from surfaces in the sediment, as imple-
mented in the model for the Cape Cod system (Parkhurst
et al., 2003). However there are no data available to support
one or the other hypothesis, and considering that site 1 had
a high TIC increasing the TIC seems reasonable.
Using SI = 0.75 for siderite leads to a model pH that is
too high in the lower part, resulting in Ca
2+
, becoming too
low, in the lower part. The simultaneous equilibrium with
siderite, calcite and Fe-sulde xes the Fe/Ca and the
HCO
3

=HS

activity ratios, the end result being that the


modeled Ca becomes too low and the modeled pH too high
(Figs. 2 and 3). Setting the SI for siderite higher would give
a better t of Ca and pH, but would imply that siderite
would not precipitate in the model.
In general, imposing all the equilibrium constraints list-
ed to some extent locks the water composition in place and
it could seem that the good ts shown for pH, alkalinity
and Ca in Fig. 2 and the redox parameters in Fig. 3 are
merely a result of xing all these equilibria. However,
CH
4
and sulfate are not xed by equilibria and the siderite
saturation discussed above shows that xing these equilib-
ria does not necessarily lead to closer ts. The tted rate
of organic matter oxidation is close to the measured value
for sulfate reduction (Fig. 6), and, since this is an indepen-
dent measurement, the good t indicates that the descrip-
tion of the system by a partial equilibrium approach is a
good approximation for the upper part. In the lower part
the model rates are lower than the measured rates. Simply
increasing the overall rate would increase the modeled
ammonium and the methane concentrations. The modeled
ammonium could be lowered by a higher C:N ratio, but
to avoid too high methane concentrations requires that
the extra added carbon is used to increase the amount of
Fe-oxide being reduced, which would also t better with
the high measured rates of acetate oxidation (Fig. 6). How-
ever to increase the amount of Fe-oxides being reduced
would require increasing the solubility of the Fe-oxides in
the lower part, which again due to the coupled equilibria
would lead to an even higher pH. Still, the close to constant
measured Fe
2+
concentration below 7 mbs, together with
the calculated supersaturation for siderite does presumably
conceal a concomitant reduction of Fe-oxides and metha-
nogenesis as reported by Berner (1981), Jakobsen and Post-
ma (1999) and Ferro and Middelburg (2003), and as the
appreciable acetate oxidation rates in this zone (Fig. 6)
together with the lack of available Gibbs energy for the pro-
duction of H
2
from acetate (Fig. 9), would indicate. It is
interesting to note that if siderite is precipitating while
Fe-oxide reduction and methanogenesis proceed concomi-
tantly at a ratio of 1:6 in terms of C(0) transformed, one
may have a net transformation of Fe-oxides to siderite, con-
currently with methanogenesis, with no direct trace of the
Fe-oxide reduction in the porewater chemistryexcept
for a constant SI for siderite according to Eqs. (7)(9):
Fe-oxide reduction :
4FeOOHCH
2
O3H
2
CO
3
4FeCO
3
6H
2
O (7)
Methanogenesis :
6CH
2
O3H
2
O 3H
2
CO
3
3CH
4
(8)
Combined reaction :
4FeOOH7CH
2
O 4FeCO
3
3CH
4
3H
2
O (9)
this scenario would be ideal from a microbiological system
point of view as the energy available for the Fe-oxide reduc-
tion would be maintained, and the energy available for
methanogenesis would stay higher than without the precip-
itation of the produced carbonate. This `deality could cause
the system to approach the ideal ratio and stay in this state
as long as the kinetics of the Fe-oxide reduction could keep
up. In the model, the ratio is in the interval 1:6.47.8 below
a depth of 8.4 mbs, where there is also siderite precipitation
in the model taking place at a rate matching the Fe-oxide
reduction, indicating a thermodynamic optimum.
Considering that the values for the threshold energy
yields for sulfate reduction and methanogenesis entered in
the model were calculated from observed H
2
values, one
would expect the nice t to the H
2
data where these process-
es occur (Fig. 4). So for these two reactions the applied
shifting of the equilibrium constants will yield a good pre-
diction of the H
2
concentrationsespecially if H
2
measure-
ments for calibrating the energy yields of the processes in
the given system are present. The model values for H
2
in the upper part dominated by Fe-oxide reduction are
Geochemistry at the sulfate reductionmethanogenesis transition zone 1963
underestimated, but could not be increased to the measured
values using reasonable t parameters. If the threshold
energy yield for Fe(III) reduction to occur is set higher,
to increase the modeled H
2
concentration, the Fe(III)-oxide
would have to be more soluble than ferrihydrite, in order
for enough Fe
3+
to be available for reduction. The explana-
tion could be that stagnant zones play a role as discussed in
Sect. 4.4. The 1D dual porosity model used by Jakobsen
(2002) could give a better t of the H
2
concentrations than
a single porosity model, but there is currently no dual
porosity option in PHAST. Still, the fact that the Fe-oxides
are solids could also imply that the reactivity of the Fe-ox-
ides is more limiting than the reactivity of the organic mat-
ter as discussed in Larsen et al. (2006). Or, that a simple
partial equilibrium description is in fact too simple and
needs to be expanded with kinetic terms as suggested, more
generally, by Jin and Bethke (2005), for reactions occurring
at more oxic conditions than sulfate reduction, as shown to
be the case for the reduction of chlorinated solvents
(Heimann and Jakobsen, 2006). Also several recent studies
suggest that Fe-oxide reduction of the solid phase oxide is
facilitated by either chelators assisting in the dissolution
of the Fe-oxides (Luu and Ramsay, 2003), electron shuttles
(Nevin and Lovley, 2002), removing the need for direct
contact, or nano-wires that transfer the electrons directly
(Reguera et al., 2005).
For formate and acetate, the simplied model applied
here, where acetate and formate are transformed via H
2
,
is obviously not adequate (Fig. 5), and denitely an approx-
imation as all of the redox processes in the system can,
depending on the microbiology, in fact proceed with acetate
as electron donor. A more comprehensive approach model-
ing several electron pathways was not made, as this would
involve tting numerous unknown parameters controlling
the entire microbial ecology. Another complication is that
for example acetate may both form from, and transform
into H
2
and CO
2
, and the simple shifting of the logK value,
to increase the modeled acetate concentration, implies that
the model may form acetate when there is actually no ener-
gy available for this. As mentioned above this may work in
a system where electron donor is constantly added every-
where, but it is a simplication. The radiotracer measure-
ments, however, indicate that at least for methanogenesis
the CO
2
reduction by H
2
is in fact the preferred pathway.
Ideally this type of microbial reaction, that can go both
ways should be implemented so that it only occurs with a
positive energy yield in either direction as it was done for
methanogenesis by Jakobsen (2002). Still the ability of the
model to reasonably predict the H
2
level can be important
when evaluating the potential for reduction of pollutants,
e.g. Cr(VI) reduction (Marsh and McInerney, 2001). So,
though the model t is not perfect it does indicate that in
more general terms the release of inorganic carbon, Fe
2+
and sulde leads to a state where the groundwater compo-
sition is controlled by several mineral equilibria which
again aect the ongoing redox processes forming a closely
coupled system. An example scenario shows the implica-
tions of this: if the ux of sulfate into the system decreases,
methanogenesis will occur higher up in the system and the
net production of carbonic acid from this will lead to the
release of FeII and Ca from siderite and calcite and FeII
and sulde from Fe-suldes. The higher H
2
concentration
occurring during methanogenesis and any decrease in pH
that is not neutralized by the mineral dissolution would in-
crease the potential for Fe-oxide reduction, but on the other
hand the increase in Fe
2+
will lower this (Eq. 2, Table 1). If
the net eect is an increase in Fe-oxide reduction this will
lead to an increase in pH, again aecting the mineral disso-
lution and eventually a new quasi-steady state is reached.
The model indicates that the long term development of
the system will be a function of how the input of sulfate,
the bioavailability of the Fe-oxides and the reactivity of
the organic matter develop over time. Sulfate input could
decrease due to, e.g. better pollution control, initiating the
scenario described above. Sulfate input could also increase
due to more extreme weather conditions in the future, lead-
ing to a higher dry deposition, which would force the zone
of methanogenesis deeper into the system again leading to a
cascade of eects towards a new steady-state.
5. CONCLUSIONS
5.1. A comparison with the Rm aquifer
The Asserbo aquifer is in many ways geochemically sim-
ilar to the Rm aquifer, with some noteworthy dierences.
(1) There is no pool of AVS below the sulfate reducing zone
indicating that there has been enough sulde available for
the conversion, which could be related to the higher mea-
surable sulde concentrations in the Asserbo system com-
pared to the non-measurable concentrations in the Rm
aquifer. (2) The H
2
level in the lower part of the system is
high enough to sustain methanogenesis removing the need
for stagnant microniches, the reason for this is not obvious.
(3) The Asserbo data on DOC clearly indicate an inux of
organic matter to the system from the soil above, presum-
ably supplementing the sediment bound organic matter,
which appears to be necessary for sustaining redox process-
es in the system. This is probably also the case on Rm but
data on this were less clear.
5.2. Modelling
The partial equilibrium approach using shifted K-values
for the redox processes assuming H
2
is the electron donor
comes a long way in describing the observed redox chemis-
try including the H
2
concentrations when sulfate reduction
or methanogenesis are dominating. When going in to more
detail it is clear that in order to model the concentrations of
acetate and formate in general and the H
2
concentration
when Fe-oxide reduction is dominating, a model with a
closer description of the microbial ecology, the kinetics of
Fe-oxide reduction, and perhaps able to handle dual poros-
ity is needed. The relatively simple setup using just three dif-
ferent inltration chemistries and a 1D variation in
sediment characteristics does produce most of the complex
variation and shows that in the very probable case where
inltration composition, geology and geochemistry are all
less simple, quantitative interpretations of observations
based on less densely spaced groundwater and sediment
1964 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966
samples will be dicult to obtain. The overall conclusion is
that anaerobic aquifer systems may develop to a state where
the groundwater chemistry is controlled by a set of inorgan-
ic dissolution and precipitation reactions that are closely
coupled with terminal electron acceptor processes driven
by the oxidation of organic matter in the system. In such
a system, small changes in one input parameter may lead
to considerable changes in the whole system.
ACKNOWLEDGMENTS
We thank the editor and two anonymous reviewers whose help-
ful input has led to substantial improvement of the manuscript.
REFERENCES
Amend J. P., and Teske A. (2005) Expanding frontiers in deep
subsurface microbiology. Palaeogeogr. Palaeoclimat. Palaeo-
ecol. 219(12), 131155.
Atkins P., and de Paula J. (2002) Atkins Physical Chemistry,
seventh ed. Oxford University Press, New York.
Berner R. A. (1981) Authigenic mineral formation resulting from
organic matter decomposition in modern sediments. Fortschr.
Miner. 59, 117135.
Brun A., and Engesgaard P. (2002) Modelling of transport and
biogeochemical processes in pollution plumes: literature review
and model development. J. Hydrol. 256, 211227.
Busenberg E., and Plummer L. N. (1989) Thermodynamics of
magnesium calcite solid-solutions at 25 and 1 atm total
pressure. Geochim. Cosmochim. Acta 53, 11891208.
Caneld D. E., Thamdrup B., and Hansen J. W. (1993) The
anaerobic degradation of organic matter in Danish Coastal
sediments: iron reduction, manganese reduction and sulfate
reduction. Geochim. Cosmochim. Acta 57, 38673883.
Chapelle F. H., and McMahon P. B. (1991) Geochemistry of
dissolved inorganic carbon in a coastal plain aquifer. 1. Sulfate
from conning beds as an oxidant in microbial CO
2
production.
J. Hydrol. 127, 85108.
Cline J. D. (1967) Spectrophotometric determination of hydrogen
sulde in natural waters. Limnol. Ocean. 14, 454458.
Crill P. M., and Martens C. S. (1986) Methane production from
bicarbonate and acetate in an anoxic marine sediment.
Geochim. Cosmochim. Acta 50, 20892097.
Curtis G. P. (2003) Comparison of approaches for simulating
reactive solute transport involving organic degradation reac-
tions by multiple electron acceptors. Comp. Geosci. 29, 319329.
Darling W. G., and Goody D. C. (2006) The hydrogeochemistry of
methane: evidence from English groundwaters. Chem. Geol.
229, 293312.
Davison W. (1991) The solubility of iron suldes in synthetic and
natural-waters at ambient-temperature. Aquat. Sci. 53,
309329.
Ferro I., and Middelburg J. J. (2003) Factors limiting iron oxide
reduction in a tidal freshwater and a Mediterranean deep sea
sediment. In Ferro I. (2003) Cycling of iron and manganes in
freshwater, estuarine and deep sea sediment, Ph.D. Disserta-
tion, University of Groningen, Netherlands.
Hansen L. K. (1998) Biogeochemistry of methane in a shallow
sandy aquifer. Ph.D. dissertation, Department of Geol. and
Geotechn. Eng. Technical University of Denmark.
Hansen L. K., Jakobsen R., and Postma D. (2001) Methanogenesis
in a shallow sandy aquifer, Romo, Denmark. Geochim.
Cosmochim. Acta 65, 29252935.
Heimann A. C., and Jakobsen R. (2006) Experimental evidence for
a lack of thermodynamic control on hydrogen concentrations
during anaerobic degradation of chlorinated ethenes. Environ.
Sci. Technol. 40, 35013507.
Hoehler T. M. (1998). Thermodynamics and the role of hydrogen in
anoxic sediments, Ph.D. dissertation. University of North
Carolina, USA.
Hoehler T. M., Alperin M. J., Albert D. B., and Martens C. S.
(2001) Apparent minimum free energy requirements for
methanogenic Archaea and sulfate-reducing bacteria in an
anoxic marine sediment. FEMS Microb. Ecol. 38, 3341.
Howarth R. W., and Merkel S. (1984) Pyrite formation and
measurements of sulfate reduction in salt marsh sediments.
Limnol. Oceanogr. 29, 598608.
Hunter K. S., Wang Y., and van Cappelen P. (1998) Kinetic
modeling of microbially-driven redox chemistry of subsurface
environments. J. Hydrol. 209(14), 5380.
Jakobsen R., and Postma D. (1989) Formation and solid-solution
behavior of Ca-rhodochrosites in marine muds of the baltic
deeps. Geochim. Cosmochim Acta 53, 26392648.
Jakobsen R., and Postma D. (1994) In situ rates of sulfate
reduction in an aquifer (Rm, Denmark) and implications
for the reactivity of organic matter. Geology 23,
11031106.
Jakobsen R., and Postma D. (1999) Redox zoning, rates of sulfate
reduction and interactions with Fe-reduction and methanogen-
esis in a shallow sandy aquifer, Rm, Denmark. Geochim.
Cosmochim. Acta 63, 137151.
Jakobsen R. (2002) Including stagnant zones in order to model
overlapping redox processes occurring in a shallow sandy
aquifer. Geochim. Cosmochim. Acta 66(15A), A362, Suppl. 1
Aug.
Jensen D. L., Boddum J. K., Tjell J. C., and Christensen T. H.
(2002) The solubility of rhodochrosite (MnCO
3
) and siderite
(FeCO
3
) in anaerobic aquatic environments. Appl. Geochem.
17, 503511.
Jin Q., and Bethke C. (2005) Predicting the rate of microbial
respiration in geochemical environments, Geochim. Cosmo-
chim. Acta 69, 11331143.
Kipp K. L., Jr. (1997) Guide to the revised heat and solute
transport simulator: HST3D. U.S. Geological Survey Water-
Resources Investigations Report 97-4157, 149p.
Larsen O., Postma D., and Jakobsen R. (2006) The reactivity of
iron oxides towards reductive dissolution with ascorbic acid in a
shallow sandy aquifer (Rm, Denmark). Geochim. Cosmo-
chim. Acta 70, 427435.
Luu Y., and Ramsay J. A. (2003) Review: Microbial mechanisms
of accessing insoluble Fe(III) as an energy source. W. J. Microb.
Biotechnol. 19, 215225.
Marsh T. L., and McInerney M. J. (2001) Relationship of hydrogen
bioavailability to chromium reduction in aquifer sediments.
Appl. Environ. Microbiol. 2001, 15171521.
Nevin P. K., and Lovley D. R. (2002) Mechanisms for Fe(III) oxide
reduction in sedimentary environments. Geomicrobiol. J. 19,
141159.
Park J., Sanford R. A., and Bethke C. M. (2006) Geochemical and
microbiological zonation of the Middendorf aquifer, South
Carolina. Chem. Geol. 230, 88104.
Parkhurst D. L., Kipp K. L., Engesgaard P., Charlton S. C. (2005)
PHASTa program for simulating ground-water ow, solute
transport, and multicomponent geochemical reactions. U.S.
Geological Survey Techniques and Methods, 6, A8, 154p.
Parkhurst D. L., Stollenwerk K. G., Colman J. A. (2003) Reactive-
transport simulation of phosphorus in the sewage plume at the
Massachusetts Military Reservation, Cape Cod, Massachusetts.
Geochemistry at the sulfate reductionmethanogenesis transition zone 1965
U.S Geological Survey Water-Resources Investigations Report,
03-4017, 33p.
Parkhurst D. L., Appelo C. A. J. (1999) Users guide to PHREEQC
(Version 2), U.S Geological Survey Water-Resources Investiga-
tions Report, 99-4259.
Pedersen H. D., Postma D., Jakobsen R., and Larsen O. (2005)
Fast transformation of iron oxyhydroxides by the catalytic
action of aqueous Fe(II). Geochim. Cosmochim. Acta 69, 3967
3977.
Pedersen K. (2000) Exploration of deep intraterrestrial microbial
life: current perspectives. FEMS Microbiol. Lett. 185(1), 916.
Postma D. (1981) Formation of siderite and vivianite and the pore-
water composition of a recent bog sediment in Denmark. Chem.
Geol. 31, 225244.
Postma D., and Jakobsen R. (1996) Redox zonation: Equilibrium
constraints on the Fe(III)/SO4-reduction interface. Geochim.
Cosmochim. Acta 60, 31693175.
Reguera G., McCarthy K. D., Mehta T., Nicoll J. S., Tuominen
M. T., and Lovley D. R. (2005) Extracellular electron
transfer via microbial nanowires. Nature 435, 10981101.
Schulz S., and Conrad R. (1997) Inuence on temperature on
pathways to methane production in the permanently cold
profundal sediment of Lake Constance. FEMS Microb. Ecol.
20, 114.
Starr R. C., and Ingleton R. A. (1992) A new method for collecting
core samples without a drilling rig: Ground Water Mon. Rev.
Winter 1992, 9195.
Stookey L. L. (1970) Ferrozinea new spectrophotometric reagent
for iron. Anal. Chem. 42, 779781.
Thullner M., van Cappellen P., and Regnier P. (2005) Modeling the
impact of microbial activity on redox dynamics in porous
media, Geochim. Cosmochim. Acta 69, 50055019.
Watson I. A., Oswald S. E., Banwart S. A., Crouch R. S., and
Thornton S. F. (2005) Modeling the dynamics of fermentation
and respiratory processes in a groundwater plume of phenolic
contaminants interpreted from laboratory- to eld-scale. Envi-
ron. Sci. Technol. 39, 88298839.
Whiticar M. J., Faber E., and Schoell M. (1986) Biogenic methane
formation in marine and freshwater environments: CO
2
reduc-
tion vs. acetate fermentationisotopic evidence. Geochim.
Cosmochim. Acta 50, 693709.
Associate editor: Jack J. Middelburg
1966 R. Jakobsen, L. Cold / Geochimica et Cosmochimica Acta 71 (2007) 19491966

You might also like