Ecological Engineering 110 (2018) 54–66

Contents lists available at ScienceDirect

Ecological Engineering
journal homepage: www.elsevier.com/locate/ecoleng

Determination of major biogeochemical processes in a denitrifying MARK

woodchip bioreactor for treating mine drainage
⁎
Albin Nordström , Roger B. Herbert
Uppsala University, Department of Earth Sciences, Villavägen 16, SE-752 36, Uppsala, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: At the Kiruna iron ore mine in northern Sweden, mine drainage and process water contain elevated con-
Denitrification centrations of nitrate (NO3−) from the use of ammonium nitrate fuel oil explosives. In order to investigate the
Sulfate reduction treatment capacity of a denitrifying woodchip bioreactor technique for the removal of NO3− through deni-
DNRA trification, a bioreactor was installed at the mine site in 2015 and operated for two consecutive years. Neutral-pH
Woodchip bioreactor
mine drainage and process water containing 22 mg NO3−-N and 1132 mg SO42− (average) was passed through
Temperature
Biogeochemical processes
the bioreactor which was filled with a reactive mixture of pine woodchips and sewage sludge, at treatment
temperatures ranging between 0.8 and 17 °C. At bioreactor temperatures above ∼5 °C, NO3− removal pro-
ceeded to below detection limits (0.06 mg N L−1) without substantial production of nitrite (NO2−), ammonium
(NH4+), nitrous oxide (N2O), or methane (CH4). The relative production of NH4+ and N2O to the NO3− reduced
increased as bioreactor temperatures decreased below ∼5 °C. Based on the resultant changes in alkalinity and
pH from the production of bicarbonate (HCO3−) and carbonic acid (H2CO3), a stoichiometric mass balance
model indicated that denitrification, nitrate reduction to ammonium (DNRA), sulfate reduction, and fermen-
tation were the major biogeochemical processes controlling pH, alkalinity and nitrogen, sulfur and carbon
concentrations in the system. It is suggested that fermentation changed from being mainly butyrate producing to
acetate producing with time, triggering a decline in biogeochemical process diversity and leaving denitrification
as the sole major electron accepting process.

1. Introduction
Ammonium nitrate (NH4NO3) is the most common explosive used in
the mining industry (Forsyth et al., 1995) and is highly soluble in water.
Due to spillage during storage, transport, and loading of the explosives,
as well as the incomplete detonation of the explosives (Revey, 1996),
NH4NO3 dissolves in mine drainage and process water and is eventually
discharged to the environment, primarily in the form of nitrate (NO3−)
(Lindeström, 2012). Excess release of NO3− to aquatic ecosystems can
be hazardous as it may induce eutrophication leading to hypoxia, or
may be transformed into ammonium (NH4+) or ammonia (NH3). NH4+
is more easily incorporated into biomass than NO3− (Rittmann and
McCarty, 2001), leading to increased risks of eutrophication, and may
further be transformed into NO2−/NO3− through nitrification, an
oxygen consuming reaction, both leading to increased risks of hypoxia
in aquatic ecosystems. NH3 is additionally toxic to aquatic biota at high
concentrations (EPA, 2013).
Denitrification provides the possibility of completely transforming
NO3− into nitrogen gas (N2), a comparatively inert form of nitrogen (N)
that is unavailable for uptake by most living organisms. The stoichio-
metric definition of net denitrification is exemplified in reaction (1)
using glucose (C6H12O6) as the carbon substrate/electron donor
5 1 1 1
NO3− + C6 H12 O6 → N2 + HCO3− + H2 CO3 + H2 O
24 2 4 2 (1)
Denitrification occurs under the sequential reduction of NO3−,
NO2−, NO, and N2O to N2, and is known to be “leaky”, such that in-
termediate NOx species may accumulate.
Woodchip bioreactors have been identified as effective technologies
for the removal of NO3− from contaminated water (e.g. Moorman et al.,
2010; Robertson and Merkley, 2009; Warneke et al., 2011a). For a
general overview of the woodchip bioreactor technology, the reader is
refered to Schipper et al. (2010). In terms of the reactive material in the
bioreactor, wood products (e.g. sawdust and woodchips) have in-
creasingly been used as an electron source for denitrification due to the
high content of carbon (electron donor) available at low cost (Schipper
et al., 2010), and due to the longevity of the material (Moorman et al.,
2010; Robertson and Merkley, 2009; Robertson, 2010; Warneke et al.,
2011a).

⁎
Corresponding author.
E-mail addresses: albin.nordstrom@geo.uu.se (A. Nordström), roger.herbert@geo.uu.se (R.B. Herbert).

http://dx.doi.org/10.1016/j.ecoleng.2017.09.018
Received 29 June 2017; Received in revised form 25 September 2017; Accepted 29 September 2017
0925-8574/ © 2017 Elsevier B.V. All rights reserved.
A. Nordström, R.B. Herbert
In this study, a pilot-scale woodchip bioreactor was installed in the
subarctic climate at the Kiruna iron ore mine (northern Sweden) with
the purpose of removing NO3− in mine drainage and process water
originating from the use of ammonium-nitrate fuel oil explosives
(ANFO). To the best of our knowledge, this is the first woodchip bior-
eactor that has been tested for the removal of NO3− from mine water.
The objectives of the study were to determine the predominating
biogeochemical processes controlling pore water composition in the
system. Furthermore, a mass balance approach was applied to nitrogen
and sulfur in order to calculate changes in the carbonate speciation, and
hence pH, based on reaction stoichiometry.
2. Material and methods
2.1. Study site
This study was conducted at the Kiruna iron ore mine, operated by
the mining company Luossavaara-Kiirunavaara Aktiebolag (LKAB), and
located in the subarctic climate of northern Sweden, Kiruna (67°51′N
20°13′E). The mean annual air temperature in the region of the study
site was −3 °C for the period 1961–1990 (SMHI, 2017).
Mine drainage and process water are released through a clarifica-
tion pond into the recipient Mettä Rakkurijärvi lake that drains through
the Rakkuri river system into the Kalix River (LKAB, 2016), and ulti-
mately to the Baltic Sea. The ore and waste rock has a low sulfide
mineral content, and hence the pH of the site drainage is close to
neutral at 8.05 ± 0.04. The water has relatively high concentrations of
nitrogen (N) from the use of ammonium-nitrate fuel oil (ANFO) ex-
plosives (LKAB, 2016). In 2015, ∼5,7 Mm3 of mine and process water
was released into the Mettä Rakkurijärvi lake, with an average NO3−
concentration of 27.8 mg N L−1 (LKAB, 2016).
The bioreactor in this study was constructed near the point where
water was discharged from the clarification pond to the recipient.
Construction took place during May and June 2015.
2.2. Bioreactor design
The bioreactor was designed as a subsurface system: a ∼1.1 m deep
trench with trapezoidal cross-section was excavated and enclosed by
∼1 m high mounds of waste rock material originating from the Kiruna
iron ore mine. The total depth was 2.1 m. At the ground surface, the
bioreactor was 44 m long and 6.65 m wide (see dimensions in Fig. 1a
and b). The bottom of the trench was covered with a 1.5 mm thick
impermeable geomembrane (high-density polyethylene, HDPE) serving
as an impermeable boundary and preventing the leakage of water to the
surrounding soil.
Five “inner walls” of plywood (W1–W5, Fig. 1a and b) were con-
structed and placed at 8.5 m intervals, demarcating six compartments
(C1–C6, Fig. 1a), with the purpose of forcing water flow below the
surface of the bioreactor (i.e. avoiding surface flow). The inner walls
were fixated by their own weight and by the additional weight of steel
I-beams that were placed on top of the walls (Fig. 1b).
W1–W4 measured 1.2 m in height, extending from approximately
0.1 m above the bioreactor surface to 1 m above the trench floor
(Fig. 1b). W5 covered the entire cross section of the bioreactor and
featured a rectangular weir (Fig. 1b) with an adjustable cross-sectional
area intended for regulation of the water table in the bioreactor.
Water from the clarification pond was delivered to the bioreactor
using a submersible pump placed at the clarification pond outlet, and a
ball valve was used to adjust the input flow rate (Q) to the bioreactor
system. The inlet water from the clarification pond was spread across
the width of the bioreactor using a perforated PVC drainage pipe so that
flow would be distributed as evenly as possible over the width of the
bioreactor.
The outlet was constructed as a basal drain connected to a polyvinyl
chloride (PVC) pipe extending to the surface of the bioreactor,
55
Ecological Engineering 110 (2018) 54–66
emerging at 0.9 m from W5 (in the direction of flow, see Fig. 1a). The
elevation at the top of the vertical drainage pipe was adjustable and
regulated the water level in the bioreactor. The distribution of water
over the width of the bioreactor at the inlet, and the broad-crested weir
near the outlet, were both efforts to minimize preferential flow and
reduce the advective velocity within the bioreactor; these are two fac-
tors that have been shown to maximize contact between the treated
solution and the reactive material (Herbert et al., 2014; Nordström and
Herbert, 2017).
The bioreactor featured 20 observation points referred to as
A1–A20. A1 and A20 was the inlet and the outlet, respectively. A2-A19
were PVC pipes attached in different configurations on the inner walls
of the bioreactor, and of two different designs used for pore water
sampling. A3, A7, A11, A15, and A18 were sampled according to the
BAT sampling technique (see Torstensson, 1984). The remaining PVC
pipes ended in a 10 cm screen covered in a polyester plastic net filled
with medium grained sand that worked as a filter to prevent smaller
particles from passing through the well screens.
2.3. Bioreactor material
Previous column studies have shown that a mixture of pine wood-
chips and sewage sludge created a suitable biogeochemical environ-
ment for the promotion of denitrification at temperatures relevant at
the study site (see Nordström and Herbert, 2017). Decorticated pine
woodchips (porosity 0.54, estimated in the laboratory) were retrieved
from a nearby sawmill, and digested sewage sludge was shipped from
the Uddebo waste water plant, Luleå, northern Sweden. A digested form
of sewage sludge was used as a source of denitrifying bacteria and was
selected in preference of activated sewage sludge as it was easier to
transport and handle.
Woodchips and sewage sludge were deposited in layers when filling
the bioreactor. In total, 2.5 m3 of digested sewage sludge and ∼210 m3
of decorticated pine woodchips were added to the bioreactor, yielding a
sludge-woodchip ratio of 1:84 on a volume basis. The woodchips were
emplaced and gently compressed using an excavator, while the sewage
sludge was deposited manually. A sewage sludge slurry (sludge:water
volume ratio ∼1:10) was poured over the woodchips in layers located
50 cm and 100 cm above the geomembrane (S1–S2; Fig. 1c). Approxi-
mately 10 Liters of slurry were added per square meter of the woodchip
surface. In addition, sludge (not slurry) was dispersed on the wood chip
surfaces in the first three compartments (C1–C3, Fig. 1a). Note that no
sewage sludge (or slurry) was added to compartment C6 in the bior-
eactor.
The accumulated thickness of the layers of pine woodchips and
sewage sludge in C1, C5, C6 and C2–C4, was ∼1.7 m and 1.2 m, re-
spectively. The mixture of pine woodchips and sewage sludge was then
covered with a geotextile as to minimize the contact with the atmo-
sphere. In compartments C2–C4, 0.9 m of glacial till (primarily silty-
sand with detrital organic matter, but also with larger cobbles < 300
mm) was placed, and compressed, on top of the geotextile using an
excavator. The purpose of the till layer was primarily for forcing the
flow to pass through the more permeable woodchip material and
minimize gas exchange (O2, CO2, N2O, CH4) between the bioreactor
interior and the surrounding atmosphere. The till layer also decreased
the infiltration of precipitation into the permeable woodchip material,
which would lead to a dilution effect. The geotextile underlying the till
prevented the till from migrating into the woodchip material and
thereby risking a decrease in permeability. During the placement of the
glacial till, the bioreactor was saturated with water from the clarifica-
tion pond at a flow rate of 2 L s−1.
2.4. Bioreactor operation
After the saturation of the bioreactor system, the pore water was
slowly recirculated for 10 days (cf. Nordström and Herbert, 2017) to
A. Nordström, R.B. Herbert Ecological Engineering 110 (2018) 54–66

Fig. 1. Construction and instrumentation of

the bioreactor. The longitudinal and cross-
sectional dimensions of the bioreactor
trench are shown in (a) and (b). The di-
mensions of the five inner walls (W1-W5)
are shown in (b). The material layering is
shown in (c).

allow for the development of the microbial community. After this
period, water was pumped to the bioreactor inlet and the outlet was
opened. The bioreactor was thereafter operated for two consecutive
field seasons. The first operational year extended from 22nd of June
2015 until the 20th of November 2015 when air temperatures dropped
to below −20 °C. The second operational year commenced on the 9th
of May 2016 when air temperatures approached 0 °C, and the bior-
eactor was operated until the 21st of October 2016. Flow was stopped
to the bioreactor during the winter months because the piping system
leading to the bioreactor would have otherwise frozen. Indeed, tem-
perature measurements in the bioreactor (see below) indicated that
temperatures were well above 0 °C in the active regions of the bior-
eactor (1–2 m depth) during the entire period June 2015–October
2016.
The hydraulic residence time (HRT) in the bioreactor system was
adjusted several times during the operation of the bioreactor (see
Table 1) by changing the pump discharge Q. Water flow to the bior-
eactor was varied between 43.2 and 109.6 m3 d−1, with 43.2 m3 d−1
being the most common flow rate used.
Table 1
Theoretical hydraulic residence times (HRT) calculated by dividing the estimated pore
volume with the pump flow rate (Q).
Days Q [m3 day−1] Theoretical HRT [days]
2.5. Sampling and analysis
In this study, data collected from the inlet (A1), outlet (A20), and
the observation points located along the bottom middle profile in the
bioreactor (A2, A6, A10, A14, A17) were used for analysis.
2.5.1. Water sampling
Water samples were collected from the inlet and outlet approxi-
mately two times a week. Once a month, an extended sampling took
place where bioreactor pore water was also sampled. Pore water was
sampled using a peristaltic pump. Before sampling the porewater, ∼2 L
of water was purged from the sampling tube to ensure the collection of
a representative sample (see Appelo and Postma, 2005, p.12). Samples
intended for analysis of NO3−, NO2−, NH4+, pH, alkalinity, total or-
ganic carbon (TOC), and other inorganic anions were collected in clean
2 L PVC containers that were refrigerated at 4° +/− 2 °C within two
hours of collection. The chemical analyses on the water samples were
performed by LKAB’s accredited laboratory services, following stan-
dardized procedures as presented in table S1.1 (see supplementary
materials).
Whenever sample analysis could not be guaranteed within the
standard time for analysis, samples were sent by ground transport to the
accredited laboratory ALS Scandinavia in Luleå for analysis, which in-
cluded ∼1 day of additional storage at room temperature during
transport.

−10-0 14.0 Recirculation

0–45 43.2 2.6
2.5.2. Gases
46–52 0 No flow – Pump malfunction After water sampling, a 50 mL sample was injected into a pre-
53–151 59.4 1.9 sealed, non-evacuated (air-filled), 100 mL serum vial containing 1 mL
152–321 0 No flow – winter intermission ZnCl2 (50% w/v). Zinc chloride was used as a preservative and was
322–370 43.2 2.6
prepared from milli-Q water and anhydrous, reagent grade (≥98%)
371–428 48.9 2.3
429–490 109.6 1 ZnCl2 (Redi-Dri™). In total 130 dissolved gas samples were collected,
whereof 33 were technical replicates.

56
A. Nordström, R.B. Herbert
The dissolved gas samples were shipped from Kiruna to Uppsala via
ground transport (4-7 days), and stored at room temperature (22 °C) for
24–72 h for equilibration. The headspace in the 100 mL serum vials was
sampled using a 60 mL luer lock syringe equipped with a stop cock and
a hypodermic needle and injected into a pneumatically pre-sealed, non-
evacuated (air-filled), 22 mL PerkinElmer™ glass vial, so that the at-
mosphere was replaced in the vial. Sampled headspace gas was ana-
lyzed for N2O and CH4 on a gas chromatograph equipped with an
electron capture detector (Clarus 500 GC, Perkin Elmer, CT, USA).
Surface gas fluxes (N2O, CH4) were measured using the static gas
chamber technique, with the chambers constructed with reference to the
guidelines provided by Parkin and Venterea (2010). The chambers (60 L)
were constructed in polyethylene plastic and equipped with water locks,
vent tubes, fans, and thermometers. The chamber anchors were perma-
nently installed, and when not in use, the anchors were covered with
geotextile. Samples were withdrawn from the headspace in the top of
chambers at 0, 15, 30, 45, and 60 min after emplacement using a 60 mL
syringe and a hypodermic needle. The collected sample was injected into a
non-evacuated (air-filled), 22 mL PerkinElmer™ glass vial and shipped and
analyzed together with dissolved gas samples (see above).
2.6. Determination of gas concentrations and surface gas emissions
Dissolved gas concentrations were determined via headspace equi-
librium using Henry’s law and the Bunsen coefficient which was cal-
culated and adjusted to the increased pressure (2 atm) in the sampling
vials for dissolved gases (see Section 2.5.2), based on Wilhelm et al.
(1977) and Breibarth et al. (2004). The R (R Core Team, 2017) package
quantchem (Komsta, 2012a, 2012b) was used produce the calibration
curves and to account for heteroscedasticity and non-linearity. Tech-
nical replicates showed a significant correlation with the original
samples (see Fig. S1.1, supplementary material). Surface gas fluxes
were calculated using the R package HMR (see Pedersen, 2015), ac-
counting for nonlinearity in the change in gas concentration in the
chamber over time. This resulted in six surface gas flux measurements
for each sampling occasion, and the average of these were used as re-
presentative of the surface gas flux for the entire bioreactor.
2.7. Determination of the N2O emission factor
The N2O emission factor is here defined as the ratio of emitted N2O
to the NO3− removed. The total mass of emitted N2O-N was divided by
the total mass of removed NO3−-N on a daily basis, for each of the
sampling occasions, yielding the N2O emission factor. The total mass of
emitted N2O-N was calculated as the sum of dissolved and surface gas
emissions of N2O. The emitted N2O-N mass in gas phase was de-
termined from N2O-N surface emission rates multiplied with the bior-
eactor surface area, under the assumption that there were no gas flux
from the areas of the bioreactor covered by the glacial till. The emitted
N2O-N mass in dissolved phase was determined by multiplying ob-
served dissolved N2O-N effluent concentrations with the input flow
rate. The daily total removal of NO3−-N was calculated as the differ-
ence between the inlet and outlet concentrations for each profile sam-
pling occasion multiplied with the input flow rate.
3. Results and discussion
3.1. Influent and effluent concentrations
Time series of influent and effluent nitrate, nitrite, ammonium, pH,
alkalinity, TOC, and sulfate concentrations are shown in Fig. 2.
Fig. 2a–c shows that NO3− is the major nitrogen species in the incoming
mine water. Over the two year operational period, the average NO3−
concentration in the influent and effluent waters were 22.0 and
5.4 mg N L−1, respectively, with effluent concentrations ranging be-
tween the detection limit (0.06 mg N L−1) and 23.3 mg N L−1 (Fig. 2a).
57
Ecological Engineering 110 (2018) 54–66
Generally, at temperatures above 5 °C and low flow (Table 1), the in-
coming NO3− is removed to below the detection limit (0.06 mg N L−1).
Effluent NO2− concentrations were generally lower than the influent
NO2− concentrations, with the exception of days 1–7 following the
start-up of the bioreactor and during the second operational year during
periods of lower temperature (< 5 °C) and high flow (i.e. after day
429). Higher NO2− concentrations are observed in the effluent under
these conditions because denitrification kinetics are either slower (i.e.
at low temperature) or the HRT is shorter than the time required for
complete NO2− removal (i.e. high flow).
Effluent NH4+ concentrations generally surpassed influent con-
centrations, with the average ingoing and outgoing NH4+ concentra-
tions ranging from below detection limit (0.015 mg N L−1) to
0.4 mg N L−1. During the first operational year, NH4+ production was
intermittent; concentrations were of greater magnitude than during the
second year when effluent NH4+ concentrations remained compara-
tively stable and always greater than influent concentrations (Fig. 2c).
Hence, the bioreactor was a net source of low levels of NH4+.
Influent and effluent sulfate concentrations were on average 1132
and 1083 mg SO42− L−1, respectively. Based on the difference between
influent and effluent SO42− concentrations, there was a clearly-notable
loss of SO42− from the system between days 25–63 when effluent
SO42− concentrations were consistently lower than influent con-
centrations (Fig. 2d). In addition, a strong odor of H2S was noted during
the entire first operational year, indicating sulfate-reducing conditions
in the bioreactor:
1
SO42 − + C6 H12 O6 → H2 S + 2HCO3−
3 (2)
SO42− reduction is also noted between days 378 and 409 the second
year, but at a lesser magnitude. Unfortunately, analyses for H2S were
not performed during the study. Since the analytical uncertainty in
SO42− concentrations was ± 15% (LKAB, pers. comm.), the exact level
of H2S production cannot be calculated based solely on differences in
SO42− concentrations.
Effluent pH was consistently lower than influent pH with the lowest
pH observed following the start-up of the bioreactor (Fig. 2e), during
the first 100 days of operation. After start-up, the effluent pH increased
and stabilized with time, varying within the neutral to slightly basic pH
range. Effluent alkalinity was consistently greater than the influent al-
kalinity (on average 44.4 mg L−1) except for days 1–28 following the
start-up of the bioreactor when the system was alkalinity-consuming
(see Fig. 2f). Hence, the system went from alkalinity-consuming to al-
kalinity-producing at the same time as sulfate-reducing conditions were
initiated during the first year of operations (see Fig. 2d and f). Except
for the high alkalinity observed on day 322 (890 mg L−1) following the
winter intermission, maximum alkalinity concentrations were similar
for both years of operation. As the HRT was decreased to 24 h on day
429 (Table 1), there was an immediate decrease in alkalinity because of
the lower removal of NO3−.
3.2. Concentration changes along profiles
Fig. 3 shows the change in NO3−, NO2−, N2O(aq), and NH4+ con-
centrations as observed in the pore water between the inlet and outlet
on eleven different occasions during the operational period of the
bioreactor. Overall, NO3− is the major ionic nitrogen species in the
system, with exceptions being the relatively high NO2− and NH4+
concentrations in the beginning of the bioreactor operations (Fig. 3b, d,
and f). During the first year, NO3− removal to below the detection limit
was completed in close proximity to the bioreactor inlet, whereby
conditions became especially conducive for sulfate reduction (see
above).
The sequential production of NO2− after NO3− production shows
that NO2− is the product of microbiological NO3− reduction. In the
initial three profiles from the first year, the rate of change in NO3−
A. Nordström, R.B. Herbert
concentration (with distance from inlet) is comparatively high and
occurring simultaneously as peaks in NO2− concentrations are ob-
served. Elevated NO2− concentrations are often observed when glu-
cose, a fermentable carbon substrate, is used by denitrifiers as the
carbon substrate in comparison to less fermentable carbon substrates,
e.g. acetate (e.g. Akunna et al., 1994, 1993; Wilderer et al., 1987). The
decrease in observable NO2− in the bioreactor with time (Fig. 3) could
be the result of a change in the primary carbon substrate from a fer-
mentable to a less fermentable carbon substrate (e.g. from glucose to
acetate) used by the denitrifiers in the system. For example, Wilderer
et al. (1987) concluded that during fermentative conditions when glu-
cose was supplied as a carbon substrate, the growth of bacteria capable
of only reducing NO3− to NO2− was promoted, as opposed to those
capable of further reduction to N2O and N2, such that the increase in
NO2− concentrations was due to unbalanced rates of NO3− and NO2−
reduction (e.g. Betlach and Tiedje, 1981).
58
Ecological Engineering 110 (2018) 54–66
Fig. 2. Time series of effluent (open sym-
bols) and input (full symbols) concentra-
tions of nitrate (a), nitrite (b), ammonium
(c), sulfate (d), pH (e), alkalinity (f), and
TOC (g). Note broken x-axis, separating the
2015 data from 2016 data. Daily mean
temperature in (a) as full line with no sym-
bols.
Dissolved N2O concentrations are consistently low in comparison to
the magnitude of NO3− reduction, but the concentrations increase to-
ward the outlet of the bioreactor, indicating that N2O is the product of
sequential denitrification. The observed low concentrations of dissolved
N2O imply that N2O is fully reduced to N2 in the system, that N2O
leaves the system through surface outgassing, or that N2O is never
produced due to an accumulation of NO. In denitrifying woodchip
bioreactors, N2O emissions are mainly in dissolved form (e.g. Warneke
et al., 2011b). Further, any major NO accumulation from denitrification
is unlikely based on previous studies where low emissions of NO from
denitrification have been observed (e.g. Betlach and Tiedje, 1981;
McKenney et al., 1982). Therefore, except for the observed concentra-
tions of N2O in the bioreactor, any N2O produced by sequential deni-
trification is assumed to be fully reduced into N2.
Fig. 4 shows the change in TOC, alkalinity, and pH in the pore water
between the inlet and the outlet at the same eleven occasions as the
A. Nordström, R.B. Herbert
nitrogen profiles in Fig. 3. The observed TOC concentrations in the pore
water samples show that the TOC concentrations were generally in-
creasing in a non-linear relationship toward the outlet of the bioreactor
(see Fig. 4). This nonlinear increase in TOC is most evident in the
profiles from the first year of operations (left side graphs in Fig. 4) when
NO3− removal to below detection limits occurred near the inlet, leaving
sulfate reduction as the dominating electron-accepting process. Fol-
lowing the bioreactor start-up (day 1), alkalinity was fully consumed in
the pore water (Fig. 4b). With time, the zone of net alkalinity con-
sumption gradually moved towards the outlet (Fig. 4) as the system
became net alkalinity producing (see above). Bioreactor pH changed
simultaneously with alkalinity. The initial pH decrease in the reactor is
substantial with a decrease in ∼4.5 pH units within the first 5 m of the
reactor following the start-up (Fig. 4b), where after pH gradually in-
creased in the bioreactor (Fig. 4b, d, f, h).
59
Ecological Engineering 110 (2018) 54–66
Fig. 3. Nitrate, nitrite, ammonium, and ni-
trous oxide concentration profiles sampled
on days 330 (a), 1 (b), 365 (c), 22 (d), 400
(e), 57 (f), 428 (g), 85 (h), 456 (i), 113 (j),
and 477 (k). Profiles in the same row were
sampled at similar temperatures in the
bioreactor. Profiles in the left panel column
in Fig. 3 are from the first operational year
(days 1–150), while the profiles in the right
panel column are from the second opera-
tional year (days 322–486); profiles in the
same row were sampled at similar tem-
peratures (i.e. the same time of the year).
Note that N2O-N was not determined for (b)
and (d).
3.3. Surface gas emissions
Fig. 5a and b show the N2O(g) and CH4(g) surface emission rates
determined for each occasion of profile sampling.
The average observed N2O(g) surface emission rate for the full dura-
tion of the bioreactor operations was 1.75 ± 0.97 μg N m−2 min−1. The
emission rates appear to negatively correlate with the bioreactor tem-
perature (see Fig. 2a), with the highest emission rates observed at the
lowest temperatures and the lowest emission rates observed at the highest
temperatures (Fig. 5a).
The largest CH4(g) surface emission rates (average 20.81 ± 34.67
[μg m−2 min−1]) were observed in the beginning of the bioreactor
operations (during the start-up phase). Dissolved CH4 were generally
low (< 1 mg L−1), and generally increased toward the outlet of the
bioreactor, and decreased with time from bioreactor start-up (see Fig.
A. Nordström, R.B. Herbert
S2.1, supplementary material). CH4 is the product of methanogenesis,
and may occur through two general pathways that use H2 and CO2 or
acetate as electron donors (Appelo and Postma, 2005; Muyzer and
Stams, 2008) as represented by reactions (3) and (4) below:
CO2 + H2 → 2H2 O + CH4 (3)
CH3 COO− + H2 O → CH4 + HCO3− (4)
The observed decline in methanogenesis with time in this study (cf.
Fig. 5b) occurs simultaneous as an increase in pH and could be related
to a decline in CO2/H2 production from fermentation with time (see
discussion below).
Fig. 5c shows the N2O emission factor as a function of the total
NO3− removed in the bioreactor, calculated on a daily basis using in-
fluent and effluent concentration data (see Section 2.7). The N2O
emission factor relative to the amount of NO3− reduced was mainly
60
Ecological Engineering 110 (2018) 54–66
Fig. 4. Alkalinity and TOC concentration,
and pH, profiles sampled on days 330 (a), 1
(b), 365 (c), 22 (d), 400 (e), 57 (f), 428 (g),
85 (h), 456 (i), 113 (j), and 477 (k). Profiles
in the same row were sampled at similar
temperatures in the bioreactor. Profiles in
the left panel column in Fig. 3 are from the
first operational year (days 1–150), while
the profiles in the right panel column are
from the second operational year (days
322–486); profiles in the same row were
sampled at similar temperatures (i.e. the
same time of the year).
related to the magnitude of NO3− removal (Fig. 5c). Generally, when
the NO3− removal in the bioreactor was less than 10% in the bior-
eactor, all the reduced NO3− left the bioreactor as N2O (Fig. 5c), which
coincided with increased production of NH4+ and low temperatures
(< 5 °C) in the bioreactor (see Fig. 2c and a). N2O was predominantly
leaving the system in dissolved form, with an average of 93.8 ± 3.4%
(n = 8) of the emitted N2O being dissolved. The largest percentage N2O
emissions were observed on days 329 and 476 where they were de-
termined as 139 and 111%, respectively, of the reduced NO3−. More
than 100% production of N2O is most likely due to measurement errors,
or analytical uncertainty, as the nitrate removal was very low on these
days (see Fig. 2a).
3.4. Nitrogen transformation processes
According to our general understanding of nitrate removal in anoxic
A. Nordström, R.B. Herbert
Fig. 5. N2O (a) and CH4 (b) gas emissions from the bioreactor surface [μg m−2 min−1]
represented as an average of the gas flux determined in six spatially separated chambers.
Error bars represent one standard deviation. For day 56, the average estimated CH4
ranged between 99.4 ± 316.9 [μg m−2 min−1]. Figure (c) shows the percentage of re-
moved total NO3− emitted as N2O determined on a daily basis. Note log scale on axes in
(c).
systems (e.g. Burgin and Hamilton, 2007), the processes that may be re-
sponsible for the removal of NO3− in the bioreactor system are (1) deni-
trification, (2) dissimilatory nitrate reduction to ammonium/ammonia
(DNRA), (3) anaerobic ammonium oxidation (ANAMMOX), and/or (4)
biomass assimilation. In previous studies on woodchip bioreactors (Elgood
et al., 2010; Greenan et al., 2009, 2006; Moorman et al., 2010; Robertson
and Merkley, 2009; Schipper et al., 2010; Warneke et al., 2011a,b), de-
nitrification has been determined as the main pathway for NO3− removal
based on observed N2O production, enrichment of the 15N isotope, and the
generally increased abundances of functional genes pertaining to deni-
trification. The comparatively low production of NH4+ relative to the
amount of reduced NO3− have excluded DNRA as a major process for
NO3− removal in other studies. Further, NO3− immobilization by assim-
ilation in biomass has been inferred to be a minor process, estimated to be
within the ranges of ∼2.0-3.5% of the total NO3− removed (cf. Greenan
et al., 2009, 2006; Schipper et al., 2010). Additionally, ANAMMOX is
likely not a major pathway for NO3− removal in bioreactor systems
(Herbert et al., 2014; Schipper et al., 2010), mainly as it suppressed in
61
Ecological Engineering 110 (2018) 54–66
systems with high abundance of organic carbon (Burgin and Hamilton,
2007; González-Cabaleiro et al., 2015a).
The results of this current study support these previous findings: the
generally high level of NO3− removal in parallel with a relatively low
production of NO2−, N2O and NH4+, along with alkalinity production
(see above), strongly suggest that denitrification is the primary path for
nitrate removal. Overall, effluent alkalinity varied with bioreactor
temperature and HRT, which was due to the dependence of the deni-
trification kinetics on temperature and HRT (as HRT affects the time
available for reaction and thus the completeness of the reaction; see e.g.
Nordström and Herbert, 2017). This suggests that as the effluent pH and
alkalinity stabilized during the second year of operations, the sig-
nificance of denitrification as the sole major terminal electron-ac-
cepting process (TEAP) in the bioreactor increased with time. This oc-
curred primarily because of a suppression of secondary TEAPs such as
sulfate reduction and DNRA with time, which were more prominent
during the first operational year than during the second year. Never-
theless, ammonium production by DNRA is not an insignificant process.
The net stoichiometric definition of DNRA is shown in reaction (5)
using glucose as the carbon substrate/electron donor
1
NO3− + C6 H12 O6 + H2 O → NH4+ + 2HCO3−
3 (5)
In DNRA, NO3− is sequentially reduced to NO2−, and in turn to
NH4+/NH3. The DNRA pathway represents an undesirable pathway for
NO3− removal from an environmental perspective as NH4+ oxidation is
an oxygen-consuming process. The production of NH3 in the bioreactor
was insignificant as pH values were well below the pKa of NH4+ (9.3).
In sulfidic environments, or when there is a high availability of carbon
with respect to the available NOx (high C:N ratios), the significance of
DNRA in NOx cycling increases (see Burgin and Hamilton, 2007). The
type of carbon substrate may also affect the relative prevalence of
DNRA. For example, Akunna et al. (1993) observed significant pro-
duction of NH4+ relative to the amount of NO3− reduced when fer-
mentable carbon substrates (glucose and glycerol) were used in com-
parison to less fermentable carbon substrates (acetic acid, lactic acid,
methanol). It is likely that the relatively increased effluent NH4+ con-
centrations during the first year of operations (see Section 3.1) were
due more fermentable carbon substrates being used by the NOx cycling
microbial community, which is consistent with the discussion above on
NO2− in the profile samples (Section 3.2).
NH4+ may, in addition to DNRA, be produced by aerobic miner-
alization of organic matter (MOM). The stoichiometric definition of
MOM is exemplified in reaction (6) with C5H7O2N, an empirical com-
position of a prokaryotic cell (from Rittmann and McCarty, 2001), as
the electron donor
5O2 + C5 H7 O2 N + H2 O → NH4+ + HCO3− + 4H2 CO3 (6)
MOM is an aerobic process where biomass oxidation results in the
release of NH4+, while DNRA is an anaerobic process that produces
NH4+ under the simultaneous reduction of NO3−/NO2−. On a molar
basis, the observed production of NH4+ relative to the removal of NO3−
was on average 3.8%, leaving denitrification as the main process for
reductive removal of NO3−. However, NH4+ concentrations deviated
considerably from the average pattern during days 1–7 following the
start-up of the bioreactor. During that time, relatively high NH4+
concentrations were observed to occur simultaneously with compara-
tively low effluent concentrations of NO3− and NO2− (∼8–9% of re-
duced NO3−, Fig. 2a–c), suggesting DNRA. In addition, if MOM was
significant in the system, it would be expected that there would be a
relative increase in NH4+ close to the bioreactor inlet due to the pre-
ference of O2 over NO3− as an electron acceptor (Appelo and Postma,
2005, p. 439) as the oxygenized inlet water enters the bioreactor. This
is, however, not observed in any of the profiles sampled (Fig. 3). The
increase in NH4+ occurs closer to the outlet, which is most evident from
Fig. 3f, as NO3− concentrations were low, suggesting that NH4+
A. Nordström, R.B. Herbert
production was from DNRA. The sulfidic environment and the low
NO3− concentrations in the bioreactor, during the first year of ob-
servations (Fig. 3), would have resulted in conditions favorable for
DNRA (see above). Note, however, that the NH4+ production is clearly
suppressed during the period of clear SO42− reduction between days
25–63 (i.e. based on differences in influent and effluent concentrations,
see Fig. 2c). In addition to using SO42− as the electron acceptor, sulfate
reducers can also reduce NO3−/NO2− to NH4+ through DNRA (see
Dalsgaard and Bak, 1994; López-Cortés et al., 2006). This suggests that
the sulfate reducers in the system alternated their metabolism between
sulfate reduction and DNRA. During periods of low temperature and
low NO3− removal (days 113–126 and 455–486, Fig. 2a), and following
the winter intermission (days 325–329), the produced NH4+ relative to
the amount of reduced NO3− was ∼ 8-9% and ∼50%, respectively.
The increased relative production of NH4+ during these periods are,
however, due to the low absolute NO3− removal at low temperatures,
rather than an increased significance of DNRA.
3.5. Carbon cycling in bioreactor
For natural aquatic environments in general, it may be assumed that
anaerobic degradation processes are driven by the oxidation of carbo-
hydrates (Conrad, 1999). For denitrification, the carbon contained
within the lignocellulose material (primarily as cellulose) must first be
solubilized through hydrolysis before the compounds are available as
energy sources for heterotrophic denitrifying microorganisms
(Glombitza et al., 2015; Muyzer and Stams, 2008). Once solubilized,
different microbial functional groups may compete for the labile or-
ganic carbon, which may be transformed into new organic carbon
substrates by fermentation or sulfate reduction (see Muyzer and Stams,
2008), and/or be oxidized into inorganic carbon compounds, e.g.
H2CO3/CO2, HCO3−, and CO32−; this leads to characteristic changes in
pH and alkalinity. The changes in pH, alkalinity and TOC in the bior-
eactor pore water indicate that different biogeochemical processes are
controlling the consumption of organic carbon substrates at different
times during the two year of operations.
Electron donors that are not directly produced by the hydrolysis of
the carbon sources contained in the woody biomass must be produced
by secondary reactions, for example fermentation and sulfate reduction.
Woody biomass mainly consists of cellulose (40–55%), hemicellulose
(24–40%), lignin (18–35%) (Sun and Cheng, 2002), and additional
extraneous substances such as organic extractives and ash/inorganic
minerals (see Fan et al., 1982). Typical hydrolysates of cellulose are
reducing sugars (e.g. glucose). Hemicellulose is more easily hydrolyzed
than cellulose due to their branched structure, and hemicellulose hy-
drolysates are xylose (C5H10O5), acetic acid (CH3COOH), and various
isomers of glucose (C6H12O6), e.g. galactose, and mannose (Palmqvist
and Hahn-Hägerdahl, 2000). From the above discussion, glucose is
assumed as the primary carbon substrate in the bioreactor system,
originating from the woodchips.
The effluent alkalinity generally followed the completeness of NO3−
removal in the system (see Fig. 2a and f); this is supported by the
stoichiometry of denitrification (reaction (1)), which is interpreted as
the dominating process for NO3− removal (see above). However,
during the first 80 days after the start-up of the bioreactor, alkalinity
and pH consistently increased, with alkalinity consumption being ob-
served immediately after the start-up. Since none of the biogeochemical
reactions that are indicated as being active in the bioreactor (i.e deni-
trification, DNRA, and sulfate reduction) are alkalinity-consuming, the
observed initial alkalinity consumption must have been due to some
secondary process, possibly fermentation. For example, during the
fermentation of glucose to various volatile fatty acids (e.g. valerate,
butyrate, propionate, acetate, and formate) there is a consumption of
HCO3− and a production of H2CO3 (see table S3.1, Supplementary
materials). Fermentation is an alkalinity-consuming and acidifying
process as long as there is an imbalance between the production of
62
Ecological Engineering 110 (2018) 54–66
organic carbon and its consecutive consumption by heterotrophic
electron-accepting processes (see table S3.1, Supplementary materials).
This is consistent with the observations made following the start-up of
the bioreactor when comparatively low pH, low alkalinity, and high
concentrations of TOC were observed in the bioreactor effluent (see
Fig. 2e–g). It should be noted that the initial increase in effluent TOC
concentrations is commonly seen in woodchip bioreactors, and is
usually explained as a “flush-out” of soluble organic carbon (e.g.
Hoover et al., 2016). However, the observed drop in pH and alkalinity
consumption in this study suggest a fermentative signature on the TOC
production. Alternatively, the increase in TOC under sulfidic conditions
(see Section 3.2) may indicate incomplete oxidation of high molecular
weight organic carbon substrates (e.g. glucose, butyrate) to acetate by
sulfate reducers (Muyzer and Stams, 2008). However, the simultaneous
alkalinity consumption and pH decrease during the production of TOC
suggests fermentation.
The product spectrum (i.e. the distribution of produced species of
organic carbon substrates) during anaerobic fermentation of glucose
(the primary carbon substrate in the bioreactor, see above) has been
observed to shift from mainly butyrate (C3H7COO−) and acetate
(CH3COO−) at low pH ranges (∼4–6.5) to mainly ethanol
(CH3CH2OH), propionate (C2H5COO−), and acetate at higher pH
(∼6.5–8.5) (see Fang and Liu, 2002; Horiuchi et al., 2002; Temudo
et al., 2008, 2007). Temudo et al. (2007) additionally observed a
transition from CO2/H2 to formate production during the anaerobic
fermentation of glucose as pH increased (range 4.0–8.5). The shift in
product spectrum during anaerobic fermentation of glucose has further
been confirmed in a thermodynamic modeling framework (see
González-Cabaleiro et al., 2015a). This suggests that the biogeochem-
ical changes observed in the bioreactor with time, particularly between
days 31 and 63 when effluent pH increased from 6 to 7 (see Fig. 2e),
were due to a change in the organic carbon substrates utilized by the
biogeochemical processes. Indeed, the observed elevated NH4+ and
NO2− concentrations in the bioreactor following the bioreactor start-up
could be related to the use of more fermentable carbon substrates (see
Section 3.2), e.g. butyrate. Further, the decline in methane production
could be related to a decline in CO2/H2 production (see Section 3.3),
the primary substrates for methanogenesis, from fermentation of glu-
cose.
4. Mass balance modeling
To test if the inferred biogeochemical processes (see Sections 3.4
and 3.5) indeed were responsible for the observation reduction/pro-
duction of NOx, NH4+, SO42−, and TOC, mass balance modeling was
used to predict the observed pH and alkalinity changes in the bioreactor
based on stoichiometric definitions of the respective biogeochemical
processes. Methanogenesis was excluded in the modeling as it was
considered insignificant in relation to the other biogeochemical pro-
cesses in the bioreactor. The observed alkalinity was assumed to en-
tirely consist of HCO3−. The mass changes of NO3−, NO2−, NH4+,
SO42−, and TOC per liter were calculated from the difference in influent
and effluent concentrations. From these, the production/consumption
of HCO3− and H2CO3* were calculated by propagating the calculated
mass changes through stoichiometric definitions of the inferred bio-
geochemical reactions as discussed above. The modeled production of
alkalinity (HCO3−) and carbonic acid (H2CO3*) were used to calculate
the resulting pH based on the following equilibrium reaction with a pKa
value of 6.3:
H2 CO3* ↔ H+ + HCO3− (7)
In EQ. (7), H2CO3* represents the combination of carbonic acid
(H2CO3) and dissolved carbon dioxide (CO2). For the pH ranges observed
in the bioreactor effluent (5.0–7.9, Fig. 2e), the CO32− concentration was
insignificant and it is assumed that the relative concentrations of HCO3−
(alkalinity) to H2CO3* controlled the pH in the bioreactor.
A. Nordström, R.B. Herbert
4.1. Stoichiometric definitions of the biogeochemical reactions
4.1.1. Selection of electron donor
The production/consumption of HCO3− and H2CO3* is dependent
on the electron donor (e.g. glucose, acetate, sulfides) utilized by the
biogeochemical reactions (see table S4.1 in Supplementary materials).
Glucose was inferred as the major electron donor in the bioreactor
system (see above). The net stoichiometric definitions of the secondary
reactions producing electron donors from hydrolysates, and the con-
secutive consumption of the secondary electron donors by e.g. deni-
trification, condenses into direct utilization of hydrolysates (provided
balanced production and consumption of secondary electron donors).
This is shown in section S3 in Supplementary materials. This justifies
the use of primary hydrolysates in the stoichiometric definitions of the
biogeochemical processes operating the bioreactor. However, it does
not necessarily imply that glucose is the actual carbon substrate used by
e.g. denitrification, merely that the net stoichiometric definition of the
biogeochemical processes condenses into glucose utilization. In fact,
fermentative coupling may be an energetic advantage (see González-
Cabaleiro et al., 2015b), such that the fermentation of glucose may
provide additional carbon substrates than glucose to the biogeochem-
ical processes in the bioreactor.
4.1.2. Stoichiometric definitions of denitrification and DNRA
The production of HCO3− and H2CO3 from denitrification was
modeled assuming full reduction of NO and/or N2O, but with the po-
tential for NO2− accumulation. The observed reduction in NO3− was
propagated step-wise in reactions (8.1)–(8.4).
1 1
NO3− + C6 H12 O6 → NO2− + H2 CO3
12 2 (8.1)
1 3 1 In reaction (11), the parameter α determines the quotient between
NO2− + C6 H12 O6 + H2 CO3 → NO + HCO3− + H2 O
24 4 2 (8.2)
1 1 1
NO + C6 H12 O6 → N2 O + H2 CO3
24 2 4 (8.3)
1 1 1 1
N2 O + C6 H12 O6 → N2 + H2 CO3
2 24 2 4 (8.4)
The observed production of NH4+was assumed to come from the
reduction of NO2− through the DNRA pathway (reaction (9)). The
produced NH4+ was subtracted from the theoretical production of
NO2− (reaction (8.1)), accounting for any observable net accumula-
tion/reduction of NO2− in the bioreactor system (Fig. 2b).
1 1
NO2− + C6 H12 O6 + H2 CO3 + H2 O → NH4+ + 2HCO3−
4 2 (9)
From the observed production of NH4+,the production and con-
sumption of HCO3− and H2CO3 was modeled using reaction (9). In
addition to the carbonate system, denitrification and DNRA involes the
acid-base pairs of NO2−/HNO2 (pKa 3.4) and NH4+/NH3 (pKa 9.3).
From the observed pH range in the bioreactor (see above), the pro-
duction of HNO2 and/or NH3 should be insignificant and is therefore
disregarded in the model.
4.1.3. Stoichiometric definition of sulfate reduction
The production of HCO3− and H2CO3 from heterotrophic sulfate
reduction was modeled assuming that all the observed reduction in
SO42− in the bioreactor could be attributed to sulfate reduction. In
addition to the carbonate system, heterotrophic sulfate reduction in-
volves the acid-base pair HS−/H2S (pKa 7.04), which is active in the pH
range observed in the bioreactor (see above). The quotient of produced
HS−/H2S should change during the course of bioreactor operations,
being mainly H2S prior to day 63 (pH < 7) and HS− post day 63
(pH > 7). However, since the effluent pH in the bioreactor on average
was 7.1 which is close to the pKa of H2S, sulfate reduction was modeled
with an equal production of H2S and HS− for simplicity (reaction (10)).
63
Ecological Engineering 110 (2018) 54–66
1 1 1 3 1
SO42 − + C6 H12 O6 → HS − + H2 S + HCO3− + H2 CO3
3 2 2 2 2 (10)
From the above discussion, the observed TOC production was due to
fermentation, wherefore the potential of acetate production from sul-
fate reduction was disregarded. Nevertheless, the stoichiometric defi-
nition of incomplete glucose oxidation to acetate by sulfate reduction
expands into separate stoichiometric definitions of sulfate reduction
and fermentation of glucose to acetate (see section S5, Supplementary
materials).
4.1.4. Stoichiometric definition of fermentation
From the above discussion, fermentation is a major process in the
system following the bioreactor start-up. Observations of elevated
NO2− and NH4+ concentrations following the bioreactor start-up sug-
gest that fermentable carbon substrates are used (e.g. butyrate) in
preference over less fermentable carbon substrates (e.g. acetate) (see
Section 3.2). The molar consumption and production of HCO3− and
H2CO3, respectively, from the fermentation of glucose is dependent on
the type of carbon substrate produced (see table S3.1, Supplementary
materials). Since the speciation of produced organic carbon in the
system was not determined, the stoichiometry of fermentation was
based on the observations by Fang and Liu (2002), Horiuchi et al.
(2002), and Temudo et al. (2008, 2007). It was hence assumed that
fermentation of glucose resulted in the production of a mixture of
(mainly) butyrate and acetate (reaction (11)).
9α 6α
C6 H12 O6 + ⎛3 − ⎞ HCO3− → ⎛ ⎞ C3 H7 COO− + (3 − 3α ) CH3 COO−
⎝ 5 ⎠ ⎝ 5 ⎠
3α
+ ⎛3 − ⎞ H2 CO3
⎝ 5 ⎠ (11)
butyrate and acetate production from anaerobic glucose fermentation.
If α = 1, then the inferred TOC production from the anerobic fermen-
tation of glucose purely results in the production of butyrate. If α = 0,
then acetate is the single organic carbon substrate produced. The ob-
served molar production of organic carbon (TOC) was related to the
consumption and production of HCO3− and H2CO3, respectively, per
mol organic carbon formed as predicted from reaction (11).
The parameter α was randomly sampled between 0 and 1 from a
uniform distribution 1000 times. The optimal values of α were selected
based on the absolute residuals between the observed and simulated pH
and alkalinities, respectively. Since different values of α were “optimal”
with respect to the observed alkalinity and pH, the selected value is
represented as the average between these two.
4.1.5. Accounting for retention time in the system and uncertainty in
concentrations
The influent NO3−, NO2−, NH4+, and SO42− concentrations were
linearly interpolated on a daily basis between days of observation. The
observed effluent concentrations (on day X) were then matched with
the influent concentration that was interpolated/observed on day X-
HRT (HRT in days). This was done in order to account for the variability
in influent concentrations, and the delay between influx and outflux to/
from the system.
Due to the large uncertainties in the determined sulfate concentra-
tions (see above), the occurrence and magnitude of sulfate reduction is
highly uncertain. Because of this, pH and alkalinity were first modeled
only accounting for heterotrophic denitrification and DNRA (scenario
1). Uncertainties in NO3−, NO2−, and NH4+ concentrations were not
accounted for, and the produced HCO3− was considered the only
contributor to alkalinity. Sulfate reduction was then added to the re-
actions in scenario 1 (scenario 2). Sulfate reduction was only allowed to
occur whenever there was an observed reduction of 30 mg L−1 SO42−
between influent and effluent data. Fermentation was added to scenario
2 in scenario 3.
A. Nordström, R.B. Herbert
4.1.6. CO2 outgassing
The stoichiometric definitions of the biogeochemical processes used
in the modeling implicitly assumes no CO2 outgassing from the bior-
eactor. CO2 surface fluxes from the bioreactor were not determined in
this study, but have nevertheless been observed in previous woodchip
bioreactor studies (e.g. Elgood et al., 2010; Warneke et al., 2011a). CO2
outgassing would lead to a pH increase without affecting the alkalinity
(HCO3−); however, we believe that (natural) CO2 outgassing is not
substantial in the bioreactor in this study as: (1) the covering till layer
creates confining conditions; and (2) natural waters and groundwaters
are often observed to be supersaturated with respect to atmospheric
CO2. CO2 partial pressures (pCO2) in groundwater are usually ∼10-100
times higher than the atmosphere (Macpherson, 2009), and boreal
streams, rivers, and lakes are often observed to be supersaturated with
respect to atmospheric CO2, with pCO2 ranging between
∼300–5500 μatm (Roehm et al., 2009; Teodoru et al., 2009), sug-
gesting slow/incomplete CO2 outgassing.
4.2. Biogeochemical evolution of the bioreactor
4.2.1. Simulated pH and alkalinity
Scenario 1 is generally predicting a pH below the observed pH in the
bioreactor effluent (see Fig. 6a). However, prior to day 63 the simulated
pH falls above the observed (see Fig. 6a).
Scenario 1 is generally able to predict the alkalinity changes post
day 63 in the bioreactor effluent (see Fig. 6b). Prior to day 63 as the
observed pH falls below 7, the proposed biogeochemical reactions
predict a too high alkalinity (HCO3−) production. Post day 63, Scenario
1 is able to account for approximately all of the alkalinity production
except during the high temperature periods when the predicted HCO3−
is too low. During the high temperature periods, NO3− reduction was
complete in the bioreactor (see Fig. 2a), triggering sulfidic conditions in
the bioreactor. Sulfate reduction was added to scenario 1 in scenario 2,
and from Fig. 6b it is seen that sulfate reduction is able to compensate
for the deficiency in alkalinity production from scenario 1 during the
high temperature periods. From Fig. 6b it is also seen that scenario 2
generally overpredicts alkalinity production during periods of TOC
production (see Fig. 2g), particulary during the first operational year
(Fig. 6b). This suggests that during these periods, fermentation was
active. Fermentation (reaction (11)) was added to scenario 2 in scenario
3. As seen in Fig. 6a, scenario 3 is able to predict the effluent pH from
the bioreactor, under the assumption that close to all of the produced
TOC is in the form of acetate (i.e. α in reaction (11) is close to 0). At the
pH ranges observed in the bioreactor prior to day 63 (Fig. 6a), the
studies by Fang and Liu (2002), Horiuchi et al. (2002), and Temudo
et al. (2008, 2007) suggests a production of butyrate and acetate from
glucose fermentation. Scenario 3 suggests no net production of buty-
rate, which could imply that butyrate is consumed by the biogeo-
chemical processes in the bioreactor. Indeed, if butyrate was the
64
Ecological Engineering 110 (2018) 54–66
preferred carbon substrate following the start-up of the bioreactor, then
it should not be traceable in the bioreactor effluent. With time as pH
increases in the bioreactor, scenario 3 predicts a step-like change in α
(reaction (11)) to 1, implying that fermentation almost only produces
butyrate. The simulated change in primary glucose fermentation pro-
ducts from acetate to butyrate with time indicates that the preferred
carbon substrate of the TEAP biogeochemical processes (e.g. deni-
trification) changed from butyrate to acetate/ethanol with time. The
change in the preferred carbon substrate of e.g. denitrification is sug-
gested to occur in response to the relative availability of different
carbon substrates at different pH (see e.g. Temudo et al., 2008,2007),
which ultimately is explained by a maximization of the energy harvest
rate in the system (see González-Cabaleiro et al., 2015a). Indeed, the
metabolic energy harvest rate has been identified as a selective force of
microbial activities in competitive systems (see González-Cabaleiro
et al., 2015b).
5. Conclusions
The woodchip bioreactor in this study was tested as a potential method
to remove nitrate (NO3−) from mine and process water originating from
the Kiruna iron ore mine (northern Sweden). The test period extended for
the duration of two years. The results show that the bioreactor was able to
reduce an average NO3− concentration of 22 mg N L−1 to below the de-
tection limit (0.06 mg N L−1) at temperatures above 5 °C. The (theore-
tical) hydraulic residence time required for this was 1.9–2.6 days. The
main NO3− removing mechanism at temperatures above 5 °C was com-
plete denitrification. As the bioreactor temperatures decreased to below
5 °C, dissimilatory nitrate reduction to ammonium (DNRA) and in-
complete denitrification to nitrite (NO2−) or nitrous oxide gas (N2O) in-
creased in significance, resulting in a substantial production of ammonium
(NH4+) and N2O relative to the amount of NO3− reduced in the system.
Following bioreactor start-up, sulfate reduction (as noted by H2S odor) and
the production of organic carbon from fermentation were substantial.
Methanogenesis, resulting in the production of methane (CH4), was ob-
servable yet insignificant in relation to the other biogeochemical processes
operating in the system.
With time the diversity of biogeochemical processes controlling
major redox-active species decreased: sulfate reduction and fermenta-
tion decreased in importance with time from bioreactor start-up,
leaving denitrification as the sole major terminal electron accepting
process. The decrease in biogeochemical diversity occurred simulta-
neously as the decline in NO2−, NH4+, and TOC production, and an
increase and stabilization of the effluent pH and alkalinity. Based on a
stoichiometric mass balance confined by observed changes in pH and
alkalinity in the bioreactor, the trigger of the decline in biogeochemical
diversity is suggested to have been a change in the carbon substrate
utilized by the nitrate and sulfate-reducing communities from primarily
butyrate to acetate, both provided by the fermentation of glucose.
Fig. 6. Modeled and observed pH (a) and alkalinity
(b) for scenario 1, 2, and 3. Denitrification and
DNRA was considered in scenario 1, sulfate reduc-
tion was added to scenario 1 in scenario 2, and fer-
mentation was added to scenario 2 in scenario 3.
A. Nordström, R.B. Herbert
Conflicts of interest
None.
Acknowledgements
This work was supported by the Luossavaara-Kiirunavaara
Aktiebolag (LKAB), Boliden Aktiebolag (Boliden AB), and VINNOVA,
The Swedish Innovation Agency, under Grant 2014-01134. The work
presented herein is part of the larger project “miNing – Reduction of
nitrogen discharges in mining processes and mitigating its environ-
mental impact”, a collaboration between LKAB, Boliden AB, Rock Tech
Centre, the Swedish University of Agricultural Sciences, Luleå
University of Technology, and Uppsala University.
Special thanks are extended to Johan Johansson for assistance
during the construction of the bioreactor, Christopher Jones and
Joachim Audet for discussions and help during planning and prepara-
tions of gas sampling, and to two anonymous reviewers for their va-
luable comments on the manuscript.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.ecoleng.2017.09.018.
References
Akunna, J.C., Bizeau, C., Moletta, R., 1993. Nitrate and nitrite reductions with anaerobic
sludge using various carbon sources: glucose, glycerol, acetic acid, lactic acid and
methanol. Water Res. 27, 1303–1312. http://dx.doi.org/10.1016/0043-1354(93)
90217-6.
Akunna, J.C., Bizeau, C., Moletta, R., 1994. Nitrate reduction by anaerobic sludge using
glucose at various nitrate concentrations: ammonification, denitrification and me-
thanogenic activities. Environ. Technol. 15, 41–49. http://dx.doi.org/10.1080/
09593339409385402.
Appelo, C.A.J., Postma, D., 2005. Geochemistry, Groundwater and Pollution, Second ed.
A.A. Balkema, Leiden.
Betlach, M.R., Tiedje, J.M., 1981. Kinetic explanation for accumulation of nitrite, nitric
oxide, and nitrous oxide during bacterial denitrification. Appl. Environ. Microb. 42,
1074–1084.
Breibarth, E., Mills, M.M., Friedrichs, G., LaRoche, J., 2004. The Bunsen gas solubility
coefficient of ethylene as a function of temperature and salinity and its importance
for nitrogen fixation assays. Limnol. Oceanogr-Method 2, 282–288. http://dx.doi.
org/10.4319/lom.2004.2.282.
Burgin, A.J., Hamilton, S.K., 2007. Have we overemphasized the role of denitrification in
aquatic ecosystems? A review of nitrate removal pathways. Front. Ecol. Environ. 5,
89–96. http://dx.doi.org/10.1890/1540-9295(2007)5[89:HWOTRO]2.0.CO;2.
Conrad, R., 1999. Contribution of hydrogen to methane production and control of hy-
drogen concentrations in methanogenic soils and sediments. FEMS Microbiol. Ecol.
28, 193–202. http://dx.doi.org/10.1111/j.1574-6941.1999.tb00575.x.
Dalsgaard, T., Bak, F., 1994. Nitrate reduction in a sulfate-reducing bacterium, desulfo-
vibrio desulfuricans, isolated from rice paddy soil sulfide inhibition, kinetics, and
regulation. Appl. Environ. Microb. 60, 291–297.
EPA, 2013. Aquatic Life Ambient Water Quality Criteria for Ammonia –freshwater. Report
EPA 822-R-13-001. United States Environmental Protection Agency, Washington, DC.
Elgood, Z., Robertson, W.D., Schiff, S.L., Elgood, R., 2010. Nitrate removal and green-
house gas production in a stream-bed denitrifying bioreactor. Ecol. Eng. 36,
1575–1580. http://dx.doi.org/10.1016/j.ecoleng.2010.03.011.
Fan, L.T., Lee, Y.H., Gharpuray, M.M., 1982. The nature of lignocellulosics and their
pretreatment for enzymatic hydrolysis. Adv. Biochem. Eng. 23, 158–187.
Fang, H.H.P., Liu, H., 2002. Effect of pH on hydrogen production from glucose by a mixed
culture. Bioresour. Technol. 82, 87–93. http://dx.doi.org/10.1016/S0960-8524(01)
00110-9.
Forsyth, B., Cameron, A., Miller, S., 1995. Explosives and water quality. In: Proceedings of
Sudbury ’95 Mining and the Environment. Montreal, Quebec, Canada: MEND (Mine
Environment Neutral Drainage). pp. 795–803.
Glombitza, C., Jaussi, M., Røy, H., Seidenkrantz, M.S., Lomstein, B.A., Jørgensen, B.B.,
2015. Formate, acetate, and propionate as substrates for sulfate reduction in sub-
arctic sediments of Southwest Greenland. Front. Microb. 6, 846. http://dx.doi.org/
10.3389/fmicb.2015.00846.
González-Cabaleiro, R., Ofiţeru, I.D., Lema, J.M., Rodríguez, J., 2015a. Microbial cata-
bolic activities are naturally selected by metabolic energy harvest rate. ISME J. 9,
2630–2641. http://dx.doi.org/10.1038/ismej.2015.69.
González-Cabaleiro, R., Lema, J.M., Rodríguez, J., 2015b. Metabolic energy-based mod-
elling explains product yielding in anaerobic mixed culture fermentations. PLoS One
10, 1–17. http://dx.doi.org/10.1371/journal.pone.0126739.
Greenan, C.M., Moorman, T.B., Kaspar, T.C., Parkin, T.B., Jaynes, D.B., 2006. Comparing
65
Ecological Engineering 110 (2018) 54–66
carbon substrates for denitrification of subsurface drainage water. J. Environ. Qual.
35, 824–829. http://dx.doi.org/10.2134/jeq2005.0247.
Greenan, C.M., Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., 2009.
Denitrification in wood chip bioreactors at different water flows. J. Environ. Qual. 38,
1664–1671. http://dx.doi.org/10.2134/jeq2008.0413.
Herbert, R.B., Winbjörk, H., Hellman, M., Hallin, S., 2014. Nitrogen removal and spatial
disitribution of denitrifier and anammox communities in a bioreactor for mine
drainage treatment. Water Res. 66, 350–360. http://dx.doi.org/10.1016/j.watres.
2014.08.038.
Hoover, N.L., Bhandari, A., Soupir, M.L., Moorman, T.B., 2016. Woodchip denitrification
bioreactors: impact of temperature and hydraulic retention time on nitrate removal.
J. Environ. Qual. 45, 803–812. http://dx.doi.org/10.2134/jeq2015.03.0161.
Horiuchi, J.I., Shimizu, T., Tada, K., Kanno, T., Kobayashi, M., 2002. Selective production
of organic acids in anaerobic acid reactor by pH control. Bioresour. Technol. 82,
209–213. http://dx.doi.org/10.1016/S0960-8524(01)00195-X.
Komsta, L., 2012a. Chemometric and statistical evaluation of calibration curves in
pharmaceutical analysis: a short review on trends and recommendations. J. AOAC
Int. 95, 669–672.
Komsta, L., 2012b. Quantchem: Quantitative Chemical Analysis: Calibration and
Evaluation of Results. R Package Version 0.13. http://CRAN.R-project.org/
package=quantchem.
López-Cortés, A., Fardeu, M.L., Fauque, G., Joulian, C., Ollivier, B., 2006. Reclassification
of the sulfate- and nitrate-reducing bacterium Desulfovibrio vulgaris subsp. Oxamicus
as Desulfovibrio oxamicus sp. Nov., comb. Nov. Int. J. Syst. Evol. Microbiol. 69,
1495–1499. http://dx.doi.org/10.1099/ijs.0.64074-0.
LKAB, 2016. Miljörapport LKAB Kiruna [Environmental Report LKAB Kiruna 2015].
Luossavaara-Kiirunavaara AB (in Swedish).
Lindeström, L., 2012. Kväveutsläpp från gruvindustrin: Risker för miljöproblem, krav på
utsläppsbegränsningar och möjliga åtgärder [Nitrogen releases from the mining in-
dustry: risks for environmental problems, requirements for release limits, and pos-
sible solutions]. SveMin. (in Swedish), Stockholm.
Macpherson, G.L., 2009. CO2 distribution in groundwater and the impact of groundwater
extraction on the global C cycle. Chem. Geol. 264, 328–336. http://dx.doi.org/10.
1016/j.chemgeo.2009.03.018.
McKenney, D.J., Shuttleworth, K.F., Vriesacker, J.R., Findlay, W.I., 1982. Production and
loss of nitric oxide from denitrification in anaerobic brookston clay. Appl. Environ.
Microb. 43, 534–541.
Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., 2010. Denitrification activity,
wood loss, and N2O emissions over 9 years from a wood chip bioreactor. Ecol. Eng.
36, 1567–1574. http://dx.doi.org/10.1016/j.ecoleng.2010.03.012.
Muyzer, G., Stams, A.J.M., 2008. The ecology and biotechnology of sulphate-reducing
bacteria. Nat. Rev. Microbiol. 6, 441–454. http://dx.doi.org/10.1038/nrmicro1892.
Nordström, A., Herbert, R., 2017. Denitrification in a low temperature bioreactor system
at two different hydraulic residence times: laboratory column studies. Environ.
Technol. 38, 1362–1375. http://dx.doi.org/10.1080/09593330.2016.1228699.
Palmqvist, E., Hahn-Hägerdahl, B., 2000. Fermentation of lignocellulosic hydrolysates. II:
inhibitors and mechanisms of inhibition. Bioresour. Technol. 74, 25–33. http://dx.
doi.org/10.1016/S0960-8524(99)00161-3.
Parkin, T.B., Venterea, R.T., 2010. Chamber-based trace gas flux measurements. In:
Follett, R.F. (Ed.), Sampling Protocols. United States Department of Agriculture, Fort
Collins 3.1–3.39.
Pedersen, A.R., 2015. HMR: Flux Estimation with Static Chamber Data. R Package Version
0.4.1. http://CRAN.R-project.org/package=HMR.
R Core Team, 2017. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org.
Revey, G., 1996. Practical methods to control explosives losses and reduce ammonia and
nitrate levels in mine water. Min. Eng. Tech. Pap. 61–64.
Rittmann, B.E., McCarty, P., 2001. Biotechnology: Principles and Applications. McGraw-
Hill Book Co., New York.
Robertson, W.D., Merkley, L.C., 2009. In-Stream bioreactor for agricultural nitrate
treatment. J. Environ. Qual. 38, 230–237. http://dx.doi.org/10.2134/jeq2008.0100.
Robertson, W.D., 2010. Nitrate removal rates in woodchip media of varying age. Ecol.
Eng. 36, 1581–1587. http://dx.doi.org/10.1016/j.ecoleng.2010.01.008.
Roehm, C.L., Prairie, Y.T., del Giorgio, P.A., 2009. The pCO2 dynamics in lakes in the
boreal region of northern Québec, Canada. Global Biogeochem. Cy. 23, 1–9. http://
dx.doi.org/10.1029/2008GB003297.
SMHI, 2017. Swedish Meteorological and Hydrological Institute. https://www.smhi.se/
klimatdata/meteorologi/temperatur (Accessed 17.06.21) (in Swedish).
Schipper, L.A., Robertson, W.D., Gold, A.J., Jaynes, D.B., Cameron, S.C., 2010.
Denitrifying bioreactors –an approach for reducing nitrate loads to receiving waters.
Ecol. Eng. 36, 1581–1587. http://dx.doi.org/10.1016/j.ecoleng.2010.04.008.
Sun, Y., Cheng, J., 2002. Hydrolysis of lignocellulosic materials for ethanol production: a
review. Bioresour. Technol. 83, 1–11. http://dx.doi.org/10.1016/S0960-8524(01)
00212-7.
Temudo, M.F., Kleerebezem, R., van Loosdrecht, M., 2007. Influence of the pH on (Open)
mixed culture fermentation of glucose: a chemostat study. Biotechnol. Bioeng. 98,
69–79. http://dx.doi.org/10.1002/bit.21412.
Temudo, M.F., Muyzer, G., Kleerebezem, R., van Loosdrecht, M.C.M., 2008. Diversity of
microbial communities in open mixed culture fermentations: impact of the pH and
carbon source. Appl. Microbiol. Biotechnol. 80, 1121–1130. http://dx.doi.org/10.
1007/s00253-008-1669-x.
Teodoru, C.R., del Giorgio, P.A., Prairie, Y.T., Camire, M., 2009. Patterns in pCO2 in
boreal streams and rivers of northern Quebec, Canada. Global Biogeochem. Cy.
23http://dx.doi.org/10.1029/2008GB003404. GB2012.
Torstensson, B.A., 1984. A new system for ground water monitoring. Ground Water
Monit. Rem. 4, 131–138. http://dx.doi.org/10.1111/j.1745-6592.1984.tb00904.x.
A. Nordström, R.B. Herbert
Warneke, S., Schipper, L.A., Matiasek, M.G., Cameron, S., Bruesewitz, D.A., McDonald,
I.R., 2011a. Nitrate removal, communities of denitrifiers and adverse effects in dif-
ferent carbon substrates for use in denitrification beds. Water Res. 45, 5463–5475.
http://dx.doi.org/10.1016/j.watres.2011.08.007.
Warneke, S., Schipper, L.A., Bruesewitz, D.A., McDonald, I., Cameron, S., 2011b. Rates,
controls and potential adverse effects on nitrate removal in a denitrification bed.
66
Ecological Engineering 110 (2018) 54–66
Ecol. Eng. 37, 511–522. http://dx.doi.org/10.1016/j.ecoleng.2010.12.006.
Wilderer, P.A., Jones, W.L., Dau, U., 1987. Competition in denitrification systems af-
fecting reduction rate and accumulation of nitrite. Water Res. 21, 239–245. http://
dx.doi.org/10.1016/0043-1354(87)90056-X.
Wilhelm, E., Battino, R., Wilcock, R.J., 1977. Low-pressure solubility of gases in liquid
water. Chem. Rev. 77, 219–261. http://dx.doi.org/10.1021/cr60306a003.