Page 1 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
1
Temperature and Substrate Controls Woodchip Bioreactor Performance in
2
Reducing Tile Nitrate Loads in East-central Illinois
3
Mark B. David1, Lowell E. Gentry1, Richard A. Cooke2, and Stephanie M. Herbstritt2
4
1
5
Hall, 1102 S. Goodwin Ave., Urbana, IL 61801
6
2
7
Engineering Building, 1304 W. Pennsylvania Ave., Urbana, IL 61801
8
*Corresponding author ([email protected])
9
Abstract
University of Illinois, Dep. of Natural Resources and Environmental Sciences, W503 Turner
University of Illinois, Dep. of Agricultural and Biological Engineering, 338 Agricultural
10
Tile drainage is the major source of nitrate in the upper Midwest, and end-of-tile removal
11
techniques such as wood chip bioreactors have been installed that allow current farming
12
practices to continue, with nitrate removed through denitrification. There have been few multi-
13
year studies of bioreactors examining controls on nitrate removal rates. We evaluated the nitrate
14
removal performance of two wood chip bioreactors during the first 3 years of operation and
15
examined the major factors that regulated nitrate removal. Bioreactor 2 was subject to river
16
flooding and performance was not assessed. Bioreactor 1 had average monthly nitrate removal
17
rates of 23 to 44 g N m-3 d-1 in year 1, which decreased to 1.2 to 11 g N m-3 d-1 in years 2 and 3.
18
The greater N removal rates in year 1 and early in year 2 was likely due to highly degradable C
19
in the woodchips. Only late in year 2 and in year 3 was there a strong temperature response in the
20
nitrate removal rate. Less than 1% of the nitrate removed was emitted as N2O. Due to large tile
21
inputs of nitrate (729 to 2127 kg N) at high concentrations (~30 mg nitrate-N L-1) in years 2 and
22
3, overall removal efficiency was low (3 and 7% in years 2 and 3, respectively). Based on a
23
process-based bioreactor performance model, bioreactor 1 would have needed to be 9 times as
24
large as the current system to remove 50% of the nitrate load from this 20 ha field.
1
Page 2 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
25
Introduction
26
Nitrate loss to streams and rivers from tile drained corn (Zea mays L.) and soybean (Glycine max
27
(L.) Merr.) fields in the upper Midwest of the United States is a well-established environmental
28
issue (e.g., David et al., 2010). Many in-field and edge-of-field techniques have been proposed
29
and evaluated for reducing these nitrate losses, including fertilizer management, cover crops,
30
perennial crops, water table management, constructed wetlands, buffer strips, two-stage ditches,
31
saturated buffers, and bioreactors (e.g., USEPA, 2008; Schipper et al., 2010b; Skaggs et al.,
32
2012; Jaynes and Isenhart, 2014; David et al., 2015; Groh et al., 2015). Effectiveness of these
33
practices has been found to be highly variable, both from location to location as well as
34
temporally due to differing tile flow amounts and nitrate concentrations. Edge-of-field
35
techniques have particular difficulty reducing nitrate loads during high flow periods, and this is
36
when most transport occurs (Royer et al., 2006; Ikenberry et al., 2014).
37
Wood chip bioreactors have been installed at the end of tile lines from many fields in the
38
Midwest. A wide range of nitrate removal effectiveness (up to 98% nitrate reductions) in both
39
laboratory and field studies has been reported in the literature (Greenan et al., 2009; Chun et al.,
40
2009, 2010; Moorman et al., 2010; Verma et al., 2010; Woli et al., 2010; Christianson et al.,
41
2012b; Bell et al., 2015). Most have been designed as plastic lined trenches, filled with mixed
42
species wood chips, that receive tile flow from fields ranging from 1 to 20 ha in size (e.g., Woli
43
et al., 2010; Christianson et al., 2012b). At low flow most of the tile water passes through the
44
bioreactor, but at high flows they are designed so that water will bypass the wood chip bed
45
(Christianson et al., 2012a). The published literature on the effectiveness of this technique is
46
limited and additional results are needed as bioreactors have been proposed in two state nutrient
2
Page 3 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
47
loss reduction strategies as a primary method for nitrate reductions (Iowa Nutrient Reduction
48
Strategy, 2013; Illinois Nutrient Loss Reduction Strategy, 2015).
49
One concern with any engineered denitrification technique is that nitrous oxide (N2O) could
50
be released (Groh et al., 2015), trading a water quality problem for a greenhouse gas problem
51
given the global warming effect of each N2O molecule (310 times that of carbon dioxide).
52
Previous studies on wood chip bioreactors have found that N2O emissions are small, but these
53
reports are limited (Moorman et al., 2010; Woli et al., 2010; Warneke et al., 2011a; Christianson
54
et al., 2013b).
55
Studies are needed to evaluate the effectiveness of wood chip bioreactors under a range of
56
tile flows, nitrate concentrations and loads, tile water temperature, retention times, and bioreactor
57
and wood chip age. Each of these factors can interact in complex ways that may affect the
58
overall effectiveness of a bioreactor in removing nitrate. Few studies to date have examined
59
bioreactor performance past the first year or two of operation, and given that they are proposed
60
to function for as many as 10-20 years, long-term studies are critically needed. Our objective was
61
to evaluate the nitrate removal performance of two wood chip bioreactors installed at the end of
62
subsurface drain lines during the first 3 years of operation, examining the major factors that
63
regulated nitrate removal. An additional objective was to determine how much N2O was emitted
64
from the bioreactors as nitrate was denitrified.
65
Materials and Methods
66
Study Site and Bioreactors
67
We installed two bioreactors in the Embarras River watershed in east-central Illinois, the site
68
of extensive previous work on tile drainage and nitrate loss (e.g., David et al., 1997, 2015;
69
Gentry et al., 2014). The watershed has extensive poorly and very poorly drained Mollisols that
3
Page 4 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
70
are typically tile drained, and would be an excellent candidate watershed for using wood chip
71
bioreactors to reduce riverine nitrate loads that average 30 kg N ha-1 y-1 (Gentry et al., 2014).
72
Daily precipitation data were obtained from two nearby locations located to the SW (Tuscola)
73
and the NE (Philo) of the bioreactor 1 field and averaged (MRCC, 2015). We installed bioreactor
74
1 in March of 2012 on a pattern-drained 20 ha field with Drummer series soils (fine-silty, mixed,
75
superactive, Mesic Endoaquolls) in a corn and soybean rotation; year 1 results were reported in
76
David et al. (2015). Seed corn fertilized with 168 kg N ha-1 as anhydrous ammonia at planting
77
was grown during 2012 but due to the severe drought and poor crop yield, the seed corn was not
78
harvested. Hybrid corn was grown in 2013 (spring fertilization of 200 kg N ha-1 as anhydrous
79
ammonia), followed by soybean in 2014. The bioreactor was installed near the end of a 30 cm
80
tile main within a riparian buffer, and was 6 by 15 m by 1.3 m deep and lined with 6 mil plastic
81
sheeting. Wood chips were initially from two sources, with the bottom layer of chips having
82
more fine material than the top layer with the bioreactor filled to just above the soil surface. No
83
soil was placed over the wood chips, as they were left exposed for easy application of additional
84
woodchips. The fine material in the bottom layer had some mulch like appearance along with
85
wood chips, but leaves, twigs, and other identifiable plant parts were not observed. The upper
86
chips were larger sized, fresh wood chips. Each source of wood chips was from local arborists
87
who were cutting a variety of trees in the area. The source of the wood chips with fine material
88
had stockpiled the wood chips for some period of time before we obtained them. Additional
89
wood chips were added after two years as extensive settling (about 30 cm) and degradation had
90
occurred. We used a four-chamber Agri Drain water level control structure fitted with three
91
flashboard risers and V-notch boards, pressure transducers, and dataloggers to allow
92
measurement of water that went through the bioreactor as well as bypass flow (tile water that
4
Page 5 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
93
flows over the middle V-notch boards and does not pass through the woodchips and does not get
94
treated). We used Motorola pressure transducer sensors (model MPX5010DP) combined with
95
custom designed and built dataloggers during 2012 and 2013 to record water levels at 30 min
96
intervals. In spring 2013 we installed Solinst Levelogger Junior Edge model 3001 water level
97
pressure transducers and dataloggers, and found that they compared well with our previous
98
system during a several month period where both were in operation. During 2014, only the
99
Solinst instruments were utilized. During 2013 and 2014, an Onset Hobo temperature
100
sensor/dataloggers were inserted in the tile inlet and bioreactor outlet to measure water
101
temperatures at 15 min intervals.
102
Bioreactor 2 was installed during December 2012, and had the same dimensions as bioreactor
103
1 and the same type of Agri Drain structure and monitoring. It also drained a field of 20 ha. The
104
major difference was that bioreactor 2 was installed in the main channel floodplain of the
105
Embarras River, where river water often backs up and compromises the bioreactor outlet flow
106
measurement. During major flood events the bioreactor can be completely submerged.
107
Bioreactor Sampling and Analyses
108
Weekly grab samples from bioreactor 1 were supplemented with ISCO automatic water
109
samplers during high flow events to determine water quality. Nitrate, chloride, and sulfate
110
concentrations on filtered samples (0.45 µm pore size) in all tile inlet and outlet samples were
111
measured by ion chromatography (Dionex DX-120). Ammonium and dissolved reactive P (DRP)
112
were analyzed colorimetrically on filtered samples by flow injection analysis with a Lachat
113
QuikChem 8000 using the automated sodium salicylate and ascorbic acid methods, respectively.
114
Samples for total P were digested with sulfuric acid (11.2 N) and ammonium persulfate (0.4 g
115
per 50 ml sample) and then analyzed for DRP as described previously. Dissolved organic carbon
5
Page 6 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
116
(DOC) was determined using a Shimadzu total organic carbon analyzer. All methods followed
117
APHA (2005) procedures. Linear interpolation was used to estimate daily and annual nutrient
118
loads. Nitrous oxide fluxes were measured using static chambers following Woli et al. (2010)
119
and Groh et al. (2015) based on the GRACEnet protocol (Blowes et al., 2003). See supplemental
120
material for further details. We did not measure dissolved N2O in water leaving the bioreactor.
121
Hydraulic Conductivity Analysis and Process Model Development
122
We used flow and depth measurements from 2013 in bioreactor 1 to determine the in-situ
123
hydraulic conductivity in bioreactor 1. Assuming that flow in a bioreactor is Darcian, then it is
124
described by
125
q=
(h − hd ) = K × i
Q
= K eff × u
eff
A
L
[1]
126
where q is the specific discharge or Darcian velocity (m s-1), Q is the flow rate through the
127
bioreactor (m3 s-1), A is the area of the cross section through which flow takes place (m2), Keff is
128
the in-situ or effective hydraulic conductivity, hu and hd are the upstream and downstream flow
129
depths (m), respectively, L is the length of the bioreactor (m), and i is the hydraulic gradient, the
130
ratio of the head loss through the bioreactor to the length over which the head loss occurred. A
131
plot of q against i should yield a straight line with slope Keff.
132
We also developed a process-based model with a daily time step to characterize bioreactor 1
133
performance. The model assumes that the bioreactor is a completely stirred reactor with constant
134
concentration throughout. The flow rate through the system each day is the lesser of tile flowrate
135
from the field and the capacity of the bioreactor given by Equation 1, but with hu and hd being
136
the height of the upstream and downstream stop logs, respectively. If the tile flow rate exceeds
137
the bioreactor capacity, the excess tile flow bypasses the system. If the nitrate concentration in
6
Page 7 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
138
the bioreactor exceeds the threshold value for a first order reaction, set at 1 mg N L-1, based on
139
the work of Robertson (2010), the mass of nitrate is the function of the woodchip volume and
140
temperature dependent removal rate. The removal rate function was based on the relationship
141
between temperature and observed nitrate removal rates during 2014 in bioreactor 1. If the
142
concentration is less than the threshold value then the amount of nitrate removed is a percentage
143
of the amount of nitrate in the bioreactor, in accordance with first order kinetics. The
144
concentration is updated each day and was given by
145
Cd =
V p × C d −1 + I × C I − M d
Vp + I
[2]
146
where Cd is the concentration at the end of the day, Cd-1 is the concentration at the end of the
147
previous day, CI is the inflow concentration, Vp is the pore volume, I is the inflow volume, and
148
Md is the mass of nitrate removed. The model was applied to 30 years of DRAINMOD (Skaggs,
149
1980, 1982; Sanoja et al., 1990) simulated daily drain flow for the 20 ha field, using historical
150
weather data and the dominant soil type in the field (Drummer series).
151
Results
152
Flow and Nutrient Balances
153
Tile flow varied greatly during the three year study period, from 8.1 cm in 2012 to 33.3 cm in
154
2013 (Table 1) in response to precipitation (Figure 1). January through June precipitation was
155
310, 641, and 507 mm for 2012, 2013, and 2014, respectively. There were many large
156
precipitation and resulting flow events during 2013, with about half the flow bypassing the wood
157
chip bed (Figure 1). During 2012 and 2014, most of the flow (85-89%) passed through the
158
bioreactor. Tile nitrate concentrations increased from about 16 mg N L-1 in 2012 to >30 mg N L-1
159
in 2013, remaining high throughout 2014. The high nitrate concentrations were a result of the
7
Page 8 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
160
drought of 2012, which led to a failed seed corn crop that was not harvested and likely large
161
residual soil N pools that fall. This nitrate was then leached from the field during the following
162
two years, with nitrate yields of 106 and 36.5 kg N ha-1 yr-1 in 2013 and 2014, respectively.
163
Bioreactor 1 had an input of 2127 kg N as nitrate in 2013, a large load that was reduced by only
164
3% (72 kg N) (Table 1). This was much less than the 81% removal in 2012, where 208 kg N
165
were removed. In 2014, the percentage removal was 7%, with 47 kg N removed. The chloride
166
budget was nearly balanced, indicating little leakage of water in the bioreactor or around the Agri
167
Drain structure. Sulfate-S was nearly balanced in 2013 and 2014, but the tile input was reduced
168
by 25% in 2012 suggesting much greater reducing conditions in the first year of operation. Both
169
DRP and total P had much larger outputs from the bioreactor than the tile input each year, as did
170
DOC. The flow weighted outlet DOC concentration in 2012 was 18.8 mg C L-1, which decreased
171
to 3.7 and 3.0 mg C L-1 in 2013 and 2014, respectively. The nitrate removal rate was highly
172
correlated with the DOC concentration (r=0.62, n=102, p < 0.0001).
173
Tile Temperature and Bioreactor Nitrate Removal Rates, Nitrous Oxide Emissions
174
Tile water temperature into bioreactor 1 was between 6-8°C from January through April of
175
2013, and then slowly increased to a maximum of 17°C just as tile flow stopped for the year in
176
early July (Figure S1). In 2014, tile water was only 3-4°C in February and March, and then
177
increased to 17°C by early July. The 2014 water temperatures lagged 2013 by approximately 2
178
weeks. Nitrate removal rates for wood chip bioreactors are typically expressed on a g N m-3 d-1
179
basis. Note: wood chip volumes used for removal rates were based on average depth of water in
180
the bioreactor, not total bioreactor volume. During 2012, daily nitrate removal rates in bioreactor
181
1 were quite high, and decreased with increasing retention times, suggesting nitrate limitation
182
(Figure 2). During 2013 and 2014, there was no relationship of retention time with nitrate
8
Page 9 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
183
removal rates. We plotted both daily and average monthly nitrate removal rates versus tile water
184
temperature for 2013 and 2014, and observed that beginning in March of 2013 through July of
185
2013, and through all of 2014, removal rates increased linearly with increasing water
186
temperatures (Figures 3 and 4). Bioreactor 1 had January and February 2013 removal rates that
187
were well above subsequent months, and 2012 average monthly removal rates were not plotted
188
as we did not measure water temperature that year, but rates were much larger than 2013-2014
189
(23 to 44 g N m-3 d-1). Bioreactor 2 does function, but monitoring was not possible most days due
190
to either flooding or more typically, the outlet backing up due to high river water. We have not
191
reported any effectiveness data for this bioreactor given the partial record available.
192
Nitrous oxide fluxes from the surfaces of bioreactors 1 and 2 in 2013 and 2014 were nearly
193
all <0.2 mg N2O-N m-2 h-1 (Figure S2). Rates were greatest during warmer months when there
194
was tile flow. For bioreactor 1, 0.32 and 0.42 kg N as N2O was emitted in 2013 and 2014,
195
respectively. This was 0.44 and 0.89% of the nitrate N removed by bioreactor 1 in 2013 and
196
2014.
197
Flow Analysis
198
For bioreactor 1 in 2013 the plot of specific discharge versus the hydraulic gradient had
199
clusters of points that defined linear features, but there was not a unique linear relationship
200
between hydraulic gradient and specific discharge, except at hydraulic gradients greater than
201
0.008 (Figure 5). At lower hydraulic gradients, specific discharges between 0.1 and 0.3 cm s-1
202
appear to be independent of hydraulic gradient. Specific discharges ranging from near 0 up to 0.3
203
cm s-1 occur at a hydraulic gradient of approximately 0.006. Such a result could be due to short
204
circuiting, with water flowing along the sides and bottom of the bioreactor. Almost every
205
instance where specific discharge exceeded 0.1 cm s-1 was during periods of bypass flow, that is,
9
Page 10 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
206
when the tile flow exceed the transport capacity of the bioreactor. It would be during such events
207
that short circuiting is more likely to occur.
208
The near parallel linear features at hydraulic gradients less than 0.006 may be indicative that
209
the porosity in the bioreactor is not constant, as is likely with bioclogging, a decrease in the
210
porosity of a porous medium due to the growth of microbial biomass. The serial cross
211
correlogram of Q versus A×i, the product of area and hydraulic gradient, has a peak correlation
212
at a lag time of approximately 16 days (Figure S3). This correlation, and the nature of the
213
hydraulic gradient plot may be indicative that biofilm growth peaks in 16 days, and is reduced
214
during periods of high flow. The peak cross correlation value was 0.11, so this process does not
215
explain much of the variation in specific discharge. The highest hydraulic gradient in the
216
bioreactor during 2013 was approximately 0.026. Ghane et al. (2015) found this value to be the
217
upper limit for Darcian flow in fresh woodchips.
218
An analysis of the relationship between specific discharge and hydraulic gradient for the
219
linear features in the plot yields in-situ hydraulic conductivity of 6.7 and 3.1 cm s-1, respectively,
220
for hydraulic gradients less than and greater than 0.015. These values are not dissimilar to the
221
range of 2.5 to 5.5 cm s-1 that Chun et al. (2009, 2010) obtained from the analysis of
222
breakthrough curves from laboratory-scale and field-scale bioreactors in Illinois, and to values of
223
8.4 and 2.4 cm s-1, reported by Ghane et al. (2015) for fresh and old woodchips, respectively.
224
The reduction of the slope at higher hydraulic gradients is consistent with the deviation from
225
Darcian flow reported by Ghane et al. (2015).
10
Page 11 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
226
Discussion
227
Bioreactor Performance and Controls on Nitrate Removal
228
We installed two similar bioreactors on 20 ha fields in the Embarras River watershed of east-
229
central Illinois. Bioreactor 1 was located along a small tributary of the Embarras River and was
230
rarely flooded; however, bioreactor 2 was located along the main channel of the Embarras River
231
and was often flooded. Bioreactor 2 did function much of the time, but accurate monitoring that
232
would allow for an input and output budget was not possible on most days due to the outlet
233
backing up with river water. A wetland would have been more suitable at this location, however
234
the landowner was not interested in removing land from row crop production and the NRCS only
235
allows bioreactors to be installed in riparian buffers and not constructed wetlands.
236
Bioreactor 1 had excellent removal during the first year of operation in 2012, removing 81%
237
of the tile nitrate however this was during a year with little tile flow (8.1 cm). The nitrate
238
removal percentage was similar to many reports in the literature such as Christianson et al.
239
(2012b) where they reported a range of 12 to 76% for four bioreactors in Iowa, and Verma et al.
240
(2010) who reported a range of 42 to 98% for three bioreactors in central Illinois. The percentage
241
effectiveness and mass of nitrate removed decreased greatly in year 2 (2013), with only 72 kg N
242
removed from a large tile input of 2127 kg N. Both the mass and tile concentration of nitrate-N
243
in 2013 was high due to leftover nitrate on the field following the drought and failed seed corn in
244
2012. However, half of the flow bypassed the bioreactor in 2013, and the capacity was not
245
enough to remove much of the substantial nitrate-N load, the largest load of nitrate we are aware
246
of entering a bioreactor from a tile drained field in the Midwest.
247
248
Our results demonstrate that performance and response to drivers such as wood chip quality,
wood chip age, and temperature can vary greatly during the first three years of bioreactor
11
Page 12 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
249
operation. In addition, nitrate removal rates may be overestimated when tile flow temporarily
250
ceases during dry weather and samples are collected during the overturn and flushing of old
251
water out of the wood chips. Care must be taken to avoid using samples that do not represent
252
nitrate removal but are indicative of the flushing of stagnant water out of the bioreactor during
253
extremely low flow or no flow periods.
254
In bioreactor 1, the bottom layer of woodchips contained fine organic materials, and when
255
combined with low flows in 2012 this led to large N removal rates (daily average of 43 g N m-3
256
d-1, range of 0.7 to 116 g N m-3 d-1), much beyond those reported in the literature (e.g., Schipper
257
et al., 2010ab). Dissolved organic C concentrations were high in 2012, which decreased rapidly
258
through 2013. At times during year 1 (2012) the bioreactor bed had removed all nitrate and
259
sulfate reduction occurred. Therefore, removal rates of nitrate were high throughout 2012, and
260
nitrate load was the controlling factor. The relationship between nitrate removal and retention
261
time was opposite what has been found in previous studies, where removal increases with longer
262
retention times (Christianson et al., 2012a). For bioreactor 1, it was likely that there was so much
263
available C in the first year, that nitrate limitation at low flows (and longer retention times)
264
reduced the N removal rate.
265
At the start of year 2 when flow began in January, removal rates remained high as there
266
likely was easily degradable C following the drought conditions of the summer of 2012 with the
267
long dry period and limited flows. By March daily and average monthly N removal rates were
268
responding to temperature, and from that month onward through 2014 N removal rates were
269
related to tile water temperature (Figure 4). Due to the high nitrate concentrations in the tile inlet
270
water and little removal, no sulfate was reduced in 2013 and 2014. Retention time was again not
271
related to N removal rates, as temperature clearly was the regulating factor. Warneke et al.
12
Page 13 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
272
(2011a) reported a Q10 (the factor of the reaction rate increase with every 10°C increase in
273
temperature) of 2.0 for temperatures between 17 and 24°C, and Warneke et al. (2011b) reported
274
a Q10 of 1.2 for experimental temperatures of 16.8 and 27.1°C. Our calculated Q10 was 3.8
275
between 6 and 16°C for 2014. This suggests a strong response to temperature and rapid increase
276
in the denitrification rate with increasing temperature, and that at low tile water temperatures in
277
January through March, denitrification rates will be slow.
278
Currently, wood chip bioreactors are an NRCS interim standard practice (605) for removing
279
nitrate from tile water. The NRCS has sized some bioreactors using a 20°C water temperature
280
that was published by Chun et al. (2010). However, this was a two-day, pulse addition of nitrate
281
experiment conducted at the end of June at a temperature of 20 to 21.7°C, not representative of
282
typical tile water temperatures in east-central Illinois. Tiles in east-central Illinois typically flow
283
from January through July most years, and water temperatures will be low during much of this
284
period (Figure 3), reaching 10°C in May, while likely never reaching 20°C most years.
285
Nitrate Removal Rates in Comparison to Published Values
286
Overall, bioreactor 1 did a poor job (on a percentage basis) of reducing nitrate-N from a tile
287
with unlimited nitrate due to low water temperatures in years 2 and 3. However, the nitrate
288
removal rate per m3 of wood chips was in the range of those reported in the literature. Many
289
studies report typical values of 5 to 10 g N m-3 d-1 when nitrate is non-limiting (Schipper et al.,
290
2010a), similar to rates we measured during May and June of 2013 and 2014. However, our
291
average monthly rates were only 1-3 g N m-3 d-1 when water temperature was colder in February
292
and March. Christianson et al. (2013c) reported rates of 0.38 to 1.06 g N m-3 d-1 for a bioreactor
293
in Iowa, and determined that internal hydraulics limited the removal rates. They found that
294
temperature was not a good predictor of removal rates, with measurements from May through
13
Page 14 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
295
August and water temperatures varying from 8.9 to 17.1°C (Christianson et al., 2013c). Our
296
results and those of Christianson et al. (2013c) suggest that bioreactors may have to be much
297
larger to effectively remove nitrate from a typical tile line as the retention time would have to be
298
substantially longer. On the same farm as bioreactor 1 we monitored two constructed wetlands
299
intercepting tile lines, and found they were much more effective at reducing nitrate-N loads
300
during 2012 and 2013 (Groh et al., 2015).
301
Nitrous Oxide Emissions
302
As had been previously reported (Moorman et al., 2010; Woli et al., 2010; Warneke et al.,
303
2011a; Christianson et al., 2013b), we found small N2O emissions from either bioreactor 1 or 2
304
wood chip beds. However, our rates were generally greater than those reported by Christianson
305
et al. (2013b) for bioreactors with soil covers, but similar to those without. They suggested that
306
soil surfaces limited N2O losses and could be useful for mitigating losses (Christianson et al.,
307
2013b). Less than 1% of the removed nitrate in 2013 and 2014 was as N2O (0.44 and 0.89% of
308
nitrate removed in 2013 and 2014, respectively), somewhat larger than the 0.4% N2O emissions
309
of nitrate removed reported by Christianson et al. (2013b). Christianson et al. (2013b) found
310
more N2O loss dissolved in the water leaving their bioreactors than the amount emitted from the
311
wood chip or soil surfaces. Warneke et al. (2011a) also found substantial loss of dissolved N2O
312
from a denitrification bed that increased the fraction of nitrate lost as N2O from 1% (loss from
313
bed surface) to 4.3% overall. We did not measure dissolved N2O in water leaving the bioreactor,
314
which likely led to an underestimation of overall N2O losses.
315
When N2O losses are placed in the context of the overall field-bioreactor system, the losses
316
are likely small and not an important source of N2O. Bioreactor 1 had N2O losses of 0.32 and
317
0.41 kg N in 2013 and 2014, respectively. A fertilized corn and soybean field in the watershed
14
Page 15 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
318
had N2O losses of 2.2 to 7.7 kg N ha-1 yr-1 during a three year study (Smith et al., 2013), an
319
average of 4.4 kg N ha-1 yr-1. If our 20 ha corn field had a similar N2O emission rate, it would
320
average 88 kg N yr-1, and the average of 0.37 kg yr-1 emitted from the bioreactor would be
321
unimportant. Even adding substantial losses of dissolved N2O in the water leaving the bioreactor
322
would likely still be small in comparison.
323
The decomposing wood chips did release substantial amounts of DOC, and with that C was
324
both reactive and organic P. This is a potential water quality problem from installation of
325
bioreactors, as P is an important freshwater nutrient that leads to algal blooms.
326
Implications of Nitrate Removal and Flow Analysis Results on Bioreactor Size
327
Our results demonstrate the importance of longer term monitoring of newly installed
328
bioreactors. For bioreactor 1, the wood chips had substantial available C in the first year that led
329
to high rates of nitrate removal, beyond those reported in the literature. Robertson (2010) also
330
reported that fresh wood chips had nitrate removal rates about double those in years 2 through 7.
331
By the middle of the 2nd year, our N removal rates were closer to what has been typically
332
reported in the literature, but demonstrated that the bioreactor was too small for the large N load.
333
No other study has had such a large nitrate-N load into a bioreactor. If the load in 2013, as an
334
example, had been 15 kg N ha-1 (300 kg N), the removal efficiency would have been 24% given
335
the 72 kg N removed. This is similar to many removal rates previously published. Therefore,
336
removal rate based on percentage is relative and not as useful as calculating total mass removed
337
when comparing across other studies.
338
The bioreactors were sized using protocols and a routine developed by Cooke and Bell
339
(2014). In this routine, hydraulic conductivity was assumed to be 4.5 cm s-1. The empirical
340
cumulative distribution function for in-situ hydraulic conductivity, assuming Darcian flow, in
15
Page 16 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
341
bioreactor 1 in 2013, and fitted Gaussian and log-Gaussian distribution functions are given in
342
Figure S4. The design value has an almost 99% probability of exceedance, and is, therefore, not
343
representative of the performance of the system. The log-Gaussian distribution provides a better
344
fit, and thus the median conductivity (17.9 cm s-1) is more representative than the mean
345
conductivity (25.1 cm s-1). This median conductivity was used in the process model described in
346
the methods. Based on this process model, the resulting long-term annual load reduction with the
347
current bioreactor configuration was 9%. In order to have a 50% load reduction, the bioreactor
348
would have to be nearly 9 and 3.7 times as large, if the inflow nitrate concentration were 30 and
349
12 mg N L , respectively.
350
Conclusions
351
-1
During three years of operation bioreactor 1 demonstrated variable nitrate removal from the
352
tile system, driven by age of wood chips, retention time, and water temperature. Year 1 and early
353
year 2 results (January to February) had high nitrate removal rates, beyond those reported in the
354
literature, likely due to a large amount of soluble C that was easily degraded. From March on in
355
year 2, and in year 3, water temperature appeared to be the major factor determining nitrate
356
removal rates and removal rates per m3 of wood chip were consistent with those in the published
357
literature. Due to the large input of nitrate in year 2 (2127 kg N) and the amount of bypass flow,
358
the removal rate was low (72 kg N, 3% of the tile nitrate load) following excellent removal in
359
year 1 (208 kg N, 81% of the tile nitrate load). Although bypass flow was reduced in year 3, low
360
water temperature with older wood chips led to only 47 kg of nitrate-N removed, 7% of the tile
361
load. For this tile system where water temperatures average ~9°C for the year, the wood chip bed
362
would need to be 9 times the current size to provide 50% removal of the measured tile nitrate
363
loads. Little N2O was emitted from bioreactor 1, equivalent to <1% of the nitrate removed.
16
Page 17 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
364
Bioreactor 2 had no useable data to assess removal efficiencies due to its location in the
365
floodplain of the Embarras River, where high water levels led to flooding and backup of the tile
366
system.
367
Acknowledgements
368
We thank the farm cooperator who made this study possible and Corey Mitchell for laboratory
369
analysis and data summaries. This article is based upon work that is supported by the
370
National Institute of Food and Agriculture, U.S. Department of Agriculture, under award number
371
No. 2011-039568-31127. Any opinions, findings, conclusions, or recommendations expressed in
372
this publication are those of the authors and do not necessarily reflect the views of the USDA.
373
References
374
American Public Health Association. 1998. Standard methods for the examination of water and
375
376
wastewater. 20th ed. APHA, Washington, DC.
Bell, N., R.A.C. Cooke, T. Olsen, M.B. David, and R. Hudson. 2015. Characterizing the
377
performance of denitrifying bioreactors during simulated subsurface drainage events. J.
378
Environ. Qual. (in press).
379
Blowes, D.W., W.D. Robertson, J. Baker, G. Doyle, G. McCarty, A. Mosier, T. Parkin, D.
380
Reicosky, J. Smith and R. Venterea. 2003. GRACEnet: chamber-based trace gas flux
381
measurement protocol. Trace Gas Protocol Development Committee.
382
Christianson, L.E., A. Bhandari, and M.J. Helmers. 2011. Pilot-scale evaluation of denitrification
383
drainage bioreactors: reactor geometry and performance. J. Environ. Eng. 137:213-220.
384
385
Christianson, L.E., A. Bhandari, and M.J. Helmers. 2012a. A practice-oriented review of
woodchip bioreactors for subsurface agricultural drainage. Appl. Eng. Agric. 28:861-874.
17
Page 18 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
386
Christianson, L., A. Bhandari, M. Helmers, K. Kult, T. Sutphin, and R. Wolf. 2012b.
387
Performance evaluation of four field-scale agricultural drainage denitrification bioreactors in
388
Iowa. Trans. ASABE 55:2163-2174.
389
Christianson, L., R. Christianson, M. Helmers, C. Pederson, and A. Bhandari. 2013a. Modeling
390
and calibration of drainage denitrification bioreactor design criteria. J. Irrig. Drain. Eng.
391
139:699-709.
392
Christianson, L., J. Hanly, N. Jha, S. Saggar, and M. Hedley. 2013b. Denitrification bioreactor
393
nitrous oxide emissions under fluctuating flow conditions. ASABE Meeting Paper Number:
394
131597821, St. Joseph, Mich.
395
396
397
398
399
Christianson, L., M. Helmers, A. Bhandari, and T. Moorman. 2013c. Internal hydraulics of an
agricultural drainage denitrification bioreactor. Ecol. Eng. 52:298-307.
Chun, J.A., and R.A. Cooke. 2008. Calibrating Agri Drain water level control structures using
generalized weir and orifice equations. Appl. Eng. Agric. 24:595-602.
Chun, J.A., R.A. Cooke, J.W. Eheart, and J. Cho. 2010. Estimation of flow and transport
400
parameters for woodchip-based bioreactors: II. Field-scale bioreactor. Biosyst. Eng. 105:95-
401
102.
402
Chun, J.A., R.A. Cooke, J.W. Eheart, and M.S. Kang. 2009. Estimation of flow and transport
403
parameters for woodchip-based bioreactors: I. Laboratory-scale bioreactor. Biosyst. Eng.
404
104:384-395.
405
406
407
408
Cooke, R.A., and N.L. Bell. 2014. Protocol and interactive routine for the design of subsurface
bioreactors. Appl. Eng. Agric. 30:761-771.
David, M.B., L.E. Drinkwater, and G.F. McIsaac. 2010. Sources of nitrate yields in the
Mississippi River Basin. J. Environ. Qual. 39:1657-1667.
18
Page 19 of 33
409
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
David, M.B., C.G. Flint, L.E. Gentry, M.K. Dolan, G.F. Czapar, R.A. Cooke, and T. Lavaire.
410
2015. Navigating the socio-bio-geo-chemistry and engineering of nitrogen management in
411
two Illinois tile-drained watersheds. J. Environ. Qual. 44:368-381.
412
413
David, M.B., L.E. Gentry, D.A. Kovacic, and K.M. Smith. 1997. Nitrogen balance in and export
from an agricultural watershed. J. Environ. Qual. 26:1038-1048.
414
Gentry, L.E., M.B. David, F.E. Below, T.V. Royer, and G.F. McIsaac. 2009. Nitrogen mass
415
balance of a tile-drained agricultural watershed in east-central Illinois. J. Environ. Qual.
416
38:1841-1847.
417
418
419
420
421
422
Gentry, L.E., M.B. David, and G.F. McIsaac. 2014. Variation in riverine nitrate flux and fall
nitrogen fertilizer application in east-central Illinois. J. Environ. Qual. 43:1467-1474.
Ghane, E., N.R. Fausey, and L.C. Brown. 2015. Modeling nitrate removal in a denitrification
bed. Water Res. 71:294-305.
Greenan, C.M., T.B. Moorman, T.B. Parkin, T.C. Kaspar, and D.B. Jaynes. 2009. Denitrification
in wood chip bioreactors at different water flows. J. Environ. Qual. 1664-1671.
423
Groh, T.A., L.E. Gentry, and M.B. David. 2015. Nitrogen removal and greenhouse gas emissions
424
from constructed wetlands receiving tile drainage water. J. Environ. Qual. 44:1001-1010.
425
Ikenberry, C.D., M.L. Soupir, K.E. Schilling, C.S. Jones, and A. Seeman. 2014. Nitrate-nitrogen
426
export: Magnitude and patterns from drainage districts to downstream river basins. J.
427
Environ. Qual. 43:2024-2033.
428
Illinois Nutrient Loss Reduction Strategy. 2015. Illinois Environmental Protection Agency.
429
http://www.epa.illinois.gov/topics/water-quality/watershed-management/excess-
430
nutrients/index (last accessed 16 June 2015).
19
Page 20 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
431
Iowa Nutrient Reduction Strategy. 2013. A Science and Technology-based Framework to Assess
432
and Reduce Nutrients to Iowa Waters and the Gulf of Mexico. Iowa Department of
433
Agriculture and Land Stewardship, Iowa Department of Natural Resources, and Iowa State
434
University College of Agriculture and Life Sciences.
435
436
Jaynes, D.B., and T.M. Isenhart. 2014. Reconnecting tile drainage to riparian buffer hydrology
for enhanced nitrate removal. J. Environ. Qual. 43:631-638.
437
Kovacic, D.A., M.B. David, L.E. Gentry, K.M. Starks, and R.A. Cooke. 2000. Effectiveness of
438
constructed wetlands in reducing nitrogen and phosphorus export from agricultural tile
439
drainage. J. Environ. Qual. 29:1262-1274.
440
441
Midwestern Regional Climate Center. 2015. http://mrcc.isws.illinois.edu/CLIMATE/ (last
accessed 19 August 2015).
442
Moorman, T.B., T.B. Parkin, T.C. Kaspar, and D.B. Jaynes. 2010. Denitrification activity, wood
443
loss, and N2O emissions over 9 years from a wood chip bioreactor. Ecol. Eng. 36:1567-
444
1574.
445
Moorman, T.B., M.D. Tomer, D.R. Smith, and D.B. Jaynes. 2015. Evaluating the potential role
446
of denitrifying bioreactors in reducing watershed-scale nitrate loads: A case study
447
comparing three Midwestern (USA) watersheds. Ecol. Eng. 75:441-448.
448
449
450
Robertson, W.D. 2010. Nitrate removal rates in woodchip media of varying age. Ecol. Eng.
36:1581-1587.
Royer, T.V., M.B. David, and L.E. Gentry. 2006. Timing of riverine export of nitrate and
451
phosphorus from agricultural watersheds in Illinois: Implications for reducing nutrient
452
loading to the Mississippi River. Environ. Sci. Technol. 40:4126-4131.
20
Page 21 of 33
453
454
455
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
Salas, J.D., J.W. Delleur, V.M. Yevjevich, and W.L. Lane. 1980. Applied modeling of
hydrologic time series: Littleton, Colorado, Water Resources Publications, 484 pp.
Sanoja, J., R.S. Kanwar, and S.W. Melvin. 1990. Comparison of simulated (DRAINMOD) and
456
measured tile outflow and water table elevations from two field sites in Iowa. Trans. ASAE
457
33:827-833.
458
459
460
Schipper, L.A., S.C. Cameron, and S. Warneke. 2010a. Nitrate removal from three different
effluents using large-scale denitrification beds. Ecol. Eng. 36:1552-1557.
Schipper, L.A., W.D. Robertson, A.J. Gold, D.B. Jaynes, and S.C. Cameron. 2010b. Denitrifying
461
bioreactors – An approach for reducing nitrate loads to receiving waters. Ecol. Eng.
462
36:1532-1543.
463
Skaggs, R.W. 1980. A water management model for artificially drained soils. North Carolina
464
Agricultural Research Service Technical Bulletin No. 267, North Carolina State University,
465
Raleigh, NC. 54 pp.
466
467
468
469
470
Skaggs, R.W. 1982. Field evaluation of a water management simulation model. Trans. ASAE
25:666-674.
Skaggs, R.W., N.R. Fausey, and R.O. Evans. 2012. Drainage water management. J. Soil Water
Conserv. 67:167-172.
Smith, C.M., M.B. David, C.A. Mitchell, M.D. Masters, K.J. Anderson-Teixeira, C.J. Bernacchi,
471
and E.H. DeLucia. 2013. Reduced nitrogen losses following conversion of row crop
472
agriculture to perennial biofuel crops. J. Environ. Qual. 42:219-228.
473
474
USEPA. 2008. Hypoxia in the northern Gulf of Mexico, an update by the EPA Science Advisory
Board. EPA-SAB-08-003. USEPA, Washington, DC.
21
Page 22 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
475
Verma, S., R. Bhattarai, G. Goodwin, R. Cooke, and J.A. Chun. 2010. Evaluation of
476
conservation drainage systems in Illinois - Bioreactors. ASABE Meeting Paper Number:
477
1009894, St. Joseph, Mich.
478
Warneke, S., L.A. Schipper, D.A. Bruesewitz, I. McDonald, and S. Cameron. 2011a. Rates,
479
controls and potential adverse effects of nitrate removal in a denitrification bed. Ecol. Eng.
480
37:511-522.
481
Warneke, S., L.A. Schipper, M.G. Matiasek, K.M. Scow, S. Cameron, D.A. Bruesewitz, and I.R.
482
McDonald. 2011b. Nitrate removal, communities of denitrifiers and adverse effects in
483
different carbon substrates for use in denitrification beds. Water Res. 45:5463-5475.
484
Woli, K.P., M.B. David, R.A. Cooke, G.F. McIsaac, and C.A. Mitchell. 2010. Nitrogen balance
485
in and export from agricultural fields associated with controlled drainage systems and
486
denitrifying bioreactors. Ecol. Eng. 36:1558-1566.
22
Page 23 of 33
Table 1. Annual tile and bypass flow through bioreactor 1 during the 2012-2014 water years, along with input and output nitrate-N,
Cl, sulfate-S, dissolved reactive P (DRP), total P, and dissolved organic carbon (DOC) loads.
Tile flow
(bypass)
cm
In
Out
Cl
In
Sulfate-S
Out
In
Out
DRP
In
Total P
Out
In
Out
DOC
In
Out
---------------------------------------------------------------- kg ------------------------------------------------------------------
2012
8.1 (1.2)
257
49 (81)
627
603
110
82
0.1
1.3
2013
33.3 (15.6)
2127
2055 (3)
1996
1975
387
381
1.0
4.7
1.9
2014
13.0 (1.5)
729
682 (7)
777
771
157
154
0.3
1.0
0.5
15
305
6.4
83
247
1.4
28
79
23
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
-- cm--
Nitrate-N
(% removal)
Page 24 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
mm d
-1
0
20
40
60
80
16
(a)
(b)
Bioreactor
Bypass
-1
Daily flow (mm d )
14
12
10
8
6
4
2
(c)
Tile inlet
Bioreactor outlet
40
-1
Nitrate (mg N L )
0
50
30
20
10
O
ct
/1
4
Ju
l/1
4
O
ct
/1
3
Ja
n/
14
Ap
r/1
4
Ju
l/1
3
O
ct
/1
2
Ja
n/
13
Ap
r/1
3
Ju
l/1
2
Ja
n/
12
Ap
r/1
2
0
Figure 1. Daily precipitation (a), tile flow intro bioreactor 1 along with daily bypass flow (b), and
nitrate concentrations in the tile inlet and bioreactor 1 outlet (c) from 2012 through
2014.
24
Page 25 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
2012
2013
2014
-3
-1
N removal rate (g N m d )
100
80
60
40
20
0
0
10
20
30
40
50
Retention time (hr)
Figure 2. Nitrate removal rates for bioreactor 1 plotted against retention time during 2012
through 2014.
25
Page 26 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
2013
2014
-3
-1
N removal rate (g m d )
40
30
20
10
0
0
2
4
6
8
10
12
14
16
18
20
Water temperature (°C)
Figure 3. Daily nitrate removal rates and tile water temperatures for bioreactor 1 in 2013 and
2014.
26
Page 27 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
10
-3
-1
N removal rate (g m d )
12
8
6
4
2
2013
2014
0
0
2
4
6
8
10
12
14
16
18
20
Water temperature (°C)
Figure 4. Average monthly nitrate removal rates and tile water temperatures for bioreactor 1 in
2013 and 2014.
27
Page 28 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
-1
Specific discharge (cm s )
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0.000
0.005
0.010
0.015
0.020
0.025
0.030
Hydraulic gradient (cm cm-1)
Figure 5. Specific discharge versus hydraulic gradient for bioreactor 1 during 2013.
28
Page 29 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
Supplemental Material
Temperature and Substrate Controls Woodchip Bioreactor Performance in Reducing Tile
Nitrate Loads in East-central Illinois
Mark B. David1, Lowell E. Gentry1, Richard A. Cooke2, and Stephanie M. Herbstritt2
1
University of Illinois, Dep. of Natural Resources and Environmental Sciences, W503 Turner
Hall, 1102 S. Goodwin Ave., Urbana, IL 61801
2
University of Illinois, Dep. of Agricultural and Biological Engineering, 338 Agricultural
Engineering Building, 1304 W. Pennsylvania Ave., Urbana, IL 61801
Nitrous Oxides fluxes
Three PVC rings (20 cm diameter, 10 cm height inserted ~ 5 cm into the wood chips) were
installed along the flow path of bioreactors 1 and 2. End caps of PVC with weather stripping,
septa, and ventilation tubes were made to fit tightly on the PVC rings to complete the enclosed
chamber, which had overall had space volumes of 5 to 6 L (actual headspace was determined for
each measurement on each ring). Fifteen mL samples were collected at 0, 10, 20, and 30 min
with a syringe. The samples were placed into evacuated 10 mL glass vials with gray butyl septa.
Gas samples were analyzed using a Shimadzu GC-2014 gas chromatograph with autosampler to
determine N2O concentrations. Linear regression was used to determine the rate of N2O
emissions during the 30-min incubation period. Linear interpolation was used to determine daily
flux values between field measurements.
S1
Page 30 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
20
Daily temperature (°C)
18
16
14
12
10
8
6
4
2
Ju
l-1
4
M
ay
-1
4
M
ar
-1
4
Ja
n14
No
v13
Se
p13
Ju
l-1
3
M
ay
-1
3
M
ar
-1
3
Ja
n13
0
Figure S1. Daily temperature of tile water entering bioreactor 1 in 2013 and 2014.
S2
Page 31 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
Bioreactor 1
Bioreactor 2
0.4
-2
-1
N2O-N flux (mg m h )
0.5
0.3
0.2
0.1
Ju
l-1
4
Se
p14
M
ay
-1
4
Ja
n14
M
ar
-1
4
Ju
l-1
3
Se
p13
No
v13
M
ar
-1
3
M
ay
-1
3
Ja
n13
0.0
Figure S2. Nitrous oxide fluxes of bioreactors 1 and 2 during 2013 and 2014.
S3
Page 32 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
Cross correlation coefficient
0.12
0.10
0.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
0
5
10
15
20
25
30
35
Lag (d)
Figure S3. Serial cross correlogram for flowrate versus the product of flow cross sectional area
and hydraulic gradient for bioreactor 1 during 2013. The dashed lines are the upper
and lower 95% confidence limits.
S4
Page 33 of 33
J. Environ. Qual. Accepted Paper, posted 10/15/2015. doi:10.2134/jeq2015.06.0296
Probability of exceedance
1.0
Empirical distribution function
Fitted Gaussian distribution
Fitted log Gaussian distribution
Design conductivity
Median conductivity
0.8
0.6
0.4
0.2
0.0
0
20
40
60
80
100
120
140
160
-1
Hydraulic conductivity (cm s )
Figure S4. Cumulative distribution functions for in-situ hydraulic conductivity, based on the
assumption of Darcian flow, in bioreactor 1 during 2013.
S5