190 A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197
sampling began on 15 July in conjunction with SI. In 1997, however, the soil sampling procedure was
slightly modified and started immediately after plant- ing. Denitrification rates were measured bi-weekly
during the two growing seasons. On each sampling date, aluminum cylinders 50 mm id × 150 mm long
were used to collect undisturbed soil cores in dupli- cate from randomly selected locations in the center
between the two middle rows of each plot. The cylin- ders used were perforated along the sides horizontal
and vertical at 50 mm intervals to enhance acetylene gas diffusion. Samples were placed in 2 l plastic jars
fitted with rubber stoppers for gas sampling with 5 of acetylene C
2
H
2
to block further transformation of N
2
O to N
2
, allowing measurement of total denitri- fication as accumulated N
2
O and also inhibit nitrifi- cation process Yoshinari et al., 1977. To represent
field conditions, samples were incubated outdoors overnight.
The concentration of N
2
O produced through deni- trification was determined following the procedure of
Liang and Mackenzie 1997. Briefly, before gas sam- pling, the air in the plastic jars was thoroughly mixed
by inserting a syringe and pumping several times. About 4 ml of gas were removed from the jars and
injected into a gas chromatograph [GC, 5870 series II Hewlett Packard] equipped with a
63
Ni electron capture detector ECD to measure the concentration
of N
2
O. Values for N
2
O emissions by denitrification were calculated on a per area basis g N ha
− 1
. In 1997, there was a problem with the GC and the gas samples
could not be analyzed immediately after the incubation period. Therefore, 7 ml of head space gas were with-
drawn from incubating jars after the 24 h of incubation
Table 2 Monthly precipitation mm in the growing seasons of 1996 and 1997 as compared to long term 1961–1990 averages measured at Côteau-
du-Lac weather station Month
1961–1991 1996
1997 Rain mm
Rain mm Deviation
Rain mm Deviation
May 76.3
103.8 36
64.8 −
15.1 June
90.1 81.8
− 9.2
98 8.8
July 94.6
133.9 41.5
97 2.5
August 93.9
40.8 −
56.6 86.3
− 8.1
September 90.6
140.6 55.2
81.4 −
10.2 October
76.7 66
− 13.9
41.4 −
46 Total
522.2 566.9
8.6 468.9
− 10.2
period and stored in vacuutainers Vacuutainers Brand, Beckon Dickson company, Rutherford, NJ. Standards
of N
2
O in N
2
were also transferred to vacuutainers at that time and they were used for calibration of the
analysis of N
2
O at each sampling date. After denitrifi- cation measurement, soil cores were dried at 65
◦
C for 3 days and the soil then ground. Soil moisture content
to depth of 0.15 m was determined gravimetrically. To monitor NO
3 −
levels in the soil, triplicate soil samples were collected up to a 0.20 m depth on the
same sampling dates as for denitrification. The soil samples were thoroughly mixed, then 10 g moist sub-
sample was taken and shaken with 100 ml of 1M KCl for 60 min. The extracted solution was filtered through
Whatman 5 filter papers, and then frozen until anal- ysis. NO
3 −
and NH
4 +
were determined colorimetri- cally using an autoanalyzer Quickchem, Milwaukee,
WI and then converted into kg ha
− 1
using bulk den- sities from respective soil samples.
Significance of main treatment effects on NO
3 −
and denitrification rates in the soil and their interac- tions were investigated using General Linear Models
GLM procedure of the Statistical Analysis System SAS Institute, Version 6.12.
3. Results and discussion
3.1. Climatic data Monthly rainfall for the 1996 and 1997 growing
seasons May–October, and the long term 30 year averages 1961–1990 are shown in Table 2. During
the 1996 growing season, total amount of rainfall was
A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197 191
8.6 greater than the long term average. However, the month of August was exceptionally dry, with 56 less
precipitation than average, whereas July and Septem- ber were both very wet, with rainfalls of 41 and 55
above average, respectively Table 2 increasing the risk of NO
3 −
leaching. In comparison, 1997 rainfall differed from that of
1996 in two main respects. First, deviation of monthly rainfall from the long term 30 years average was gen-
erally smaller, and therefore rainfall distribution was relatively more uniform. Secondly, in spite of rainfall
in June and July being slightly above normal Table 2, the total 1997 growing season rainfall was 10.2
below average. Variations in the amount and distribu- tion of rainfall have a strong influence on water table
fluctuations, and consequently on soil moisture levels and N dynamics in the soil profile.
Soil temperature was on average higher in 1996 than 1997 Fig. 2. During the experimental period, average
soil temperature in 1996 was 20.5
◦
C, whereas it was 17.9
◦
C in 1997. Soil temperature directly influences microbial activity in the soil, which is responsible
for denitrification Bergstrom and Beauchamp, 1993; Granli and Bøckman, 1994; MacKenzie et al., 1997.
3.2. Water table depth WTD Water table level fluctuated throughout both grow-
ing seasons, responding primarily to rainfall events.
Fig. 2. Mean soil temperature
◦
C at 0–0.20 m depth at the time of soil sampling.
For example, in 1996, the shallowest WTD was ob- served on 15 July. In that month, high amounts of
rainfall Table 2 saturated the soil resulting in rise of water table. In contrast, WTD dropped significantly
on 22 August and 3 September compared to previous sampling dates Fig. 3. This decrease had two main
reasons. First, in spite of the fact that 1996 was a wet year, August was extremely dry with rainfall 56 be-
low the long term monthly average. Second, although September was very wet, 55 above average, only
1.4 mm of rainfall occurred between the 22 August and 3 September sampling dates. As a result, WTD in
SI plots were as deep as in FD plots Fig. 3. Overall, as shown in Fig. 3, average WTDs in SI plots were
deeper in the drier season of 1997 0.8 m than 1996 0.7 m.
3.3. Water table and soil NO
3 −
Freely draining plots had higher soil NO
3 −
lev- els than subirrigated plots. This trend was consistent
over the two seasons of this study except for 11 July 1997 Fig. 4. Due to the large experimental error
Fig. 4a, 22 August was the only sampling date when the effect of the water table was not statistically sig-
nificant p0.05 during the 1996 growing season. Comparatively, the effect of water table on soil NO
3 −
in the 1997 growing season was statistically variable. Nonetheless, trends appear to suggest treatment effect
192 A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197
Fig. 3. Mean water table depth m fluctuations in a 1996 and b 1997 as influenced by free drainage FD and subirrigation SI treatments.
Fig. 4b. Overall NO
3 −
concentrations were reduced by 42 and 16 in 1996 and 1997, respectively, in the
SI treatment compared to the FD treatment. This is probably due to the shallower water table enhancing
denitrification in SI plots. This is an indication that maintaining a shallow water table depth could be a
useful tool in reducing NO
3 −
pollution to the ground- water. Similarly, Fogiel and Belcher 1991 found
that controlled drainagesubirrigation reduced NO
3 −
loading through drainage by 25–59 over a 2-year period compared with conventional drainage. Gilliam
and Skaggs 1986 predicted a 32 decrease in NO
3 −
losses due to controlled drainage. It should, however, be pointed out that the decrease in NO
3 −
concen- trations under controlledsubirrigation plots may be
accompanied by an increase in N
2
O production. On the other hand, Kliewer and Gilliam 1995 estimated
that N
2
O accounted for only 2 of the measured denitrification potential for each water table treat-
ment. They concluded that water table level had no impact on percentage of total denitrification evolving
as N
2
O to the atmosphere. This and similar findings appear to suggest that the ecological impact of N
2
O produced during the denitrification process may not
A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197 193
Fig. 4. Soil NO
3 −
concentration differences between free drainage and subirrigated plots during a 1996 and b 1997 growing seasons. Vertical bars represent standard errors.
be as serious as was previously thought. To confirm this under natural conditions, field trials are needed
to quantify the proportion of N
2
O:N
2
ratio evolution. 3.4. Nitrogen rate effect on soil NO
3 −
level and denitrification rate
The effect of N fertilization rate on the level of NO
3 −
in the soil was evident Fig. 5. Because of the higher rate of N fertilizer, soil NO
3 −
concentrations were higher in the N
200
treatment than N
120
during both growing seasons; the 11 and 26 June sampling
dates in 1997 being the only exceptions Fig. 5. These observations suggest that NO
3 −
concentra- tions would be expected to increase with N inputs
regardless of whether WTM is present or not. It is well documented that NO
3 −
accumulation in the soil profile and subsequent potential for leaching losses
increases with increasing N application Roth and Fox, 1990; Angle et al., 1993; Drury et al., 1996. It
is interesting to note that N fertilizer rate had little Table 3 or no significant effect Table 4 on the to-
tal denitrification rate. This was unexpected because denitrification rates would be expected to be higher in
plots receiving higher N application rates MacKenzie et al., 1997; Ellis et al., 1998; Henault et al., 1998.
194 A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197
Fig. 5. Soil NO
3 −
concentration differences between 120 and 200 kg ha
− 1
of N fertilizer rate in a 1996 and b 1997 growing seasons. Vertical bars are standard errors.
One plausible explanation could be that mineral N content was not the main limiting factor of soil N
2
O emission Henault et al., 1998.
One important feature that must be emphasized when comparing N
200
and N
120
treatments in 1997 is the higher NO
3 −
in the latter treatment on 11 and 26 June sampling dates. These high NO
3 −
concentrations were unexpected, and may be erroneous values. It may
be due to NO
3 −
influx from recharge areas or seep- age through confining plots. The higher cumulative
soil NO
3 −
in 1997 than 1996 34.8 and 9.58 kg ha
− 1
, respectively, may be due to the relatively dry con-
ditions in 1997 during which denitrification was not enhanced. This speculative relationship is consistent
with the lower denitrification rate in 1997 than 1996 Tables 3 and 4.
3.5. Denitrification and water table management WTD showed a strong influence on denitrification.
There was no significant interaction between any of the treatment factors, therefore, main effects were exam-
ined independently. In both growing seasons, with the exception of the 22 August 1996 sampling date very
A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197 195
Table 3 Denitrification rates g N ha
− 1
per day as influenced by water table depth and N fertilization rate and analysis of variance during 1996 growing season
Treatments Sampling dates
15 July 06 August
22 August 03 September
20 September 05 October
FD
a
184.7 31.36
27.3 5.5
14.32 4.64
SI
b
225 112.95
21.8 30.51
25.78 15.3
N
120 c
124 62.5
15.74 20.75
26.95 9.47
N
200 d
285 82
33.38 15.27
13.3 10.47
Mean 204
72.25 24.56
18 20.2
10 Summary of analysis of variance
WTM
e
ns
f ∗∗
ns
∗
ns
∗
N-rate
∗∗
ns ns
∗∗
ns ns
WTM × N ns
ns ns
ns ns
ns
a
Free drainage.
b
Controlledsubirrigation.
c
120 kg N ha
− 1
.
d
200 kg N ha
− 1
.
e
Water table management.
f
ns=not significant at 5.
∗ ,∗∗
= statistically significant at 5 and 1 probability level, respectively.
dry month, denitrification rates were always higher in SI than FD treatments Tables 3 and 4. Higher deni-
trification losses were associated with higher moisture content in SI treatment plots as compared to FD treat-
Table 4 Denitrification rates g N ha
− 1
per day as influenced by water table depth and N fertilization rate and analysis of variance during 1997 growing season
Treatments Sampling days
28 May 11 June
26 June 11 July
23 July 06 August
18 August 03 September
17 September 03 October
Fd
a
ni
g
ni ni
36.1 6.7
6.77 1.1
4.84 7.84
4.93 SI
b
ni ni
ni 38.1
14.15 7.2
8.78 12
20 11.15
N
120 c
36.76 150
143 34.83
11.45 9.1
7 10.62
17.8 9.81
N
200 d
25.36 64
140 39.45
9.36 4.87
2.87 6.2
10.64 6.27
Mean 31.54
107.2 141.5
37.14 10.41
7 4.94
8.42 14.2
8.02 Summary of analysis of variance
WTM
e
ni ni
ni ns
h ∗
ns
∗∗ ∗∗
∗∗ ∗
N-rate ns
∗
ns ns
ns ns
ns ns
ns ns
WTM × N na
f
na na
ns ns
ns ns
ns ns
ns
a
Free drainage.
b
Controlledsubirrigation.
c
120 kg N ha
− 1
.
d
200 kg N ha
− 1
.
e
Water table management.
f
na=not applicable.
g
ni=not initiated.
h
ns=not significant.
∗ ,∗∗
= significant at 5 and 1 probability level, respectively.
ment data not shown. Increases in N
2
O evolution rates in the June–July period Tables 3 and 4 were
enhanced by N fertilizer application combined with warming soils. Similar to our finding, Christensen and
196 A.A. Elmi et al. Agriculture, Ecosystems and Environment 79 2000 187–197
Tiedje 1990, Beauchamp et al. 1996 and Fan et al. 1997 concluded that these peaks were due to warm-
ing of saturated soils, and enhanced microbial activity. Denitrification rates during both growing seasons
appeared to be regulated largely by climatic factors such as soil temperature, and amount and distribution
of rainfall. In relatively dry periods, for example in the 1997 growing season, water table dropped sharply
and denitrification N loss was not promoted resulting in NO
3 −
accumulation in the soil profile. In rainy periods, on the other hand, the water table rose and
denitrification losses were promoted. In 1996, for example, the highest average denitrification flux was
measured on the 11 July sampling date Table 3. Heavy rainfall occurring at the beginning of this month
saturated the soil and caused the rise of the WTD to about 0.45 m below the soil surface Fig. 3. In 1997,
denitrification peaked on the 26 June sampling date Table 4. Four days before this sampling 22 June,
the highest amount of daily rainfall during that season was measured daily rainfall data not shown creating
an anaerobic environment favourable for denitrifica- tion process.
3.6. Seasonal variability and denitrification Consistently higher denitrification losses were
measured in 1996 in all treatments, compared to 1997. Several environmental and field management
practices may be responsible for the large differences in seasonal denitrification. This seasonal differences
may be attributable to soil temperatures which were generally higher in 1996 than 1997 Fig. 2. Tem-
perature is considered as one of the most influential factors on the magnitude of denitrification Granli
and Bøckmam, 1994. Bergstrom and Beauchamp 1993 and Liang and Mackenzie 1997 asserted that
lower temperatures can result in a reduction in the denitrification rate. Rainfall events in May and July
1996 36 and 41.5 above average, respectively, fol- lowing the first and second N fertilizer applications,
respectively, might have increased moisture content beyond the saturation limit, and hence enhanced
denitrification. Additionally, considerable amount of NO
3 −
might have been leached from the surface layer to deeper depths, leaving less NO
3 −
in the soil surface.
As shown in Table 1, herbicide application in 1997 was somewhat later. As a result, tremendous weed
growth was observed in all plots. It is therefore reason- able to assume that significant amount of NO
3 −
which would have been lost as denitrification might have
been taken up by weeds. This, however, contradicts the larger amount of NO
3 −
remaining in the soil in 1997 than 1996 34.8 and 9.58 kg ha
− 1
, respectively. Therefore, another plausible explanation could be the
enrichment of NO
3 −
through mineralization from the previous hot and wet growing season. Nitrogen min-
eralization is an important source of NO
3 −
and can supply from 30 to 100 of N nutritional needs and
increases with precipitation Douglas et al., 1998. Since denitrification measurements in 1996 started
mid July, it is likely that denitrification peak was not captured. Therefore, one should be cautious when
comparing the two seasons. Similarly, since measure- ments of denitrification rate were carried out only
during the growing season, cumulative denitrification losses reported in Tables 3 and 4 should not be as-
sumed as being annual losses. If annual losses were to be estimated, additional sampling must be contin-
uing during spring thaw when denitrification may be vigorous Christensen and Tiedje, 1990; Ellis et al.,
1998; Henault et al., 1998.
4. Conclusions
