Introduction
Nitrous oxide (N2O) is of major importance as a greenhouse gas and
precursor of ozone (O3) destruction in the stratosphere (Crutzen, 1970).
Agriculture is a major source of greenhouse gases (GHGs), such as carbon
dioxide (CO2), methane (CH4) and also N2O (IPCC, 2006). The
application of organic and inorganic fertiliser N to agricultural soils
enhances the production of N2O (Baggs et al., 2000). This soil-emitted
N2O is predominantly derived from denitrification and to a smaller
extent, nitrification in soils (Davidson and Verchot, 2000). Denitrification
is a microbial process in which reduction of nitrate (NO3-) occurs to
produce N2O, and N2 is the final product of this process, benign
for the environment, but represents a loss of N in agricultural systems.
Nitrification is an oxidative process in which ammonium (NH4+) is
converted to NO3- (Davidson and Verchot, 2000). Both processes are
controlled by environmental factors and their interactions, and are
influenced by agricultural management (Firestone and Davidson, 1989). It is
well recognised that soil water content expressed as water-filled pore space
(WFPS) is a major controlling factor, and as Davidson (1991) illustrated,
nitrification is a source of N2O until WFPS values reach about 70 %,
after which denitrification dominates. In fact, Firestone and Davidson (1989)
gave oxygen supply a ranking of 1 in importance as a controlling factor in
fertilised soils, above C and N. At WFPS between 45 and 75 % a mixture of
nitrification and denitrification act as N2O sources. Davidson also
suggested that at WFPS values above 90 % only N2 is produced. Several
studies have later proposed models to relate WFPS with emissions (Schmidt
et al., 2000; Dobbie and Smith, 2001; Parton et al., 2001;
del Prado et al., 2006; Castellano et al., 2010) but the “optimum” WFPS for
N2O emissions varies from soil to soil (Davidson, 1991). Soil structure
could be influencing this effect and it has been identified to strongly
interact with soil moisture (Ball et al., 1999; van Groenigen
et al., 2005) through changes in WFPS. Particularly soil compaction
due to livestock treading and the use of heavy machinery affect soil
structure and emissions as reported by studies relating bulk density to
fluxes (Klefoth et al., 2014) and degrees of tillage to emissions (Ludwig
et al., 2011).
Compaction is known to affect the size of the larger pores (macropores)
thereby reducing the soil air volume and therefore increasing the WFPS (for
the same moisture content) (van der Weerden et al., 2012). However, little is
known about the effect of compaction on the smaller soil pores (micropores),
and this could provide valuable information for understanding the
simultaneous behaviour of the dynamics of water in the various pore sizes in
soil. Such an understanding would lead to the development of better N2O
mitigation strategies via dealing with soil compaction issues.
The role of water in soils is closely linked to microbial activity but also
relates to the degree of aeration and gas diffusivity in soils (Morley and
Baggs, 2010). Water facilitates nutrient supply to microbes and restricts gas
diffusion, thereby increasing the residence time of gases in soil, and the
chance of further N2O reduction before it can be released to the
atmosphere. This is further aided by the restriction of the diffusion of
atmospheric O2 (Dobbie and Smith, 2001), increasing the potential for
denitrification. In consequence, counteracting effects (high microbial
activity vs. low diffusion) occur simultaneously, making it difficult to
predict net processes and corresponding outputs (Davidson, 1991). Detailed
understanding of the sources of N2O and the influence of physical
factors, i.e. soil structure and its interaction with moisture, is a powerful
basis for developing effective mitigation strategies.
Isotopocules of N2O represent the isotopic substitution of the O and/or
the two N atoms within the N2O molecule. The isotopomers of N2O,
are those differing in the peripheral (β) and central N positions
(α) of the linear molecule (Toyoda and Yoshida, 1999) with the
intramolecular 15N site preference (SP; the difference between δ15Nα and δ15Nβ) used to identify production
processes at the level of microbial species or enzymes involved (Toyoda et
al., 2005; Ostrom and Ostrom, 2011). Moreover, δ18O, δ15N and SP of
emitted N2O depend on the denitrification product ratio
(N2O / (N2+N2O)) and hence provide insight into the
dynamics of N2O reduction (Well and Flessa, 2009; Lewicka-Szczebak et
al., 2014, 2015). Koster et al. (2013), for example,
recently reported δ15Nbulk values of N2O between
-36.8 and -31.9 ‰ under the conditions of their experiment,
which are indicative of denitrification according to Perez et al. (2006) and
Well and Flessa (2009), who proposed the range -54 to -10 ‰
relative to the substrate. Baggs (2008) summarised that values between -90
and -40 ‰ are indicative of nitrification. Determination of these
values is normally carried out in pure culture studies or in conditions
favouring either production or reduction of N2O (Well and Flessa, 2009).
The SP is, however, considered a better predictor of the N2O source due to
its independence from the substrate signature (Ostrom and Ostrom, 2011).
Simultaneous occurrence, production and reduction of N2O as in natural
conditions present a challenge for isotopic factors determination due to
uncertainty on N2O reduction and the co-existence of different microbial
communities producing N2O (Lewicka-Szczebak et al., 2014). Recently,
using data from the experiment reported here, where soil was incubated under
aerobic atmosphere and the complete denitrification process occurs,
Lewicka-Szczebak et al. (2015) determined fractionation factors associated
with N2O production and reduction using a modelling approach. The
analysis comprised measurements of the N2O and N2 fluxes combined
with isotopocule data. Net isotope effects (η values) are variable to a
certain extent as they result from a combination of several processes causing
isotopic fractionation (Well et al., 2012). The results generally confirmed
the range of values of η (net isotope effects) and η18O / η15N ratios reported by previous studies for N2O
reduction for that part of the soil volume were denitrification was enhanced
by the N+C amendment. This did not apply for the other part of the soil
volume not reached by the N+C amendment, showing that the validity of
published net isotope effects for soil conditions with low denitrification
activity still needs to be evaluated.
Lewicka-Szczebak et al. (2015) observed a clear relationship between 15N
and 18O isotope effects during N2O production and denitrification
rates. For N2O reduction, differential isotope effects were observed for
two distinct soil pools characterised by different product ratios
N2O / (N2+N2O). For moderate product ratios (from 0.1 to 1.0) the range
of isotope effects given by previous studies was confirmed and refined,
whereas for very low product ratios (below 0.1) the net isotope effects were
much smaller. In this paper, we present the results from the gas emissions
measurements from soils collected from a long-term permanent grassland soil
to assess the impact of different levels of soil saturation on N2O and
N2 and CO2 emissions after compaction. CO2 emissions were
measured in addition as an estimate of aerobic respiration and thus of
O2 consumption, which indicates denitrification is promoted. The
measurements included the soil isotopomer (15Nα,
15Nβ and site preference) analysis of emitted N2O, which
in combination with the bulk 15N and 18O was used to distinguish
between N2O from bacterial denitrification and other processes (e.g.
nitrification and fungal denitrification) (Lewicka-Szczebak et al., 2017).
We conducted measurements at defined saturation of pores size fractions as a
prerequisite to model denitrification as a function of water status
(Butterbach Bahl et al., 2013; Müller and Clough, 2014). We have under
controlled conditions created a single compaction stress of 200 kPa (typical
of soils compacted after grazing) in incremental layers using a uniaxial
pneumatic piston to simulate a grazing pressure. We hypothesised that at high
water saturation, spatial heterogeneity of N emissions decreases due to more
homogeneous distribution of the soil nutrients and/or anaerobic microsites.
We also hypothesised that even at high soil moisture a mixture of
nitrification and denitrification can occur. We base this on the creation of
pockets of aerobicity as well of anaerobicity at high soil moisture, mainly
driven by soil respiration after application of N and C (using up O2)
and further recovery after nutrients are used becoming limiting (increasing
aeration). We also aimed to assess how these effects (spatial heterogeneity
and source processes) occur in a relatively narrow range of moisture
(70–100 %). As far as we know there no other studies going to this level
of detail. They mostly rely on the knowledge of the effect of moisture on
soil processes, whilst in our study we combined direct measurements of both
N2O and N2 with isotopomers of N2O to verify the source
processes. In addition, the packing of the cores in our study was of great
precision, increasing our potential to achieve reproducibility in the
replicates where a mixture of aerobic/anaerobic pores might have occurred. We
aimed to understand changes in the ratio N2O / (N2O + N2)
at the different moisture levels studied in a controlled manner on soil
micro-
and macropores. The N2 emissions were based on direct measurements from
the incubated soils, avoiding methodologies that rely on inhibitors such as
acetylene with limitations in diffusion in soil and causing oxidation of NO
(Nadeem et al., 2013). Moreover, we used isotopocule values of N2O to
evaluate whether the contribution of bacterial denitrification to the total
N2O flux was affected by moisture status.
Materials and methods
Soil used in the study
An agricultural soil, under grassland management since at least 1838
(Barré et al., 2010), was collected from a location adjacent to a
long-term ley-arable experiment at Rothamsted Research in Hertfordshire
(Highfield; see soil properties in Table 1 and further details in Rothamsted
Research, 2006; Gregory et al., 2010). The mixed sward is dominated by Lolium and Trifolium species and is cut two–three times a year.
The soil was sampled as described in Gregory et al. (2010).
Briefly, it was sampled from the upper 150 mm of the profile, air-dried in
the laboratory, crumbled and sieved (< 4 mm), mixed to make a bulk
sample, and equilibrated at a pre-determined water content (37 g 100 g-1;
Gregory et al., 2010) in airtight containers at 4 ∘C for at least
48 h.
Highfield soil properties.
Property
Units
Highfield
Location Grid reference Soil type Land use pH Sand (2000–63 µm) Silt (63–2 µm) Clay (< 2 µm) Texture Particle density Organic matter Water content for packing
GB National Grid Longitude Latitude SSEWa groupc SSEWa seriesd FAObc g g-1 dry soil g g-1 dry soil g g-1 dry soil SSEWa classc g cm-3 g g-1 dry soil g g-1 dry soil
Rothamsted Research Herts. TL129130 00∘21′48′′ W 51∘48′18′′ N Palaeo-argillic brown earth Batcombe Chromic Luvisol Grass; unfertilised; cut 5.63 0.179 0.487 0.333 Silty clay loam 2.436 0.089 0.37
a Soil Survey of England and Wales classification
system.
b United Nations Food and Agriculture Organization World Reference Base
for Soil Resources classification system (approximation).
c Avery (1980).
d Clayden and Hollis (1984).
Preparation of soil blocks
The equilibrated soil was then packed into 12 stainless-steel blocks
(145 mm diameter; h: 100 mm), each of which contained three cylindrical
holes (i.d: 50 mm; h: 100 mm each). The cores were packed to a single
compaction stress of 200 kPa in incremental layers using a uniaxial
pneumatic piston. The three hole blocks were used to facilitate the
compression of the cores. The 200 kPa stress was analogous to a severe
compaction event by a tractor (Gregory et al., 2010) or livestock
(Scholefield et al., 1985). The total area of the upper surface of soil in
each block was therefore 58.9 cm2 (3 × 19.6 cm2) and the
target volume of soil was set to be 544.28 cm3
(3 × 181.43 cm3) with the objective of leaving a headspace of
approximately 45 cm3 (3 × 15 cm3) for the subsequent
experiment. The precise height of the soil (and hence the volume) was
measured using the displacement measurement system of a DN10 Test Frame
(Davenport-Nene, Wigston, Leicester, UK) with a precision of 0.001 mm.
Equilibration of soil cores at different saturations
The soil was equilibrated to four different initial saturation conditions or
treatments (t0) which were based on the likely distribution of water between
macropores and micropores. The first treatment was where both the macro- and
micropores (and hence the total soil) were fully saturated; the second
treatment was where the macropores were half-saturated and the micropores
remained fully saturated; the third treatment was where the macropores were
fully unsaturated and the micropores again remained fully saturated; and the
fourth treatment was where the macropores were fully unsaturated and the
micropores were half-saturated. These four treatments are hereafter referred
to as SAT/sat, HALFSAT/sat, UNSAT/sat and UNSAT/halfsat, respectively, where
upper-case refers to the saturation condition of the macropores and
lower-case refers to the saturation condition of the micropores. In order to
set these initial saturation conditions, we referred to the gravimetric soil
water release characteristic for the soil, as given in Gregory et al. (2010)
(see Supplement). To achieve target water contents during the
incubation, the amount of liquid added with the C / N amendment (15 mL)
was considered in the total volume of water added. For the SAT/sat and
HALFSAT/sat conditions, two sets of three replicate blocks were placed on two
fine-grade sand tension tables connected to a water reservoir. For the
UNSAT/sat condition a set of three replicate blocks was placed on a tension
plate connected to a water reservoir, and the final set of three replicate
blocks was placed in pressure plate chambers connected to high-pressure air.
All blocks were saturated on their respective apparatus for 24 h, and were
then equilibrated for 7 days at the adjusted target matric potentials which
were achieved by either lowering the water level in the reservoir (sand
tables and tension plate) or by increasing the air pressure (pressure
chambers). At the end of equilibration period, the blocks were removed
carefully from the apparatus, wrapped in airtight film, and maintained at
4 ∘C until the subsequent incubation.
Incubation
The study was carried out under controlled laboratory conditions, using a
specialised laboratory denitrification (DENIS) incubation system (Cardenas
et al., 2003). Each block containing three cores was placed in an
individual incubation vessel of the automated laboratory system in a
randomised block design to avoid effect of vessel. The lids for the vessels
containing three holes were lined with the cores in the block to ensure that
the solution to be applied later would fall on top of each soil core.
Stainless steel bulkheads fitted (size for 1/4 in. tubing) on the lids had a
three-layered Teflon coated silicone septum (4 mm thick × 7 mm
diameter) for supplying the amendment solution by using a gas tight
hypodermic syringe. The bulkheads were covered with a stainless-steel nut and
only open when amendment was applied. The incubation experiment lasted
13 days from the time the cores started to be flushed until the end of the
incubation. The incubation vessels with the soils were contained in a
temperature controlled cabinet and the temperature set at 20 ∘C. The
incubation vessels were flushed from the bottom at a rate of
30 mL min-1 with a He / O2 mixture (21 % O2, natural
atmospheric concentration) for 24 h, or until the system and the soil
atmosphere were emitting low background levels of both N2 and N2O
(N2 can get down to levels of 280 ppm much smaller than atmospheric
values). Subsequently, the He / O2 supply was reduced to
10 mL min-1 and directed across the soil surface and measurements of
N2O and N2 carried out at approximately 2-hourly cycles to sample
from all the 12 vessels. Emissions of CO2 were simultaneously measured.
Application of amendment
An amendment solution equivalent to 75 kg N ha-1 and
400 kg C ha-1 was applied as a 5 mL aliquot a solution containing
KNO3 and glucose to each of the three cores in each vessel on day 0 of
the incubation. Glucose is added to optimise conditions for denitrification
to occur (Morley and Baggs, 2010). The aliquot was placed in a
stainless-steel container (volume 1.2 L) which had three holes drilled with
bulkheads fitted: two to connect stainless-steel tubing for flushing the
vessel and the third one to place a septum on a bulkhead to withdraw
solution. Flushing was carried out with He for half an hour before the
solution was required for application to the soil cores and continued during
the application process to avoid atmospheric N2 contamination (a total
of 1.5 h). The amendment solution was manually withdrawn from
the container with a glass syringe fitted with a three-way valve onto the
soil surface; care was taken to minimise contamination from atmospheric
N2 entering the system. The syringe content was injected to the soil
cores via the inlets on the lids consecutively in each lid (three cores) and
all vessels, completing a total of 36 applications that lasted about 45 min.
Incubation continued for 12 days, and the evolution of N2O, N2
and CO2 was measured continuously. At the end of each incubation
experiment, the soils were removed from the incubation vessels for further
analysis. The three cores in each incubation vessel were pooled in one sample
and subsamples taken and analysed for mineral N, total N and C, and moisture
status.
Gas measurements
Gas samples were directed to the relevant analysers via an automated
injection valve fitted with two loops to direct the sample to two gas
chromatographs. Emissions of N2O and CO2 were measured by gas
chromatographs (GCs), fitted with an electron capture detector (ECD) and
separation achieved by a stainless-steel packed column (2 m long, 4 mm
bore) filled with “Porapak Q” (80–100 mesh) and using N2 as the
carrier gas. The detection limit for N2O was equivalent to
2.3 g N ha-1 d-1. The N2 was measured by GC with a He
ionisation detector (HID) and separation was achieved by a PLOT column (30 m
long 0.53 mm i.d.), with He as the carrier gas. The detection limit was
9.6 g N ha-1 d-1. The response of the two GCs was assessed by
measuring a range of concentrations for N2O, CO2 and N2.
Parent standards of the mixtures 10 133 ppm N2O + 1015.8 ppm
N2, 501 ppm N2O + 253 ppm N2, and 49.5 ppm N2O +
100.6 ppm N2 were diluted by means of mass flow controllers with He to
give a range of concentrations of up to 750 ppm for N2O and
1015 ppm for N2. For CO2, a parent standard of 30 100 ppm was
diluted down to 1136 ppm (all standards were in He as the balance gas).
Daily calibrations were carried out for N2O and N2 by using the low
standard and doing repeated measurements. The temperature inside the
refrigeration cabinet containing the incubation vessels was logged on an
hourly basis and checked at the end of the incubation. The gas outflow rates
were also measured and recorded daily, and subsequently used to calculate the
flux.
Measurement of N2O isotopic signatures
Gas samples for isotopocule analysis were collected in 115 mL serum bottles
sealed with grey butyl crimp-cap septa (part no. 611012, Altmann, Holzkirchen,
Germany). The bottles were connected by a Teflon tube to the end of the
chamber vents and were vented to the atmosphere through a needle in order to maintain
flow through the experimental system. Dual isotope and isotopocule signatures
of N2O, i.e. δ18O of N2O (δ18O-N2O),
average δ15N (δ15Nbulk) and δ15N
from the central N position (δ15Nα) were analysed after
cryo-focussing by isotope ratio mass spectrometry as described previously
(Well et al., 2008). 15N site preference (SP) was obtained as SP =2×(δ15Nα – δ15Nbulk). Dual
isotope and isotopocule ratios of a sample (Rsample) were
expressed as per mille deviation from 15N / 14N and
18O/16O ratios of the reference standard materials
(Rstd), atmospheric N2 and standard mean ocean water (SMOW),
respectively:
δX=(Rsample/Rstd-1)×1000,
where X=15Nbulk, 15Nα,
15Nβ, or 18O.
Data analysis and additional measurements undertaken
The areas under the curves for the N2O, CO2 and N2 data were
calculated by using GenStat 11 (VSN International Ltd, Hemel Hempstead,
Herts, UK). The resulting areas for the different treatments were analysed by
applying analysis of variance (ANOVA). The isotopic
(15Nbulk, 18O), and site preference (SP) differences
between the four treatment for the different sampling dates were analysed by
two-way ANOVA. We also used the Student's t test to check for changes in
soil water content over the course of the experiments.
Calculation of the relative contribution of the N2O derived from
bacterial denitrification (%BDEN) was done according to
Lewicka-Szczebak et al. (2015). The isotopic value of initially produced
N2O, i.e. prior to its partial reduction (δ0), was determined
using a Rayleigh model (Mariotti et al., 1982), were δ0 is
calculated using the fractionation factor of N2O reduction (ηN2O-N2) for SP and the fraction of residual N2O
(rN2O) which is equal to the N2O / (N2+N2O)
product ratio obtained from direct measurements of N2 and N2O flux.
An endmember mixing model was then used to calculate the percentage of
bacterial N2O in the total N2O flux (%BDEN) from
calculated δ0 values and the SP and δ18O endmember values
of bacterial denitrification and fungal denitrification/nitrification. The
range in endmember and ηN2O-N2 values assumed (adopted from
Lewicka-Szczebak, 2017) to calculated maximum and minimum estimates of
%BDEN is given in Table 4. We also fitted three functions through
these data (SP vs. N2O / (N2+N2O)) including a
second-degree polynomial and a linear and logarithmic function.
Because both, endmember values and ηN2O-N2 values are not
constant but subject to the given ranges, we calculated here several
scenarios using combinations of maximum, minimum and average endmember and
ηN2O-N2 values (Table 4) to illustrate the possible range of
%BDEN for each sample. For occasional cases where
%BDEN>100 % the values were set to 100 %.
At the same time as preparing the main soil blocks, a set of replicate
samples was prepared in exactly the same manner, but in smaller cores (i.d:
50 mm; h: 25 mm). On these samples, we analysed soil mineral N, total N and
C and moisture at the start of the incubation. The same parameters were
measured after incubation by doing destructive sampling from the cores.
Mineral N (NO3-, NO2- and NH4+) was analysed after
extraction with KCl by means of a segmented flow analyser using a
colorimetric technique (Searle, 1984). Total C and N in the air-dried soil
were determined using a thermal conductivity detector (TCD, Carlo Erba, model
NA2000). Soil moisture was determined by gravimetric analysis after drying at
105 ∘C.
Discussion
N2O and N2 fluxes
Effect of soil moisture
The observed decrease in total N emissions with decreasing initial soil
moisture reflects the effect of soil moisture as reported in previous studies
(Well et al., 2006). The differences when comparing the cumulative fluxes,
however, were only marginally (p<0.1) significant (Table 3) mostly due to
large variability within replicates in the drier treatments (see Fig. 1b).
Davidson et al. (1991) provided a WFPS threshold for determination of source
process, with a value of 60 % WFPS as the borderline between nitrification
and denitrification as source processes for N2O production. The WFPS in
all treatments in our study was larger than 70 %, above this 60 %
threshold, and referred to as the “optimum water content” for N2O by
Scheer et al. (2009), so we can be confident that denitrification was likely
to have been the main source process in our experiment. In addition, Bateman
et al. (2004) observed the largest N2O fluxes at 70 % WFPS on a silty
loam soil, lower than the 80 % value for the largest fluxes from the clay
soil in our study (Fig. 2), suggesting that this optimum value could change
with soil type. Further, the maximum total measured N lost
(N2O + N2) in our study occurred at about 95 % WFPS (Fig. 2), but
not many studies report N2 fluxes for comparison and we are still
missing measurements of nitric oxide (NO) (Davidson et al., 2000) and ammonia
(NH3) to account for the total N losses. It is, however, possible that the
N2O + N2 fluxes in the SAT/sat treatment were underestimated due to
low diffusivity in the water-filled pores (Well et al., 2001). Gases would
have been trapped (particularly in the higher saturation treatments) due to
low diffusion and thus possibly masked differences in N2 and N2O
production since this fraction of gases was not detected (Harter et al.,
2016). It is worth mentioning that there was some drying during the
incubation. The flow of the gas is very slow (10 mL min-1), simulating
a low wind speed, so normally this would dry the soil in field conditions too.
It would represent a rainfall event where the initial moisture differs
between treatments but some drying occurs due to the wind flow. We believe,
however, that the effect of drying will be more relevant (and significant
relative to the initial moisture) later in the incubation.
The smaller standard errors in both N2O and N2 data for the larger
soil moisture levels (Table 3 and Fig. 1) could suggest that, at high moisture
contents, nutrient distribution (N and C) on the top of the core is more
homogeneous, causing replicate cores to behave similarly. At the lower soil
moisture for both N2O and N2, it is possible that some cracks
appear on the soil surface, causing downwards nutrient movement, resulting in
heterogeneity in nutrient distribution on the surface and increasing
variability between replicates, reflected in the larger standard errors of
the fluxes. Laudone et al. (2011) studied, using a biophysical model, the
positioning of the hot-spot zones away from the critical percolation path
(described as “where air first breaks through the structure as water is
removed at increasing tensions”) and found it slowed the increase and
decline in emission of CO2, N2O and N2. They found that
hot-spot zones further away from the critical percolation path would reach
the anaerobic conditions required for denitrification in shorter time, and the
products of the denitrification reactions take longer to migrate from the
hot-spot zones to the critical percolation path and to reach the surface of
the system. The model and its parameters can be used for modelling the effect
of soil compaction and saturation on the emission of N2O. They suggest
that, having determined biophysical parameters influencing N2O
production, it remains to be determined whether soil structure, or simply
saturation, is the determining factor when the biological parameters are
constrained. Furthermore, Clough et al. (2013) indicate that microbial-scale
models need to be included in larger models linking microbial processes and
nutrient cycling in order to consider spatial and temporal variation.
Kulkarni et al. (2008) refer to “hot spots” and “hot moments” of
denitrification as scale-dependant and highlight the limitations for
extrapolating fluxes to larger scales due to these inherent variabilities. In
addition, in order to understand heterogeneity of added amendment, we assumed
(for modelling purposes) multiple pools after N and glucose amendment. In
Bergstermann et al. (2011), for example, we presumed they occupied 10 % of
the pore volume of the core (pool 1), because this resulted in a good fit for measured
and modelled N2 and N2O fluxes as well as δ15Nbulk values. In the current study, we could assume that in
the wettest treatment this (proportional) volume was smaller, i.e. similar to
the pore volume displaced by the added 5 mL of amendment, since pores were
almost completely filled with water. Furthermore, we could assume that it would have been the
largest in the driest treatment, where the amendment solution was also able to
infiltrate air-filled pores in the partly saturated pore space and thereby increase the water
content in the infiltrated volume. With regards to leaching, it was minimal
(< 0.5 mL water in the core) and so significant leaching of amendment can
thus be excluded. Other techniques such as X-ray and MRI could help determine
the distribution of added nutrients in the soil matrix.
Relationship with soil parameters to determine processes
Our results, for the two highest water contents (SAT/sat and HALFSAT/sat),
indicated that N2O only contributed 20 % of the total N emissions, as
compared to 40–50 % at the lowest water contents (UNSAT/sat and
UNSAT/halfsat, Table 3). This was due to reduction to N2 at the high
moisture level, confirmed by the larger N2 fluxes, favoured by low gas
diffusion, which increased the N2O residence time and the chance of
further transformation (Klefoth et al., 2014). We should also consider the
potential underestimation of the fluxes in the highest saturation treatment
due to restricted diffusion in the water-filled pores (Well et al., 2001). A
total of 99 % of the soil NO3- was consumed in the two high-water
treatments, whereas in the drier UNSAT/sat and UNSAT/halfsat treatments there
still was 35 and 70 % of the initial amount of NO3- left in the
soil, at the end of the incubation, respectively (Table 3). The total amount
of gas lost compared to the NO3- consumed was almost 3 times greater for the
wetter treatments, and less than twice for the two drier ones. This agrees with
denitrification as the dominant process source for N2O with larger
consumption of NO3- at the higher moisture and larger N2 to
N2O ratios (5.7, 4.7 for SAT/sat and HALFSAT/sat, respectively), whereas
at the lower moisture, ratios were lower (1.5 and 1.0 for UNSAT/sat and
UNSAT/halfsat, respectively) (Davidson, 1991). This also indicates that with
WFPS above the 60 % threshold for N2O production from denitrification,
there was an increasing proportion of anaerobic microsites with increase in
saturation controlling NO3- consumption and N2/N2O ratios
in an almost linear manner. With WFPS values between 71 and 100 % and
N2 / N2O between 1.0 and 5.7, a regression can be estimated:
Y=0.1723 X-11.82 (R2=0.8585), where Y is N2 / N2O
and X is %WFPS. In summary, we propose that heterogeneous distribution of
anaerobic microsites could have been the limiting factor for complete
depletion of NO3- and conversion to N2O in the two drier
treatments. In addition, in the UNSAT/halfsat treatment there was a decrease
in soil NH4+ at the end of the incubation (almost 50 %; Table 3),
suggesting nitrification could have been occurring at this water content
which also agrees with the increase in NO3-, even though WFPS was
relatively high (> 71 %) (Table 3). It is important to note that as we
did not assess gross nitrification, the observed net nitrification based on
lowering in NH4+ could underestimate gross nitrification since there
might have been substantial N mineralisation during the incubation. However,
under conditions favouring denitrification at high soil moisture the typical
N2O produced from nitrification is much lower compared to that from
denitrification (Lewicka-Szczebak et al., 2017) with the maximum reported
values for the N2O yield of nitrification of 1–3 % (e.g. Deppe et
al., 2017). If this is the case, nitrification fluxes could not have exceeded
1 kg N with NH4+ loss of < 30 kg × 3 %
∼ 1 kg N. This would have represented for the driest treatment, if
conditions were suitable only for one day, that nitrification-derived
N2O would have been 6 % of the total N2O produced. Loss of
NH3 was not probable at such low pH (5.6). The corresponding rate of
NO3- production using the initial and final soil contents and
assuming other processes were less important in magnitude would have been
< 1 mg NO3--N kg dry soil-1 d-1, which is a
reasonable rate (Hatch et al., 2002). The other three treatments lost similar
amounts of soil NH4+ during the incubation (23–26 %), which could
have been due to some degree of nitrification at the start of the incubation
before O2 was depleted in the soil microsites or due to NH4+
immobilisation (Table 3) (Geisseler et al., 2010).
A mass N balance, considering the initial and final soil NO3-,
NH4+, added NO3- and the emitted N (as N2O and N2),
results in unaccounted-for N loss of 177.2, 177.6, 130.6 and
110.8 mg N kg-1 for SAT/sat, HALFSAT/sat, UNSAT/sat and
UNSAT/halfsat, respectively, that could have been emitted as other N gases
(such as NO), and some immobilised in the microbial biomass. NO fluxes
reported by Loick et al. (2016), for example, result in a ratio
N2O / NO of 0.4. In summary, unaccounted-for N loss is 2 to 3
times the total measured gas loss (Table 3). In addition, in the SAT/sat
treatment there was probably an underestimation of the produced N2 and
N2O due to restricted diffusion at the high WFPS (e.g. Well et al.,
2001).
Implications for field distribution of fluxes
Well et al. (2003) found that under saturated conditions there was good
agreement between laboratory and field measurements of denitrification, and
attributed deviations, under unsaturated conditions, to spatial variability
of anaerobic microsites and redox potential. Dealing with spatial variability
when measuring N2O fluxes in the field remains a challenge, but the
uncertainty could be potentially reduced if water distribution is known. Our
laboratory study suggests that soil N2O and N2 emission for higher
moisture levels would be less variable than for drier soils and suggests that
for the former a smaller number of spatially defined samples will be needed
to get an accurate field estimate. This applied to a lesser extent to the
CO2 fluxes.
Isotopocule trends
Trends of isotopocule values of emitted N2O coincided with those of
N2 and N2O fluxes. The results from the isotopocule data (Table 6
and Fig. 3) also indicated that generally there were more isotopic
similarities between the two wettest treatments when compared to the two
contrasting drier soil moisture treatments.
Isotopocule values of emitted N2O reflect multiple processes where all
signatures are affected by the admixture of several microbial processes, the
extent of N2O reduction to N2 and the variability of the
associated isotope effects (Lewicka-Szczebak et al., 2015). Moreover, for
δ18O and δ15Nbulk the precursor signatures
are variable (Decock and Six, 2013), and for δ18O the O exchange with
water can be also variable (Lewicka-Szczebak et al., 2017). Since the number
of influencing factors clearly exceeds the number of isotopocule values,
unequivocal results can only be obtained if certain processes can be excluded
or be determined independently (Lewicka-Szczebak et al., 2015, 2017). The two latter conditions were fulfilled in this
study, i.e. N2O fluxes were high and several orders of magnitude above
possible nitrification fluxes, since the N2O – to – NO3- ratio
yield of nitrification products rarely exceeds 1 % (Well et al., 2008). Moreover, N2 fluxes and thus N2O reduction rates
were exactly quantified.
The estimated values of %BDEN indicate that, in the period
immediately after amendment application, all moisture treatments were similar,
reflecting that the microbial response to N and C added was the same and
denitrification dominated. This was the same for the rest of the period for
the wetter treatments. In the drier treatments, proportions decreased
afterwards and were similar to values before amendment application, possibly
due to recovery of more aerobic conditions that could have encouraged other
processes to contribute. As N2 was still produced in the driest
treatment (but in smaller amounts), this indicated ongoing denitrifying
conditions and thus large contributions to the total N2O flux from
nitrification were not probable, but some occurred as suggested by
NH4+ consumption.
The trends observed reflect the dynamics resulting from the simultaneous
application of NO3- and labile C (glucose) on the soil surface as
described in previous studies (Meijide et al., 2010; Bergstermann et al.,
2011) where the same soil was used, resulting in two locally distinct
NO3- pools with differing denitrification dynamics. In the soil
volume reached by the NO3- / glucose amendment, denitrification
was initially intense with high N2 and N2O fluxes and rapid
isotopic enrichment of the NO3--N. When the NO3- and/or
glucose of this first pool was exhausted, N2 and N2O fluxes were
much lower and dominated by the initial NO3- pool that was not
reached by the glucose / NO3- amendment and that is less
fractionated due to its lower exhaustion by denitrification, causing
decreasing trends in δ15Nbulk of emitted N2O.
This is also reflected in Fig. 4, where N2O fluxes from both pools
exhibited correlations (and mostly significant) between δ15Nbulk and δ18O due to varying N2O reduction,
but δ15Nbulk values in days 1 and 2 – i.e. the phase
when pool 1 dominated – were distinct from the previous and later phase.
The fit of 15Nbulk / 18O data to two distinct and
distant regression lines can be attributed to two facts: firstly, in the wet
treatment (Fig. 4a, b) pool 1 was probably completely exhausted and there was
little NO3- formation from nitrification (indicated by final
NO3- values close to 0, Table 3), whereas the drier treatment
exhibited substantial NO3- formation and high residual NO3-.
Hence, there was probably still some N2O from pool 1 after day 2 in the
dry treatment but not in the wetter ones. Secondly, the product ratios after
day 2 of the drier treatments were higher (0.13 to 0.44) compared to the
wetter treatments (0.001 to 0.09). Thus the isotope effect of N2O
reduction was smaller in the drier treatments, leading to a smaller upshift
of δ15Nbulk and thus more negative values after day 2,
i.e. with values closer to days 1 + 2.
This finding further confirms that δ15N / δ18O
patterns are useful to identify the presence of several N pools, e.g.
typically occurring after application of liquid organic fertilisers, which has
been previously demonstrated using isotopocule patterns (Koster et al.,
2015).
Interestingly, the highest δ15Nbulk and δ18O
values of the emitted N2O were found in the soils of the HALFSAT/sat
treatment, although it may have been expected that the highest isotope values
from the N2O would be found in the wettest soil (SAT/sat) because
N2O reduction to N2 is favoured under water-saturated conditions
due to extended residence time of produced N2O (Well et al., 2012).
However, N2O / (N2+N2O) ratios of the SAT/sat and
SAT/halfsat treatments were not different (Table 5). Bol et al. (2004) also
found that some estuarine soils under flooded conditions (akin to our
SAT/sat) showed some strong simultaneous depletions (rather than enrichments)
of the emitted N2O δ15Nbulk and δ18O
values. These authors suggested that this observation may have resulted from
a flux contribution of an “isotopically” unidentified N2O production
pathway. Another explanation could be complete consumption of some of the
produced N2O in isolated micro-niches in the SAT/sat treatment due to
inhibited diffusivity in the fully saturated pore space. N2 formation
in these isolated domains would not affect the isotopocule values of emitted
N2O and this would thus result in lower apparent isotope effects of
N2O reduction in water saturated environments as suggested by Well et
al. (2012).
The SP values obtained were generally below 12 ‰, in agreement with
reported ranges attributed to bacterial denitrification: -2.5 to
1.8 ‰ (Sutka et al., 2006), 3.1 to 8.9 ‰ (Well and Flessa,
2009) and -12.5 to 17.6 ‰ (Ostrom and Ostrom, 2011). The SP, believed to be a
better predictor of the N2O source as it is independent of the substrate
isotopic signature (Ostrom and Ostrom, 2011), has been suggested as it can be used to
estimate N2O reduction to N2 in cases when bacterial
denitrification can be assumed to dominate N2O fluxes (Koster et al.,
2013; Lewicka-Szczebak et al., 2015). There was a strong correlation between
the SP and N2O / (N2O + N2) ratios on the first 2 days of
the incubation for all treatments up until the N2O reached its maximum
(Fig. 3), which reflects the accumulation of δ15N at the alpha
position during ongoing N2O reduction to N2. Later on in the
experiment, beyond day 3, this was not observed probably because in that
period the product ratio remained almost unchanged and very low (Table 6).
Similar observations have been reported by Meijide et al. (2010) and
Bergstermann et al. (2011), as they also found a decrease in SP during the
peak flux period in total N2+N2O emissions, but only when the soil
had been kept wet prior to the start of the experiment (Bergstermann et al.,
2011). These results confirm from two independent studies (Lewicka-Szczebak et
al., 2014) that there is a relationship between the product ratios and
isotopic signatures of the N2O emitted. The δ18O vs. SP
regressions indicate more similarity between the three wettest treatments as
well as high regression coefficients, suggesting this SP / δ18O ratio
could also be used to help identify patterns for emissions and their sources.
Link to modelling approaches.
Since isotopocule data could be compared to N2 and N2O fluxes, the
variability of isotope effects of N2O production and reduction to
N2 by denitrification could be determined from this dataset
(Lewicka-Szczebak et al., 2015), and this included modelling the two pool
dynamics discussed above. It was demonstrated that net isotope effects of
N2O reduction (ηN2O-N2) determined for both NO3-
pools differed. Pool 1 representing amended soil and resulting in high fluxes
but moderate product ratio, exhibited ηN2O-N2 values and the
characteristic η18O / η15N ratios similar to those
previously reported, whereas for pool 2 (amendment-free soil), characterised
by lower fluxes and very low product ratio, the net isotope effects were much
smaller and the η18O / η15N ratios, previously accepted
as typical for N2O reduction processes (i.e. higher than 2), were not
valid. The question arises of whether the poor coincidence of pool 2 isotopologue
fluxes with previous N2O reduction studies reflects the variability of
isotope effects of N2O reduction or whether the contribution of other
processes like fungal denitrification could explain this (Lewicka-Szczabak et
al., 2017). The latter explanation is evaluated in Sect. 4.4.
Liu et al. (2016) noted that on the catchment-scale potential N2O
emission rates were related to hydroxylamine and NO3-, but not
NH4+ content in soil. Zou et al. (2014) found high SP (15.0 to
20.1 ‰) values at WFPS of 73 to 89 %, suggesting that fungal
denitrification and bacterial nitrification contributed to N2O
production to a degree equivalent to that of bacterial denitrification.
To verify the contribution of fungal denitrification and/or hydroxylamine
oxidation we can first look at the ηSPN2O-NO3 values
calculated in the previous modelling study applied on the same dataset
(Table 1, the final modelling step; Lewicka-Szczebak et al., 2015). For pool
1 there are no significant differences between the values of various
treatments, SP0 ranges from (-1.8 ± 4.9) to
(+0.1 ± 2.5). Pool 1 emission was mostly active in days 1–2; hence,
these values confirm the bacterial dominance in the emission at the beginning
of incubation, which originates mainly from the amendment addition and
represents a similar pathway for all treatments. However, for the pool 2
emission we could observe a significant difference when compared the two wet
treatments (SAT/sat and HALFSAT/sat: (-5.6 ± 7.0)) with the UNSAT/sat
treatment (+3.8 ± 5.8). This represents the emission from unamended
soil which was dominating after the third day of the incubation and indicates
higher nitrification contribution for the drier treatment.
Contribution of bacterial denitrification
An endmember mixing approach has been previously used to estimate the
fraction of bacterial N2O (%BDEN), but without independent
estimates of N2O reduction (Zou et al., 2014), but due to the unknown
isotopic shift by N2O reduction, the ranges of minimum and maximum
estimates were large, showing that limited information is obtained without
N2 flux measurement.
In an incubation study with two arable soils, Koster et al. (2013) used
N2O / (N2+N2O) ratios and isotopocule values of gaseous
fluxes to calculate SP of N2O production (referred to as SP0),
which is equivalent to SP0 using the Rayleigh model and published values
of ηN2O-N2. The endmember mixing approach based on SP0
was then used to estimate fungal denitrification and/or hydroxylamine
oxidation, giving indications for a substantial contribution in a clay soil,
but not in a loamy soil. Here we presented for the first time an extensive
dataset with large range in product ratios and moisture to calculate the
contribution of bacterial denitrification (%BDEN) of emitted
N2O from SP0. The uncertainty of this approach arises from three
factors: (i) from the range of SP0 endmember values for bacterial
denitrification of -11 to 0 ‰ and 30 to 37 for hydroxylamine
oxidation/fungal denitrification, (ii) from the range of net isotope effect
values of N2O reduction (ηN2O-N2) for SP which vary from
-2 to -8 ‰ (Lewicka-Szczebak
et al., 2015), and (iii) system condition (open vs. closed) taken to
estimate the net isotope effect (Wu et al., 2016).
The observation that %BDEN of emitted N2O was generally
high (63–100 %) in the wettest treatment (SAT/sat) was not unexpected.
However, interestingly, %BDEN in the HALFSAT/sat treatment was
very similar (71–98 %), pointing to the role of the wetter areas of the
soil microaggregates contributing to high %BDEN values. The
slightly lower values, i.e. down 60 % in UNSAT/sat %BDEN range
of 60–100 %, suggest that the majority of N2O derived from bacterial
denitrification still results from the wetter microaggregates of the soils,
despite the fact that the macropores are now more aerobic. Only when the
micropores become partially wet, as in the UNSAT/halfsat treatment, do the
more aerobic soil conditions allow a higher contribution of
nitrification/fungal denitrification, ranging from 0 to 46 % (1 –
%BDEN, Table 6) on days 3–12 (Zhu et al., 2013). Differences in
the contribution of nitrification/fungal denitrification between the flux
phases when different NO3- pools were presumably dominating are only
indicated in the driest treatment, since 1 – %BDEN was higher
after day 2 (14 to 46 %) compared to days 1+2 (0 to 33 %). This larger
share of nitrification/fungal denitrification can be attributed to the
increasing contribution from pool 2 to the total flux as indicated by the
modelling of higher SP0 for pool 2 (see previous section and
Lewicka-Szczebak et al. (2015). In addition, indications for elevated
contribution of processes other than bacterial denitrification were only
evident in the drier treatments during phases before and after N2, and
N2O fluxes were strongly enhanced by glucose amendment. The data supply
no clue whether the other processes were suppressed during the anoxia induced
by glucose decomposition or just masked by the vast glucose-induced bacterial
N2O fluxes.