Introduction
The contribution of managed grasslands to reducing atmospheric
greenhouse gas (GHG) concentrations through net uptake of CO2 (Ammann
et al., 2007) may be at least partially offset by net emissions of
N2O (Conant et al., 2005; Fléchard et al., 2005). These emissions
may be substantial, with N2O emission factors of as large as 3 %
measured in intensively managed grasslands with fertilizer rates of
25–30 gNm-2yr-1 (Imer et al., 2013; Rafique et
al., 2011). These emissions are highly variable temporally and spatially
because they are determined by complex interactions among short-term weather
events (warming, precipitation), land management practices (N amendments,
defoliation) and soil properties (e.g. bulk density, water retention). The
N2O driving these emissions in managed grasslands is thought to be
generated within the upper 2 cm of the soil profile (van der Weerden
et al., 2013) and in surface litter left by grazing or harvesting (Pal et
al., 2013) so that diurnal heating and precipitation events that cause rapid
warming and wetting of the litter and soil surface may cause large but brief
emission events. These events are thought to be driven by increased demand
for electron acceptors by nitrification and denitrification, a reduced supply
of O2 by which these demands are preferentially met, and therefore
increased demand for alternative acceptors NO3-, NO2- and
N2O by autotrophic nitrifiers and heterotrophic denitrifiers.
The magnitude of N2O emission events in managed grasslands generally
increases with the amount of N added as urine, manure or fertilizer and with
the intensity of defoliation by grazing or cutting (Ruzjerez et al., 1994).
Thus, Imer et al. (2013) found a negative correlation between leaf area index (LAI) and
N2O emissions at intensively managed grasslands in Switzerland. The
increase in emissions with defoliation has been attributed to increased urine
and manure deposition and soil compaction with defoliation by grazing and to
slower uptake of N and water by slower-growing plants with defoliation by
harvesting (Jackson et al., 2015). Both N additions and defoliation are
thought to raise these emissions by increasing the supply of NH4+
and NO3- to autotrophic nitrifiers and heterotrophic denitrifiers.
This increase raises the demand for alternative e- acceptors by these
microbial populations if the supply of O2, the preferred e-
acceptor, fails to meet demand, as may occur when soil water content
(θ) after defoliation rises with precipitation or reduced
transpiration. This supply is governed by physical and hydrological
properties (porosity, water retention) of the near-surface soil.
Consequently, land use practices and soil properties must be considered when estimating
N2O emissions from managed grasslands.
Recognition of the effects of precipitation events, N amendments and soil
properties on N2O emissions has led to empirical models in which
annual emission inventories are calculated directly from annual precipitation
and N inputs (Lu et al., 2006) or in which monthly emission events are
calculated from monthly precipitation, air temperature Ta, and
θ (Fléchard et al., 2007). However, the soil depth at which most
emitted N2O is generated (0–2 cm) is much shallower than
that at which θ used in these models is measured (5–10 cm)
(Fléchard et al., 2007), and the soil temperature Ts at this
depth may differ from Ta. This is particularly so for
grasslands in which N additions are
necessarily left on the soil surface without incorporation. Thus, large
N2O emissions may be caused by surface wetting from precipitation on
dry soils following fertilizer application, so that deeper θ is
sometimes found to be of little explanatory value in empirical models
(Fléchard et al., 2007). Furthermore, the response of denitrification to
θ has been found in experimental studies to rise sharply with
Ts, likely through the combined effects of Ts on
increasing demand and reducing supply of O2 at microbial microsites
(Craswell, 1978). The interaction between Ts and θ on
N2O emissions is clearly apparent in the meta-analysis of N2O
emissions from European grasslands by Fléchard et al. (2007). This
interaction has been represented in empirical models by fitting
interdependent threshold values of Ts and θ above which
emissions have been measured in field experiments (Smith and Massheder,
2014). However, a more robust simulation of this interaction with N2O
emissions should be built from basic biological and physical processes that
are independent of site-specific measurements.
Process models used to simulate N2O emissions from managed grasslands
must therefore explicitly represent the effects of short-term weather events
on near-surface Ts and θ, as well as the effects of
N additions and defoliation on near-surface NH4+ and NO3-.
These models must also explicitly represent the effects of mineral N,
Ts and θ, and of soil physical and hydrological
properties, on the demand for vs. supply of O2 and alternative e-
acceptors NO3-, NO2- and N2O, and on the
oxidation–reduction reactions by which these e- acceptors are reduced.
However, earlier process models have usually simulated N2O emissions
as Ts-dependent functions of nitrification and denitrification
rates, modified by texture-dependent functions of water-filled pore space
(WFPS) (e.g. Li et al., 2005). In some models additional empirical functions
of Ts (Chatskikh et al., 2005), or of Ts and WFPS
(Schmid et al., 2001), are used to calculate the fraction of nitrification
that generates N2O and the fraction of heterotrophic respiration
Rh that drives denitrification (Schmid et al., 2001), thereby
avoiding the explicit simulation of O2 and its control on N2O
emissions. A more detailed summary of functions of the mineral N,
Ts and WFPS currently used to model N2O emissions is
given in Fang et al. (2015). These functions have many model-dependent
parameters and function independently of each other, so that key interactions
among reduced C and N substrates, Ts and θ on N2O
production may not be simulated. In none of these approaches are the
oxidation–reduction reactions by which N2O is generated or consumed
explicitly represented. Furthermore, the effects of defoliation and surface
litter on N2O emissions have not been considered in earlier process
models.
Process models used to simulate N2O emissions must also accurately
represent the key processes of C cycling that drive those of N cycling, from
which N2O is generated and consumed. These include gross and net
primary productivity (GPP and NPP), which drive mineral N uptake and
assimilation with plant growth. GPP and consequent plant growth also drive
autotrophic respiration (Ra), the below-ground component of which
contributes to soil O2 demand. NPP drives litterfall and root
exudation, which in turn drive heterotrophic respiration (Rh)
that also contributes to litter and soil O2 demand and thereby to
demand for alternative e- acceptors which drive N2O generation.
Heterotrophic respiration also drives key N transformations such as
mineralization or immobilization, thereby controlling availability of these
alternative e- acceptors. Land use practices, such as defoliation from
grazing or harvesting, and soil properties, such as porosity and water
retention, alter these key C cycling processes and thereby N2O
emissions. Therefore these emissions are best simulated by comprehensive
ecosystem models.
In the mathematical model ecosys, the effects of weather and
N amendments on Ts, θ and mineral N, and hence on the
demand for vs. supply of O2, NO3-, NO2- and
N2O, and thereby on N2O emissions, are simulated by
explicitly coupling the transport processes with the oxidation–reduction
reactions by which these e- acceptors are known to be generated,
transported and consumed in soils (Grant and Pattey, 1999, 2003, 2008; Grant
et al., 2006; Metivier et al., 2009). The development of model algorithms for
these processes was guided by two key principles:
all algorithms in the model must represent physical, biochemical and
biological processes studied in basic research programs (e.g.
convective–diffusive transport, oxidation–reduction reactions) so that these
algorithms can be parameterized independently of the model;
this parameterization must be conducted on spatial and temporal scales
smaller than those of prediction (in this case seasonal N2O fluxes)
so that site-specific effects on predicted values are not incorporated into
the algorithms, limiting their robustness.
These principles are designed to avoid as much as possible the use of site-
and model-specific algorithms that may lack application in sites and models
other than those for which they were developed. Although models based on
these principles appear complex, they can be better constrained than simpler
models because they are parameterized from independent experiments. The
resulting detail that the application of these principles brings to the model
enables better-constrained tests of model output against more comprehensive
and diverse site data than are possible with simpler models.
In an extension of earlier work with ecosys, we propose that
temporal and spatial variation in N2O emissions from an intensively
managed grassland can be largely explained from the modelled effects of N
amendments (fertilizer, manure), plant management (e.g. harvest intensity and
timing), soil properties (e.g. bulk density) and weather (Ts,
precipitation events) on the demand for vs. supply of O2,
NO3-, NO2- and N2O in surface litter and
near-surface soil (0–2 cm). Testing this explanation requires
frequent measurements to characterize the large temporal variation in
N2O emissions found in managed ecosystems. Such measurements were
recorded from 2004 to 2009 using automated chambers in intensively managed
grass–clover grassland at Oensingen, Switzerland, and used here to test our
modelled explanation of these fluxes.
Summary of key processes governing generation and emission of
N2O as represented in ecosys.
Model development
General overview
The hypotheses for N2O oxidation–reduction reactions and their
coupling with gas transport in ecosys are represented in Fig. 1 and
described further below with reference to equations and definitions listed in
Sects. A, C, D, E and H of the Supplement (indicated by
square brackets in the text below; e.g. [H1] refers to Eq. 1 in Sect. H), as
well as in earlier papers (Grant and Pattey, 1999, 2003, 2008; Grant et
al., 2006; Metivier et al., 2009). These hypotheses are part of a larger
model of soil C, N and P transformations (Grant et al., 1993a, b), coupled to
one of soil water, heat and solute transport in surface litter and soil
layers, which are in turn components of the comprehensive ecosystem model
ecosys (Grant, 2001).
Mineralization and immobilization of ammonium by all microbial
populations
Heterotrophic microbial populations m (obligately aerobic bacteria,
obligately aerobic fungi, facultatively anaerobic denitrifiers, anaerobic
fermenters, acetotrophic methanogens, and obligately aerobic and anaerobic
non-symbiotic diazotrophs) are associated with each organic substrate i
(i: animal manure, coarse woody plant residue, fine non-woody plant
residue, particulate organic matter, or humus). Autotrophic microbial
populations n (aerobic NH4+ and NO2- oxidizers,
hydrogenotrophic methanogens and methanotrophs) are associated with inorganic
substrates. These populations grow with energy generated from the coupled
oxidation of reduced dissolved organic C (DOC) by heterotrophs or of mineral N
(NH4+ and NO2-) by nitrifiers and a reduction of e-
acceptors O2 and NOx. These populations decay according to
first-order rate constants with provision for internal recycling of limiting
nutrients (N, P). During growth, each functional component j (j:
nonstructural, labile, resistant) of these populations seeks to maintain a
set C : N ratio by mineralizing NH4+ ([H1a]) from, or by
immobilizing NH4+ ([H1b]) or NO3- ([H1c]) to, microbial
nonstructural N. Nitrogen limitations during growth may cause C : N
ratios to rise above set values and a greater recovery of microbial N from
structural to nonstructural forms to reduce N loss during decay but at a cost
to microbial function. These transformations control the exchange of N
between organic and inorganic states and hence affect the availability of
alternative e- acceptors for nitrification and denitrification.
Oxidation of DOC and reduction of oxygen by heterotrophs
Constraints on heterotrophic oxidation of DOC imposed by O2 uptake are
solved in four steps:
DOC oxidation under non-limiting O2 is calculated from active
biomass, DOC concentration and an Arrhenius function of Ts [H2];
O2 reduction to H2O under non-limiting O2
(O2 demand) is calculated from (1) using a set respiratory quotient
[H3];
O2 reduction to H2O under ambient O2 is calculated
from radial O2 diffusion through water films of thickness determined
by soil water potential [H4a] coupled with active uptake at heterotroph
surfaces driven by (2) [H4b]. O2 diffusion and active uptake is
calculated for each heterotrophic population associated with each organic
substrate, allowing [H4] to calculate lower O2 concentrations at
microbial surfaces associated with more biologically active substrates (e.g.
manure, litter). Localized zones of low O2 concentration (hotspots)
are thereby simulated when O2 uptake by any aerobic population is
constrained by O2 diffusion to that population. O2 uptake by
each heterotrophic population also accounts for competition for O2
uptake with other heterotrophs, nitrifiers, roots and mycorrhizae, calculated
from its O2 demand relative to those of other aerobic populations;
DOC oxidation to CO2 under ambient O2 is calculated
from (2) and (3) [H5]. The energy yield of DOC oxidation drives the uptake of
additional DOC for the construction of microbial biomass Mi,h
according to construction energy costs of each heterotrophic population
[A21]. Energy costs of denitrifiers are larger than those of obligately
aerobic heterotrophs, placing denitrifiers at a competitive disadvantage for
growth and hence DOC oxidation that declines with greater use of e-
acceptors other than O2.
Oxidation of DOC and reduction of nitrate, nitrite and nitrous oxide by
denitrifiers
Constraints imposed by NO3- availability on DOC oxidation by
denitrifiers are solved in five steps:
NO3- reduction to NO2- under non-limiting NO3-
is calculated from electrons demanded by DOC oxidation to CO2 but not met by O2
reduction to H2O because of diffusion limitations to O2
supply, and hence transferred to NO3- [H6];
NO3- reduction to NO2- under ambient NO3-
is calculated from (1), accounting for relative concentrations and affinities
of NO3- and NO2- [H7];
NO2- reduction to N2O under ambient NO2- is
calculated from demand for electrons not met by NO3- reduction in
(2), accounting for relative concentrations and affinities of NO2-
and N2O [H8];
N2O reduction to N2 under ambient N2O is
calculated from demand for electrons not met by NO2- reduction in
(3) [H9];
additional DOC oxidation to CO2 enabled by NOx
reduction in (2), (3) and (4) is added to that enabled by O2
reduction from [H5], the energy yield of which drives additional DOC uptake
for the construction of Mi,n. This additional uptake offsets
the disadvantage incurred by the larger construction energy costs of
denitrifiers.
Oxidation of ammonia and reduction of oxygen by nitrifiers
Constraints on nitrifier oxidation of NH3 imposed by O2
uptake are solved in four steps:
substrate (NH3) oxidation under non-limiting O2 is
calculated from active biomass, NH3 and CO2 concentrations
and an Arrhenius function of Ts [H11];
O2 reduction to H2O under non-limiting O2 is
calculated from (1) using set respiratory quotients [H12];
O2 reduction to H2O under ambient O2 is
calculated from radial O2 diffusion through water films of thickness
determined by soil water potential [H13a] coupled with active uptake at
nitrifier surfaces driven by (2) [H13b]. O2 uptake by nitrifiers also
accounts for competition for O2 uptake with heterotrophic DOC
oxidizers, roots and mycorrhizae;
NH3 oxidation to NO2- under ambient O2 is
calculated from (2) and (3) [H14]. The energy yield of NH3 oxidation
drives the fixation of CO2 for the construction of microbial biomass
Mi,n according to construction energy costs of nitrifier
populations.
Oxidation of nitrite and reduction of oxygen by nitrifiers
Constraints on nitrifier oxidation of NO2- to NO3- imposed
by O2 uptake [H15–H18] are solved in the same way as are those of
NH3 [H11–H14]. The energy yield of NO2- oxidation drives
the fixation of CO2 for construction of microbial biomass Mi,o according to the construction energy costs of each nitrifier
population.
Oxidation of ammonia and reduction of nitrite by nitrifiers
Constraints on nitrifier oxidation of NH3 imposed by NO2-
availability are solved in three steps:
NO2- reduction to N2O under non-limiting
NO2- is calculated from electrons demanded by NH3 oxidation
but not accepted for O2 reduction to H2O because of diffusion
limitations to O2 supply and hence transferred to NO2- [H19];
NO2- reduction to N2O under ambient NO2-
and CO2 is calculated from (1) [H20], competing for NO2-
with denitrifiers [H8] and nitrifiers [H18];
additional NH3 oxidation to NO2- enabled by
NO2- reduction in (2) [H21] is added to that enabled by O2
reduction from [H14]. The energy yield from this oxidation drives the
fixation of additional CO2 for the construction of Mi,n.
Uptake of ammonium and reduction of oxygen by roots and
mycorrhizae
NH4+ uptake by roots and mycorrhizae under non-limiting
O2 is calculated from mass flow and radial diffusion between adjacent
roots and mycorrhizae [C23a] coupled with active uptake at root and
mycorrhizal surfaces [C23b]. Active uptake is subject to inhibition by root
nonstructural N : C ratios [C23g], where nonstructural N is the active
uptake product and nonstructural C is the CO2 fixation product
transferred to roots and mycorrhizae from the canopy.
O2 reduction to H2O is calculated from (1) plus oxidation
of root and mycorrhizal nonstructural C under non-limiting O2 using a
set respiratory quotient [C14e].
O2 reduction to H2O under ambient O2 is calculated
from mass flow and radial diffusion between adjacent roots and mycorrhizae
[C14d] coupled with active uptake at root and mycorrhizal surfaces driven by
(2) [C14c]. O2 uptake by roots and mycorrhizae also accounts for
competition with O2 uptake by heterotrophic DOC oxidizers and
autotrophic nitrifiers calculated from their O2 demands relative to
those of other populations.
Oxidation of root and mycorrhizal nonstructural C to CO2 under
ambient O2 is calculated from (2) and (3) [C14b].
NH4+ uptake by roots and mycorrhizae under ambient O2 is
calculated from (1), (2), (3) and (4) [C23b].
Cation exchange and ion pairing of ammonium
A Gapon selectivity coefficient is used to solve cation exchange of
NH4+ vs. Ca2+ [E10] as affected by other cations
[E11]–[E15] and CEC (cation exchange capacity) [E16]. A solubility product is used to equilibrate
soluble NH4+ and NH3 [E24] as affected by pH [E25] and other
solutes [E26–E57].
Soil transport and surface – atmosphere exchange of gaseous substrates
and products
Exchange of all modelled gases γ (γ=O2, CO2,
CH4, N2, N2O, NH3 and H2) between
aqueous and gaseous states is driven by disequilibrium between aqueous and
gaseous concentrations according to a Ts-dependent solubility
coefficient, constrained by a transfer coefficient based on an air–water
interfacial area that depends on air-filled porosity [D14–D15] (Fig. 1).
These gases undergo convective–dispersive transport through soil in gaseous
[D16] and aqueous [D19] states driven by soil water flux and by gas
concentration gradients. Dispersive transport is controlled by gaseous
diffusion [D17] and aqueous dispersion [D20] coefficients calculated from
gas- and water-filled porosity. Exchange of all gases between the atmosphere
and both gaseous and aqueous states at the soil surface are driven by
atmosphere–surface gas concentration differences and by boundary layer
conductance above the soil surface, calculated from wind speed and from the
structure of vegetation and surface litter [D15].
Key soil properties of the Eutri-Stagnic Cambisol at Oensingen as
used in ecosys.
Depth
BDa
TOC
TON
FCb
WPb
Ksat b
pH
Sanda
Silta
Claya
CF
(m)
(Mgm-3)
(gkg-1)
(m3m-3)
(mmh-1)
(gkg-1)
(m3m-3)
0.01
1.21
27.2
2.9
0.38
0.22
3.4
7
240
330
430
0
0.03
1.21
27.2
2.9
0.38
0.22
3.4
7
240
330
430
0
0.07
1.21
27.2
2.9
0.38
0.22
3.4
7
240
330
430
0
0.13
1.24
27.2
2.9
0.39
0.23
3.4
7
240
330
430
0
0.28
1.28
20.2
2.1
0.40
0.24
2.4
7
180
380
440
0
0.6
1.28
11.6
1.1
0.40
0.24
1.4
7
180
380
440
0
0.7
1.28
11.6
1.1
0.40
0.24
1.4
7
180
380
440
0
0.9
1.28
9
0.9
0.40
0.24
1.4
7
180
380
440
0
1.5
1.28
6
0.6
0.40
0.24
1.4
7
180
380
440
0.1
Abbreviations: BD: bulk density; TOC: total organic C; TON: total
organic N; FC: field capacity; WP: wilting point; Ksat: saturated
hydraulic conductivity; CF: coarse fragments. a BD,
TOC and texture were determined from soil cores taken in 2001 and 2006.
Details are given in Leifeld et al. (2011). b FC, WP
and Ksat were estimated from BD, TOC and texture according to
Saxton et al. (1996) and Saxton and Rawls (2006).
Plant and soil management operations at the Oensingen intensively
managed grassland from 2001 to 2009.
Year
Plant management
Soil management
Date
Management
Date
Management
Amount (gm-2)
NH4+
NO3-
ON
OC
7 May
tillage
10 May
tillage
11 May
planting
15 Jun
mineral fertilizer
1.5
1.5
2001
1 Jul
harvest
12 Jul
mineral fertilizer
1.5
1.5
8 Aug
harvest
16 Aug
mineral fertilizer
1.15
1.15
12 Sep
harvest
31 Oct
harvest
2002
12 Mar
mineral fertilizer
1.5
1.5
15 May
harvest
22 May
manure slurry
4.2
2.8
31.2
25 Jun
harvest
1 Jul
mineral fertilizer
1.75
1.75
15 Aug
harvest
18 Aug
manure slurry
5.9
5.3
49.6
18 Sep
harvest
30 Sep
mineral fertilizer
1.5
1.5
7 Dec
harvest
2003
18 Mar
manure slurry
5.9
5.3
61.1
30 May
harvest
2 Jun
mineral fertilizer
1.5
1.5
4 Aug
harvest
18 Aug
manure slurry
6.3
1.9
19.0
13 Oct
harvest
17 Mar
manure slurry
5.0
1.5
19.5
11 May
harvest
17 May
mineral fertilizer
1.5
1.5
2004
25 Jun
harvest
1 Jul
manure slurry
5.5
0.5
9.9
28 Aug
harvest
31 Aug
mineral fertilizer
1.5
1.5
3 Nov
harvest
29 Mar
manure slurry
6.7
3.1
42.0
10 May
harvest
17 May
mineral fertilizer
1.5
1.5
2005
27 Jun
harvest
5 Jul
manure slurry
5.0
3.5
59.6
29 Aug
harvest
16 Sep
mineral fertilizer
1.5
1.5
24 Oct
harvest
2006
24 May
harvest
5 Jul
harvest
13 Jul
manure slurry
4.7
1.4
12.5
12 Sep
harvest
27 Sep
manure slurry
4.4
1.3
13.6
26 Oct
harvest
30 Oct
manure slurry
6.4
3.2
57.8
2007
3 Apr
manure slurry
5.2
4.6
75.1
26 Apr
harvest
3 May
mineral fertilizer
1.5
1.5
6 Jul
harvest
13 Jul
manure slurry
4.9
1.8
45.9
23 Aug
harvest
28 Aug
mineral fertilizer
1.5
1.5
11 Oct
harvest
24 Oct
manure slurry
4.6
3.0
38.9
19 Dec
terminate
19 Dec
plowing
1 May
tillage
4 May
tillage
5 May
planting
2008
1 Jul
harvest
10 Jul
mineral fertilizer
1.5
1.5
29 Jul
harvest
7 Aug
mineral fertilizer
1.5
1.5
8 Sep
harvest
19 Sep
manure slurry
2.9
0.5
8.6
7 Nov
harvest
2009
7 Apr
mineral fertilizer
1.5
1.5
1 May
harvest
12 May
manure slurry
4.4
1.6
26.0
16 Jun
harvest
6 Aug
manure slurry
3.3
1.2
19.0
29 Jul
harvest
7 Sep
harvest
15 Sep
mineral fertilizer
6.5 (urea)
20 Oct
harvest
Field experiment
Site description
The Oensingen field site is located in the central Swiss lowlands
(7∘44′ E, 47∘17′ N) at an altitude of 450 m.
The climate is temperate with an average annual rainfall of about
1100 mm and a mean air temperature of 9.5 ∘C. The soil is
classified as a Eutri-Stagnic Cambisol developed on clayey alluvial deposits,
key properties of which are given in Table 1. Prior to the experiment, the
field site was managed as a ley–arable rotation. In December 2000, the field
was ploughed and left in fallow until 11 May 2001. The field was then sown
with a grass–clover mixture typical for permanent grassland under intensive
management. The field was ploughed again on 19 December 2007, left in fallow
until 5 May 2008, when it was tilled and resown with the same grass–clover
mix as in 2001. The period of study extended from sowing in 2001 to the end
of 2009, during which the field was cut between three and five times per year
and harvested as hay, silage or fresh grass; it was fertilized two to three
times per year with manure as liquid cattle slurry and two to three times per
year with mineral fertilizer as ammonium nitrate (NH4NO3) pellets,
for an average annual N application of 23 gNm-2. All key
management operations during this period are summarized in Table 2.
Soil, plant and meteorological measurements
Soil θ and Ts were recorded continuously using TDR (time
domain reflectometry, ThetaProbe ML2x, Delta-T Devices, Cambridge, UK) and
thermocouples at 5, 10, 30 and 50 cm for θ and at 2, 5, 10, 30
and 50 cm for Ts. Leaf area index (LAI) was measured
weekly with an optical leaf area meter (LI-2000, Li-Cor, Lincoln, NB, USA).
Plants were collected every 2–4 weeks, and the samples were dried for
48 h at 80 ∘C and weighed and analysed for C, N, P and K by
using an elemental analyzer. Hourly climatic data were recorded continuously
with an automated meteorological station, including air temperature
(∘C), rainfall (mm), relative humidity (%), global radiation
(Wm-2) and wind speed (ms-1).
Nitrous oxide flux measurements
N2O fluxes were measured with a fully automated system consisting of
up to eight stainless steel chambers (30cm×30cm×25 cm) (Fléchard et al., 2005; Felber et
al., 2014) fixed on PVC frames permanently inserted 10 cm deep into the
soil. The positions of the chambers were changed about every 2 months. During
measurements, the lids of the chambers were sequentially closed for
15 min every 2 h to allow N2O accumulation in the
chamber headspace. During closure the chamber atmosphere was recirculated at
a rate of 1000 mLmin-1 through polyamide tube lines (4 mm ID)
to analytical instruments installed in a temperature-controlled field cabin
adjacent to the field plots (10 m) and then back to the chamber
headspace. Until autumn 2006, concentrations of N2O, CO2 and
H2O in the head space were measured once per minute with an INNOVA
1312 photoacoustic multi-gas analyzer (INNOVA Air Tech Instruments, Ballerup,
Denmark; www.innova.dk). Interferences in the measurements caused by
overlaps in the absorption spectra of the different gases and by temperature
effects were corrected with a calibration algorithm described in detail by
Fléchard et al. (2005). In autumn 2006, the system was changed to the gas
filter correlation technique for N2O (Model 46C, Thermo 279
Environmental Instruments Inc., Sunnyvale, CA, USA). This system was
calibrated every 8 hours using certified standard gas mixtures (Messer
Schweiz AG, Lenzburg, Switzerland) (Felber et al., 2014).
These measurements were used to calculate N2O fluxes from the rate of
change in concentration by using a linear or non-linear approach determined
by the HMR R-package (Pedersen et al., 2010). The first three of the fifteen
1 min measurements were omitted from the flux calculation to exclude gas
exchange during closing that did not result from changes in emission or
production in the soil. This procedure caused a mean increase of about
30 % in the fluxes compared to values published in Fléchard et
al. (2005) and Ammann et al. (2009), which were evaluated using linear
regression. Fluxes from all chambers were averaged over 4-hourly intervals
and resulting values attributed to the midpoints of the intervals. Standard
errors of these averages were calculated from all fluxes measured during each
interval and thus included both spatial and temporal variation. The fluxes
measured from 2002 to 2003 were summarized in Fléchard et al. (2005).
Those from 2004 to 2007 were re-evaluated from values described in Ammann et
al. (2009). Those from 2008 and 2009 were reprocessed from the EU Project
NitroEurope-IP database using the HMR algorithm.
CO2 and energy flux measurements
CO2 and energy fluxes were measured by an eddy covariance (EC) system
consisting of three-axis sonic anemometers (models R2 and HS, Gill
instruments, Lymington, UK) and an open-path infrared CO2/H2O
gas analyzer (model LI-7500, Li-Cor, Lincoln, USA). The EC system used in
this study is described in Ammann et al. (2007). The EC tower was located in
the centre of the field (52m×146m), whereas the
chambers were located in the south-east corner. For most meteorological
conditions, the chambers were not within the footprint of the EC towers,
although for the main wind directions 80 % or more of the footprint was
within the field (Neftel et al., 2008). The management of the entire field
was uniform throughout the experiment.
Model experiment
Ecosys was initialized with the biological properties of plant
functional types (PFTs) representing the ryegrass and clover planted at
Oensingen. These properties were identical to those in an earlier study
(Grant et al., 2012) except for a perennial rather than annual growth habit.
These PFTs competed for common resources of radiation, water and nutrients,
based on their vertical distributions of leaf area and root length driven by
uptake and allocation of C, N and P in each PFT. Ecosys was also
initialized with the physical and chemical properties of the Eutri-Stagnic
Cambisol at Oensingen (Table 1). The model was then run from model dates
1 January 1931–31 December 2000 under repeating sequences of land management
practices and continuous hourly weather data (radiation, Ta, RH,
wind speed and precipitation) recorded at Oensingen from 1 January 2001 to
31 December 2007 (i.e. 10 cycles of 7 years). This run was long enough for C,
N and energy cycles in the model to attain equilibrium under the Oensingen
site conditions well before the end of the spinup run. The modelled site was
plowed on 19 December 2000, terminating all PFTs.
The model run was then continued from model dates 1 January 2001 to
31 December 2009 under continuous hourly weather data recorded at Oensingen
from 1 January 2001 to 31 December 2009 with the same PFTs and land
management practices as those at the field site listed in Table 2. For each
manure application in the model, an irrigation of 4 mm was added to
account for the water in the slurry. For each harvest in the model, the
fraction of canopy LAI to be cut (usually 0.85–0.95) was calculated from
measurements of LAI before and after the corresponding harvest in the field.
In ecosys, leaves of each PFT are aggregated into a common canopy
which is dynamically resolved into a selected number of layers (10 in this
case) of equal LAI for calculating irradiance interception. The leaf fraction
to be cut was removed from successive leaf layers from the top of the
combined canopy downwards until the cumulative removal attained the set
fraction, so that the LAI cut from each PFT depended on the leaf area of the
PFT in these layers. Of the phytomass cut with the LAI, 0.76 was removed as
harvest and the remainder was added to surface litter, as determined in the
intensively managed grassland at Oensingen by Amman et al. (2009).
N2O emissions modelled from 2004 through 2009 were compared with
those measured by the automated chambers by regressing log-transformed 4 h
averages of modelled on measured values during each year of the study and
also by regressing total emissions modelled vs. measured during emission
events following each fertilizer or manure application. These comparisons
were supported by ones with thermistor and TDR measurements of Ts
and θ and with EC measurements of CO2 and energy exchange.
Model sensitivity studies
Modelled N2O emissions may be affected by three general
sources of uncertainty in model inputs: land management practices, soil
properties and model parameters. To examine the possible effects of some
different land management practices on N2O emissions, the model run
from 2001 to 2009 (field) was repeated with (1) increased harvest intensity,
in which canopy LAI remaining after each harvest was reduced to half of that
in the first run (1/2), and (2) increased harvest intensity with each harvest
delayed by 5 days (1/2 + 5 d). These alternative practices caused canopy
regrowth and hence N uptake to be slower during emission events following
subsequent manure and fertilizer applications.
To examine the possible effects of spatial variability in soil properties on
N2O emissions, the model run from 2001 to 2009 (field) was repeated
with bulk density (BD) of the upper 3 cm in the soil profile
(Table 1) increased by 5 or 10 %. These larger BDs reduced soil porosity
in the upper 3 cm of the soil, thereby slowing gas exchange with the
atmosphere, particularly when the soil was wet (Fig. 1). All other soil
properties used in the model remained unchanged (Table 1).
To examine an effect of uncertainty in model parameterization, the model run
from 2001 to 2009 (field) was repeated with the values of two key parameters
governing N2O emissions, the Michaelis–Menten constants for the
reduction of O2 (KO2 in [H4]) or of NO3- and
NO2- (KNOx in [H7], [H8] and [H20]), halved or doubled
from those used in the model. Halving or doubling KO2 hastened or
slowed the reduction of O2 by nitrifiers and denitrifiers and hence
slowed or hastened the transfer of electrons to reduce NO2- and
NO3- during nitrification and denitrification. Halving or doubling
KNOx hastened or slowed the reduction of NO2- by
nitrifiers and of NO3- and NO2- by denitrifiers. All other
parameters in the model remained unchanged.
Results
LAI modelled vs. measured from 2002 to 2009
Accurate modelling of ecosystem C cycling and hence N2O emissions
requires accurate modelling of plant growth as determined by land management
practices. LAI modelled and measured from 2002 to 2009 rose rapidly from low
values remaining in spring and after each harvest (Table 1) to
4–6 m2m-2 before the next harvest, except during 2003
(Fig. 2). Regrowth of LAI in ecosys was driven by plant
nonstructural C, N and P pools replenished from storage reserves remobilized
after harvests and from products of current C, N and P uptake, those of C
being governed by irradiance interception calculated from regrowing LAI.
Regrowth in the model was less rapid than that measured in 2009 (Fig. 2)
because more frequent cutting forced more frequent replenishment of plant
nonstructural C, N and P pools, which gradually depleted storage reserves and
hence slowed subsequent regrowth. Hence, rates of regrowth modelled after
harvests were affected by harvest timing and intensity, as represented by the
fractions of LAI removed at harvest.
N2O fluxes modelled vs. measured from 2004 to 2009
During peak emissions, standard deviations of N2O fluxes measured
within each 4-hourly interval were found to be as much as 85 % relative
to mean values. These deviations were largely attributed to small-scale
spatial variation in land management (manure and fertilizer application,
surface litter from harvesting) and in soil properties (bulk density, water
retention), which was not represented in the model run, rather than to
temporal variation in environmental conditions (θ, Ts),
which was represented in the model run. Therefore, only a limited fraction of
variation in the measured values was amenable to correlation with modelled
values. Consequently, slopes and coefficients of determination (R2) from
regressions of modelled on measured log-transformed fluxes varied from 0.5 to
1.0 and from 0.1 to 0.5 respectively, while intercepts remained close to zero
(Table 3a). However, ratios of mean squares for regression vs. error (F)
were highly significant (P<0.001) in all years of the study, indicating
some agreement in the timing and magnitude of modelled and measured emission
events. Improved agreement would require that more detailed information about
land management and soil properties at each chamber site be provided to the
model.
LAI measured (symbols) and modelled (lines) from 2002 through 2009
at the Oensingen intensively managed grassland.
Intercepts (a), slopes (b), coefficients of determination
(R2), ratios of mean squares for regression vs. error (F), and number of
data pairs from regressions of (a) log-transformed 4 h averages of
N2O fluxes (mgNm-2h-1) modelled vs. measured
during each year from 2004 to 2009 and (b) total N2O fluxes
(mgNm-2) modelled vs. measured during emission events following
each fertilizer or manure application from 2004 to 2009 (see Fig. 3) at the
Oensingen intensively managed grassland.
Year
a
b
R2
F*
n
(a)
2004
1.25±0.88×10-5
0.49±0.06
0.08
69
818
2005
1.63±0.43×10-5
0.59±0.03
0.24
368
1173
2006
4.28±0.44×10-5
1.04±0.08
0.14
155
948
2007
1.21±0.33×10-5
0.67±0.02
0.35
989
1794
2008
1.44±0.51×10-5
0.44±0.03
0.08
157
1703
2009
-0.03 ± 0.25×10-5
0.71±0.02
0.49
1574
1614
(b)
2004–2009
28±9 mgNm-2
0.67±0.13
0.54
27
23
* All values of F were highly significant (P<0.001).
Daily aggregated N2O emissions measured (symbols) and
N2O and N2 emissions modelled (lines) from 2004 through 2009
at the Oensingen intensively managed grassland. Numbers above and beside each
fertilizer or manure addition indicate total measured or modelled
N2O-N emitted during emission events (mgNm-2) and total
N applied (gNm-2). Negative values indicate effluxes to the
atmosphere.
Daily aggregated N2O fluxes modelled vs. measured from 2004 to
2009
Daily aggregations of both measured and modelled N2O emissions
indicated that emission events during the study period were confined to
intervals of no longer than 5 days when precipitation followed manure or
fertilizer applications (Fig. 3). Outside of these intervals emissions
remained very small except for a period of emissions modelled but not
measured after manure application in autumn 2006 (Fig. 3c) and measured but
not modelled before fertilizer application in spring 2008 (Fig. 3e).
The largest emissions followed manure applications in July and August, but
their magnitudes did not vary with the amount of manure N applied. For
example, emissions during an event in August 2009 (244 vs.
185 mgNm-2 measured vs. modelled in Fig. 3f) were greater than
those during an event in July 2007 (86 vs. 112 mgNm-2 measured
vs. modelled in Fig. 3d), which in turn were greater than those during an
event in July 2005 (54 vs. 96 mgNm-2 measured vs. modelled in
Fig. 2b). However, manure N application preceding the event in August 2009
(4.5 gNm-2) was less than that in July 2007
(6.7 gNm-2), which in turn was less than that in July 2005
(8.5 gNm-2) (Table 2), so that smaller applications were
followed by greater emissions, precluding a simple emission factor for manure
N application.
The magnitude of emission events following fertilizer application also
varied. For example, emissions during an event in late August 2007 (105 vs.
82 mgNm-2 measured vs. modelled in Fig. 3d) were greater than
those during events in September 2004 (24 vs. 2 mgNm-2
measured vs. modelled in Fig 2a) and 2005 (6 vs. 11 mgNm-2
measured vs. modelled in Fig. 3b), although the fertilizer N applications of
3.0 gNm-2 preceding each event were the same (Table 2). These
differences in emissions indicated important differences in ecological
controls imposed by environmental conditions (θ and Ts)
and plant management during each event.
(a) Precipitation and soil temperature at 0.05 m,
(b) soil water content (θ) at 0.05, 0.10, 0.30 and
0.50 m, (c) energy, and (d) CO2 fluxes
measured (closed symbols), gap-filled (open symbols) and modelled (lines)
during 20 days from harvest (cut) to the end of the emission event following
manure application (manure) in July 2007. (e) θ,
(f, g) aqueous concentrations of O2 and N2O modelled
in the surface litter and at 0.01 and 0.02 m in the soil, and
(h) N2O and N2 fluxes measured (symbols) and
modelled (lines) during the last 10 days of this period when the emission
event occurred. For fluxes, positive values represent influxes to the soil,
negative values effluxes to the atmosphere.
The standard deviations of ∼ 85 % relative to the mean values of
fluxes measured within each 4-hourly interval during emission events was used
to estimate an uncertainty in daily aggregated fluxes of ca. 30 %.
Uncertainty in daily fluxes measured during emission events was smaller than
the severalfold differences among the events indicating that the magnitude of
these events likely differed significantly. Regressions of modelled on
measured total emissions during the events following each fertilizer or
manure application from 2004 to 2009 (Fig. 3) gave better agreement than did
those of the 4-hourly averaged fluxes (Table 3b), indicating that modelling
the precise timing of fluxes during these events remains a challenge.
(a) Precipitation and soil temperature at 0.05 m,
(b) soil water content (θ) at 0.05, 0.10, 0.30 and
0.50 m, (c) energy, and (d) CO2 fluxes
measured (closed symbols), gap-filled (open symbols) and modelled (lines)
during 20 days from harvest (cut) to the end of the emission event following
manure application (manure) in August 2008. (e) θ,
(f, g) aqueous concentrations of O2 and N2O modelled
in the surface litter and at 0.01 and 0.02 m in the soil, and
(h) N2O and N2 fluxes measured (symbols) and
modelled (lines) during the last 10 days of this period when the emission
event occurred. Positive flux values represent influxes to the soil, negative
values effluxes to the atmosphere.
Relationships between N2O fluxes and environmental conditions during
selected emission events
Environmental conditions measured and modelled from harvest to the end of the
two largest emission events following manure applications in July 2007
(Fig. 3d) and August 2009 (Fig. 3f) were examined in greater detail to
investigate relationships among near-surface Ts, θ,
aqueous gas concentrations, and surface fluxes of energy, CO2 and
N2O (Figs. 4, 5). In July 2007, several small precipitation events
wetted and cooled the soil between harvesting on DOY 187 and manure
application on DOY 194 (Fig. 4a, b). The soil then dried during several days
without precipitation and warmed with reduced shading from defoliation
(Fig. 2) until DOY 200, after which the soil wetted with further
precipitation and cooled with increased shading from plant regrowth
(Fig. 4a, b). The higher θ measured during this period (Fig. 4b) may
have been caused by difficulties in maintaining the calibration of the TDR
probes over long periods in the high-clay soil at Oensingen (Table 1). This
higher θ was not likely caused by overestimated evapotranspiration
because modelled latent heat (LE) fluxes, reduced by low LAI after harvesting but
increasing with subsequent regrowth, were close to those measured (Fig. 4c),
suggesting that total water uptake was accurately modelled. Comparison of
modelled and measured θ was further complicated by soil cracking which
altered infiltration at low θ. The effects of θ-dependent
macroporosity on preferential flow are explicitly modelled in ecosys
but have not yet been tested in detail.
CO2 influxes were also reduced by low LAI after cutting but recovered
to pre-cut levels by the end of the emission event (Fig. 4d), driving rapid
regrowth of LAI (Fig. 2). Large CO2 effluxes measured and modelled
after manure application indicated rapid Rh and hence O2
demand that persisted for several days. Influxes measured in the field were
reduced from those in the model for several days after manure application,
suggesting temporary interference of CO2 fixation by the manure
application which was not accounted for in the model.
Litterfall from plant growth [C18, C19] and cutting as well as from manure
application caused a litter layer of 1–2 cm to develop on the soil
surface in the model. During the N2O emission event from DOY 200 to
DOY 205 in 2007 (Fig. 3d), several precipitation events (Fig. 4a) wetted the
modelled surface litter and near-surface soil (layers 1 and 2 in Table 1)
(Fig. 4e) without increasing θ at 5 cm (Fig. 4b). This surface
wetting slowed gas exchange with the atmosphere, sharply reducing aqueous
O2 concentrations [O2(s)] (Fig. 4f) and thereby raising
aqueous N2O concentrations [N2O(s)] (Fig. 4g). Between
precipitation events, drying of the surface litter and near-surface soil in
the model allowed the recovery of [O2(s) ] and forced declines in
[N2O(s)]. These rises and declines in [N2O(s)] drove
rises and declines in N2O emissions that tracked those measured in
the chambers (Fig. 4h). These emissions rose immediately with the onset of
precipitation on DOY 200 (Fig. 4a) before wetting occurred at 5 cm
(Fig. 4b), indicating that emissions were driven by surface wetting (Fig. 4e)
combined with rapid O2 demand (Fig. 4d). The net generation of
N2O modelled in each soil zone, calculated from [H8] + [H20] -
[H9], indicated that 0.21 of surface emissions
originated in the surface litter and the remainder in the 0–1 cm soil layer
as indicated by higher [N2O(s)] (Fig. 4g), while the deeper soil
layers were a very small net sink of N2O. Rises and declines in
[N2O(s)] also drove rises and declines in N2 emissions
that persisted until DOY 205, after which more rapid mineral N uptake with
recovering plant growth, driven by rising LAI (Fig. 2) and hence CO2
influxes (Fig. 4d), caused both emissions to return to background levels
(Fig. 4h).
LAI modelled from 2002 through 2009, with LAI after each cut reduced
to half of that estimated from the field experiment without or with a
delay of 5 days at the Oensingen intensively managed grassland.
In 2009, a period of low precipitation with soil drying and warming occurred
between harvesting in late July and manure application on DOY 218 in early
August, followed by heavy precipitation with soil wetting and cooling on
DOY 220 (Fig. 5a, b). LE effluxes and CO2 influxes declined sharply
with LAI after cutting, and did not recover to pre-cut levels by the end of
the subsequent emission event on DOY 224 (Fig. 5c, d), indicating a slow
recovery of plant growth. Slurry application caused brief surface wetting on
DOY 218 (Fig. 5e) and heavy precipitation on DOY 220 caused prolonged soil
wetting at the surface (Fig. 5e) and at 5 cm (Fig. 5b). Wetting
caused declines in [O2(s)] (Fig. 5f) and thereby rises in
[N2O(s)] (Fig. 5g) that were sustained over 3 days. These rises
drove particularly rapid N2O emissions in the model which were
consistent in magnitude with those measured in the chambers (Fig. 5h).
Diurnal variation modelled with soil warming and cooling (Fig. 5a) was not
apparent in the measurements, although modelled values remained within the
large uncertainty of the measured values during the emission event. These
large emissions were enabled in the model by slow plant uptake of manure N
(Table 2) caused by the slow recovery of plant CO2 uptake and hence
growth after cutting (Fig. 5d). The rises in [N2O(s)] also drove
rises in modelled N2 emissions (Fig. 5h). Emissions declined with
surface litter drying on DOY 223 (Fig. 5e), which allowed surface
[O2(s)] to rise (Fig. 5f) and [N2O(s)] to fall (Fig. 5g)
while θ at 5 cm remained high (Fig. 5b), again indicating that
N2O emissions were largely determined by ecological controls in the
surface litter and soil. The net generation of N2O modelled in each
soil zone indicated that 0.48 of surface emissions
originated in the surface litter, 0.48 in the 0–1 cm soil layer and 0.05 in
the 1–3 cm soil layer, while the deeper soil layers were a very small net
sink of N2O, as indicated by near-surface gradients of
[N2O(s)] (Fig. 5g).
Greater N2O emissions were modelled and measured during the event in
August 2009 than in July 2007 (Fig. 5h vs. Fig. 4h), in spite of smaller N
addition (Fig. 3f vs. Fig. 3d; Table 2) and similar θ and
Ts modelled and measured at 5 cm (Fig. 5a, b vs.
Fig. 4a, b). These greater emissions were attributed in the model to
(1) earlier and heavier precipitation after manure application (2 days after
application in Fig. 5a vs. 6 days in Fig. 4a) and (2) slower recovery of
CO2 fixation after defoliation, indicated by slower rises in diurnal
amplitude of CO2 fluxes (Fig. 5d vs. Fig. 4d). Heavier precipitation
in 2009 vs. 2007 drove sustained vs. intermittent surface and near-surface
wetting (Fig. 5e vs. Fig. 4e) and hence sustained vs. intermittent declines
in [O2(s)] and rises in [N2O(s)] (Fig. 5f, g vs.
Fig. 4f, g). Slower recovery of CO2 fixation after cutting in 2009
vs. 2007 slowed removal of added NH4+ and NO3- from soil.
This slower removal, combined with the shorter period between manure
application and precipitation, left larger NO3- concentrations
([NO3-]) in litter and surface soil to drive N2O production
following precipitation [H7]. These model findings indicated the importance
to N2O emissions of surface and near-surface θ after
precipitation and of plant management (intensity and timing of defoliation in
relation to N application) and its effect on subsequent plant CO2
fixation and N uptake.
(a, g) CO2 fluxes, (b, h) cumulative
NH4+ (dashed) and NO3- (solid) uptake since manure
application, (c, i) aqueous NO3- concentrations at
0–1 cm, (d, j) Ts and (e, k) θ
at 5 cm, and (f, l) N2O fluxes measured (symbols)
and modelled (lines) with LAI after each cut reduced to half of that
estimated from the field experiment without or with a delay of 5 days during
emission events following manure applications on DOY 194 in
(a–f) 2006 and (g–l) 2007 (see Table 2). For fluxes,
positive values represent influxes to the soil, negative values effluxes to
the atmosphere.
Effects of intensity and timing of defoliation on N2O emission
events
Increasing harvest intensity and delaying harvest dates slowed LAI regrowth
modelled after harvests (Fig. 6). The effects of this slowing on N2O
emissions during selected events modelled after subsequent fertilizer and
manure applications were examined under diverse θ and Ts
(Figs. 7, 8). Following manure application on DOY 194 in 2006 (Table 2),
slower LAI regrowth from increasing and delaying defoliation slowed the
recovery of CO2 fixation (Fig. 7a) and of NH4+ uptake
(Fig. 7b), allowing more nitrification of manure N and hence greater surface
[NO3-] (Fig. 7c). Slower LAI regrowth (Fig. 6) also reduced shading
and ET, raising Ts (Fig. 7d) and θ (Fig. 7e). N2O
emissions modelled under field management remained small because of soil
drying, in spite of high Ts, consistent with measurements
(Figs. 3c, 7f). Increases in emissions modelled with slower LAI regrowth,
particularly from delayed harvesting (Fig. 7f), were attributed to slower N
uptake (Fig. 7b) and hence larger [NO3-] in litter and surface soil
(Fig. 7c) and to warmer and wetter soil (Fig. 7d, e), which increased
O2 demand while reducing O2 supply.
Following a similar manure application on DOY 194 in 2007 (Table 2; Fig. 6),
slower LAI regrowth from increasing and delaying defoliation also caused
reductions in CO2 fixation (Fig. 7g), which slowed NH4+ and
NO3- uptake (Fig. 7h), allowing more nitrification of manure N and
hence greater [NO3-] (Fig. 7i). Lower LAI also caused increases in
Ts (Fig. 7j) and θ (Fig. 7k). Emissions modelled and
measured under field management in 2007 (Fig. 7l) were greater than those in
2006 (Fig. 7f), in spite of lower Ts (Fig. 7j vs. Fig. 7d),
because near-surface wetting from several precipitation events (Fig. 4a, e)
reduced [O2(s)] and increased [N2O(s)] (Fig. 4f, g).
Emissions modelled with increased and delayed harvesting rose from those with
field harvesting as the emission event progressed (Fig. 7l) because elevated
[NO3-] from the manure application persisted longer during the event
(Fig. 7i).
(a, g) CO2 fluxes, (b, h) cumulative
NH4+ (dashed) and NO3- (solid) uptake since fertilizer
application, (c, i) aqueous NO3- concentrations at
0–1 cm, (d, j) Ts and (e, k) θ
at 5 cm, and (f, l) N2O fluxes measured (symbols)
and modelled (lines) with LAI after each cut reduced to half of that
estimated from the field experiment without or with a delay of 5 days during
emission events following fertilizer applications on DOY 259 in
2005 (a–f) and DOY 240 in 2007 (g–l) (see Table 2). For
fluxes, positive values represent influxes to the soil, negative values
effluxes to the atmosphere.
Following fertilizer application on DOY 259 in 2005 (Table 2), modelled and
measured emissions remained small after soil wetting (Fig. 8f) because lower
Ts (Fig. 8d) slowed soil respiration after wetting, manifested as
smaller measured and modelled CO2 effluxes (Fig. 8a), and so slowed
demand for e- acceptors. Under these conditions, increasing and delaying
defoliation had little effect on modelled N2O emissions (Fig. 8f),
while CO2 fixation (Fig. 8a) and N uptake (Fig. 8b) were only
slightly reduced and surface NO3- only slightly increased (Fig. 8c).
Following the same fertilizer application on DOY 240 in 2007, modelled and
measured emissions were greater than those in 2005 (Fig. 8l) because soils
were warmer (Fig. 8j) with more rapid respiration (Fig. 8g), and because
fertilizer application and subsequent wetting occurred sooner after cutting
(Table 2). Consequently, recovery of CO2 fixation was less advanced
(Fig. 8g), reducing cumulative N uptake (Fig. 8h) and leaving larger
[NO3-] to drive N2O generation during the event (Fig. 8h).
However, reducing LAI remaining after each harvest did not raise N2O
emissions after this application (Fig. 8l) because slower LAI regrowth from
earlier harvests had reduced primary productivity and consequently litterfall
and hence the mass of the surface litter from which much of the emitted
N2O was generated. Consequently, more intense harvests could cause
surface litter later in the year to decline to levels at which the
N2O generation modelled in the litter was reduced.
Annual productivity, N2O emissions, and the effects of defoliation
intensity and timing
In the model, plant management practices affected LAI regrowth (Fig. 6),
CO2 fixation, N uptake, and hence soil [NO3-] and
N2O emissions (Figs. 7, 8). These effects were summarized on an
annual timescale in Table 4. Modelled and EC-derived gross primary
productivity (GPP) remained close to 2000 gCm-2yr-1
during most years except with low precipitation in 2003 and replanting in
2008, indicating a highly productive ecosystem with rapid C cycling and hence
rapid demand for e- acceptors (Table 4). Larger modelled vs. measured GPP
caused larger modelled vs. measured net ecosystem productivity (NEP) in 2003,
2005 and 2007. Harvest removals in the model varied with NEP except during
replanting in 2008 but tended to exceed those recorded in the field,
particularly with low EC-derived NEP in 2005 and 2006. Modelled values were
determined in part by the assumed constant harvest efficiency of 0.76.
Including C inputs from manure applications, modelled and estimated net biome
productivity (NBP) were positive except during replanting in 2008, indicating
that this intensively managed grassland was a C sink unless replanted.
Average annual NBP modelled vs. measured from 2002 to 2009 was 30 vs.
58 gCm-2, with the lower modelled value attributed to greater
modelled harvest removals, particularly in 2006.
Annual gross primary productivity (GPP), ecosystem respiration
(Re), net ecosystem productivity (NEP=GPP-Re), harvest, net biome productivity (NBP) and N2O
emissions derived from EC or chambers and modelled (M) with current land
management (Table 2) and with defoliation increased so that LAI remaining
after harvesting was reduced by half (1/2), with defoliation increased and
delayed by 5 days (1/2 + 5 d). Positive values indicate uptake, negative
values emissions.
Year
2002
2003
2004
2005
2006
2007
2008
2009
Precip. (mm)
1478
817
1158
966
1566
1328
1188
1004
MAT (∘C)
9.56
9.58
8.92
8.67
9.30
9.59
9.30
9.48
GPP (gCm-2yr-1)
EC
2159
1773
2058
1766
1817
2102
1455
2119
M: current
2214
1836
2220
2111
1953
2539
1419
1852
M: 1/2
2064
1764
2054
1969
1865
2285
1305
1705
M: 1/2 + 5 d
2014
1774
2076
1966
1771
2277
1225
1686
Re (gCm-2yr-1)
EC
-1490
-1558
-1541
-1565
-1577
-1684
-1450
-1657
M: current
-1560
-1421
-1704
-1679
-1680
-1935
-1366
-1373
M: 1/2
-1457
-1345
-1569
-1572
-1579
-1714
-1212
-1259
M: 1/2 + 5 d
-1458
-1350
-1541
-1517
-1519
-1679
-1183
-1235
NEP (gCm-2yr-1)
EC
669
215
517
201
240
418
5
462
M: current
654
415
516
432
273
604
53
479
M: 1/2
607
419
485
397
286
571
93
446
M: 1/2 + 5 d
556
414
535
449
252
598
42
451
Harvest (gCm-2yr-1)
field
462
241
401
247
232
448
293
532
M: current
570
314
525
460
421
690
308
487
M: 1/2
561
360
465
497
455
678
314
484
M: 1/2 + 5 d
537
353
579
513
446
686
262
473
C inputs
81
80
29
102
84
160
9
45
NBP (gCm-2yr-1)
field
288
54
145
56
92
130
-279
-25
M: current
165
181
20
74
-64
74
-246
37
M: 1/2
127
139
49
2
-85
53
-212
7
M: 1/2 + 5 d
101
141
-15
38
-110
72
-211
23
N inputs
27.6
22.5
18.5
24.3
21.4
30.1
9.4
20.0
N2O (gNm-2yr-1)
chamber
upper bound
-0.130
-0.050
-0.060
-0.230
-0.020
-0.280
-0.480
-0.510
lower bound
-0.450
-0.180
-0.180
-0.320
-0.060
-0.350
-0.620
-0.680
M: current
-0.302
-0.209
-0.183
-0.193
-0.220
-0.281
-0.326
-0.366
M: 1/2
-0.269
-0.215
-0.250
-0.249
-0.318
-0.312
-0.335
-0.318
M: 1/2 + 5d
-0.284
-0.234
-0.347
-0.352
-0.273
-0.348
-0.327
-0.395
(a, c) Aqueous O2 concentrations, and
(b, d) N2O fluxes measured (symbols) and modelled (lines)
with bulk density (BD) from field measurements and with BD
(0–3 cm) raised by 5 or
10 % following (a, b) manure application on DOY 194 and
(c, d) fertilizer application on DOY 240 in 2007 (see Table 2). For
fluxes, positive values represent influxes to the soil, negative values
effluxes to the atmosphere.
Slower LAI regrowth from increasing and delaying defoliation (Fig. 6) reduced
modelled GPP, Re and hence NEP by 5–10 % during years with
greater productivity. However, increasing and delaying defoliation did not
much affect harvest removals because reduced NEP was offset by greater
harvest intensity, so that NBP was reduced except with replanting in 2008.
Annual N2O emissions were estimated from chamber measurements for
each year of the study by scaling the mean measured fluxes to annual values.
These values are presented in Table 4 as upper boundaries for annual
emissions because flux measurements from which means were calculated were
more frequent during emission events. A lower boundary for annual emissions
was also estimated in Table 4 by replacing missing flux measurements with
zero. Average lower and upper boundaries for annual emissions estimated from
2002 to 2009 were 0.220 and 0.355 gNm-2 respectively vs. an
average annual emission in the model of 0.260 gNm-2 (Table 4).
Modelled emissions were nearer to upper boundaries during years with lower
measured emissions (2003, 2004, 2006) and to lower boundaries during years
with higher measured emissions (2007, 2008, 2009). There was no significant
correlation between annual N inputs and measured or modelled emissions.
Although annual emissions in the model were close to 1 % of annual
N inputs during most years, they were greater in 2008 and 2009 in spite of
smaller N inputs because of the large emission events modelled after summer
applications of fertilizer and manure (Figs. 3e, f, 5h). Annual N inputs
(Table 4), supplemented by 3–6 gNm-2yr-1 modelled from
symbiotic fixation by clover [F1–F26]) were only slightly larger than annual
N removals with harvesting, supplemented by losses of
2–3 gNm-2yr-1 from all other gaseous and aqueous
emissions (N2 from denitrification, NH3 from volatilization,
NO3- from leaching). Consequently, residual soil NO3-,
while present in the model, did not accumulate during the study period, and
so did not drive increasing N2O emissions with sustained N
applications. Modelled and measured annual N2O emissions, if
expressed in C equivalents (∼ 130 gCgN-1), largely
offset net C uptake expressed as NBP (Table 4).
Increasing harvest intensity and delaying harvest dates had little effect on
annual N2O emissions modelled during the first 2 years after planting
in 2001 and 2008 but raised them substantially thereafter (2003–2007)
(Table 4). During this period, annual emissions rose by an average of
24 % with increased harvest intensity and by an average of 43 % with
increased harvest intensity combined with delayed harvest dates. These
increases were attributed to reduced N uptake and to increased Ts
and θ (Figs. 7, 8).
Effects of increased bulk density on N2O emissions
Increasing near-surface (0–3 cm) soil BD by 5 or 10 % at the
beginning of 2001 in the model reduced [O2(s)] after rainfall
events and slowed recovery of [O2(s)] during subsequent drying as
shown following the manure application in July 2007 (Fig. 9a) and the
fertilizer application in late August 2007 (Fig. 9c). These reductions caused
increases in modelled N2O effluxes that varied during emission events
(Fig. 9b, d). Effluxes modelled with increases of 10 % in near-surface BD
were at times double those modelled without (e.g. DOY 201 and 240 in Fig. 9),
indicating that relatively small changes in soil surface properties could at
times cause large changes in emissions. The effects of increased BD on
modelled Ts, θ, CO2 exchange, crop production and
N uptake during these events were small (results not shown). Increasing
near-surface BD by 10 % raised annual N2O emissions by amounts
that increased with annual precipitation from ca. 10 % in drier years
(e.g. 2003) to ca. 50 % in wetter years (e.g. 2006) (Table 5).
Annual N2O emissions modelled with current field management
(Table 2) and soil properties (Table 1) (current), with soil bulk density
(BD) increased by 5 and 10 % to a depth of 3 cm, and with the
Michaelis–Menten constants for reduction of O2 (KO2) and
of NO3- and NO2- (KNOx) halved or doubled from
those used in the model.
Year
2002
2003
2004
2005
2006
2007
2008
2009
Precip. (mm)
1478
817
1158
966
1566
1328
1188
1004
MAT (∘C)
9.56
9.58
8.92
8.67
9.30
9.59
9.30
9.48
current
-0.302
-0.209
-0.183
-0.193
-0.220
-0.281
-0.326
-0.366
BD + 5 %
-0.352
-0.213
-0.218
-0.199
-0.309
-0.332
-0.358
-0.372
BD + 10 %
-0.334
-0.235
-0.231
-0.236
-0.336
-0.374
-0.424
-0.371
N2O (gNm-2yr-1)
KO2 x 0.5
-0.250
-0.179
-0.154
-0.159
-0.160
-0.216
-0.276
-0.349
KO2 x 2.0
-0.390
-0.263
-0.221
-0.247
-0.315
-0.385
-0.381
-0.468
KNOx x 0.5
-0.382
-0.261
-0.265
-0.267
-0.262
-0.378
-0.432
-0.457
KNOx x 2.0
-0.234
-0.163
-0.126
-0.132
-0.126
-0.208
-0.232
-0.288
Effects of changes in KO2 and KNOx on N2O emissions
Lowering KO2 to half that used in ecosys reduced annual
N2O emissions modelled from 2004 to 2009 by 16 % to an average of
0.218 gNm-2yr-1, near the average lower boundary of the
measured values (Table 5). Raising KO2h to double that
used in ecosys increased these emissions by 28 % to an average
of 0.334 gNm-2yr-1, near the average upper boundary of
the measured values. Lowering KNOx to half that used in
ecosys increased annual N2O emissions modelled from 2004 to
2009 by 30 % to an average of 0.338 gNm-2yr-1, near
the average upper boundary of the measured values (Table 5). Raising
KNOx to double that used in ecosys reduced these
emissions by 27 % to an average of 0.189 gNm-2yr-1,
near the average lower boundary of the measured values. In years with lower
annual emissions (2003, 2004, 2006 in Table 4), the lower KO2 or
higher KNOx gave modelled values that were closer to measured
values. However, in years with higher annual emissions (2008 and 2009 in
Table 4), the higher KO2 or lower KNOx gave modelled
values that were closer.
Discussion
Modelled vs. measured N2O emissions
Most N2O emission events measured from 2004 to 2009 were simulated
within the range of measurement uncertainty, estimated to be about 30 %
of mean daily values (Fig. 3). However, some deviations between modelled and
measured N2O emissions were apparent, such as the larger emissions
modelled in autumn 2006 (Fig. 3c) and the smaller emissions modelled in
spring 2008 (Fig. 3e). These deviations may be attributed to uncertainties in
both the measurements and the model. In the automated measurement system, the
static chambers were rotated about every 2 months among fixed positions in a
corner of the field. During these periods, surface conditions in the chamber
could deviate from the mean field conditions represented in the model.
However, we do not have an explanation for the very small emissions measured
after the three manure slurry applications in 2006. The chambers had been
removed before the applications and were reinstalled within 2 h, during
which the cut grass was removed so that the surface litter in the chambers
may have been reduced from that outside. In the model, emissions following
manure or fertilizer applications were sensitive to the amount of surface
litter as noted earlier. The absence of emission events measured after slurry
applications in 2006 was unusual (Fig. 3) given the large precipitation that
year (Table 4), demonstrating that large variability on small spatial scales
inevitably affects these measurements. Such variability adversely affects
agreement between modelled and measured emissions (Table 3).
During spring 2008 sustained emissions of about
5 mgNm-2d-1 were measured by the chambers in the absence
of any manure or fertilizer applications (Fig. 3e). These emissions were
related to the ploughing of the field to a depth of 25 cm in December
2007 (Table 2), which hastened soil organic matter decomposition and hence
N mineralization that increased mineral N substrate for nitrification and
denitrification and possibly for microbial nitrifier and denitrifier
populations. These increases must remain conjectural as the Oensingen study
did not include a stratified analysis of N2O production factors (e.g.
microbial biomass, potential denitrification) within the chamber soils.
Although ecosys simulates hastened soil organic matter (SOM)
decomposition with tillage (Grant et al., 1998), large amounts of above- and
below-ground plant litter with relatively high C : N ratios were
incorporated into the model with tillage in December 2007, which slowed net
N mineralization and hence accumulation of mineral N products in the model
during spring 2008. Consequently, modelled N2O emissions remained
small until mineral N was raised by fertilizer applications in July
(Fig. 3c).
Modelling controls on N2O emissions by litter and
near-surface θ and Ts
In the model, almost all the N2O emissions originated in the surface
litter and in the near-surface (0–1 cm) soil layer, so that
emissions were strongly controlled by litter and near-surface θ and
Ts (Figs. 3, 4). This model finding is consistent with the
experimental finding of Pal et al. (2013) from 15N enrichment studies
that approximately 70 % of N2O measured during emission events in
a managed grassland originated in the surface litter. Similarly van der
Weerden et al. (2013) inferred from diurnal variation in Ts and
N2O emissions measured after urine amendments on a managed grassland
that N2O production was at or near the soil surface
(0–2 cm). Also Fléchard et al. (2007) inferred in a
meta-analysis of N2O emissions from grasslands in Europe that
θ measured at 5 cm was not in some cases an adequate scaling
factor for N2O source strength because N2O production and
emission took place at or near the soil surface. Ecosys simulated
little net production, and even a small net consumption, of N2O in
soil below 2 cm during emission events, as may be inferred from peak
[N2O(s)] modelled in the 0–1 cm soil layer and much
lower [N2O(s)] modelled in the 1–3 cm soil layer below
(Figs. 3g, 4g). This model finding was consistent with the experimental
finding of Neftel et al. (2000) that N2O concentrations below
near-surface soil layers in a managed grassland remained below atmospheric
values during emission events, from which they inferred that any N2O
generated at depths greater than ∼ 3 cm would not likely reach
the soil surface. Thus, attempts to relate N2O emissions to
Ts and θ measured at greater depths than 3 cm in
grasslands are unlikely to be informative if these differ from near-surface
values. These emissions should rather be related to conditions in the litter
and near-surface soil, which need to be better characterized in future
studies.
Consequently, modelled N2O emissions were highly sensitive to surface
wetting and drying (e.g. Fig. 4e, h) modelled from precipitation vs. ET (e.g.
Fig. 4a, c) or to surface warming and cooling (e.g. Fig. 8j, l) modelled from
surface energy balance (e.g. Fig. 4c). The sensitivity to surface wetting and
drying was modelled from the effects of θ on air- vs. water-filled
porosity and hence on the diffusivity of gases in gaseous [D17] and aqueous
[D20] phases and on gaseous volatilization–dissolution transfer coefficients
and hence gas exchange between gaseous and aqueous phases [D14, D15]. These
transfers controlled O2 supply, and hence demand for alternative
e- acceptors as the O2 supply fell below O2 demand, which
drove N2O generation from denitrification [H6–H8] and nitrification
[H19]. The control of O2 supply on e- acceptors used in
nitrification thereby simulated the effect of WFPS on the fraction of
N2O generated during nitrification identified by Fang et al. (2015)
as necessary to modelling N2O emissions, while avoiding the
model-specific parameterization needed in simpler models. The sensitivity to
surface wetting in ecosys enabled sharp rises in N2O
emissions to be modelled from surface litter and near-surface soil after
small precipitation events during DOY 200–201 in 2007 (Fig. 4a, h) and after
slurry application during DOY 218 in 2009 (Fig. 5a, h), even when the soil at
5 cm remained dry (Figs. 4b, 5b). Such rises were consistent with the
experimental findings of Fléchard et al. (2007) that precipitation on dry
soil can cause substantial N2O emissions after fertilizer application
in grasslands.
The sensitivity to surface warming and cooling was modelled from the effects
of Ts on the diffusivity of gases in gaseous [D17] and aqueous
[D20] phases and on the solubility of gases and hence the exchange of gases
between gaseous and aqueous phases [D14, D15], both parameterized from basic
physical relationships independently of the model. These transfers controlled
[O2(s)] in the surface litter and soil (Figs. 3f, 4f) and
hence O2 uptake by aerobic heterotrophs [H4] and autotrophs [H13]
through a Michaelis–Menten constant [H4b, H13b]. The sensitivity to surface
warming and cooling was also modelled from the effects of Ts on
soil organic carbon (SOC) oxidation [H2] and hence O2 demand by
aerobic heterotrophs [H3] and on NH4+ and NO2- oxidation
[H11, H15] and hence O2 demand by aerobic autotrophs [H12, H16].
These effects were driven by a single Arrhenius function used for all
biological transformations [A6] parameterized from basic research conducted
independently of the model. Under sustained high surface θ, this
combination of physical and biological processes drove large diurnal
variation in N2O emissions modelled with diurnal surface warming and
cooling during emission events (e.g. DOY 221 in Fig. 5h, DOY 243 in Fig. 8l),
as observed experimentally by van der Weerden et al. (2013). By explicitly
simulating the diverse processes that determine N2O emissions,
ecosys could model the large sensitivity of emissions to
Ts without the use of unrealistically large parameters for
temperature sensitivity inferred from controlled temperature studies of
N2O emissions (e.g. Dobbie and Smith, 2001). This large sensitivity
to Ts has been inadequately represented in simpler models,
causing the underestimation of large emissions measured from warm soils (e.g.
Saggar et al., 2004). On a seasonal timescale higher Ts could
cause large increases in N2O emissions modelled with comparable
θ after the same fertilizer application (Fig. 8l vs. Fig. 8f).
However, the effects of Ts on N2O emissions were
dominated by those of θ during surface wetting and drying (e.g.
Figs. 4h, 7l).
Values of both θ and Ts thus determined O2 demand
not met by O2 uptake, which drove demand for alternative e-
acceptors by heterotrophic denitrifiers [H6] and autotrophic nitrifiers
[H19]. This demand drove the sequential reduction of NO3-,
NO2- and N2O to NO2-, N2O and N2
respectively by heterotrophic denitrifiers [H7, H8, H9] and the reduction of
NO2- to N2O by autotrophic nitrifiers [H20]. The consequent
production of N2O (Figs. 4g, 5g) and N2 drove emissions of
both N2O and N2 (Figs. 4h, 5h) through volatilization [D14,
D15] and through gaseous and aqueous diffusion [D16, D19]. Ratios of
N2O and N2 emissions in ecosys (Fig. 4h, 5h) were
not parameterized as done in other models but rather were determined by
relative affinities determined from basic research [H8, H9] and by
environmental conditions. When demand from heterotrophic denitrifiers for
alternative e- acceptors was small relative to their availability, the
preferential reduction of more oxidized e- acceptors generated larger
emissions of N2O [H7, H8] relative to N2 [H9]. Such
conditions occurred during the early part of an emission event when surface
[NO3-] rose with nitrification of fertilizer or manure NH4+
after application (e.g. DOY 200–201 in Fig. 4h). However, when demand for
alternative e- acceptors was large relative to their availability, this
same reduction sequence forced a more rapid reduction of N2O to
N2 and hence smaller emissions of N2O relative to N2.
Such conditions occurred during the later part of emission events when
surface [NO3-] declined with plant uptake (e.g. DOY 202–205 in
Fig. 4h and DOY 222 in Fig. 5h) or when greater surface wetting reduced
O2 supply (e.g. DOY 220 in Fig. 5h). This greater demand for
alternative e- acceptors with wetting provided a process-based explanation
for declines in N2O emissions frequently found at higher θ in
field studies (e.g. Rafique et al., 2011) without explicit parameterization
of N2O : N2 ratios.
Nitrification and denitrification were also driven by the concentrations of
NH4+ [H11], NO3- [H7], NO2- [H8, H15, H20] and
N2O [H9] relative to Michaelis–Menten constants evaluated from basic
research. The concentrations of NH4+ and NO3- in
ecosys were increased by N additions from manure and fertilizer N
applications (Table 2) and by net mineralization soil organic N from
oxidation of litterfall, manure and SOM [A26] as indicated by soil
CO2 effluxes. These concentrations were reduced by root uptake of
NH4+ and NO3- [C23] and consequent plant N assimilation
with growth, indicated by more rapid CO2 fixation with time after
cutting (Figs. 3, 4, 6, 7). In the model, more rapid CO2 fixation
drove a more rapid production of nonstructural C, and hence a more rapid
exchange of nonstructural C and N between canopy and roots [C50], and so
hastened root active N uptake by increasing Ra driving root
growth [C14b] and by hastening the removal of N uptake products and hence
reducing their inhibition of active uptake [C23g]. The diversity of controls
on key substrates for N2O generation suggests that robust simulations
of N2O emissions require comprehensive ecosystem models in which
these controls are fully represented.
Modelling effects of defoliation intensity and timing on N2O emissions
The control of NH4+ and NO3- availability by root N uptake
indicated that plant management practices determining uptake would thereby
affect N2O emissions. In the model, increasing harvest intensity and
delaying harvest dates both slowed N uptake (Figs. 7b, h, 8b, h) by slowing
the recovery of LAI (Fig. 6) and CO2 fixation (Figs. 7a, g, 8a, g).
Both thereby increased [NO3-] (Figs. 7c, i, 8c, i), Ts
(Figs. 7d, j, 8d, j) and θ (Figs. 7e, k, 8e, k), raising N2O
effluxes modelled during most emission events (Figs. 7f, l, 8f, l) and hence
annually (Table 4). This model finding was consistent with the field
observations of Jackson et al. (2015) that increased N2O emissions
after defoliation in grasslands were caused by the reduced uptake of N and
water by slower-growing plants.
The effects of defoliation on N2O emissions during modelled emission
events were similar to, or greater than, those of Ts and θ
(e.g. Fig. 7f, l), consistent with the experimental finding of Imer et
al. (2013) that plant management, as represented by its effects on LAI, had a
larger effect on N2O fluxes than did the environment, as represented
by Ta, at an intensively managed grassland in Switzerland.
Reducing LAI remaining after harvest by half and delaying harvest by 5 days
had little effect on modelled harvest removals (Table 4), suggesting that
N2O emissions from managed grasslands are more sensitive to plant
management practices than are yields. The intensity and timing of harvests
should therefore be selected to avoid slow regrowth of LAI following N
additions by avoiding excessive defoliation and by allowing as much time as
possible between defoliation and subsequent fertilizer or manure application.
Neftel et al. (2010) reported enhanced N2O emissions after cuts in
managed grassland and hypothesized that a simple mitigation option would be
to optimize the timing of the fertilizer applications. To our knowledge this
option has not been systematically investigated.
Modelling effects of soil bulk density on N2O emissions
The small increases in near-surface BD included in this study were typical of
those arising from natural variation in soil properties or from compaction by
vehicular traffic during field management operations. In the model, these
increases reduced soil porosity and hence gaseous diffusivity [D17] which
slowed O2 uptake from the atmosphere [D15] and O2 transfer
through the soil profile [D16]. Consequent reductions in near-surface
[O2(s)] (Fig. 9a, c) slowed O2 reduction by denitrifiers
[H4] and nitrifiers [H13], forcing more rapid e- transfer to NO3-
by denitrifiers [H6] and to NO2- by nitrifiers [H19] and hence more
rapid emissions of N2O following applications of manure (Fig. 9b) and
fertilizer (Fig. 9d).
In a study of soil compaction effects on N2O emissions from a
fertilized agricultural field in a climate similar to that at Oensingen,
Bessou et al. (2010) found that increasing the BD of the upper 30 cm
of the soil profile by ca. 15 % raised annual N2O emissions
measured with automated chambers by at least 50 % during each of two
growing seasons. These rises were similar to those modelled with a smaller
increase in BD of the upper 3 cm during the wettest year of this
study (Table 5). During emission events, Bessou et al. (2010) measured peak
fluxes from compacted soil that were double those from uncompacted, as also
modelled here (Fig. 9b, d).
The detailed algorithms from which ecosys was constructed enabled
increases in N2O emissions from surface compaction to be simulated
from specified changes to surface BD, a measurable site characteristic,
without further model parameterization. The marked increases in N2O
emissions modelled with these increases in BD (Table 5) indicated that some
of the large spatial variation in these emissions commonly found in field
measurements could arise from relatively small variation in physical
properties of near-surface soil. In future studies of N2O emissions,
near-surface soil properties could be determined at each measurement site to
establish the extent to which variation in these properties is associated
with those in emissions.
Modelling effects of KO2 and KNOx on N2O
emissions
The value of KO2 used in ecosys (= 2 µM)
was taken from the upper range of values determined experimentally for intact
cells of heterotrophic bacteria by Longmuir (1954). Halving or doubling
KO2 changed modelled N2O emissions (Table 5) by amounts
similar to uncertainty in measured emissions expressed as lower and upper
boundaries of likely values (Table 4), although the doubled value of
KO2 was larger than those derived from experiments. The value of
KNOx used in ecosys (= 100 µM) was
within the range of values determined experimentally by Yoshinari et
al. (1977). As for KO2, halving or doubling KNOx
changed modelled N2O emissions (Table 5) by amounts similar to
uncertainty in measured emissions expressed as lower and upper boundaries of
likely values (Table 4). The halved value of KNOx was closer to
those measured by Betlach and Tiedje (1981) and Khalil et al. (2005), while
the doubled value was closer to that measured by Klemedtsson et al. (1977).
These changes indicate that key parameters used in process models must be
capable of being constrained by accurate evaluation in independent
experiments.