Response of soil nitrous oxide flux to nitrogen fertiliserapplication and legume rotation in a semi-arid climate,identified by smoothing spline models

Sally Jane OfficerA, Frances PhillipsB,F, Gavin KearneyC, Roger ArmstrongD,E,John GrahamA, and Debra PartingtonA

ADepartment of Economic Development, Jobs, Transport and Resources, PMB 105, Hamilton,Vic. 3300, Australia.

BSchool of Chemistry, Faculty of Science, Medicine and Health, University of Wollongong,Wollongong, NSW 2522, Australia.

C36 Paynes Road, Hamilton, Vic. 3300, Australia.DDepartment of Economic Development, Jobs, Transport and Resources, PMB 260, Horsham,Vic. 3401, Australia.

ESchool of Life Sciences, La Trobe University, Melbourne Campus, Bundoora, Vic. 3086, Australia.FCorresponding author. Email: francesp@uow.edu.au

Abstract. Although large areas of semi-arid land are extensively cropped, few studies have investigated the effect ofnitrogen (N) fertiliser on nitrous oxide (N2O) emissions in these regions (Galbally et al. 2010). These emissions need tobe measured in order to estimate N losses and calculate national greenhouse gas inventories. We examined the effect ofdifferent agronomic management practices applied to wheat (Triticum aestivum) grown on an alkaline Vertosol in south-eastern Australia on N2O emissions. In 2007, N2O emissions were measured over 12 months, during which N fertiliser(urea) was applied at sowing or N fertiliser plus supplementary irrigation (50mm) was applied during the vegetative stageand compared with a treatment of no N fertiliser or irrigation. In a second experiment (2008), the effect of source of N onN2O emissions was examined. Wheat was grown on plots where either a pulse (field peas, Pisum sativum) or pasturelegume (barrel medic, Medicago truncatula) crop had been sown in the previous season compared with a non-legumecrop (canola, Brassica napus). To account for the N supplied by the legume phase, N fertiliser (50 kgN ha–1 as urea) wasapplied only to the wheat in the plots previously sown to canola. Fluxes of N2O were measured on a sub-daily basis (upto 16 measurements per chamber) by using automated chamber enclosures and a tuneable diode laser, and treatmentdifferences were evaluated by a linear mixed model including cubic smoothing splines. Fluxes were low and highlyvariable, ranging from –3 to 28 ng N2O-Nm–2 s–1. The application of N fertiliser at sowing increased N2O emissions for~2 months after the fertiliser was applied. Applying irrigation (50mm) during the vegetative growth stage produced atemporary (~1-week) but non-significant increase in N2O emissions compared with plots that received N fertiliser atsowing but were not irrigated. Including a legume in the rotation significantly increased soil inorganic N at sowing of thefollowing wheat crop by 38 kgN ha–1 (field peas) or 57 kg ha–1 (barrel medic) compared with a canola crop. However, N2Oemissions were greater in wheat plots where N fertiliser was applied than where wheat was sown into legume plots whereno N fertiliser was applied. Over the 2 years of the field study, N2O emissions attributed to fertiliser ranged from 41 to 111 gN2O-N ha–1, and averaged of 75 g N2O-N ha–1 or 0.15% of the applied N fertiliser. Our findings confirm that theproportion of N fertiliser emitted as N2O from rainfed grain crops grown in Australian semi-arid regions is less than theinternational average of 1.0%.

Additional keywords: chamber, cubic smoothing spline, fertiliser emissions factor, greenhouse gas, N2O, nitrous oxide,

soil, south-eastern Australia, wheat.

Received 2 March 2012, accepted 17 December 2014, published online 7 May 2015

Introduction

Nitrous oxide (N2O) is a potent greenhouse gas with ~300 timesthe climate-warming effect of an equivalent quantity of carbondioxide (Dalal et al. 2003). The concentration of N2O in the

atmosphere has been steadily increasing since the industrialrevolution, an observation largely attributed to the increasedconcentration of N in agricultural soils (Davidson 2009).Applying synthetic N fertilisers to soil stimulates the soil

Journal compilation � CSIRO 2015 www.publish.csiro.au/journals/sr

CSIRO PUBLISHING

Soil Research, 2015, 53, 227–241http://dx.doi.org/10.1071/SR12049

microbial processes (e.g. nitrification, denitrification) thatproduce N2O emissions and are thought to have generatedmore than half of the present-day atmospheric N2O (Mosieret al. 1998; Kroeze et al. 1999). The Intergovernmental Panelon Climate Change estimates that 1% of N fertiliser applied isdirectly lost as N2O (IPCC 2006), although the estimate islargely based on measurements from temperate or subtropicalclimates in the northern hemisphere (Bouwman et al. 2002a).

Emissions of N2O from rainfed crops in semi-arid regions aretypically lower than those reported for more moist temperateregions (Galbally et al. 2010), with only a small proportion ofthe applied N fertiliser emitted as N2O (Barker-Reid et al.2005; Barton et al. 2008). For example, in the south-west ofAustralia, a fertiliser emissions factor of 0.02% was estimatedfor rainfed wheat grown on an acidic sandy soil (Barton et al.2008). Comprehensive datasets of N2O emissions from rainfed,semi-arid cropping land are rare despite the large area of cropsgrown in semi-arid regions (Barton et al. 2008, 2011; Galballyet al. 2008). Measurements of N2O emission from semi-aridtemperate environments are critical for calculating nationalgreenhouse gas inventories and for developing strategies tomitigate these losses in countries with substantial areas ofsemi-arid cropping.

Accurate quantification of small and variable N2O emissionsfrom semi-arid environments over time necessitates frequentmonitoring and appropriate statistical techniques. This has ledto the development of a continuously operating, automatedchamber system connected to a tuneable diode laser (TDL) tomeasure N2O fluxes on a sub-daily basis. The accuratemeasurement of N2O emissions also requires a relativelylarge gas-collection chamber operating on a short closuretime to minimise the chamber feedback effects on fluxes(Conen and Smith 2000; Venterea et al. 2009) and tointegrate the large spatial variability of N2O emissions thatcan occur over relatively small distances (Parkin 1987).Connecting a chamber system to a TDL provides sufficientsensitivity and precision to measure accurately the small N2Ofluxes expected in relatively dry, semi-arid environments(Drewitt and Warland 2007; Phillips et al. 2007). Thecontinuous monitoring of N2O flux generates a ‘longitudinal’dataset characterised by a strong covariance betweenmeasurements made close together in time (Hyde et al.2006). Therefore, a statistical analysis technique is requiredthat incorporates covariance modelling into the comparison ofthe treatment differences, such as mixed model includingcubic smoothing splines (Orchard et al. 2000). Smoothingsplines offer a sufficiently flexible method of smoothing theextreme variability without rigid fitting to all of the variationsin the flux (Kastanek and Nielsen 2001) and form an intrinsicpart of the linear mixed-model analysis (Verbyla et al.1999;Orchard et al. 2000).

Approximately 60% of Australian wheat is grown in south-eastern Australia (Victoria, New South Wales and SouthAustralia) where ~7.8Mha of land produces 13.5Mt of wheat(5-year average to 2011–12; ABARES 2012). Grain productionin this region is based on a semi-arid rainfed winter croppingsystem and is strongly dependent on rainfall (soil water) and Nsupply (O’Leary and Connor 1997; Connor 2004; Sadras andRodriguez 2010). The N requirements of crops can be supplied

via applications of fertiliser (usually granular urea) or N2 fixationby legumes (e.g. grain legumes or pasture legumes).

We hypothesised that in the dryland cropping systems ofsouth-eastern Australia: (i) soil-based N2O emissions willincrease if N fertiliser is applied, (ii) the rate of N2Oemissions will increase further if rainfall also increases duringthe vegetative period, and (iii) soil N2O emissions will behigher in rotations where the N requirements of non-legumecrops are supplied as fertiliser rather than via a preceding, N2-fixing legume (pulse or pasture) phase. To test the first andsecond hypotheses, we measured N2O emissions in wheat plotswithout or with N fertiliser (as granular urea) applied at sowingand with or without supplementary irrigation applied duringvegetative growth stages. To test the third hypothesis, wemeasured soil-based N2O emissions in the second phase of arotation where the first phase comprised a non-legume crop(canola, Brassica napus), an N2-fixing pulse crop (field peas,Pisum sativum) or a N2 fixing annual pasture legume (barrelmedic, Medicago truncatula).

Methods and materials

Site and soil

The site was near Horsham (368450S, 142880E) in the Wimmeraregion of Victoria, Australia, in a field with a long history ofcropping. The region has a Mediterranean climate with hot,dry summers and cool to mild winters that are relatively wet(Cawood 1996). The Wimmera region is classed as semi-arid,with a precipitation : potential evapotranspiration ratio of 0.32(Saxton et al. 1992; Galbally et al. 2008) and an average annualrainfall of 447mm (Bureau of Meteorology 2012).

The soil is known locally as a ‘Kalkee clay’ (Martin et al.1996), and is classified as an Epicalcareous-Endohypersodic,Grey Vertosol (Australian Soil Classification; Isbell 2002). TheKalkee clay has a light clay topsoil (10 cm) and medium orheavy clay subsoil. The clay has a high content of smectite,which swells in response to water and shrinks as the profiledries. Over time, blocks of subsoil are forced upwards to form acharacteristic ‘gilgai’ mound and hollow micro-relief, althoughthese have subsequently been smoothed by cultivation. Selectedsoil properties determined at the commencement of the studyare listed in Table 1.

Experimental design and crop management

Two experiments were conducted over two consecutive growingseasons in 2007 and 2008. In both years, a randomised blockdesign consisting of three treatments replicated three times wasused, providing nine experimental plots (each 7m by 15m).

In 2007, wheat (cv. Caliph) was sown into broad bean (Viciafaba) stubble on 5 June, with 15 kg phosphorus (P) ha–1. Threetreatments were subsequently imposed: (1) rainfed wheat grownwith no N fertiliser or supplementary irrigation; (2) rainfedwheat with 50 kgNha–1 (as urea) side-banded at sowing andno supplementary irrigation; (3) wheat with 50 kgN ha–1 (asurea) side-banded at sowing, and supplementary drip irrigation(50mm) applied once, on 6 and 7 September at the mid-tilleringgrowth stage. The wheat crop was harvested on 12 December2007.

228 Soil Research S. J. Officer et al.

In 2008, the experiment was moved to an adjacent set of plotswithin the field used in 2007. The plots had been sown theprevious growing season (2007) to barrel medic, field peas orcanola. On 6 June 2008, winter wheat (cv. Carra) was sown intoslashed stubble with P fertiliser (15 kg P ha–1) and the followingthree treatments were imposed: (4) barrel medic stubble and noN fertiliser; (5) field pea stubble and no N fertiliser; (6)canola stubble with N fertiliser (50 kgN ha–1) side-banded.The wheat crop was harvested on 12 December. Because theunusually dry seasonal conditions (growing season rainfallJune–November 195mm compared with long-term average268mm) were likely to result in total crop failure, i.e. nograin produced, all treatments were irrigated with trickle-tapeon 3 October (14mm) and 15–16 October (30mm). This amount

of irrigation was sufficient to ensure grain was harvested, butthe total water available to the crop (rainfall + irrigation) wasbelow the long-term average.

Throughout the 2-year study, cultivation (minimum tillage)and fertiliser rates represented commercial farming practices inthe region. Herbicide and insecticide sprays were applied asrequired.

Nitrous oxide measurements

Soil N2O was measured with automated gas measurementchambers (0.8m by 0.8m wide by 0.5m high) constructedfrom stainless steel and clear polycarbonate panels (Fig. 1).Extensions (0.5m) of similar construction were added toincrease the chamber height as the crop grew. The topsurface panel was divided into two lids that opened awayfrom the top of the chamber to reduce rain shadowing. Thechambers were clamped to an open, stainless-steel base unit(0.8m by 0.8m by 150mm depth) set ~70mm into the soil.Two base units were installed in each plot, 0.5m apart,positioned ~1m and 2.3m from one end of the plot. Two80-mm-diameter fans constantly mixed the air in each of thechambers, which were pressure-vented with a 1.7m length of6-mm-diameter tubing (Meyer et al. 2001). The chambers weremoved between the two bases every week. Chamber airtemperature (thermocouple type T; OneTemp Pty Ltd,Adelaide, S. Aust.) and rainfall (CS700-L; CampbellScientific Australia, Townsville, Qld) were monitoredcontinuously. To ensure that soil–water content under thechambers was similar to the surrounding crop, the chamber

Table 1. Soil properties before sowing winter wheat in a Vertosol atthe emissions monitoring site in the Wimmera region of Victoria, south-

eastern AustraliaBD, Bulk density; PWP, permanent wilting point; FC, field capacity; EC,

electrical conductivity; ESP, exchangeable sodium percentage

Depth BD PWP FC pH EC ESP(m) (g cm–3) (%, v/v) (%, v/v) (dSm–1) (%)

0.0–0.1 1.12 23 41 7.9 0.24 4.20.1–0.2 1.23 27 46 8.0 0.28 4.90.2–0.4 1.26 27 46 8.1 0.33 6.30.4–0.6 1.32 31 53 8.4 0.46 8.50.6–0.8 1.34 33 60 8.5 0.64 10.90.8–1.0 1.39 36 63 8.6 0.86 12.71.0–1.2 1.37 36 64 8.7 0.84 14.3

Fig. 1. Nitrous oxide automated gas collection chambers with lids open, at Horsham, Victoria, 2008. Thechambers are mounted on 0.5-m-high extensions. The crop is winter wheat, just before the anthesis growth stage.

Nitrous oxide in semi-arid wheat Soil Research 229

lids opened automatically during the closure time if thetemperature in the chambers exceeded 508C or when thesite received rain.

Air was sampled from three chambers over a 30-minclosure period. At the start of a closure period, chambersfor the three replicates of one treatment would close and thechambers were sampled sequentially. This allowed the flux foreach treatment to be measured every 90min. The N2O mixingratio in the air sample was measured using a TDL (TGA100A;Campbell Scientific Inc., Logan, UT, USA). A sample deliverysystem continually drew air from the chambers into anautomatically switching manifold (Campbell Scientific,Inc.), which distributed air into tube driers (PD1T Naphion;Campbell Scientific, Inc.) connected to the TDL and to aclosed-path CO2 analyser (LI-820; LI-COR Inc., Lincoln,NE, USA).

During a closure period, the air from each closed chamberwas sampled for 30 s in a 3-min measurement cycle. Thisallowed each chamber to be sampled 10 times over the closureperiod. Each measurement cycle included a working standardgas (30 s) (instrument grade air; BOC Australia, Sydney)followed by two chamber air samples, followed by thestandard gas. The N2O (ppmV) mixing ratio of the standardgas was used to correct the mixing ratio of the sampled airfor the effect of instrument oscillation. Typical instrumentoscillation was 0.38% CV (s.d./mean, %) in 24 h, asdetermined by measuring the N2O mixing ratio in thestandard gas every 30 s for 24 h (0.3172 ppmV� 0.0012 s.d.). For the standard gas, correction factors were derivedfrom the known and measured mixing ratios. A correctionfactor for the time the air was sampled was derived from thelinear interpolation of the standard gas correction factors,measured before and after the air was sampled.

The corrected N2O sample gas mixing ratio (ppmV N2O(g))was converted to density of N(g) (mg N2O-Nm–3):

NðgÞ ¼ N2OðgÞ � ðP � 2MwÞðR � TÞ ð1Þ

where P is atmospheric standard air pressure of 101.31 kPa,Mwis the molecular weight of N, R is the universal gas constant(8.314 JK–1mol–1), and T is chamber air temperature (Kelvin).The N2O flux (ng N2O-Nm–2 s–1) was calculated by the methodof least-squares regression to determine the linear increasein N2O-N density during the chamber closure period (10measurements over 30min), adjusted for the chambervolume : surface area ratio. Based on a t-test, flux values werediscarded if there was no significant fit (a= 0.033) of the N2Odensity m–2 values to the predicted linear model (examples inFig. 2). Flux values were also discarded when the chamber lidsopened during the closure cycle because of rain or excess heat inthe chambers, or during site maintenance or equipment failure. Amaximum of 16 flux values per day could be collected froma single chamber, providing 48 values per treatment per day.Excluding days when no values were collected from anytreatment because of agronomic operations or sitemaintenance closing the site, on average, 63% of all possibleflux values were accepted for statistical analysis over the twomonitoring periods (Fig. 3). Acceptance rates were related to theamount of flux, with more values passing the significance testduring the first half of the monitoring periods, when the soil wasrelatively wet during winter and flux values were generallylarger.

Soil analyses

The soil was characterised (to a depth of 1.2m) for bulk density(BD), soil water content, pH (1 : 5w/v, CaCl2), electricalconductivity (EC; 1 : 5w/v, in deionised water), exchangeablesodium percentage, total N and carbon (LECO FP-2000;LECO Corp., St. Joseph, MI, USA) and inorganic N (air drysoil <2mm, 2M KCl extraction, NO3 +NH4) (Rayment and

Chamber 5, 2.00 am 18 July

373

374

375

376

377

378

379

Chamber 5, 0.30 am 18 July

y = –0.0003x + 375.15

R2 = 0.0446

y = 0.0028x + 373.6

R2 = 0.9587

3740 500 1000 1500 2000 0 500 1000 1500 2000

374

375

375

376

376

377

Time (s)

N2O

-N µ

g m

–3

Measured Predicted Linear trend

Fig. 2. Individual flux calculations and significance testing for one chamber on 18 July 2008 at 00 : 30 and 02 : 00,calculated by fitting 10 gas density values measured over 30min to a linear trend. Flux values were tested for a significantlinear trend based on a t-test (slope/standard error >2.12). Graphs show the 00 : 30 measurement cycle where the fluxcalculation failed the t-test, leading to a discarded value, and the following measurement at 02 : 00, in which the fluxcalculation passed the t-test, leading to acceptance of a flux value of 1.388 ng N2O-Nm–2 s–1.

230 Soil Research S. J. Officer et al.

Higginson 1992). The permanent wilting point (PWP) and fieldcapacity of the soil were determined by the pressure platemethod at –1500 kPa (Klute 1986), using a sieved sample(<2mm) to avoid difficulties of defining soil volume due toshrink–swell properties associated with Vertosols. Volumetricsoil-water content in the surface soil of each chamber base wascontinuously monitored by calibrated theta probes (60mmdepth; Theta-Probe MK2x; Delta-T Devices Ltd, BurwellUK) and converted to water-filled pore space (WFPS) basedon an assumed particle density of 2.65 (Linn and Doran 1984).

Crop measurements

Crop dry matter was determined at anthesis and grain maturity,in quadrats (four 1-m rows at two random places per plot). Grainyield components were measured at grain maturity in quadratsand in the chamber bases. Plant material was dried (808C) andweighed. The grain and straw were then separated and grainweights and screenings were measured. The plant material wasground (<2mm) and analysed for N content by the Dumasmethod (LECO FP-428 nitrogen analyser).

Statistical analyses

Nitrous oxide data (sub-daily values) were assessed by a linearmixed model including cubic smoothing splines (Verbyla et al.1999), using the software program ASReml (Gilmour et al.2006). This cubic-spline approach handles complex longitudinaldata, such as presented here, by modelling the covariancestructure at the chamber level and by partitioning trends intosmooth and non-smooth components, which catered for bothlong-term trends and short-term fluctuations, such as moving thechamber between bases at weekly intervals. The model wasfitted to test for the effects of treatment with an autoregressivecorrelated structure of time, with additional partitioning forchamber effects and measurement time. Random day wasalso used in the 2007 analysis, which accounted for some

short-term variation in chamber integrity before protocolswere finalised at the start of the experiment.

The linear mixed model analysis fitted cubic smoothingsplines to the flux data for each chamber, because there wassignificant interaction between the fixed effects of treatment andtime. The intercept derived from the linear model indicateddifferences in the total flux from each treatment, whereas theslope indicated the rate at which flux diminished during themonitoring period.

The spline-based model accurately described the changes inN2O emissions, although it reduced the temporal oscillations inthe flux. The random effects component identified a highlysignificant treatment� spline interaction in both years, whichindicated that the timing of the episodes of increased flux wereunique to each treatment (Table 2). The model predicted fluxvalues with confidence intervals for each treatment, whichprovided a basis for comparing the variations in the flux fromeach treatment over time.

Values of N2O fluxes exhibited a large skew, which is atypical feature of N2O data and N-related soil processes ingeneral (Parkin 1987; Kaiser et al.1998; Flechard et al.2005), and which necessitated log-transformation beforetreatment comparisons could be made. The data weretransformed to correct for skewness using the natural log,with 4.0 added before transformation for 2007 data and 2.7added to all 2008 data.

Agronomic data was analysed using the general ANOVAprocedure, modelling for the effect of agronomic treatment ineach year (P < 0.05) (GENSTAT Release 12.1; VSN InternationalLtd, Hemel Hempstead, UK).

Results

Rainfall

In 2007, the regional rainfall for the calendar year was slightlybelow the long-term average (decile 4). Rainfall at the site duringthe crop season (June–November) was 243mm (Fig. 4)

006-Jun-07 05-Aug-07 04-Oct-07 03-Dec-07 07-Jun-08 06-Aug-08 05-Oct-08 04-Dec-08

10

20

30

40

50

60

70

80

90

100A

ccep

tanc

e of

flux

val

ues

(%)

Treatment 1 Treatment 2 Treatment 3

Fig. 3. Daily acceptance rates of N2O flux values measured by automated gas measurement chambers at trial sites 1(2007) and 2 (2008) in south-eastern Australia, showing the daily proportion of 48 measured flux values that wereaccepted for statistical analysis for each treatment, where 100% indicates that no values were discarded from the threereplicate chambers each operating on 16 closure cycles per day.

Nitrous oxide in semi-arid wheat Soil Research 231

compared with the long-term mean for this period of 269mm(Bureau of Meteorology 2012). Rainfall was further reduced in2008 (Fig. 5). The regional annual rainfall for the calendar yearwas 315mm (decile 3), and rainfall at the site during the cropseason was only 195mm (June–December). Spring (September–November) was particularly dry, with only 44mm rain, which isone-third of the long-term average spring rainfall (Bureau ofMeteorology 2012).

Inorganic N and water

In 2007, the soil profile (0–120 cm) at sowing of the wheat crop,following the previous N2-fixing crop, contained 64 kg ha–1 ofinorganic N (Table 3). In 2008, the amount of inorganic Npresent in the soil profile before seeding varied depending on thecrop grown in the preceding growing season (Table 3). Forexample, in the surface 600mm, treatment 4 (barrel medic)and treatment 5 (field peas) contained more inorganic N thantreatment 6 (canola), by ~57 and 38 kg ha–1, respectively.Consequently, N fertiliser (50 kgN ha–1) was applied totreatment 6 at sowing to give similar inorganic N for thethree treatments.

In 2007, plant-available soil water in the profile at sowingwas limited to the surface 0.4m of soil. At 0.4–0.8m depth, thesoil water was at or below plant PWP, a result of many years ofdrought (Table 1). The WFPS was <60% until rain fell in earlyJuly (Fig. 4), then remained >60% until late August when soilwater began to decrease. In the rainfed treatments (1 and 2), thetopsoil (0–0.1m) continued to dry until PWP was reached inOctober. In treatment 3, WFPS briefly increased in Septemberto >60% for 10 days following the application of 50mm ofsupplementary irrigation.

In 2008, soil-water conditions were marginal for crop growth.The soil profile was relatively dry before sowing, with plant-available water only in the surface 0.1m. The remainder of theprofile was at or below PWP (crop lower limit). Shortly aftersowing, 26mm rain on 9 and 10 June increased the WFPS of

the surface 0.1m to >60% (Fig. 5). WFPS then decreased untilJuly, when the soil-water content again increased to reach atleast 60% WFPS, where it remained until 12 September, brieflyexceeding 75% WFPS at the beginning of August andSeptember. The topsoil subsequently dried out in the dryspring conditions. Despite supplementary irrigation on 3 and15–16 October, soil-water content declined rapidly and returnedto PWP by late October.

Crop yield and grain N content

In 2007, there were no significant treatment effects on thedry matter production of wheat at mid-tillering or at anthesis(Table 4). At maturity, grain yields were substantially reducedin treatments where N fertiliser was applied without irrigationrelative to the other treatments. There was no significantdifference in N concentration of the grain between treatments;however, N fertiliser increased the N concentration of the straw.

In 2008, seedling emergence of wheat was poor across alltreatments, reflecting marginal soil-water content at sowing.Soils remained relatively dry, despite the application of asmall quantity of irrigation at anthesis, so that plant growthwas stunted and grain-set was poor. Yields were small and notsignificantly different between treatments (Table 4). Removalof N in the grain was only about one-third of that recorded in2007, corresponding to the reduced yields.

Nitrous oxide emissions

During the 2007 crop season, flux values ranged from –2.59 to+28.48 ng N2O-Nm–2 s–1 over all treatments and replicates,with a mean flux value of 1.94 ng N2O-Nm–2 s–1. Generally,fluxes were higher in the first half of the monitoring period forall treatments (Fig. 4). Flux values were skewed and had a highdegree of kurtosis (Table 5), with the median emission rate(1.16 ng N2O-Nm–2 s–1) considerably less than the mean.

In 2007, N2O emissions were significantly influenced bythe three different management treatments, as indicated bythe standard error of the intercept of the linear models(Table 6). Applying N fertiliser (treatment 2) increasedemissions significantly relative to the control treatment withno N applied (treatment 1). There were two distinct episodesof increased flux in treatments where N was applied. First, N2Ofluxes increased in treatment 2 (compare Fig. 6a, b) at ~5 daysafter sowing and following application of the N fertiliser(5 June), with the difference between the treatments peaking~11 days after fertiliser application. Fluxes from the twotreatments then became similar until the end of June. Second,in early July, N2O fluxes increased from treatment 2 andremained high compared with treatment 1 throughout most ofJuly.

Significant differences in N2O flux were recorded betweentreatment 2 (with N fertiliser) and treatment 3 (with Nfertiliser + supplementary irrigation) despite both treatmentsreceiving identical treatment before the application ofirrigation in early September. In June, when fluxes initiallyincreased in treatment 2, the flux from treatment 3 did notsignificantly differ from that of treatment 1 (Fig. 6a, c). InJuly, with the second increase in flux from treatment 2, the fluxalso increased from treatment 3, with no significant difference

Table 2. Summary of terms in the final model of the linear mixedmodel analysis using smoothing splines of N2O emissions measuredthroughout the 2007 and 2008 crop seasons (June–December) from soil

under three agronomic treatments of wheat

Analysis of variance Degreesof freedom

Significance2007

Degreesof freedom

Significance2008

Fixed effectsIntercept 1 P< 0.001 1 P< 0.001Time 1 P< 0.001 1 P< 0.001Treatment 2 P< 0.05 2 P< 0.01Treatment� time 2 P< 0.001 2 P< 0.05

Random effectsChamber P< 0.05 P< 0.05Chamber� dateA P< 0.001 P< 0.001Chamber� dayB P< 0.001 –

Spline (time) P< 0.001 –

Treatment� spline (time) P< 0.001 P< 0.001

ADate is a factor numbered 1–2923, which represent short-term deviations ofmeasurement times at the chamber level.

BDay is a factor numbered 1–185, which represents short-term deviations ofcalendar day at the chamber level.

232 Soil Research S. J. Officer et al.

between treatments 2 and 3 (Fig. 6b, c). Following theapplication of irrigation on 6 September, the flux fromtreatment 3 initially tended to be greater than from treatment2 for ~1 week, although this difference was not significant

(P< 0.05). The period of increased flux commencedseveral days after the irrigation, and 1 day after rain, with atotal of 15.8mm recorded between 10 and 17 September. Thesupplementary irrigation appeared to magnify the effect of

0

5

10

15

20

25

30

35

Dai

ly r

ainf

all (

mm

) 50 mm irrigationapplied to

treatment 3

0

20

40

60

80

100

WF

PS

(%

)

Treatment 1

Treatment 2

Treatment 3

–202468

101214161820

Flu

x N

2O (

ng N

m–2

s–1

) F

lux

N2O

(ng

N m

–2 s

–1)

Flu

x N

2O (

ng N

m–2

s–1

)

Treatment 1

–202468

101214161820

Treatment 2

–202468

101214161820

Treatment 3

06-Jun-07 06-Jul-07 05-Aug-07 04-Sep-07 04-Oct-07 03-Nov-07 03-Dec-07

Fig. 4. Daily rainfall, water-filled pore space (WFPS, 0–60mm) and average daily N2O flux at trial site 1 in south-easternAustralia. Emissions were monitored between 6 June 2007 and 12 December 2007, from soil growing wheat in faba beanstubble, in treatments of: (1) no N fertiliser applied, (2) 50 kgNha–1 applied at sowing on 5 June 2007, and (3) 50 kgN ha–1

applied at sowing with supplementary irrigation applied on 6 September 2007. Bars on the daily N2O flux show the standarddeviation of three replicate measures around the arithmetic treatment mean.

Nitrous oxide in semi-arid wheat Soil Research 233

the rain event. The lower flux from treatment 2 during this periodprobably reflected the lower soil-water content due to a greaterrate of crop growth at anthesis in response to the N fertiliser. Thedifferences in flux were small for this period. The total flux for

September from treatment 3 was 42 g N2O-N ha–1, comparedwith 31 g N2O-N ha–1 for treatment 1 and 21 g N2O-N ha–1 fortreatment 2, with 11–22 g N2O-N ha–1 being attributed toirrigation (Fig. 6). Emissions were not significantly different

0

5

10

15

20

25

30

35

Dai

ly r

ainf

all (

mm

)

14 mm irrigation

30 mm irrigation

0

20

40

60

80

100

WF

PS

(%

)

Treatment 4

Treatment 5

Treatment 6

–101234567

Treatment 4

–101234567

Treatment 5

–101234567

Treatment 6

Flu

x N

2O (

ng N

m–2

s–1

) F

lux

N2O

(ng

N m

–2 s

–1)

Flu

x N

2O (

ng N

m–2

s–1

)

07-Jun-08 07-Jul-08 06-Aug-08 05-Sep-08 05-Oct-08 04-Nov-08 04-Dec-08

Fig. 5. Daily rainfall, water-filled pore space (WFPS, 0–60mm) and average daily N2O flux at trial site 2 in south-easternAustralia. Emissions were monitored between 7 June and 12 December 2008 from soil in agronomic treatments of wheatgrowing in: (4) medic stubble, (5) field pea stubble, and (6) canola stubble with 50 kgNha–1 applied at sowing on 6 June.Equipment breakdown caused the period of missing data in October and November. Bars on the daily N2O flux show thestandard deviation of three replicate measures around the arithmetic treatment mean. Supplementary irrigation was appliedto all treatments on 3 and 15–16 October.

234 Soil Research S. J. Officer et al.

between the three treatments from October until harvest inDecember, reflecting the dry spring.

During the 2008 crop season, flux values ranged from –1.65to +11.4 ng N2O-Nm–2 s–1, with a mean of 1.06 ng N2O-Nm–2 s–1. The sub-daily flux values were less skewed andhad less kurtosis than the 2007 fluxes, so the medianemissions rate (0.72 ng N2O-Nm–2 s–1) was closer to themean (Table 5).

Fluxes of N2O were greatest in treatment 6 (wheat grown incanola stubble with N fertiliser); emissions from the twotreatments previously sown to annual medic (treatment 4) andfield peas (treatment 5) were not significantly different fromeach other (Fig. 7). The flux from treatment 6 increased ~1 weekafter fertiliser application (6 June) (Fig. 7c). This increasepeaked 14 days after fertiliser application and remained highthroughout June, with no significant corresponding change influx recorded for treatments 4 and 5 (Fig. 7a, b). The flux fromtreatment 6 increased to a second maximum in mid-July beforedecreasing during the second half of July. The flux from

treatments 4 and 5 also increased, decreasing the relativedifference, with no significant difference between the threetreatments by the end of July. After July, there was only oneshort period of greater flux from treatment 6 (for 1 week in thefirst part of August). Flux remained low for the rest of the seasonacross all treatments, corresponding to low rainfall, increasingtemperatures and drying of the soil (Fig. 5).

Missing data in October and November 2008 were due tofailure of the TDL (Fig. 5). Although all the plots were irrigatedon 15–16 October (to ensure survival of the crop), the topsoilWFPS increased only briefly to a maximum of 65% immediatelyfollowing irrigation and then remained at <30% duringNovember. Hence, fluxes from all treatments were expectedto be low for the period data were missing, with limited impacton results.

Comparing fluxes from 2007 and 2008, the dryer conditionsin 2008 resulted in lower fluxes across all treatments. However,the total N2O flux attributed to the application of N fertiliserwas similar in 2007 and 2008. Although emissions were higherin 2007, during 2008 the reduction in N2O fluxes from theunfertilised treatments was greater than from the fertilisedtreatment (Table 7), with the duration of the relative increase

Table 4. Plant and grain characteristics of winter wheat grown in asemi-arid climate in the Wimmera region of Victoria in 2007 and 2008Treatments were wheat grown in 2007 with (1) no N fertiliser, (2) 50 kg ha–1

of N fertiliser, and (3) 50 kg ha–1 N-fertiliser and supplementary irrigation;and wheat sown in 2008 into (4) medic stubble, (5) field pea stubble, and(6) canola stubble with 50 kg ha–1 of N fertiliser applied at sowing. l.s.d.,

Least significant difference; n.s., no significant difference

2007 treatment 1 2 3 l.s.d.

Mid-tillering plant material (t ha–1) 1.08 0.89 0.96 n.s.Anthesis plant material (t ha–1) 7.5 6.8 7.6 n.s.Harvest straw (t ha–1) 5.9 5.2 6.0 n.s.Harvest straw (% N) 0.75 0.90 0.87 0.11Grain (t ha–1) 3.1 2.6 3.6 0.6Grain (%N) 2.5 2.6 2.5 n.s.Total N uptake (kg ha–1) 121 114 142 16

2008 treatment 4 5 6

Mid-tillering plant material (t ha–1) 1.2 1.1 1.0 n.s.Anthesis plant material (t ha–1) 5.0 4.3 4.5 n.s.Harvest straw (t ha–1) 5.0 4.4 4.9 n.s.Harvest straw (%N) 1.0 1.0 1.1 n.s.Grain (t ha–1) 0.82 0.48 0.70 n.s.Grain (%N) 3.3 3.3 3.3 n.s.

Total N uptake (kg ha–1) 74 61 77 n.s.

Table 5. Descriptive statistics of sub-daily nitrous oxide emissionvalues (ng N2O-Nm–2 s–1) collected from nine chambers arranged inthree replicates measuring three treatments of wheat grown in 2007 with(1) no N fertiliser, (2) 50 kg ha-1 N-fertiliser, and (3) 50 kg ha–1 of Nfertiliser and supplementary irrigation; and wheat in 2008 sown into (4)medic stubble, (5) pea stubble, and (6) canola stubble with 50 kg ha–1 of

N fertiliser applied at sowing

Treatment 2007 20081 2 3 4 5 6

Mean 1.51 2.50 1.77 0.84 0.64 1.64Geometric mean 1.25 2.00 1.50 0.77 0.55 1.42Median 0.90 1.48 1.17 0.73 0.44 1.50Lower quartile 0.35 0.56 0.46 0.33 0.16 0.46Upper quartile 1.94 3.33 2.37 1.25 0.84 2.48s.d. 1.95 3.16 1.98 0.72 0.89 1.44CV (%) 130 126 112 85 140 88Skewness 2.9 2.9 1.9 0.8 3.5 1.0Kurtosis 15.1 11.5 4.4 1.0 22.0 1.9No. of observations 4982 5692 5632 4090 3758 4347

Table 6. Intercept and slope of the fitted models of log-transformedN2O emissions from linear mixed-model analysis including cubic

smoothing splines for time and treatment effectsWithin column and year, values followed by the same letter are not

significantly different based on 2 s.e.

Treatment Intercept Slope (time)

2007(1) Wheat with no N fertiliser, after faba beans 1.97a –0.00220a(2) Wheat with N fertiliser, after faba beans 2.39c –0.00421c(3) Wheat with N fertiliser and irrigation,after faba beans

2.12b –0.00296b

2008(4) Wheat with no N fertiliser, after medic 1.32a –0.000069a(5) Wheat with no N fertiliser, after field peas 1.25a –0.000057a(6) Wheat with N fertiliser, after canola 1.61b –0.000163b

Table 3. Plant-available inorganic nitrogen in the profile beforesowing in 2007 after a crop of faba beans, and in 2008 after crops ofmedic (treatment 4) field peas (treatment 5) and canola (treatment 6)

Depth Total Inorganic N (mgNg–1 soil)(m) N (%) 2007 2008

4 5 6

0.0–0.1 0.10 21.5 36.0 27.2 16.50.1–0.2 0.06 12.5 14.2 12.6 7.70.2–0.4 0.04 5.5 9.8 12.2 9.50.4–0.6 0.04 3.5 12.3 7.4 5.70.6–0.8 0.03 3.0 5.0 6.3 7.20.8–1.0 0.03 3.4 4.6 4.2 5.71.0–1.2 0.03 3.3 5.7 5.2 6.0

Nitrous oxide in semi-arid wheat Soil Research 235

in flux from the fertilised treatments more sustained in 2008than in 2007 (~60 days in 2008, compared with 12–40 days in2007, depending on the treatment comparison). The increasedresponse time compensated for the reduced flux in 2008;therefore, the total N2O flux attributed to the application ofthe N fertiliser was similar in both years.

Nitrous oxide emission factor

A fertiliser emissions factor (based on the difference betweenN2O-N emitted from fertilised and unfertilised soil) wascalculated as the sum of the predicted treatment flux

measured during the June–December monitoring period,divided by the amount of fertiliser applied (50 kgN ha–1)(Table 7). Although emission factors are normally calculatedon the basis of treatment differences measured over 12 months,there was no evidence of increased emissions from thefertilised treatments from 4 months after the fertiliser wasapplied, indicating that a calculation based on a reducedtime period was appropriate. The estimated emissionsattributed to the fertiliser, using the cumulative predictedtreatment means, ranged between 41 and 111 g N2O-N ha–1,with an average of 75 g N2O-N ha–1, which was 0.15% ofthe applied N.

–0.50

0.51.01.52.02.53.03.54.04.55.05.56.06.57.07.5 (a)

(b)

(c)

Treatment 1, no N fertiliser

Lower confidence interval Upper confidence interval Treatment value

–0.50

0.51.01.52.02.53.03.54.04.55.05.56.06.57.07.5 Treatment 2, N fertiliser

–0.50

0.51.01.52.02.53.03.54.04.55.05.56.06.57.07.5 Treatment 3, N fertiliser and irrigation in September

Flu

x N

2O (

ng N

2O-N

m–2

s–1

)

06-Jun-07 26-Jun-07 16-Jul-07 05-Aug-07 25-Aug-07 14-Sep-07 04-Oct-07 24-Oct-07 03-Dec-0713-Nov-07

Fig. 6. Nitrous oxide flux for the 2007 crop season (6 June–12 December) for agronomic treatments of: (1) no N fertiliserapplied, (2) N fertiliser applied at sowing, and (3) N fertiliser with supplementary irrigation. Modelled back-transformedflux values and upper and lower confidence intervals were derived for each treatment from linear mixed-model analysisincluding a fitting a cubic smoothing spline to the log-transformed flux values.

236 Soil Research S. J. Officer et al.

Discussion

Nitrous oxide emissions from a nitrogen-fertilised,alkaline Vertosol

In this study, N2O emissions from an alkaline Vertosol growingwinter wheat in a semi-arid climate were successfully measuredover two cropping seasons and six treatments (three per season),including no additional N (2007, one treatment), plus N fertiliser(2007, two treatments, with and without irrigation; 2008, onetreatment), and N provided by a previous N2-fixing legumephase (2008, two treatments, barrel medic and field peas). Fluxes

were small and highly variable in this relatively dry croppingsystem. However, a tuneable diode laser combined with theautomatic operating chambers had sufficient precision to detectthe difference in fluxes, and a cubic spline model offeredflexibility required to identify treatment effects with wheatcrops with varying sources of soil N.

Fluxes of N2O varied on an annual basis and in response toseasonal conditions, with fluxes in the second year of the study(2008) 50% lower than the first year (2007), most likelyreflecting the overall drier soil conditions. In 2008, the fluxrange was 0.56–1.19 ng N2O-Nm–2 s–1 (0.48–1.03 g ha–1 day–1)

–0.5

0

0.5

1.0

1.5

2.0

2.5

3.0

3.5 Treatment 4, no N fertiliser

Lower confidence interval Upper confidence interval Treatment value

–0.5

0

0.5

1.0

1.5

2.0

2.5

3.0

3.5Treatment 5, no N fertiliser

–0.5

0

0.5

1.0

1.5

2.0

2.5

3.0

3.5Treatment 6, N fertiliser

Flu

x N

2O (

ng N

2O-N

m–2

s–1

)

07-Jun-08 26-Jun-08 17-Jul-08 06-Aug-08 26-Aug-08 15-Sep-08 05-Oct-08 25-Oct-08 04-Dec-0814-Nov-08

(a)

(b)

(c)

Fig. 7. Nitrous oxide flux for the 2008 crop season (7 June–12 Dec.) between agronomic treatments of wheat growing in:(4) medic stubble, (5) field pea stubble, and (6) canola stubble with N fertiliser applied at sowing. Equipment breakdowncaused the period of missing data in October and November. Modelled back-transformed flux values and upper and lowerconfidence intervals were derived for each treatment from linear mixed-model analysis including a fitting a cubic smoothingspline to the log-transformed flux values.

Nitrous oxide in semi-arid wheat Soil Research 237

from non-N-fertilised to N-fertilised soil, respectively,compared with the 2007 range of 1.1–1.56 ng N2O-Nm–2 s–1

(0.95–1.35 g ha–1 day–1).Reports are limited on N2O emissions from cropped soils in

semi-arid temperate climate regions (Stehfest and Bouwman2006; Barton et al. 2008, 2011; Galbally et al. 2008, 2010),with emissions generally <1 ng N2O-Nm–2 s–1 (<300 gN2O-N ha–1 year–1) (Galbally et al. 2008). This is comparable toN2O emissions reported from a rainfed, N-fertilised wheat cropin the south-west of Australia under similar rainfall (100 g N2O-N ha–1 year–1; Barton et al. 2008), and to emissions recordedin temperate north-eastern Victoria from strongly acid soilgrowing wheat (Barker-Reid et al. 2005). Similar emissionswere measured when canola was grown on a similar soil type(alkaline medium clay Vertosol) in northern New South Waleswith 80 kg ha–1 of urea fertiliser was applied (Schwenke et al.2010). Average annual rainfall in the northern grains belt ofAustralia, where the Schwenke et al. (2010) study wasundertaken, is significantly higher than in the VictorianWimmera, but the higher average temperatures of thenorthern grains region can markedly reduce the effectivenessof this rainfall. By contrast, higher emissions were reported fromthe N-poor soils of semi-arid Mali planted with continuouscereal crops (Dick et al.2008). With N fertiliser applied(50 kg ha–1), emissions were 1535 g N2O-N ha–1 year–1,compared with 595 gN2O-N ha–1 year–1 with no N fertiliseradded and 972 g N2O-N ha–1 year–1 with N applied as manure.Emissions in that study were measured for 13 months to includethe fallow period; however, the higher emissions were likelyrelated to the high rainfall, with 586mm over 136 days startingimmediately after sowing (Dick et al.2008).

Emission values measured here in 2008, when conditionswere very dry, were similar to values previously recorded inAustralian arable systems (Barker-Reid et al. 2005; Barton et al.2008). The higher emissions measured in 2007, compared withthose previous Australian studies or studies of semi-arid arablesystems in general, may be due to the relatively fine, light claysoil texture, including swelling clays that restricted drainage(Bouwman et al. 2002b; Galbally et al. 2010).

Inclusion of legumes in cropping rotations and effecton nitrous oxide emissions

In this study, the N supplied by a previous legume phase (pulseor pasture) did not result in increased N2O emissions compared

with the use of urea fertiliser. A moderate amount of soilinorganic N was measured throughout the profile in plotspreviously sown to legumes at the time of sowing of thewheat in 2008 (38–57 kgN ha–1), and presumably, additionallegume-derived N would mineralise during the growth of thewheat crop (Ladd and Amato 1986). The legume Nwas matchedby applying a similar amount of fertiliser N (50 kgN ha–1) to thewheat in plots sown to a non-legume (canola) crop in 2007.We hypothesised that N2O emissions from legume-derived Nwould be lower than from fertiliser N (for equivalent rates of N)due to be better synchronisation between legume-derived N(via mineralisation) and crop uptake during the subsequentcropping phase (Ladd and Amato 1986). Furthermore, asignificant proportion of ‘legume N’ is contributed by rootsbelow the soil surface (McNeill et al. 1997). By contrast, Nderived from fertiliser can be rapidly nitrified, resulting in thebuild-up of large quantities of nitrate-N near the soil surfacesoon after application (sowing) when crop demand is smalland the potential for gaseous loss as N2O is high. Results inthis study, where N was supplied by either a pasture legume(barrel medic) or a pulse crop (field peas) support thishypothesis, although this conclusion was likely affected bythe very dry seasonal conditions throughout the study period,which would have limited both dry matter production and thequantity of N2 fixed during the legume phase, as well as thepotential for N2O flux during the following fallow andsubsequent wheat phase.

Other studies in a semi-arid temperate cropping environment,on acidic sandy soils in the south-west of Western Australia(in contrast to the alkaline clay soils used in the present study),suggest that including a legume (lupin) in rotation with wheatinstead of applying all of the N requirements of the wheat byfertiliser can produce small but significant reductions in overallN2O fluxes (Barton et al. 2013). On the other hand, Dick et al.(2008) found no differences in N2O emissions between legumeand fertiliser N sources in a sub-Saharan environment. Therelatively small yield potential of wheat in rainfed semi-aridcropping systems of south-eastern Australia (average grainyields 1.5–4 t ha–1) compared with cropping systems in thenorthern hemisphere result in a similarly smaller requirementfor N. In both 2007 and 2008, there was no grain yield responseof wheat to the application of N fertiliser; in 2007, applying Nfertiliser actually decreased grain yields where no supplementaryirrigation was applied, a phenomenon referred to as ‘haying off’(van Herwaarden et al. 1998), which commonly occurs duringdry finishes to the growing season. In summary, the use oflegume crops to supply N for following crops may thereforemeet N demand (Peoples and Baldock 2001) and reduceurea-fertiliser-based N2O emissions from semi-arid croppingsystems.

Nitrous oxide emission sources

The increase in emissions from the fertilised soil occurred in twogeneralised periods in the 2 months following application offertiliser at sowing and tended to be short-lived. The pattern wassimilar in both years despite drier conditions in the second year.An initial flux response was found in two of three fertilisedtreatments, commencing ~6 days after the application of the

Table 7. Summary of emission factors, calculated from the predictedtreatment mean N2O flux values summed over time from sowing to

harvest (gN2O-Nha–1)The emissions factors assume that the difference in N emissions between anytwo treatments was due to the application of fertiliser and calculates thepercentage of additional N emitted from the fertilised treatments as a

percentage of the fertiliser applied (50 kgNha–1)

2007 2008

Treatment comparison 2,1 3,1 6,4 6,5Sum emissions fertilised soil (g N2O-N ha–1) 282 212 138 138Sum emissions unfertilised soil (gN2O-Nha–1) 171 171 75 54Difference (gN2O-Nha–1) 111 41 63 84Emission factor (%) 0.22 0.08 0.13 0.17

238 Soil Research S. J. Officer et al.

fertiliser (in June) and peaking ~7 days later. A second episodeof increased N2O emissions from fertilised soil occurred in July.This increase was evident in the three fertilised treatments andcorresponded to a sustained increase in soil water followingwinter rain.

This pattern of N2O production in the soil is consistentwith nitrification-based emissions occurring in relatively dryaerobic soil in the first month after application of the fertiliser,followed by a period of denitrification-based emissions asthe soil wetted up during the winter rainfall period. Thetopsoil at the time of sowing and application of fertiliserhad a WFPS of 62% and remained relatively dry until earlyJuly. Soil water conditions were similar in 2008, when thetopsoil had a WFPS of 40% at sowing. Rain in thefollowing week increased this to 60%, with very little rainfor the rest of June. Soil conditions when the fertiliser wasapplied may also contribute to aerobic conditions, because inmechanised, winter-wheat cropping systems, the seed istypically sown when the soil-water content is low andhence soil plasticity is relatively low. A similar laggedresponse in the production of N2O has also been reportedwhen nitrification was measured after the addition of anammonium source to soil (Rochester et al. 1992).

The principle pathway for nitrification may have beennitrifier denitrification. When urea fertiliser is applied tomoist alkaline soil, it is expected that the urea will rapidlyconvert to ammonium, forming a zone of ammonium diffusinginto the soil surrounding the fertiliser granules (Magalhãeset al.1987). The nitrifier population is expected to respondto the food source, and in the alkaline environment,to rapidly convert the ammonium to nitrite (Freneyet al.1979; Blackmer et al.1980; Dalal et al. 2003). In thisstudy, the alkaline pH of the soil (7.9 in 0–100mm) wouldhave been raised further by hydroxide production duringhydrolysis of the urea (Magalhães et al. 1987). Nitrite maythen build up in the soil near the granules because the second-stage nitrite oxidisers, Nitrobacter, are inhibited by highconcentrations of ammonium. The concentrated growth ofmicrobial activity in the zone surrounding the fertilisergranules will also rapidly deplete the available oxygen(Parkin 1987). Consequently, the soil conditions around thegranules may become enriched in nitrite, low in oxygen andrelatively high pH, and therefore conducive to nitrifierdenitrification (Shaw et al. 2006; Wrage et al. 2001). Thischain of processes would also cause the characteristic delaybetween the application of the urea fertiliser and the initialresponse in N2O emissions.

By the second month after fertiliser application, much of theammonium around the fertiliser granules would be expectedto be converted to nitrate (Freney et al. 1985; Rochester et al.1992). The second month, July, corresponded to wetter soilconditions, with topsoil WFPS 60–80%, suggesting that theanaerobic process of denitrification was increasing andnitrification decreasing (Dalal et al. 2003; Dobbie andSmith 2003). In this study, the final transformation of N2Ointo N2 could be inhibited by both the enriched concentrationsof nitrate below the fertiliser band and the topsoil beingpartially oxygenated, increasing the proportion of fertiliserN lost as N2O compared with N2.

Nitrous oxide emission factor

The calculated fertiliser emission factors measured over twowinter-wheat crop seasons and based on the difference betweenthe total flux from fertilised and unfertilised treatments rangedbetween 0.08% and 0.22%, with an average over the twoseasons of 0.15%. The range was similar to the uncertaintyrange calculated by Bouwman et al. (2002b) when collatingthe N2O fertiliser emission factors from a large number ofstudies. Although the proportion of N2O emitted followingthe application of urea fertiliser in the present study was2–10 times greater than values reported for similar arableenvironments in Australia (0.02–0.1; Barker-Reid et al. 2005;Barton et al. 2008), it is small compared with the internationaldefault value (1%) used by the IPCC (2006). Our research,although conducted under below-average seasonal rainfallconditions, suggests that a further reduction needs to beconsidered in the emission factor (0.3%) currently used inAustralia for estimating N2O emissions from the applicationof N fertiliser to rainfed crops.

Acknowledgments

The authors thank Gary Howard, Mel Munn, Graham Price and RobertBaigent for their excellent assistance during this study. We also thankCampbell Scientific Inc. for their support. Comments made by twoanonymous reviewers and the editor (Dr Louise Barton) havesignificantly improved the quality of the manuscript. This research waspart of a joint initiative involving the Department of Environment andPrimary Industries, Victoria, the Australian Greenhouse Office (now theDepartment of Environment) and the Grains Research and DevelopmentCorporation (GRDC).

References

ABARES (2012) Australian Crop Report, June 2012. No. 162. AustralianBureau of Agricultural Resources Economics and Sciences, Canberra,ACT.

Barker-Reid F, Gates WP, Wilson K, Baigent R, Galbally IE, Meyer CP,Weeks IA, Eckard RJ (2005) Soil nitrous oxide emissions from rain-fedwheat in SE Australia. In ‘Science, control, policy and implementation.Fourth International Symposium Non-CO2 Greenhouse Gases (NCGG-4)’.pp. 25–32. (Millpress: Rotterdam, the Netherlands)

Barton L, Kiese R, Gatter D, Butterbach-Bahl K, Buck R, Hinz C, Murphy D(2008) Nitrous oxide emissions from a cropped soil in a semi-aridclimate. Global Change Biology 14, 177–192. doi:10.1111/j.1365-2486.2007.01474.x

Barton L, Butterbach-Bahl K, Kiese R, Murphy D (2011) Nitrous oxidefluxes from a grain–legume crop (narrow-leafed lupin) grown in a semi-arid climate. Global Change Biology 17, 1153–1166. doi:10.1111/j.1365-2486.2010.02260.x

Barton L, Murphy DV, Butterbach-Bahl K (2013) Influence of croprotation and liming on greenhouse gas emissions from semi-arid soil.Agriculture, Ecosystems & Environment 167, 23–32. doi:10.1016/j.agee.2013.01.003

Blackmer AM, Bremner JM, Schmidt EL (1980) Production of nitrousoxide by ammonia-oxidizing chemmoautotrophic microorganisms insoil. Applied and Environmental Microbiology 40, 1060–1066.

Bouwman AF, Boumans LJM, Batjes NH (2002a) Emissions of N2Oand NO from fertilized fields. Summary of available measurementdata. Global Biogeochemical Cycles 16, 6-1–6-13. doi:10.1029/2001GB001811

Nitrous oxide in semi-arid wheat Soil Research 239

Bouwman AF, Boumans LJM, Batjes NH (2002b) Modeling globalannual N2O and NO emissions from fertilized fields. GlobalBiogeochemical Cycles 16, 28-1–28-9. doi:10.1029/2001GB001812

Bureau of Meteorology (2012) Summary statistics for Horsham PolkemmetRd weather station. Bureau of Meteorology. Available at: www.bom.gov.au/climate/averages/tables/cw_079023.shtml (accessed 9 July 2012)

Cawood RJ (1996) ‘Climate, temperature and crop production in south-eastern Australia.’ Principals of Sustainable Agriculture 7. (AgricultureVictoria: Melbourne)

Conen F, Smith KA (2000) An explanation of linear increases in gasconcentration under closed chambers used to measure gas exchangebetween soil and the atmosphere. European Journal of Soil Science 51,111–117. doi:10.1046/j.1365-2389.2000.00292.x

Connor DJ (2004) Designing cropping systems for efficient use of limitedwater in southern Australia. European Journal Agronomy 21, 419–431.

Dalal RC, Wang W, Robertson GP, Parton WJ (2003) Nitrous oxideemission from Australian agricultural lands and mitigation options: areview. Australian Journal of Soil Research 41, 165–195. doi:10.1071/SR02064

Davidson EA (2009) The contribution of manure and fertilizer nitrogen toatmospheric nitrous oxide since 1860. Nature Geoscience 2, 659–662.doi:10.1038/ngeo608

Dick J, Kaya B, Soutoura M, Skiba U, Smith R, Niang A, Tabo R (2008)The contribution of agricultural practices to nitrous oxide emissions insemi-arid Mali. Soil Use and Management 24, 292–301. doi:10.1111/j.1475-2743.2008.00163.x

Dobbie KE, Smith KA (2003) Nitrous oxide emission factors for agriculturalsoils in Great Britain: the impact of soil water-filled pore space andother controlling variables. Global Change Biology 9, 204–218.doi:10.1046/j.1365-2486.2003.00563.x

Drewitt G, Warland JS (2007) Continuous measurements of belowgroundnitrous oxide concentrations. Soil Science Society of America Journal71, 1–7. doi:10.2136/sssaj2005.0410

Flechard CR, Neftel A, Jocher M, Ammann C, Fuhrer J (2005) Bi-directionalsoil/atmosphere N2O exchange over two mown grassland systemswith contrasting management practices. Global Change Biology 11,2114–2127. doi:10.1111/j.1365-2486.2005.01056.x

Freney JR, Denmead OT, Simpson JR (1979) Nitrous oxide emission fromsoils at lowmoisture contents. Soil Biology & Biochemistry 11, 167–173.doi:10.1016/0038-0717(79)90096-8

Freney JR, Simpson JR, Denmead OT, Muirhead WA, Leuning R (1985)Transformations and transfers of nitrogen after irrigating a cracking claysoil with a urea solution. Australian Journal of Agricultural Research36, 684–694. doi:10.1071/AR9850684

Galbally IE, Kirstine WV, Meyer CP, Wang YP (2008) Soil–atmospheretrace gas exchange in semiarid and arid zones. Journal of EnvironmentalQuality 37, 599–607. doi:10.2134/jeq2006.0445

Galbally IE, Meyer CP, Wang YP, Kirstine WV (2010) Soil–atmosphereexchange of CH4, CO, N2O and NOx and the effects of land-use changein the semiarid Mallee system in Southeastern Australia. Global ChangeBiology 16, 2407–2419.

Gilmour AR, Gogel BJ, Cullis BR, Thompson R (2006) ‘ASReml UserGuide 2.0.’ (VSN International: Hemel Hempstead, UK)

Hyde BP, HawkinsMJ, Fanning AF, Noonan D, RyanM, O’Toole P, CartonOT (2006) Nitrous oxide emissions from a fertilized and grazedgrassland in the South East of Ireland. Nutrient Cycling inAgroecosystems 75, 187–200. doi:10.1007/s10705-006-9026-x

IPCC (2006) N2O Emissions from managed soils, and CO2 emissionsfrom lime and urea application. In ‘2006 IPCC Guidelines forNational Greenhouse Gas Inventories. Vol. 4. Agriculture forestryand other land use’. Ch. 11. Prepared by the National GreenhouseGas Inventories Programme, International Panel on Climate Change(Eds HS Eggleston, L Buendia, K Miwa, T Ngara, K Tanabe) p. 54(IGES: Hayama, Japan)

Isbell RF (2002) ‘The Australian Soil Classification.’ Revised edn (CSIROPublishing: Melbourne)

Kaiser E-A, Kohrs K, KückeM, Schnug E, Heinemeyer O,Munch JC (1998)Nitrous oxide release from arable soil: importance of N-fertilization,crops and temporal variation. Soil Biology & Biochemistry 30,1553–1563. doi:10.1016/S0038-0717(98)00036-4

Kastanek FJ, Nielsen DR (2001) Description of soil water characteristicsusing cubic spline interpolation. Soil Science Society of America Journal65, 279–283. doi:10.2136/sssaj2001.652279x

Klute A (1986) Water retention: laboratory methods. In ‘Methods of soilanalysis. Part 1. Physical and mineralogical methods’. 2nd edn (Ed. AKlute) pp. 635–662. (American Society of Agronomy,Madison,WI, USA)

Kroeze C, Mosier A, Bouwman L (1999) Closing the N2O budget: aretrospective analysis 1500–1994. Global Biogeochemical Cycles 13,1–8. doi:10.1029/1998GB900020

Ladd JN, Amato M (1986) The fate of nitrogen from legume and fertilizersources in soils successively cropped with wheat under field conditions.Soil Biology & Biochemistry 18, 417–425. doi:10.1016/0038-0717(86)90048-9

Linn DM, Doran JW (1984) Effect of water-filled pore space on carbondioxide and nitrous oxide production in tilled and non-tilled soils. SoilScience Society of America Journal 48, 1267–1272. doi:10.2136/sssaj1984.03615995004800060013x

Magalhães AM, Nelson DW, Chalk PM (1987) Nitrogen transformationsduring hydrolysis and nitrification of urea. 1. Effect of soil propertiesand fertiliser placement. Fertilizer Research 11, 161–172. doi:10.1007/BF01051059

Martin J, ImhofM, Lourey R, Nink R, De Plater K, Rampant P, Thompson S,Alexander S (1996) ‘Major agricultural soils of the WimmeraIrrigation Area.’ Natural Resources Monitoring and Assessment(NRMA) Team (Department of Natural Resources and Environment,Victoria: Melbourne)

McNeill AM, Zhu C, Fillery IRP (1997) Use of in situ 15N-labelling toestimate the total below-ground nitrogen of pasture legumes in intactsoil-plant systems. Australian Journal of Agricultural Research 48,295–304. doi:10.1071/A96097

Meyer CP, Galbally IE, Wang Y, Weeks IA, Jamie I, Griffith DWT (2001)Two automatic chamber techniques for measuring soil-atmosphereexchanges of trace gases and results of their use in the OASIS fieldexperiment. CSIRO Atmospheric Research Technical Paper No. 51.

Mosier AR, Duxbury JM, Freney JR, Heinemeyer O, Minami K (1998)Assessing and mitigating N2O emissions from agricultural soils.Climatic Change 40, 7–38. doi:10.1023/A:1005386614431

O’Leary GJ, Connor DJ (1997) Stubble retention and tillage in a semi-aridenvironment: 2. Soil mineral nitrogen accumulation during fallow.Field Crops Research 52, 221–229.

Orchard BA, Cullis BR, Coombes NE, Virgona JM, Klein T (2000) Grazingmanagement studies within the Temperate Pasture Sustainability KeyProgram: experimental design and statistical analysis.Australian Journalof Experimental Agriculture 40, 143–154. doi:10.1071/EA98005

Parkin TB (1987) Soil microsites as a source of denitrification variability.Soil Science Society of America Journal 51, 1194–1199. doi:10.2136/sssaj1987.03615995005100050019x

Peoples MB, Baldock JA (2001) Nitrogen dynamics of pastures: nitrogenfixation inputs, the impact of legumes on soil nitrogen fertility, and thecontributions of fixed nitrogen to Australian farming systems. AustralianJournal of Experimental Agriculture 41, 327–346. doi:10.1071/EA99139

Phillips FA, Leuning R, Baigent R, Kelly KB, Denmead OT (2007) Nitrousoxide flux measurements from an intensively managed irrigated pastureusing micrometeorological techniques. Agricultural and ForestMeteorology 143, 92–105. doi:10.1016/j.agrformet.2006.11.011

Rayment GE, Higginson FR (1992) ‘Australian laboratory handbook of soiland water chemical methods.’ (Inkata Press: Melbourne)

240 Soil Research S. J. Officer et al.

Rochester IJ, Constable GA, Macleod DA (1992) Preferential nitrateimmobilization in alkaline soils. Australian Journal of Soil Research30, 737–749. doi:10.1071/SR9920737

Sadras VO, Rodriguez D (2010) Modelling the nitrogen-driven trade-offbetween nitrogen utilisation efficiency and water use efficiency of wheatin eastern Australia. Field Crops Research 118, 297–305.

Saxton KE, Porter MA, McMahon TA (1992) Climatic impacts on drylandwinter wheat yields by daily soil water and crop stress simulations.Agricultural and Forest Meteorology 58, 177–192. doi:10.1016/0168-1923(92)90060-H

Schwenke G, Haigh B, McMullen G, Herridge D (2010) Soil nitrousoxide emissions under dryland N-fertilised canola and N2-fixingchickpea in the northern grains region, Australia. In ‘Proceedings19th World Congress of Soil Science: Soil Solutions for a ChangingWorld’. 1–6 August 2010, Brisbane, Australia. (International Union ofSoil Sciences)

Shaw LJ, Nicol GW, Smith Z, Fear J, Prosser JI, Baggs EM (2006)Nitrosospira spp. can produce nitrous oxide via a nitrifierdenitrification pathway. Environmental Microbiology 8, 214–222.doi:10.1111/j.1462-2920.2005.00882.x

Stehfest E, Bouwman L (2006) N2O and NO emission from agriculturalfields and soils under natural vegetation: summarizing availablemeasurement data and modeling of global annual emissions. NutrientCycling in Agroecosystems 74, 207–228. doi:10.1007/s10705-006-9000-7

van Herwaarden AF, Farquhar GD, Angus JF, Richards RA, Howe GN(1998) ‘Haying-off, the negative grain yield response of drylandwheat to nitrogen fertiliser. I. Biomass, grain yield, and water use.Australian Journal of Agricultural Research 49, 1067–1081.doi:10.1071/A97039

Venterea RT, Spokas KA, Baker JM (2009) Accuracy and precisionanalysis of chamber-based nitrous oxide gas flux estimates. SoilScience Society of America Journal 73, 1087–1093. doi:10.2136/sssaj2008.0307

Verbyla AP, Cullis BR, Kenward MG, Welham SJ (1999) The analysis ofdesigned experiments and longitudinal data by using smoothing splines.Applied Statistics 48, 269–311. doi:10.1111/1467-9876.00154

Wrage N, Velthof GL, van Beusichem ML, Oenema O (2001) Role ofnitrifier denitrification in the production of nitrous oxide. Soil Biology &Biochemistry 33, 1723–1732. doi:10.1016/S0038-0717(01)00096-7

Nitrous oxide in semi-arid wheat Soil Research 241

www.publish.csiro.au/journals/sr