Journal list menu

Volume 45, Issue 4 p. 1133-1143
Atmospheric Pollutant and Trace Gas
Open Access

Lower Nitrous Oxide Emissions from Anhydrous Ammonia Application Prior to Soil Freezing in Late Fall Than Spring Pre-Plant Application

Mario Tenuta

Corresponding Author

Mario Tenuta

Dep. of Soil Science, Univ. of Manitoba, Winnipeg, MB, Canada, R3T 2N2

Corresponding author ([email protected]).Search for more papers by this author
Xiaopeng Gao

Xiaopeng Gao

Dep. of Soil Science, Univ. of Manitoba, Winnipeg, MB, Canada, R3T 2N2

State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi, China, 830011

Search for more papers by this author
Donald N. Flaten

Donald N. Flaten

Dep. of Soil Science, Univ. of Manitoba, Winnipeg, MB, Canada, R3T 2N2

Search for more papers by this author
Brian D. Amiro

Brian D. Amiro

Dep. of Soil Science, Univ. of Manitoba, Winnipeg, MB, Canada, R3T 2N2

Search for more papers by this author
First published: 01 July 2016
Citations: 29

Assigned to Associate Editor Claudia Wagner-Riddle.

All rights reserved.


Fall application of anhydrous ammonia in Manitoba is common but its impact on nitrous oxide (N2O) emissions is not well known. A 2-yr study compared application before freeze-up in late fall to spring pre-plant application of anhydrous ammonia on nitrous oxide (N2O) emissions from a clay soil in the Red River Valley, Manitoba. Spring wheat (Triticum aestivum L.) and corn (Zea mays L.) were grown on two 4-ha fields in 2011 and 2012, respectively. Field-scale flux of N2O was measured using a flux-gradient micrometeorological approach. Late fall treatment did not induce N2O emissions soon after application or in winter likely because soil was frozen. Application time did alter the temporal pattern of emissions with late fall and spring pre-plant applications significantly increasing median daily N2O flux at spring thaw and early crop growing season, respectively. The majority of emissions occurred in early growing season resulting in cumulative emissions for the crop year being numerically 33% less for late fall than spring pre-plant application. Poor yield in the first year with late fall treatment occurred because of weed and volunteer growth with delayed planting. Results show late fall application of anhydrous ammonia before freeze-up increased N2O emissions at thaw and decreased emissions for the early growing season compared to spring pre-plant application. However, improved nitrogen availability of late fall application to crops the following year is required when planting is delayed because of excessive moisture in spring.

Core Ideas

  • Late-fall ammonia application prior to freeze-up did not induce N2O emissions over winter.
  • Late-fall application did increase N2O emissions during thaw the following year.
  • Despite lower area-based emission in the first study year, poor yield increased yield-scaled emissions.
  • For the second study year, yield-scaled emissions were similar for application treatments.


  • TGA
  • trace gas analyzer
  • TGAS
  • -Man, Trace Gas Manitoba research site
  • WFPS
  • water-filled pore space
  • Increases in the atmospheric concentration of nitrous oxide (N2O) are linked to increased global radiative forcing and destruction of stratospheric ozone (Myhre et al., 2013). Agriculture accounts for approximately 72% of the anthropogenic N2O emissions in Canada, with the application of synthetic N fertilizers being the largest source (Environment Canada, 2013). The Northern Great Plains of Canada that includes southern regions of the provinces of Manitoba, Saskatchewan, and Alberta is the main grain production area of Canada accounting for about 80% of all synthetic fertilizer N used in the country (Statistics Canada, 2014). Developing fertilizer management strategies to reduce N2O emissions for the Northern Great Plains of Canada is important to reduce national net emissions of greenhouse gases.

    Anhydrous ammonia (82% N) accounted for 24% of all synthetic N fertilizer used in 2013 in the Northern Great Plains of Canada (Statistics Canada, 2014). However, from meta-analyses, Bouwman et al. (2002) and Eagle and Olander (2012) concluded anhydrous ammonia induces the largest emissions of N2O from cropped soils of all fertilizer N products. In Manitoba, however, with a lower application rate than other studies included in meta-analyses, Burton et al. (2008) reported no greater emissions with 80 kg N ha−1 anhydrous ammonia than urea applied at planting to spring wheat (Triticum aestivum L.). These studies highlighted the importance of developing strategies to reduce N2O emissions from anhydrous ammonia.

    Anhydrous ammonia is commonly applied in the fall in the Northern Great Plains of Canada because of lower retail cost (MAFRD, 2007), greater availability over time, and generally drier soil conditions than at spring pre-plant or at spring planting. However, nitrification of early fall applied anhydrous ammonia can occur before winter freeze-up (Malhi and McGill, 1982; Tiessen et al., 2006). To minimize nitrification and thus nitrate accumulation before freeze-up, it is recommended in Manitoba that fall N fertilizer applications are banded and delayed until soil temperatures fall below 5°C (MAFRD, 2007).

    Applying N in late fall when soil is cool and close to the onset of freezing may not result in increased emissions compared to spring pre-plant or at-planting applications. Soil conditions before freeze-up are less conducive to N2O emissions compared to relatively high moisture and warmer temperature conditions in spring at time of planting. There are few studies, however, comparing N2O emissions from fertilizers applied in fall relative to pre-planting, especially using anhydrous ammonia. For example, Burton et al. (2008) reported similar overall N2O emissions for fall and pre-plant applications of anhydrous ammonia to spring wheat in Manitoba but had a low frequency of measurements using the static vented chamber method, and missed the spring thaw, which can account 30% of the annual total (Glenn et al., 2012; Maas et al., 2013). Nitrate is a major factor for N2O emissions during thawing of soil through denitrification (Tenuta and Sparling, 2011). Since ammoniacal fertilizers applied late fall may still be subject to some nitrification (Nyborg et al., 1997; Tiessen et al., 2006), the contribution of fall application to spring-thaw is important to determine. Continuous measurement approaches are needed to capture episodic thaw emissions to reliably compare N2O emissions from fall and pre-plant or at planting applications of fertilizer N.

    The objective of this study was to evaluate the effect of application timing (late fall vs. spring pre-plant) of anhydrous ammonia on the magnitude and temporal pattern of N2O emissions from a cropped field near Winnipeg, located in the Red River Valley of Manitoba. Near continuous measurements of N2O emissions were made using the micrometeorological flux-gradient method. Our hypothesis was that application in late fall close to the time soil freezes would result in little late fall, over-winter and spring-thaw emissions, resulting in similar overall emissions to pre-plant application. This is because N2O production processes on N applied before freeze-up would not be operative.

    Materials and Methods

    Site Description

    A field experiment was conducted over 2 yr from 15 Oct. 2010 to 14 Oct. 2012 at the Trace Gas Manitoba (TGAS-Man) study site (49.64° N, 97.16° W) at the University of Manitoba Glenlea Research Station, located 16 km south of Winnipeg. The site is within the Red River Valley that is a glaciolacustrine clay floodplain of near-level topography (0 to <2% slope) with an extreme humid-continental climate (Köppen Dfb). The soils are of the Red River association Osborne and Red River series (Ehrlich et al., 1953; Michalyna et al., 1975), classified as Gleyed Humic Vertisols in the Canadian soil classification system and Typic Humicryerts in the US classification system. The soil (0–20 cm) is of clay texture (clay 600 g kg−1, silt 350 g kg−1, and sand 50 g kg−1) with pHH2O 6.2, bulk density 1.2 Mg m−3, and organic carbon content 32 g kg−1. Imperfect drainage results in periodic flooding and excessive soil moisture often around snow melt in spring and occasionally with early growing season rains.

    Treatment and Experimental Design

    The TGAS-Man site was initiated in the fall of 2005 with the establishment of a 2 × 2 square grid of fields within a 30-ha cropped area, the four fields being 200 m × 200 m (4 ha) each. The current study compared late fall and pre-plant applications of anhydrous ammonia using the two eastern located fields (Field 2 northeast and Field 3 southeast; Fig. A1). The two western fields were not used in this study as they had alfalfa and/or forage grass receiving no fertilizer. Spring wheat was grown in 2011 and corn for grain in 2012 on Fields 2 and 3. The application rate was 112 and 180 kg N ha−1 in 2011 and 2012, respectively. Allocations of treatments for the two fields were switched in the second year to reduce possible field bias.

    Cultivation of fields occurred in fall as typically done for the Red River Valley using a chisel plow on 8 Oct. 2010 and chisel plow followed by heavy harrow on 11 Oct. 2011. Late fall anhydrous ammonia application occurred 15 Oct. 2010 and 4 Nov. 2011 and in spring occurred on 10 June 2011 and 12 Apr. 2012. The same application system was used for late fall and pre-plant treatments with anhydrous ammonia injected 10 cm below the soil surface at 50 cm spacing using a Raven AccuFlow Vortex Cooler and SCS330 flow-meter (Raven Industries, Sioux Falls, SD). The fields were seeded to spring wheat (variety A.C. Kane, Canada Western Red) on 10 June 2011 and corn (hybrid DKC26-79 RT–BT) on 3 May 2012 using an air-seeder and planter, respectively. The seeding rates were 168 kg ha−1 (4,500,000 seeds ha−1) for spring wheat and 79,040 seeds ha−1 for corn. Herbicides were applied to both crops for weed control. On 20 June 2011, Buctril M (a.i. bromoxynil and MCPA) and Refine (a.i. MCPA) were applied at rates of 1.0 L ha−1 and 10 g ha−1, respectively. On 23 May 2012, Roundup (a.i. glyphosate) was applied at a rate of 1.7 L ha−1. The relatively late seeding and herbicide application in 2011 was due to wet soil conditions.

    Grain was harvested by combine for spring wheat and corn on 27 Sept. 2011 and 18 Oct. 2012, respectively. For both crops, chopped residues were left on the fields. Before the combine harvesting operation, grain yields were obtained by hand harvesting 10 randomly chosen 0.25 m2 areas followed by mechanical threshing for wheat and hand shelling for corn with dry grain weight reported on an area basis.

    N2O Flux and Ecosystem Respiration Measurements

    A micrometeorological flux-gradient system was used for near continuous determination of N2O (FN) and ecosystem respiration (FR) fluxes from the TGAS-Man site as fully described in the Appendix. Briefly, a tunable-diode-laser trace gas analyzer was set inside a trailer located at the junction of four experimental plots. Atmospheric samples were drawn from two sample intakes separated vertically by 0.65 m at the center of each field and the mean concentration of N2O and CO2 determined every 30 min. The flux was calculated as the product of the concentration difference with height and a transfer coefficient based on measurements of turbulent transfer using a sonic anemometer. Data coverage of fluxes was generally very good, covering about 50% of possible 30-min measurement periods for the treatments in both crop years (See Appendix). Data gaps were normally less than 2 d, except between 26 Oct. 2010 and 10 Jan. 2011 when data were rejected because of uncertainty in proper switching of sample intakes for the late fall application field (Field 3).

    The FR was derived from partitioning CO2 measurements using a modified version of the standard Fluxnet-Canada protocol (See Appendix). The FR was used to determine if temperature may have restrained microbial processes leading to N2O emissions after the late fall N application. The lack of plant growth meant FR and net ecosystem production were similar from fall through to the early-growing season but diverged with growth of the crop. Mean daily 30-min FN and FR are reported as g N ha−1 d−1 and kg C ha−1 d−1, respectively. The flux measurement system can detect 30-min FN at rates greater than 0.05 nmol N2O m−2 s−1 (Glenn et al., 2012; see Appendix), which is 0.025 g N ha−1 over a 30-min period. For all measured 30-min fluxes in the 2011 and 2012 crop years, 45 and 51% were above detection, respectively. During emission events that lasted days to a couple of weeks at thaw and early growing seasons, measured fluxes were all above the detection limit. Seasonal and annual cumulative FN budgets (ΣFN, kg N ha−1) for the treatments were estimated by averaging 30-min fluxes for a day and scaling to kg N ha−1, then gap-filling daily fluxes by linear interpolation and then finally summing daily flux values. The reporting years for ΣFN were based on crop year starting with late fall anhydrous ammonia application: spring wheat (15 Oct. 2010 to 3 Nov. 2011) and corn (4 Nov. 2011 to 14 Oct. 2012). In each crop year, three seasonal periods were defined as post-fall (from the date of fall anhydrous ammonia application to the first date of daily average soil temperature at 5 cm <0°C), thaw (from the first date of average air temperature >0°C to the first date of soil temperature at 5 cm >5°C), and early growing season (from the date of pre-plant anhydrous ammonia application to the date of maximum soil temperature at 5 cm in that year). The ΣFN are reported for each of these periods as well for each of the crop years.

    Weather and Soil Measurements

    A weather station was located 25 m southwest of the instrument trailer on a mowed grassed (Poa pratensis L., Lolium perenne L., Festuca spp. mix) area at the site. Mean daily air temperature (HMP45C, Vaisala Inc., Woburn, MA) and total daily precipitation (T-200B Series Precipitation Gauge, Geonor Inc., Milford, PA) were recorded. Soil temperature at 5-cm depth was measured using thermistors (107B soil temperature probe, Campbell Scientific Inc., Logan, UT). Soil volumetric water content at 10-cm depth was measured using capacitance probes (ECHO-10, Decagon Devices Inc., Pullman, WA). Water-filled pore space (WFPS; m3 m−3) was estimated from soil volumetric water content using a bulk density of 1.15 Mg m−3 and particle density of 2.65 Mg m−3.

    Composite soil samples (0–30 cm) from six randomly chosen 5-m diameter areas in each field were obtained at bi-weekly to monthly intervals between April and November of each year. Each composite sample was a combination of 10 cores (2.5 cm i.d.) within each area. The samples were extracted fresh with 2 M KCl solution and the concentrations of NH4+ and NO3 in extracts determined by the phenate and copper cadmium reduction to nitrite methods, respectively, using a Technicon Autoanalyzer II (Pulse Instrumentation Inc., Milwaukee, WI) and results reported as mg N kg−1 dry soil.

    Statistical Analysis

    Effect of treatment on the distribution of emission fluxes for periods during the crop years was examined using the Wilcoxon signed-rank test with the univariate procedure of SAS (SAS Institute, 2004). This test compared the median of sample populations and was used because FN (g N2O-N ha−1 d−1) values were not normally distributed (P < 0.05 Shapiro–Wilk test) and had non-homogeneity of variances (P < 0.05 Levene's test). Treatment FN values were compared for the periods of post-fall, thaw, early-growing season, other days not in these periods, and for each and across both crop years. An additional period for the thaw plus early-growing season was examined because the vast majority of fluxes comprising cumulative emissions (>78%) occurred in these two periods. Nongap-filled means of FN were used. Cumulative emissions for post-fall, thaw, early growing season, thaw plus early growing season, remaining days and all days in each and across both crop years were determined by summation of gap-filled FN values in the respective periods. Percent contribution of the periods to emissions of each and across both crop years was determined from cumulative emissions of each period divided by cumulative whole crop year emission multiplied by 100. Grain yield in each year from treatments was compared by t test. Differences in treatment means were declared significant at P < 0.05.


    2011 Crop Year

    Late fall application of anhydrous ammonia in 2010 did not result in an episode of N2O emission before freeze-up (Fig. 1). Soon after application, temperatures declined so that by 13 November and 2 December, air and soil temperatures, respectively, were below 0°C (Fig. 1). Throughout the post-fall periods, FN was very low, never attaining more than 10 g N2O-N ha−1 d−1 on any day with the median daily emissions being similar (P > 0.05) for treatments (Table 1).

    Details are in the caption following the image

    Mean daily N2O emission (FN), ecosystem respiration (FR), total daily precipitation, mean daily air temperature and soil temperature at 5-cm depth, water-filled pore space (WFPS) at 10-cm depth, and soil NH4+ and NO3 concentrations from 2010 to 2012. Arrows indicate day of treatment with anhydrous ammonia (gray for very late fall applications and black for spring pre-plant applications). Means ± SE are presented for soil NH4+ and NO3 concentrations. Data are presented with field allocations for treatments switched on 4 Nov. 2011. Shown at top are periods for emission comparison between treatments as defined in the Materials and Methods section.

    Table 1. Summary of daily non gap-filled and gap-filled cumulative emissions of N2O with late fall and pre-plant application of anhydrous ammonia for emission event periods during the 2011 and 2012 growing seasons. Post-fall, date of fall application to the first date of daily average soil temperature at 5 cm <0°C; Thaw, first date of average air temperature >0°C to first date of soil temperature at 5 cm >5°C; Early growing, date of pre-plant application to date of maximum soil temperature at 5 cm in that year.
    Crop year Period Fall application Spring application
    Mean Med SE§ Σ All# Mean Med SE Σ All n†† P‡‡
    2011 Post-fall 0 1 2 0.0 0 2 2 1 0 0 6 ns
    Thaw 65 31 15 2.5 69 20 11 5 0.7 13 36 <0.001
    Early growing 3 1 2 0.1 4 97 20 38 4.0 73 41 <0.001
    Thaw + early growing 33 8 8 2.6 72 61 13 21 4.7 88 77 ns
    Remainder 5 1 1 1.0 27 2 1 1 0.5 10 206 ns
    Whole crop year 12 2 2 3.6 18 3 6 5.4 289 ns
    2012 Post-fall 0 0 1 0.0 0 1 1 2 0 0 7 ns
    Thaw 153 107 56 1.2 14 33 32 13 0.3 2 7 ns
    Early growing 85 42 10 8.0 86 148 53 23 13.8 99 93 <0.001
    Thaw + early growing 90 42 11 9.3 100 138 41 21 14.1 101 100 <0.001
    Remainder 0 0 0 0.0 1 −1 −2 1 −0.2 −1 213 <0.001
    Whole crop year 28 2 4 9.3 42 1 7 13.9 320 ns
    All Post-fall 0 0 1 0.0 0 1 2 1 0 0 13 ns
    Thaw 80 35 16 3.8 29 22 12 4 1.1 6 43 <0.001
    Early growing 60 21 8 8.1 62 132 34 20 17.7 92 134 <0.001
    Thaw + early growing 65 25 7 11.9 92 105 26 15 18.8 98 177 <0.001
    Remainder 2 1 1 1 8 1 0 1 0 2 419 <0.001
    Whole crop year 21 2 2 12.9 31 2 5 19.3 609 ns
    • Mean = non gap-filled mean daily flux (g N2O-N ha−1 d−1).
    • Med = non gap-filled median daily flux (g N2O-N ha−1 d−1).
    • § SE = non gap-filled standard error.
    • Σ = gap-filled cumulative N2O (kg N2O-N ha−1) of whole crop year.
    • # All = percent Σ of whole crop year Σ.
    • †† n = number of non gap-filled daily flux values.
    • ‡‡ P = probability |z| > zcritical for Wilcoxon Signed-Rank Test; NS = median values not significantly different at the 0.05 probability level.

    From the time of late fall application, FR declined with air and soil temperatures, being very low and less than 7 kg CO2–C ha−1 d−1 by the onset of frozen soil conditions by 2 December and throughout the winter period (Fig. 1). During this time, average daily air and soil temperature were −10.3 and −0.2°C, respectively. Water-filled pore space was fairly high at the time of late fall application being above 0.80 m3 m−3 (Fig. 1).

    The thawing of soil induced an emission episode for the late fall anhydrous ammonia treatment over the period of 20 March to 28 April with a daily maximum of 505 g N2O-N ha−1 d−1 on 2 April (Fig. 1). In contrast, the maximum FN during spring thaw in the pre-plant anhydrous ammonia treatment was only 85 g N2O-N ha−1 d−1. Right after thaw, the late fall anhydrous ammonia treatment had greater nitrate concentrations than the pre-plant treatment (Fig. 1). For the late fall treatment, concentrations were 9.7 and 13.9 mg N kg−1 on 26 April and 26 May, respectively compared to 2.6 and 2.5 mg N kg−1, respectively, for the pre-plant treatment. The average daily FN in the thaw period was greater (P < 0.001) for the late fall than pre-plant anhydrous ammonia treatment (Table 1).

    Soil moisture remained high through April and May 2011, delaying planting until the near surface dried enough to use field equipment. The majority of N2O emissions for the pre-plant treatment occurred in early June soon after anhydrous ammonia application. This emission episode occurred from 10 June to 19 July with a very high daily maximum of 1254 g N2O-N ha−1 d−1 on 23 June, coinciding with a temporary depression in WFPS from 0.83 to 0.78 m3 m–3. Over this time, FR was increasing toward the yearly maximum of 54 kg CO2–C ha−1 d−1, air and soil temperature was at or above 20°C. High emissions from the pre-plant treatment also coincided with higher soil concentrations of nitrate. An increase first and then a sharp decline in soil ammonium immediately following pre-plant anhydrous ammonia application occurred, indicating rapid nitrification (Fig. 1). Rapid drying of soil occurred with the cessation of the pre-plant anhydrous ammonia treatment emission episode. In contrast, the late fall anhydrous ammonia treatment did not have a N2O emission episode for the early-growing season. In the post-fall and early-growing season periods, average daily FN was greater (P < 0.0001 and P = 0.02, respectively) for the pre-plant treatment (Table 1). FN was very low (<22 g N2O-N ha−1 d−1) over the remainder of the growing season and fall period with well-aerated conditions and a WFPS of around 0.45 m3 m–3 prevailing (Fig. 1). Cracks wider than 3 cm were readily visible at the soil surface and penetrated to more than 60 cm deep. Even the occurrence of 44 mm of rain on 20 Sept. 2011 failed to increase WFPS and induce emission of N2O. The rainfall simply bypassed the soil matrix, infiltrating the soil through the cracks.

    Relative differences in treatment values for ΣFN were similar to those for FN with ΣFN in the thaw period being greater for the late fall treatment but that for the early-growing season being less than the pre-plant treatment (Table 1). The 2010 to 2011 annual ΣFN were 37% less for the late fall treatment (3.2 kg N2O-N ha−1) compared with the pre-plant treatment (5.1 kg N2O-N ha−1).

    2012 Crop Year

    Similar to the results for the 2011 crop year, the late fall application of anhydrous ammonia in 2011 resulted in negligible N2O emissions in the fall-winter period but induced an emission episode with maximum FN of 405 g N2O-N ha−1 d−1 during soil thaw when air and soil temperatures reached above 0°C and WFPS was high at around 0.80 m3 m–3 (Fig. 1). In 2012, the onset of spring-thaw (air temperature >0°C) occurred on March 11. The extractable ammonium and nitrate concentration of soil was higher for the late fall than pre-plant treatment going into the winter of 2011–2012 and soon after the spring-thaw period (Fig. 1). The mean daily FN in the thaw period was greater (P = 0.049) for the late fall than pre-plant treatment, being 144 g N2O-N ha−1 d−1 and 36 g N2O-N ha−1 d−1, respectively (Table 1).

    Application of anhydrous ammonia on 12 Apr. 2012 failed to result in an immediate episode of N2O emission coincident with cool air and soil temperatures (<10°C) despite high WFPS of 0.80 m3 m–3 and nitrate concentrations greater than 55 mg N kg−1 (Fig. 1). However, an emission episode occurred 6 wk following the time of pre-plant application for both treatments, when temperatures were above 10°C and rains amounting to 60 mm increased WFPS from 0.50 to 0.60 m3 m–3 (Fig. 1). The emission episode occurred over 4 wk for both treatments with maximum daily FN was 1294 g N2O-N ha−1 d−1 for the pre-plant treatment and 572 g N2O-N ha−1 d−1 for the late fall treatment on June 12 (Fig. 1). Emissions from both treatments coincided with high FR and nitrate concentrations in the early growing season period. Soil nitrate concentrations declined with crop growth. On average, daily FN in the early growing season period was 41% greater (P = 0.013) for the pre-plant than late fall treatment (Table 1). Emissions during the latter part of the growing season and late fall were very low for both treatments, being less than 10 g N2O-N ha−1 d−1.

    The ΣFN for the 2012 crop year generally agreed with the results in the previous crop year. Late fall anhydrous ammonia application increased the ΣFN in the thaw period but decreased it for the early growing season period compared to pre-plant treatment. The proportion of N2O from thaw and early growing season periods contributing to whole crop year emissions was affected by the treatments. Very late fall application resulted in 14 and 86% of the emissions having occurred at thaw and early season periods than 2 to ∼100% of emissions for the pre-plant treatment, respectively (Table 1). Clearly, emissions for the late fall treatment during the early growing season period contributed most to the whole crop year emissions. This points to the more favorable conditions for N2O production in the early growing season than very late fall. But nevertheless, lower early growing season emissions resulted in ΣFN for the 2011–2012 whole crop year being 34% less for the late fall (9.2 kg N2O-N ha−1) compared with the spring pre-plant treatment (14.0 kg N2O-N ha−1).

    Over the two study years, emissions in the early growing season accounted for over 90% of total ΣFN for the spring pre-plant treatment and 65% for the late fall treatment. Despite an increase for thaw emissions, the greater decrease in early growing season emissions resulted in 35% lower ΣFN over the two crop years for the late fall than spring pre-plant treatment.

    Yield-Scaled Emissions

    Grain yield of wheat was lower for late fall than pre-plant treatment in 2011 (P = 0.0004) but grain yields of corn were similar in 2012 (P = 0.65). Spring wheat yielded 0.5 ± 0.1 (SE) and 1.2 ± 0.1 dry Mg ha−1 and corn 6.8 ± 0.2 and 7.4 ± 0.6 Mg dry grain ha−1, for late fall and pre-plant treatments, respectively. The early growing season of 2011 was unusually wet, resulting in delayed planting and herbicide treatment. By the time herbicide was applied on 30 June 2011, volunteer rapeseed (Brassica napus L.) widely covered the field treated the previous fall with anhydrous ammonia. The 2011 average yield in the region for our spring wheat variety was also poor at 1.8 Mg dry grain ha−1, and the 2012 average regional yield for our corn variety was 5.5 Mg dry grain ha−1 (MASC, 2015).

    Yield-scaled emissions for the wheat year were 6.4 and 4.3 kg N2O-N Mg−1 dry grain for the late fall and pre-plant treatments, respectively. For the corn year, yield-scaled emissions were 1.4 and 1.9 kg N2O-N Mg−1 dry grain for the late fall and pre-plant treatments, respectively.

    Temperature and N2O Emissions

    The episodic nature of N2O emissions resulted in some clear distinctions as a non-continuous function of soil temperature at 5-cm depth (Fig. 2). Emissions occurred over the range of 0 to 5°C for the thaw periods, most noticeably for the late fall treatment. However, emissions also occurred for both treatments over the temperature range of 10 to 20°C, highlighting quite distinct events and controlling mechanisms.

    Details are in the caption following the image

    Mean daily N2O emission (FN) in relation to soil temperature at 5-cm depth for late fall and pre-plant anhydrous ammonia treatment for cropping years 2011 and 2012. Data are presented with field allocations for treatments switched on 4 Nov. 2011.


    Over two crop years, late fall application of anhydrous ammonia close to soil freezing emitted 35% less N2O than application at pre-planting. Not surprising, the temporal pattern of emissions varied with application timing and study year. Late fall application did not induce N2O emissions immediately after application but increased emissions during the following spring thaw. In contrast, spring pre-plant application, spurred by rain events, induced N2O emissions with a higher magnitude than those for late fall application. Late fall application had inconsistent emissions during the following early growing season, with no emissions in one year.

    A few other studies in cold climates with frozen soil have also reported lack of emissions with late fall application of anhydrous ammonia and other synthetic N sources. Application of anhydrous ammonia in early October in Iowa resulted in no fall and winter emissions despite application occurring at soil temperatures slightly above 10°C with freeze-up occurring two months later (Bremner et al., 1981). Also in Iowa, application of anhydrous ammonia in mid-November when air temperature was near freezing resulted in no fall and winter emissions (Parkin and Hatfield, 2010). Burton et al. (2008) reported no difference in emissions for fall and spring banded anhydrous ammonia application in Manitoba. In studies with animal manures in Manitoba (Tenuta et al., 2010; Asgedom et al., 2014), late fall application before freezing also did not increase N2O emissions. Similarly, composted and raw pig slurry applied in fall in southern Ontario did not result in fall and winter N2O emissions (Kariyapperuma et al., 2012).

    The results of the current study further add to a consensus of reports that late fall addition of N products before freeze-up result in negligible emissions of N2O in the post-application fall and winter periods in regions with below-freezing air and surface soil conditions (Bremner et al., 1981; Parkin and Hatfield, 2010; Tenuta et al., 2010; Kariyapperuma et al., 2012; Asgedom et al., 2014). In the current study, the majority of emissions from either anhydrous ammonia treatment generally occurred with soil temperatures over the range of 10 to 20°C in late spring and early summer. Thus, the current recommendation for fall application of ammoniacal N fertilizers below 5°C to minimize nitrification (e.g., MAFRD, 2007) seems appropriate to prevent N2O emissions in late fall.

    The timing of anhydrous ammonia application altered the temporal pattern of N2O emissions in the current study with increased emissions when soil temperature reached 0°C during thaw following late fall treatment. Episodes of N2O emission at thaw are commonly observed at the TGAS-Man site (Glenn et al., 2012; Maas et al., 2013), other locations in the Canadian Northern Great Plains (Nyborg et al., 1997; Lemke et al., 1998; Dunmola et al., 2010; Gao et al., 2014), and other locations with subfreezing conditions in winter (Pattey et al., 2007; Wagner-Riddle et al., 2007). In this study, thaw emissions occurred as soil near surface increased from 0 to 5°C and similar to observed by Wagner-Riddle et al. (2010) in southwestern Ontario.

    We hypothesized that frozen conditions shortly after late fall application would result in negligible N2O emissions from the time of that application to thaw the following spring. Contrary to expectation, late fall applications of anhydrous ammonia increased N2O emissions at thaw. Wagner-Riddle et al. (1997, 2008) observed that N2O field emissions during spring-thaw were associated with greater soil nitrate concentrations. Tenuta and Sparling (2011) confirmed using a laboratory assay that greater nitrate concentrations increased thaw emissions due to denitrification from the plow layer but less so for subsurface soil. In the current study, late fall application of anhydrous ammonia increased nitrate concentrations in the subsequent spring of each study year. Though anhydrous ammonia was applied in late October or early November, when soil temperature was below 5°C and subfreezing soil temperatures occurred soon after and throughout winter, it seems that nitrification produced nitrate leading to higher N2O emissions at thaw. In Saskatchewan, Malhi and Nyborg (1979) observed overwinter accumulation of nitrate from granular urea applied early October to alfalfa (Medicago sativa L.) despite frozen soil conditions having set in by the middle part of November. Clearly, an understanding of nitrification under frozen soil conditions is required.

    There was a noticeable lack of emission in the early growing season of 2011 for the field that was fertilized in late fall of 2010, coinciding with low extractable ammonium and nitrate soil concentrations. The magnitude and duration of the N2O emission episodes following spring pre-plant application in 2011 were much less than in 2012. Differences in N2O emissions between years could partly be caused by the lower N application rates for the 2011 crop year (112 kg N ha−1) compared to 180 kg N ha−1 in 2012. However, pre-plant spring applications of urea fertilizer to nine site years of fields at the TGAS-Man site from 2006 to 2010 at about the same application rate as for the 2011 cropping year (average rate 105 kg N ha−1) averaged 5.2 (range 3.8–7.2) kg N2O-N ha−1 (Glenn et al., 2012; Maas et al., 2013). Thus, soil conditions affecting N2O production and consumption processes could have been also responsible for differing total emissions of N2O between cropping years in this study. Late snow-melt delayed planting in 2011 with WFPS being relatively high at 0.82 m3 m–3 with emissions occurring with drying to 0.78 m3 m–3 two to three weeks after pre-plant application compared to rain increasing WFPS from 0.50 to 0.60 m3 m–3 over a 7-wk period following pre-plant application in 2012. Granli and Bøckman (1994) proposed that soil moisture at 0.60 m3 m–3 WFPS maximizes nitrification and is the threshold at which higher levels enhance denitrification, whereas WFPS > 0.80 m3 m–3 could result in full inhibition of nitrification and more complete reduction of nitrate to N2 by denitrification. Thus it is likely that the poor aeration of the early growing season in 2011 resulted in denitrification reduction of N2O. Interestingly, a temporary slight depression in WFPS from 0.83 to 0.78 m3 m–3 coincided with the small emission of N2O with the spring pre-plant application. It is important to note that WFPS was determined at the 10-cm depth and the surface was likely drier because machinery could use the field for fertilizer application and planting. Emissions of N2O result from production of the gas nearer to the surface than deeper (Wagner-Riddle et al., 2008; Gao et al., 2014). Thus, perhaps lower WFPS at the surface than the measured values at 10 cm depth corresponded to the emissions following spring pre-plant application in 2011.

    The current study highlights the risk of late fall applied N being unavailable to the following crop. Grain yield was lower and yield-scaled emission higher with late fall than pre-plant anhydrous ammonia application of the 2011 crop year. We believe yield-scaled N2O emission treatment differences were not because of differences in anhydrous ammonia application efficiencies. The same anhydrous ammonia application system was used for fall and pre-plant treatments, ammonia vapors were not visible from injection bands, injection bands seemed to be closed, and soil moisture level was high for late fall and pre-plant treatments of the 2011 crop year. Rather, lower nitrate concentrations in the early growing season for the late fall compared to spring pre-plant treatment likely accounted for the lower yields in the former. Nitrate concentration may have been suppressed by either denitrification under wet soil conditions or uptake by volunteer rapeseed. Rapeseed had been grown at the TGAS-Man study site in 2009. Late fall application of anhydrous ammonia seems to have broken dormancy of a volunteer rapeseed seedbank likely because of increased nitrate concentration (Bewley et al., 2006) in spring with late fall anhydrous ammonia treatment. Loss of N through leaching is unlikely in the clay soil at the site.


    Late fall application of anhydrous ammonia before freeze-up decreased cumulative emissions of N2O compared to spring pre-plant application over the 2 yr in the current study. Timing of application did change the temporal pattern of N2O emissions with median emissions rates higher at thaw and early growing season for late fall and spring pre-plant applications, respectively. The results highlight the value of near continuous whole-year measurements over large areas to estimate cumulative emissions when the temporal pattern of N2O episodes is affected by soil management practices such as timing of fertilizer N addition. Further work is required to control nitrification in frozen soil of late fall applied N to decrease emissions at thaw and agronomic effectiveness when spring conditions results in delayed planting the following spring.


    We thank Mervin Bilous, Matt Gervais, Krista Hanis, Jenna Rapai, and Brad Sparling for assistance in the field and laboratory, as well as the farm staff at the University of Manitoba Glenlea Research Station and Dustin Wiens for assistance with field operations. Funding was provided by the Natural Sciences and Engineering Council of Canada (NSERC) Discovery Grant Program, the Canada Research Chair Program in Applied Soil Ecology, the Canadian Fertilizer Institute, the Canada Agricultural Adaptation Program, and the Agricultural Greenhouse Gas Program of Agriculture and Agri-Food Canada. This study was conducted under the program activities of the National Centre for Livestock and the Environment (NCLE) with equipment and field infrastructure supported in part by a Canadian Foundation for Innovation grant to NCLE.


      Detailed Methodology

      The flux-gradient technique measures gas concentrations at two heights, with the difference proportional to the net exchange of the gas between the atmosphere and soil-plant surface. The flux (F) of either N2O or CO2 over 30-min intervals was determined as:
      where K is the turbulent transfer coefficient or eddy diffusivity, and Δ[C] is the concentration gradient of N2O or CO2 measured over the vertical height difference, Δz. Concentrations of N2O and CO2 were measured with a tunable-diode-laser absorption spectrometer (TGA100A, Campbell Scientific) set inside a trailer located at the junction of four experimental fields (Fig. A1). The TGA was further housed within an insulated, temperature-controlled enclosure. The lead-salt tunable-diode-laser of the TGA (IR-N2O/CO2, Laser Components GmbH, Olching, Germany) was operated at a cryo-cooled temperature of −189.15°C using liquid nitrogen in a dual-ramp, jump-scanning mode (Fried et al., 1993). The laser was parameterized for the concurrent measurement of atmospheric concentrations of N2O and CO2 at 10 Hz with a measurement cell of 1.5-m length. Atmospheric samples above the fields were drawn ∼125 m at 5 L min−1 from a triangular instrumentation tower located in the center of each field by a rotary vane vacuum pump (Model RS0021, Busch Vacuum Technics, Boisbriand, QC). A Nafion dryer at the instrument tower and the inlet to the TGA removed water vapor from atmospheric samples. A reference gas with concentrations of 0.2% v/v N2O and 30% v/v CO2 was continuously passed through the TGA reference cell of 4.6-cm length. The calibration of the TGA was verified by passing reference gases of ultra-pure N2 (zero) and N2O (407 ± 20 nmol mol–1) and CO2 (397.96 ± 7.96 μmol mol–1) in N2 (span) from the inlets. This procedure followed the replacement of reference tanks and sample intake lines that were previously removed for field operations.
      Details are in the caption following the image

      Field locations as outlined by Glenn et al. (2010). In this study, Fields 2 and 3 on the east side of the experimental design were used.

      Two sample intakes separated vertically by 0.65 m (Δz) were mounted to each instrumentation tower. Switching between the upper and lower intakes occurred every 12 s to calculate average differences in the concentrations of N2O and CO2. Concentration differences (between intake heights) were calculated on 30-min intervals to 0.1 nmol mol–1 for each of the four fields at the site sequentially to obtain approximately one mean 30-min difference every 2 h per field. Concentration differences values were removed if power outages resulted in incorrect wavelengths of the gases being monitored. Outlier concentration difference values were removed if greater than ±4 SD of the mean 30-min concentration difference. System leaks sometimes occurred because of rodent damage to gas lines in the field. Line damage was indicated by a reduction in vacuum pressure of the tunable diode laser and concentration difference data removed if pressure exceeded 8 kPa.

      Care was taken to ensure a balance between measuring gradients far enough above the surface to minimize roughness sub-layer effects but low enough to contain the flux footprint to each field. During the winter, bottom intakes were located ∼0.5 m above the snow-pack; for the spring and fall non-cropping seasons, the bottom intakes were kept 0.5 to 0.75 m above the stubble, residue, or bare soil. During the growing season, the intake assembly was raised periodically relative to crop height (hc). In the presence of spring wheat, the upper and lower intakes were maintained between ∼1.6 to 2.2 hc, respectively. In the presence of corn, the upper and lower intakes were maintained between ∼1.2 to 2 hc, respectively, with the shortest distance occurring at peak crop height (August to October). Snow depth, stubble, and hc were measured manually with a ruler or tape weekly.

      The K term was calculated for momentum using similarity-theory (e.g., Pattey et al., 2007; Phillips et al., 2007; Denmead, 2008), assuming that K for mass and momentum were equal. A sonic anemometer-thermometer (CSAT-3, Campbell Scientific) was mounted at a height of 2 m on each instrumentation tower to measure temperature and three-dimensional velocities. The friction velocity (u*) was calculated as the 30-min covariance between the vertical and horizontal velocities, following the coordinate rotation of Tanner and Thurtell (1969). K was then calculated as:
      where k is the von Karmann constant (0.4), z2 and z1 are the upper and lower intake heights, respectively, and φ2 and φ1 are stability correction factors for upper and lower intakes, respectively. The heights z2 and z1 were corrected for zero-plane displacement to be an effective height; for periods of snow cover, this was set to the depth of snow and during the rest of the year it was set to be 0.66 hc (Garratt, 1992; Denmead, 2008).
      Atmospheric stability was calculated from the Monin–Obukov length, L, as:
      where T is air temperature in Kelvin, g is acceleration due to gravity, and wT′ is the sensible heat flux density measured through eddy covariance. Values for φ2 and φ1 are stability dependent. For stable conditions (L > 0; Businger et al., 1971),
      whereas, for unstable conditions (L < 0; Paulson, 1970),
      where subscript n denotes either the upper (n = 2) or lower height (n = 1) in Eq. [A2].

      Gapfilling Methods and Respiration Determination

      The flux-gradient system revisits each field every 2 h, and we calculate daily total N2O flux based on the mean of the measurements for the day; nominally based on twelve 30-min measurements each day. In some instances, data were missing, and these gaps contribute to a smaller sample to arrive at daily totals. If no measurements were available for a given day, the whole day was treated as a gap, and reported as missing. Gaps that occurred when fluxes were negligible had little effect on the cumulative flux. Hence there was uncertainty only during emission events when gaps created a smaller sample size to estimate the daily total. Gaps were caused by instrument malfunction or calibration, quality-control issues, and when turbulent conditions were too calm for flux determination, that is, when the friction velocity (u*) was below the defined threshold of 0.12 and 0.15 m s−1 for the wheat and corn years, respectively (Glenn et al., 2010). The low-turbulence condition occurs on many nights, which reduces data capture at these times by as much as two-thirds at some long-term flux sites (Falge et al., 2001). Our quality-control included a stationarity test where the 30-min periods were broken into 5-min periods. If the vertical wind velocity for any 5-min period had SD outside of ±30% from the 30-min period, it was removed from the series (Foken et al., 2004). If more than one 5-min period was non-stationary, the whole 30-min period was eliminated; this typically occurred when conditions approached the friction velocity threshold. Table A1 shows the percentage of measurements retained during each N2O emission event. More than half of the period was captured with measurements for most of the events. There were missing data for the post-fall event in 2011 between 13 Nov. 2010 and 3 Jan. 2011 when data were rejected because of uncertainty in proper switching of sample intakes. The fluxes were low during this period, thus missing data tended not to have a large effect on the estimation of an integrated flux. For the full study period, data retention averaged 50%.

      Table A1. The percentage of retained measurements for N2O flux over the post-fall, thaw, and early growing season events for the two years of study. n is the number of 2-h periods sampled during each event.
      Treatment 2011 crop period 2012 crop period Study total
      Post-fall Thaw Early growing Total Post-fall Thaw Early growing Total
      Spring 6 55 56 43 61 69 62 58 50
      Fall 7 55 57 44 62 68 61 58 50
      n 588 480 480 4620 156 108 1128 4152 8772

      The respiration flux of CO2 was calculated in two steps. First, measurements of net ecosystem exchange were used to calculate respiration during periods when photosynthesis is known to be zero. This was defined as nighttime or when air temperatures <0°C. Next, respiration during daytime or when there were gaps in the CO2 flux was calculated through the Fluxnet Canada Research Network algorithm (Barr et al., 2004). This calculation used a logistic regression equation as a function of 5-cm-depth soil temperature developed on a 100-point moving window that helped account for other changing variables, such as soil moisture and substrate availability. This method was used previously for respiration and net ecosystem CO2–exchange measurements at this site (Glenn et al., 2010). Gaps in the carbon flux data are the same as those for N2O flux (Table A1).

      Detection Limits and Uncertainty

      Detection limits for our flux-gradient system were assessed through testing a “null gradient”. This was done by setting the intakes to identical heights, and measuring the difference between intakes to detect both random differences and bias. For N2O, concentrations at the same height had SD = 0.045 nmol mol−1 with essentially no bias (Glenn et al., 2012). Applying a friction velocity threshold of 0.12 m s−1 for the corn crop year gave a minimum flux detection level of about 0.05 nmol m−2 s−1. For comparison, chamber studies typically have similar detection levels (0.05–0.1 nmol m−2 s−1), although these depend on the sampling period and the gas accumulation rate model method (Parkin et al., 2012; Duran and Kucharik, 2013).

      Uncertainties and potential bias in the flux-gradient technique arise from a few technical issues. One issue is assumed constant flux layer where fluxes could be underestimated if measurements are less than about 1.6 hc (e.g., Simpson et al., 1998). At our sites, most measurements were made above this height, although we had some measurements at 1.2 hc during our corn year. A second issue is the estimation of K for scalar transfer, Kx, using the transfer coefficient for momentum (Km) or heat, which is more readily quantified. Several researchers have compared the different transfer coefficients with variable results. For example, Phillips et al. (2007) found that the transfer coefficient for heat was about 0.85 that for Km but assumed unity among transfer coefficients to estimate N2O fluxes. Flesch et al. (2002) suggested that the ratio of Km/Kx (the Schmidt number) was about 0.6, whereas Wilson (2013) suggested that this is about 0.78 for CO2; however, it is important to note that this should approach unity at higher measurement heights. Wagner-Riddle et al. (1996) and Simpson et al. (1997) increased Kx by a factor of 1.3 over Km to compensate for under-sampling of low frequencies of turbulent motions that cause energy closure imbalance. We tested our CO2–flux estimates from the flux-gradient technique directly by comparing with three eddy covariance towers located on the same field (our north-east field) from late July to early September in 2014. For a joint data set of 57 30-min daytime periods, the three eddy covariance towers had a coefficient of variation of 26% (mean of −7.3 μmol m−2 s−1), and the flux-gradient tower was within 10% of the mean for the period. Similarly, Griffis et al. (2007) had only a 6% difference between their eddy covariance and flux-gradient measurements for CO2. Hence, we have not adjusted our fluxes based on arguments related to bias in the K value. However, we recognize that both micrometeorological techniques may underestimate fluxes, based on a common inability to close energy balance, typically in the range of 15 to 20% (e.g., Wilson et al., 2002). We assume that the arguments made for CO2 flux also hold for N2O flux measurements.

      Sampling Footprint

      Micrometeorological flux measurements sample air parcels that have experienced contact with an upwind surface. Sampling over the flux measurement period (i.e., 30 min at our towers) is an ensemble measurement of the spatial flux density originating from an upwind area, known as the flux footprint. For the flux-gradient technique, we can pose such questions as: (i) How much of the treatment field is sampled spatially? (ii) What is the proportion of the measured flux that could have originated outside of the treatment field? (iii) Given that concentrations are sampled at two different heights, how different is the footprint for each height and what is the potential effect of this difference?

      We address these questions using the footprint model of Chen et al. (2009) and the analytical formulation of Kormann and Meixner (2001). Here, we set a domain of 400 m × 400 m (fourfold the field size), and look at the footprints for both upper (1.3 m height) and lower (0.7 m height) intakes. The model was run for the month of April in 2011 (30 d) coinciding with the extended period of N2O emissions that year. For this non-vegetated surface, the surface element height was set to 0.05 m to depict surface soil aggregates with a roughness height of 0.0065 m. Non-vegetated surfaces are smoother than cropped surfaces, so the resulting footprint will originate slightly further upwind than for the rougher summer conditions. Inputs were 30-min averaged friction velocity, wind direction, and sensible heat flux data from the sonic anemometer-thermometer located at the center of the field.

      Figure A2 depicts footprint contours averaged over the month of April 2011 for a given field. The north-east quadrant of the field is under-represented because of less-frequent winds from that direction. The footprint for the upper intake is slightly greater than for the lower intake. For the upper intake, 60% of the flux footprint originates from within the field (100 m extent from the tower) for all wind directions, with 90 to 99% being within 150 m. Fluxes originating from the north-east and south-west quadrants are totally from within the field. To address our initial questions, we conclude, based on the 90% contour, (i) 83% of the field is sampled from the upper intake and 70% from the lower intake, (ii) 26% of the flux originates from outside the field for the upper intake, and 2% from the lower intake, and (iii) the upper and lower intake heights measured coinciding areas, but the upper intake has a flux footprint integrated area of 4.2 ha whereas the lower intake samples integrates an area of 2.9 ha, averaged over the 30-d period.

      Details are in the caption following the image

      Flux footprints for the upper (1.3 m) and lower (0.7 m) intakes for April 2011. The contours represent footprint percentiles for the full month with the measurement tower at the center location. The treatment field extends from –100 to 100 m in each ordinate.

      Our conclusion is that the measurements sample a very large area of each field (>70%) allowing for the integration of spatial variability. Although ∼0.9 ha is sampled by the upper intake outside of the field (based on a 90% contour), this is all within 50 m and heavily weighted to a 20-m buffer that has a similar treatment outside the field.