Overwinter and Spring Thaw Nitrous Oxide Fluxes in a Northern Prairie Cropland Are Limited but a Significant Proportion of Annual Emissions

Croplands that experience seasonal soil freezing and thawing have been shown to be significant sources of N2O emissions. Yet, there is a paucity of year‐round N2O emission data for one of the most significant crop production regions that seasonally freeze, the Prairies. Here, we present micrometeorological N2O fluxes measured over 4 years in Saskatchewan, Canada, to evaluate the magnitude of freeze‐thaw N2O emissions and investigate its driving factors. Significant thaw related emissions occurred in 2 of the 4 years and were associated with relatively higher fall nitrate levels and a more gradual soil thawing period. Overall, fall soil nitrate levels were a strong explanatory variable for the differences in non‐growing season (NGS) N2O emission (r2 = 0.485). Measured cumulative N2O emissions for the NGS were 123–938 g N ha−1 and were much smaller than those obtained at other cold climate sites but amounted to 52% of annual totals on average. The November to April period contributed 30% of the annual total emissions in years without major thaw events, but 70% in years with significant thaws. NGS N2O emissions were not explained by cumulative freezing degree days unlike most other cold climate sites. We propose that NGS N2O emissions are more strongly influenced by thaw dynamics during freezing‐thawing conditions in dry regions, whereas freezing intensity is the dominant factor for wetter regions. Our results indicate that even for a semi‐arid region freeze‐thaw is an important source of N2O emissions and must be considered for more accurate reporting and development of mitigation strategies.


Introduction
Croplands are a significant source of greenhouse gas emissions in the form of nitrous oxide (N 2 O) (Tian et al., 2020), a gas that also remains as the main stratospheric ozone depleting substance (Hayashi & Itsubo, 2023).Nitrous oxide is produced by microbial processes and increased soil inorganic nitrogen (N) levels resulting from N fertilization are the main driver of N 2 O emissions from cropland soils (Butterbach-Bahl et al., 2013).A large fraction of annual emissions can occur during short events (∼days) termed "hot moments" (Wagner-Riddle et al., 2020) induced by dry-wet or freeze-thaw cycles, which are magnified with increased available soil N substrate in croplands (Congreves et al., 2018).Seasonal freeze-thaw cycles occur in large areas of intensive annual crop production in Canada, US, Northern Europe, and China and have been estimated to increase global budget estimates by 1.07 Tg N or 17%-28% based on a simple model using cumulative freezing degree-days as a predictor (Wagner-Riddle et al., 2017).Recently, inclusion of freeze-thaw processes based on freezing degreedays increased annual emissions for croplands and grasslands reported in the US national greenhouse gas emission inventory by 6%-16% (Del Grosso et al., 2022).Canada's National Inventory Report includes a 1.4 adjustment factor that is applied to the Eastern ecozones of the country to account for non-growing season (NGS) emissions based on a limited number of studies that have shown significant emissions during winter and spring thaw (Rochette et al., 2008).An analysis using recently published research recommended a revised factor of 1.55 for Western and Eastern ecozones in Canada and estimated NGS emissions to be 35.5% of annual emissions (Pelster et al., 2023).A top-down modeling approach indicated that spring thaw N 2 O emissions could be comparable in size to the growing season maxima induced by N fertilizer application in Canadian cropland (Nevison et al., 2023).Clearly, consideration of freeze-thaw emissions from croplands is important for improved national inventory emission estimates in cold climates.However, there is a paucity of data collected over winter that captures freeze-thaw emissions in these regions, adding uncertainty to national inventories despite recent updates.
The Western region in Canada is dominated by the Prairie ecozone which occurs mostly between latitudes of 49°a nd 54°N, comprises more than 80% of the national agricultural land or 52 Mha (of which ∼60% is cropland), receives most of the N fertilizer used (∼75%) and grows nearly all the canola, spring wheat, and barley produced in Canada (Awada et al., 2021;Statistics Canada, 2016, 2022).The region is characterized by its aridity with an annual average precipitation ranging from 300 to 550 mm (McGinn, 2010) and long winters with approximately 150 soil freezing days and 5-15 freeze-thaw cycles (Henry, 2008) There are five soil zones in this region (Brown, Dark-brown, Black, Gray and Dark-gray) with a general precipitation gradient along these soil types and contrasting soil organic matter values (higher for the Black zone at 5%-10% and lower for the Brown zone soils at 2%-5%) (Awada et al., 2021).The diversity in climatic and soil conditions implies that the drivers of freeze-thaw cycle emissions and the importance of freeze-thaw events to annual totals could vary considerably across the Prairies.
The most extensive N 2 O flux dataset for the North American Prairie region is based on year-round micrometeorological measurements on a Black soil site located at its Eastern edge in Manitoba (Tenuta et al., 2019).Cumulative freezing degree-days are large at this site (∼400-800) and induce significant NGS (November-April) emissions ranging from 0.94 to 2.8 kg N ha 1 (median = 2.1 kg N ha 1 ) (Wagner-Riddle et al., 2017); although growing season emissions are also high so that these NGS emissions represent 30% of the annual total.In contrast, Pelster et al. (2023) indicated much lower NGS emissions of 0.345 kg N ha 1 (34% annual total) averaged for the Prairie ecozone, flagging the Manitoba site's high clay content, poor drainage and greater precipitation as likely not representative of the majority of Prairie conditions.Indeed, Yang et al. (2022) found different dominant drivers for NGS N 2 O emissions in US Midwest regions depending on precipitation: more intensive freezing was the main driver for the wetter southeast region (NGS precipitation >300 mm) but increased soil moisture at thaw was the main driver for the drier northeast region (NGS precipitation <300 mm).These authors used a processbased model to simulate NGS emissions and cautioned that evaluation of simulations was limited and did not cover spring thaw but rather used data collected during winters with small fluxes, adding uncertainty to model forcing.Additional overwinter measurements are clearly needed for the North American Prairie region, particularly the drier part of this significant agricultural production zone.
Micrometeorological measurements provide year-round gas fluxes integrated over large areas (several hectares) with high temporal resolution-ideal for capturing episodic N 2 O events overwinter and during thaw.However, deployment of micrometeorological approaches has been limited to a few agricultural sites in cold regions (Wagner-Riddle et al., 2017).Non-growing season flux measurements using soil chambers have also not been extensive, although recently there has been an increase in the number of studies that have attempted to capture spring thaw emissions.For example, emission data collected in the Prairie ecozone using soil chambers comprised most of the database used to derive the NGS revised adjustment factor for Canadian croplands proposed by Pelster et al. (2023).However, the temporal resolution of chamber studies is low (e.g., lack of sampling during the coldest part of the winter season) and fully capturing the thaw events is difficult due to the unpredictability as to when they occur (An et al., 2021;Cambareri et al., 2017).
Here we report on a new agricultural micrometeorological N 2 O monitoring station on the Canadian prairies in Saskatchewan established in fall 2018.Our objectives were to (a) evaluate the magnitude of overwinter and spring thaw N 2 O emissions over 4 years, and (b) investigate the driving factors that impact these NGS emissions.We expected low N 2 O emissions during the cold overwinter period but a significant emission event during spring thaw.It was hypothesized that the high degree and severity of overwinter freezing would proportionally lead to high N 2 O emissions at thaw.

Site Description
The micrometeorological flux monitoring station was established in October 2018 at the University of Saskatchewan North Management Area (52°09'22.7"N106°36'28.8"W)research site in Saskatoon, SK.The soil at the site is classified as Dark Brown Chernozem of the Asquith association, with a sandy loam texture (52% sand, 32% silt, 16% clay), soil organic matter content of 4.5%, soil bulk density of 1.49 Mg m 3 , soil pH of 7.9, and a cation exchange capacity of 21.3 cmol c kg 1 in the 0-15 cm layer.
Here we report on N 2 O flux measurements conducted in four fields over four NGS from December 2018 to September 2022, according to methodology by Wagner-Riddle et al. (2007).Each field was 2.9 ha in size (110 m in the North-South direction by 265 m in the East-West direction) labeled northwest (NW), southwest (SW), northeast (NE) and southeast (SE) fields depending on its orientation in relation to the center of the area.Towers to sample air were placed in each field (Figure S1 in Supporting Information S1).Flux footprint calculations using Kljun et al. (2015) and projected onto the field area with the mapping tools found in Rey-Sanchez et al. (2022) indicated that the measured flux originated well within the 2.9 ha of each field from approximately 25% of the area on average (Figure S1 in Supporting Information S1).The fields were within a level and aerodynamically homogenous area of 68 ha, with the surrounding area planted to the same crop as the 2.9 ha fields.Cereal feed crops (primarily barley) had been grown at the site over the previous decade and barley silage was grown during the 2018 growing season prior to field establishment.All four fields received the same management from October 2018 to October 2019 to assess any inherent variability in N 2 O emission amongst fields, with barley planted on 20 May 2019 and swathed and baled on 14 October 2019.Starting in fall 2019, different management was applied during the NGS and/or growing season.In the fall of 2019, we attempted to create different soil moisture and nutrient conditions before soil freezing in late fall by applying 70 mm of supplemental irrigation to two of the fields (NW and SW) between September 30 and October 2, and by adding urea at 55 kg N ha 1 to the NW field on October 29 (Table 1).Addition of N aimed at creating two fields with high nitrate concentration and two fields with low nitrate concentration and was based on soil nitrate levels of samples taken on 22 October 2019 (Table 2).The soil analysis indicated contrasting nitrate levels between fields; hence, the NW and NE fields entered the freezing period with more than 3× the nitrate levels of SW and SE fields (Table 2).Starting in the spring of 2020, the crop N management of the fields diverged based on fall nitrate levels with the low nitrate fields (SW and SE, Table 2) receiving urea at 77 and 66 kg N ha 1 , respectively, when fields were seeded to barley on 24 May 2020 (Table 1).Barley was harvested for silage on 4 August 2020.In fall 2020, we opted for creating contrasting soil conditions by cutting the barley stubble in the NE and SE fields to 10-13 cm height (low stubble) compared to the normal 20-26 cm height in the NW and SW fields (Table 1).Lower stubble has previously been shown to trap less snow and hence was expected to result in lower soil insulation from cold air temperatures and less water content at

Global Biogeochemical Cycles
10.1029/2023GB008051 thaw.In the spring of 2021, canola was planted and fertilized on May 18, with the NW and NE fields receiving 100 kg N ha 1 of urea, whereas the SW and SE fields received 50 kg N ha 1 of urea mixed with an N stabilizer (Anvol®, Koch) containing the urease inhibitor Duromide.The rationale for contrasting rates was to avoid artificial skewing or confounding the results (2021-onward) by assigning low N rates to two fields that already had low N levels and/or the same stubble/irrigation history.Canola was harvested on 28 October 2021 and no other field operations were conducted until the end of April 2022.

Micrometeorological N 2 O Flux Measurements and Emission Calculations
Half-hourly N 2 O fluxes (F N2O ) were measured using the flux-gradient method (Wagner-Riddle et al., 1996): where u* is the friction velocity, k is the von Karman constant (=0.41),ΔC is the difference in N 2 O concentration between the sample heights z 1 and z 2 , d is the displacement height, and Ψ h2 and Ψ h1 are the integrated Monin-Obukhov similarity functions for heat for each sampling height.The difference in concentration (ΔC) was measured at a frequency of 10 Hz using a tunable diode laser trace gas analyzer (TGA200A; Campbell Scientific, Logan, UT, USA) placed inside an instrumentation trailer situated at the center of the four fields.The TGA200A received air drawn by heated sample intakes (AP200, Campbell Scientific Inc., Logan, UT, USA) placed at two heights on each of four measurement towers (one per field) with the upper (z 2 ) and lower (z 1 ) air intakes separated by 50 cm.Over the study period, the intake heights were adjusted to account for changes in snow depth or crop height, with the lower intake being at approximately 40 cm when no crop was present and 1.5-1.9times the crop canopy height during the growing season.The intakes featured rain diverters, a 3 μm inline filter with a small heater to prevent condensation, and had a mixing volume of 0.75 L. Tubing (∼100 m of 1/4 inch OD SynflexTM) connected the eight intakes to a 16-inlet manifold with integrated pressure control and flow measurement (Campbell Scientific Inc., Logan, UT, USA).The total flow into each inlet was set by orifices (0.007 mm) and a vacuum pump (XDD1, Edwards) continuously drew air from the four fields and two sample heights into the manifold.A second vacuum pump sub-sampled the manifold, pulling air into a single-tube dryer (PDIT, Campbell Scientific, Logan, UT, USA) and then to the sample inlet of the TGA200A.Analyzer operation and calibration followed the procedure described by Brown et al. (2018).Air sampling alternated between the upper and lower intakes for one of the four plots every 15 s over 30 min and was controlled by a valve control module (SDM-CD16S, Campbell Scientific, Logan, UT, USA) and a program in the CR3000 data logger.Approximately 60 of the 150 data points sampled during each of the 15 s intervals were eliminated from the calculation during the transition times between sampling levels to ensure complete flushing of sample lines, leaving ∼5,400 samples per intake per 30 min period.Switching between the four fields occurred in sequence every 30 min, so that the maximum number of half-hourly ΔC measurement per day and per plot was 12.
The other parameters needed to calculate F N2O were derived from sonic anemometer (CSAT3, Campbell Scientific) measurements and equations listed in Wagner-Riddle et al. (1996).For periods when the sonic anemometer did not yield data (e.g., due to rainy or foggy conditions) but the trace gas analyzer still yielded a ΔC value, we calculated u* based on cup anemometer data (Wagner-Riddle et al., 2007).Half-hourly N 2 O flux values were filtered, with values discarded based on analyzer performance, friction velocity, stability parameters, boundary layer height or fetch conditions and wind direction according to criteria given in Wagner-Riddle et al. (2007).
Mean daily N 2 O fluxes per field for each year were calculated by averaging the 30 min fluxes each day and converting the values to g N 2 O-N ha 1 d 1 .Non-growing season (November-April), growing season (May-October) and annual total N 2 O emissions were calculated based on daily values after gap-filling missing data.For months with emission events and short gaps (<6 days), missing values were estimated using the average daily flux for the n days preceding and following the gap (where n is the number of missing days).Cumulative values were then calculated for the whole month and each field by summing daily values.For months without emission events and no clear difference in fluxes between fields, we combined all available daily values to obtain a mean daily flux for the month that was then scaled for a total for the month by considering the number of days in a month.Monthly totals for each period were then added to obtain annual, NGS and growing season emissions.In cases where no Global Biogeochemical Cycles 10.1029/2023GB008051 data were collected for a whole month, we used the average monthly N 2 O emission from the corresponding month in other years as a substitute.For years with clear thaw emission events, we separated the NGS N 2 O emission into a Winter period (all months before thaw) and a Thaw period (month(s) when the main thaw events took place).Nitrous oxide emissions are identified for the four NGS as: Year 1 (November 2018-April 2019), Year 2 (November 2019-April 2020), Year 3 (November 2020-April2021), and Year 4 (November 2021-April2022).

Supporting Data
The trace gas analyzer used also measured carbon dioxide concentrations and these values were used to infer analyzer performance since flux values significantly larger than zero occur more often for carbon dioxide compared to N 2 O. Air temperature, precipitation, and snow depth on the ground were obtained from the Climate Reference Station (CRS) operated directly on site by the Saskatchewan Research Council.Soil moisture and temperature were measured using eight probes (CS650, Campbell Scientific) installed at 10 and 20 cm depth, two in each field.Sensors were installed every year around November 1 except in 2020 when sensors were not installed and data from the weather station was used instead.Freezing degree days (FDD) (i.e., temperature <0°C) were calculated using air temperature and soil temperature at 10 cm depth, and cumulative sums over the NGS were obtained.Composite soil samples (0-15 and 15-30 cm depths) were collected along a zig-zag transect across each field involving 11 sample points, each fall after crop harvest (September or Oct).Soil samples were kept on ice in a cooler until transported back to the laboratory.Soil samples were shipped to a commercial soil testing laboratory (for 2018 to 2020, Farmers' Edge Laboratory, Winnipeg, MB; for 2021 to 2022, AgVise Laboratories, Benson, MN, USA) for initial soil characterization (i.e., texture, soil organic matter, pH) and to determine soil nitrate levels.

Statistical Analysis
Statistical analysis was conducted using GraphPad Prism version 9, GraphPad Software, San Diego, California, and alpha values were set at 0.05.To detect significant N 2 O flux events associated with thawing, median daily N 2 O fluxes from the spring thaw periods were compared to those from the overwinter periods using Mann-Whitney tests (nonparametric t-tests).For the significant thaw events identified, daily N 2 O fluxes were compared between fields using Friedman tests (nonparametric one-way ANOVAs) according to a paired/repeated measures design, followed by Dunn's multiple mean post-test comparison.For cumulative N 2 O emissions during the overwinter and thaw periods, regression analysis was conducted to analyze the relationship to fall soil nitrate levels.

Environmental Conditions
The winters of Year 1 and 4 accumulated the largest FDD based on air temperature, with Year 2 a close second and Year 3 having the lowest FDD (Table 3).The pattern of freezing was different in Year 4 with an earlier onset of freezing despite an overall warmer November at 3.6°C, compared to the coldest November of the study years in Year 1 with 6.8°C, but this was followed by the coldest December in the measurement period with an average of 18°C and the coldest January with 16.1°C compared to other years with > 14.4°C (Table S1 in Supporting Information S1).Year 1 was not as cold during the early part of the NGS, but March recorded the lowest air temperature of the measurement period at 8.6°C compared to Year 4 at 2°C.The warmest April was recorded in Year 2 at 4.6°C and Year 4 at 4.5°C, while Year 1 and 3 had averages below freezing at 0.7°C and 1.5°C, respectively.The winters of Year 1 and 4 followed dry growing seasons with ∼145 mm from May to October (Table S1 in Supporting Information S1).In contrast, Year 2 and 3 growing seasons were much wetter with 228 and 244 mm from May to Oct, respectively.The driest NGS occurred during Years 2 and 4 with ∼40 mm, while the other two years received ∼55 mm during the November to April period.Soil nitrate concentrations were highest (>29.5 mg N kg 1 ) in the fall of Year 4 compared to other years, with the NW and NE fields showing about 2× the levels in the SW and SE fields (Table 2).Lowest soil nitrate levels were Global Biogeochemical Cycles 10.1029/2023GB008051 observed in Year 3 (all fields <15.5 mg N kg 1 ) and in Year 2 (SW and SE fields <20 kg N ha 1 ).In Year 1, all fields had soil nitrate levels between 14.5 and 26 mg N kg 1 .

Daily Nitrous Oxide Fluxes Over the Measurement Period
A total of 2373 individual daily N 2 O fluxes for the four fields were captured over the December 2018 to October 2022 period, resulting in 964 mean fluxes when averaged across fields (Figure 1).Periods when no data were collected tended to be due to equipment removal from the field for planting and harvesting or equipment (e.g., pump) failure.For example, in fall 2019 there were delays associated with setting up irrigation equipment and in fall 2021 the personnel contracted for crop harvest delayed their return to harvest the swathed canola after equipment was removed for the swathing operation.Distinct emission events occurred in April 2020 (Year 2) and March to April 2022 (Year 4) associated with the thaw period and also with fertilizer application in June of the canola year (2021, Year 3) starting on day 150 (Figure 1).Prolonged emission events did not occur in the other .They tended to be on single days rather than multiple days as seen for the emission events and are likely not significant.

Environmental Conditions and Daily Nitrous Oxide Fluxes During Overwinter and Thaw Periods
The onset of freezing soil temperatures was earliest in Years 2 and 4 (Figures 2a and 3a) but overall cumulative FDD based on soil temperature were similar for Years 1, 2 and 4 with 597, 576, and 564 degree-days, respectively, compared to only 332 degree-days in Year 3 (Table 3).This can be seen in Figure 2b, where Year 3 presents a shorter (and much later) period of very low soil temperatures (< 5°C) compared to the other years.Average snow depth (Figures 2a and 3a) was similar between years (12-14 cm on average for the NGS) and lasted from November to March.The temporal pattern varied between years with Year 3 having a deeper snowpack early in the fall season and an earlier melt in the spring (Figure 3a).In Year 2, the opposite occurred with a later onset of a snowpack and a later melt in the spring (Figure 2a).An isothermal period of soil temperature around 0°C at 10 cm depth occurred at the time of snowpack melting.Around or just before the isothermal period of 0°C at 10 cm, the soil water content probes registered an increased signal (Figures 2b and 3b).Note that the sensors used do not provide accurate readings when water in the solid phase is present; however, the decrease in signal as well as the increase serve as indicator that some soil water freezing or thawing is taking place.The isothermal period was followed by a rapid warming above 0°C in Year 1 and 3, that is, after snow depth reached zero on day 78 in Year 1, close to 7.5 days later, the soil temperature had increased to 2°C, while for Year 3, snow had melted on day 73 and the soil temperature was at 2°C 11 days afterward (Figures 2a and 3b).In contrast, for the other 2 years, soil warming was delayed for a longer period after snowpack disappearance, with soil temperature at 10 cm reaching 2°C 22.5 and 13.5 days after complete snow melting in Year 2 and 4, respectively (Figures 2a and 3a).
Averaged across the four fields, N 2 O emissions were small throughout November to mid-March or April and increased significantly during the spring thaw period in Year 2 and 4 (p < 0.001) but not in Year 1 and 3 (p > 0.05) (Figures 2b and 3b).Prolonged emission events took place starting on day of year (DOY) 107 of Year 2 and DOY 83 of Year 4 when flux values exceeded 15 g N ha 1 d 1 , which was consistent with the timing of soil water content increase indicating soil thawing (Figures 2b and 3b) and immediately after complete snow melting (Figures 2a and 3a).The emission event concluded approximately 5 and 14 days later for Year 2 and 4, respectively, when fluxes returned to values <15 g N ha 1 d 1 (Figures 2b and 3b).
Because no clear emission episodes associated with thaw were observed in Year 1 or Year 3 (Figures 2b and 3b), we inspected the CO 2 half-hourly flux data for the time of thawing and verified that emission episodes were registered in these years, confirming that there were no issues with the measurement system (Figure S2 in Supporting Information S1).Increased CO 2 fluxes were consistently observed in all the thawing periods when soil temperature stayed around 0°C for several days.In 2020 and 2022 (Years 2 and 4), half-hourly N 2 O fluxes also consistently increased during this period (Figure S2 in Supporting Information S1).In 2019 and 2021 (Years 1 and 3), a very brief period of increased fluxes was observed on average across all four fields, though not significant enough to register as a sustained emission event for these years (Figure S2 in Supporting Information S1).

Comparison of Daily N 2 O Emissions Between Fields
During the years when N 2 O fluxes increased significantly at spring thaw (Years 2 and 4), significant differences in fluxes between fields were also observed (Figures 4a and 4c).Differences were relatively small in spring 2020 when daily emissions were mostly <40 g N ha 1 d 1 (Figure 4a); nevertheless, greater median fluxes were produced in the NW and NE fields than in the SW field (p < 0.01).Emissions increased when air temperature began to rise above 0°C, soil temperatures at 10 cm depth warmed to or above 0°C and soil water content measurements indicated thawing at 10 cm depth (day 107-112, Figure 4b).Irrigation was applied to the NW and SW fields before freeze-up in fall 2019 (Year 2), and soil water content measurements at 20 cm depth on 6 November 2019 indicated a slight trend of higher values at 0.238 ± 0.015 and 0.241 ± 0.036 m 3 m 3 (mean ± standard deviation) for the NW and SW fields compared to the NE and SE fields at 0.207 ± 0.047 and 0.171 ± 0.022 m 3 m 3 , respectively.However, at thaw, maximum water content values at 10 cm depth were similar for all plots ∼0.35 m 3 m 3 (Figure 4b).Note that the soil water content measurements in the SW field indicated that full thawing at 10 cm occurred about 2 days after the NW and NE fields on day 112, but at that point in time the water content values were the same.Large differences in N 2 O emissions were observed in spring 2022 (Year 4) when the SW and NE fields had several days with emissions >40 g N ha 1 d 1 and as high as ∼100 g N ha 1 d 1 , while the NW and SE fields did not pass 30 g N ha 1 d 1 during the prolonged emission event (Figure 4c).As such, greater median fluxes were observed for the NE and SW fields than for the NW and SE fields (p < 0.001).Increased emissions were observed for close to 25 days in Year 4 when daily air temperature varied frequently between above and below freezing (Figure 4d), which contrasted with Year 2 when the thaw event had a consistent air temperature above 0°C and lasted about 10 days (Figure 4b).During the first thawing period when soil temperature reached above 0°C on day  b) and (d) air temperature (dashed line; values <0°C not displayed), soil temperature (lines with symbols) and water content (thin lines) at 10 cm depth.In fall 2019, the NW and SW fields received irrigation and the NW field received a fertilizer application (see Table 1).Data was unavailable for the SE field in 2020.No specific treatments were applied in fall 2021, but spring fertilizer applications differed in spring 2021 (see Table 1).Values in legend are cumulative N 2 O emissions for the duration of the main thaw period (i.e., 12 days from day 107 to 119 in 2020 and 25 days from day 80 to 105 in 2022).
83-84 in 2022, soil water content at 10 cm depth reached 0.376 ± 0.011, 0.396 ± 0.019, 0.409 ± 0.014, and 0.311 ± 0.027 m 3 m 3 for the NW, SW, NE and SE fields, respectively (Figure 4d).Note that at the start of the freeze up period at the beginning of November 2021, soil water contents at 10 cm and 20 depths averaged across the fields were 0.08 and 0.10 m 3 m 3 (data not shown).
Overall emissions over the 14 days from day of year 107-119 in 2020 (Year 2) were between 109 and 211 g N ha 1 , while for the longer thaw period from day 80-105 in 2022 (Year 4), N 2 O emissions amounted to 262 and 236 g N ha 1 for the NW and SE fields, and were 2 to 3× larger at 557 and 776 g N ha 1 for the SW and NE fields, respectively (Figures 4a and 4c).There were no significant thaw N 2 O emissions in spring 2021   Global Biogeochemical Cycles 10.1029/2023GB008051 indicating the low stubble height treatment applied to the NE and SE fields in Year 3 did not have an impact (data not shown).

Monthly and Annual Variability in N 2 O Emissions
The August to February period was typically characterized by daily monthly averages of 0.5-2 g N ha 1 d 1 (Figure 5).In years when a thaw-related emission episode was observed, daily averages increased several fold to 8 g N ha 1 d 1 for April 2020 and 5.5-14 g N ha 1 d 1 for March and April 2022, respectively (Figure 5).These enhanced emissions due to thaw were of similar magnitude or larger than fertilizer-related monthly averages observed in some of the years, notably in May to July 2019 and 2020, and June 2022 (Figure 5).Overall, cumulative emissions for the Winter (November to Feb, March or April) varied between 6 and 217 g N ha 1 (Table 4) and for the Thaw period (April or March to April) ranged from 0 to 928 g N ha 1 .Note that when a significant thaw period did not occur, we considered all emissions to be part of the Winter period and assigned 0 g N ha 1 to the non-existent Thaw period.In contrast, growing season emissions varied between 206 and 466 g N ha 1 .
In Year 2, between 109 and 211 g N ha 1 of N 2 O loss occurred during the short 12-day period of enhanced emissions (Figure 4a), while in Year 4, between 262 and 776 g N ha 1 of N 2 O were emitted over 25 days (Figure 4c).When compared to total emissions during the Thaw period in each of these years and for each of the fields (i.e., between 190 and 928 g N ha 1 ) the short 12-to 25-day periods contributed 57%-90% of emissions, and compared to total annual emissions (i.e., 573-1,201 g N ha 1 , Table 4) these periods contributed 18%-65% of emissions.The contribution of the Winter period to total annual emissions was between 1% and 35% with an average of 22% and of the Thaw period was between 0% and 77% with an average of 30% over 4 years.Of note is the large variability in the contribution of NGS to annual totals between years with a contribution of close to 70% in years with significant thaw events (Years 2 and 4), while in the years without a significant thaw event the contribution of NGS and GS flipped with 30% of annual emissions occurring during the NGS (Table 3).On average over the 4 years, the contributions of NGS and GS were evenly split at 52:48 (Table 3).The annual totals varied between 484 and 1,201 g N ha 1 .

Drivers of Non-Growing Season N 2 O Emissions
Differences in cumulative soil freezing degree days were not significantly related to NGS N 2 O emissions (P = 0.7) (data not shown).Soil water content at the time of thaw was close to saturation in all years and did not vary significantly between fields.However, soil N levels were higher for the NW and NE fields in the fall of 2019 (Year 2) and 2021 (Year 4).The fall soil nitrate level was a strong explanatory variable for the difference in observed total emission during the Nov-April period with an r 2 of 0.485 (P = 0.037) (Figure 6).

Temporal Dynamics During the Melting Period Impacted Duration of N 2 O Flux Thaw Events
During the 2 years when a prolonged period of soil temperatures around 0°C occurred after complete snow melting (Years 2 and 4), a significant emission event was recorded with increased N 2 O fluxes that lasted between 13.5 and 20 days until approximately the time when soil temperature at 10 cm rose to above 2°C.Peak N 2 O fluxes immediately at thaw, before significant soil warming has occurred, are difficult to observe without continuous measurements but have also been observed in other studies (Brin et al., 2018;Chantigny et al., 2016).At a humid cold site in Ontario, N 2 O flux during the soil water phase change was an exponential function of the soil surface layer temperature (average of surface and 5 cm soil temperature) but decreased once temperature reached values >5°C (Wagner-Riddle et al., 2010).As reviewed by Risk et al. (2013), it has been proposed that N 2 O reductase may be inhibited at low temperatures of 0-5°C, supporting high N 2 O fluxes because the N 2 O production enzymes are less affected by low temperature (Holtan-Hartwig et al., 2002).The lower temperature cut-off observed here could be related to our deeper soil temperature measurement (10 cm), while warmer temperatures were likely occurring at 5 cm.
More rapid soil warming at 10 cm after snowmelt was observed in Years 1 and 3 (<10.5 days) compared to Year 2, which stood out with a prolonged period when soil conditions stayed around 0°C.This isothermal period was also slightly longer for Year 4, but conditions were not substantially different compared to Years 1 and 3, although it is possible that conditions in the top few centimeters were quite different between these years.Studies have shown that conditions in the top few millimeters to centimeters are most related to the timing of N 2 O fluxes (Furon et al., 2008).A higher rate of soil warming as seen predominantly in Years 1 and 3 would affect how quickly the drainage of meltwater could move through the soil profile by removing the infiltration barrier presented by the frozen soil layers more rapidly and reducing the anoxic conditions in the melted surface layer.Even though the conditions during the thaw period could have impacted the observed differences between years, it is likely that other factors were at play, as discussed below.

Differences in Magnitude of N 2 O Flux Thaw Events Were Observed Between Fields
In years when conditions were conducive for N 2 O emissions (Years 2 and 4), the magnitude of peak fluxes and duration of the emission events were significantly different between fields.Several factors interact with soil N for N 2 O production and emission to occur, including soil properties, weather, soil moisture and temperature, plant growth, microbial and enzymatic dynamics-but because of the ephemeral, irregular, and heterogeneous nature of soil freezing and thawing, these interactions are arguably even more complex overwinter and at thaw than at application of N fertilizer in the growing-season (Congreves et al., 2018).As frozen soil thaws, the liquid water content increases, often resulting in anaerobiosis, favoring the production of N 2 O via denitrifier activity (Wagner-Riddle et al., 2008).The large differences in fall soil nitrate levels observed between fields in Years 2 and 4 were likely still present at spring thaw given that the soil temperature was consistently below freezing during the winter, which would have prevented nitrate leaching.The contrasting soil nitrate levels between fields explained the response in cumulative N 2 O emissions observed between fields as was also observed in other studies (Wagner-Riddle & Thurtell, 1998).The contrasting N fertilizer applications to canola in spring 2021, coupled with very poor growing conditions that followed due to lack of rainfall throughout the summer, directly impacted residual nitrate levels going into the fall and the subsequent differences in emissions between fields at thaw.The largest NGS emission recorded in this study (∼1 kg N ha 1 ) was for the NE field, which had 70.5 kg N ha 1 of fall residual soil nitrate (Figure 6).It is not clear why the NW field with 56.5 kg N ha 1 of residual soil nitrate had relatively low NGS emissions, but it could be that other important factors with high spatial variability such as soil water content had a role and were not captured with our measurements.
The pulse of N 2 O emission at thaw has been hypothesized to be fueled by a freezing-induced increase in substrate originating from aggregate disruption and microbial biomass (Congreves et al., 2017), though a recent study ruled out aggregate disruption as a significant driver (King et al., 2021).However, at our study site, freezing per se did not seem to be sufficient to induce large emission events, except for conditions when nitrate levels were quite high, that is, at >25 mg N kg 1 and up to 70 mg N kg 1 going into the fall (Figure 6).This is a significant finding since it implies that minimizing residual soil N levels in the fall through spring soil testing and better matching of fertilizer application rates to crop N uptake could be a tool to manage spring thaw N 2 O emissions in this region.
Global Biogeochemical Cycles  et al., 2007, 2017).Our results support the notion that NGS N 2 O emissions are heavily influenced by thawing (i.e., the significant thaw-induced emission events in Years 2 and 4 contributed to greater NGS N 2 O emissions), but we add that thawing dynamics is key in that more prolonged and gradual thaws were associated with greater emissions than abrupt and sharp thaws.Research should explore how different soil thawing dynamics including melting, vaporization, and sublimation influence the N 2 O production pathways to better predict overwinter and thaw N 2 O fluxes, especially for dry regions where soil moisture at thawing plays a large role.Another explanation for the relatively low NGS fluxes at our site is soil texture and how it interacts with the freezing intensity to induce N 2 O production.Soil texture (and likely the associated organic carbon levels) has been shown to have an impact on freeze-thaw emissions with freezing incubation treatments inducing N 2 O emissions only in a silt loam (38.0%sand, 54.5% silt, 3.1% organic carbon) but not in a loamy sand (79.2% sand, 17.5% silt, 0.5% organic carbon) (King et al., 2021).The soil at our study site was a sandy loam (52% sand, 32% silt, with high organic carbon of 2.6%), partially explaining the overall lack of significant thaw emissions.
Although overwinter and thaw N 2 O fluxes were relatively small at our site, they amounted to an important source of annual total N 2 O emissions (Table 4).This is important new information because if overwinter and thaw emissions are not accounted for in semi-arid croplands subjected to soil freezing, then GHG budgets could be underestimated.Yang et al. (2023) predicted that prairie provinces have lower N 2 O emissions per hectare than provinces in central, Atlantic, and pacific Canada; however, their estimates do not include NGS N 2 O emissions.
Because the Prairie region contains 80% of Canada's agricultural land and has experienced significant increases in N fertilizer applications (i.e., 2.1 times greater in 2016 than 1981), it is crucial that agricultural N 2 O inventories consider these NGS emissions for more accurate reporting and the development of mitigation strategies.

Conclusions
Prior to this study, the most extensive N 2 O flux dataset for the North American Prairie region was based on yearround micrometeorological measurements conducted at the very Eastern edge of Manitoba Canada.Our results fill a gap in year-round measurements of N 2 O fluxes for the North Central region representative of drier conditions and located in the middle of the Canadian Prairies-a broad region that includes 80% of Canada's agricultural land.Northern Prairie croplands cover a very extensive area in North America and are a major source of grains and oilseeds for national consumption and export.Areas with similar climate and cropland production also exist in Europe and Asia.Our findings stress the importance of measurement in the non-growing season for more accurate N 2 O budgets in cold climate regions.We show that for a semi-arid cold climate site, winter and thaw N 2 O fluxes are limited (on average 345 g N ha 1 ) compared to humid sites but amount to a significant proportion of total annual N 2 O emissions (52% on average

Figure 1 .
Figure 1.Daily nitrous oxide flux was measured in the four study years.March and April are highlighted as the main periods for soil thawing.Note that flux measurements started in December 2018 but are not shown here.

Figure 2 .
Figure 2. (a) Half-hourly soil temperature at 10 cm depth (solid line) and daily snow depth (line with symbols), and (b) daily volumetric soil water content (VWC) at 10 cm depth (solid line) and daily N 2 O flux (line with symbols) at the experimental site from November 1 (day 305) to April 30 (day 121) for Year 1 (fall 2018/spring 2019, black lines and symbols) and Year 2 (fall 2019/spring 2020, gray lines and symbols).Significant increases in median N 2 O flux from overwinter to thaw were detected in Year 2 (Mann-Whitney p < 0.001) but not Year 1 (Mann-Whitney p > 0.05).

Figure 3 .
Figure 3. (a) Half-hourly soil temperature at 10 cm depth (solid line) and daily snow depth (line with symbols), and (b) daily volumetric soil water content (VWC) at 10 cm depth (solid line) and daily N 2 O flux (line with symbols) at the experimental site from November 1 (day 305) to April 30 (day 121) for Year 3 (fall 2020/spring 2021, black lines and symbols) and Year 4 (fall 2020/spring 2021, gray lines and symbols).Significant increases in median N 2 O flux from overwinter to thaw were detected in Year 4 (Mann-Whitney p < 0.001) but not Year 3 (Mann-Whitney p > 0.05).Note that soil water content data shown for Year 3 are estimated values based on measurements at the adjacent weather station.

Figure 4 .
Figure 4. Main thaw events for (a) and (b) spring 2020 (Year 2); (c) and (d) spring 2022 (Year 4) by field.Note the different X-and Y-axis for each year.(a) and (c) show daily N 2 O fluxes which significantly differed by field (Friedman p < 0.01, Dunn's comparison NW = NE > SW in 2020; and Friedman p < 0.001, Dunn's comparison NE = SW ≥ NW = SE in 2022), (b) and (d) air temperature (dashed line; values <0°C not displayed), soil temperature (lines with symbols) and water content (thin lines) at 10 cm depth.In fall 2019, the NW and SW fields received irrigation and the NW field received a fertilizer application (see Table1).Data was unavailable for the SE field in 2020.No specific treatments were applied in fall 2021, but spring fertilizer applications differed in spring 2021 (see Table1).Values in legend are cumulative N 2 O emissions for the duration of the main thaw period (i.e., 12 days from day 107 to 119 in 2020 and 25 days from day 80 to 105 in 2022).

Figure 5 .
Figure 5. Mean daily N 2 O fluxes for each measurement month across all for fields from December 2018 to September 2022.Vertical lines delineate the non-growing season (NGS, November to April) and growing season (May to October) with crops grown each year indicated.Error bars show the standard error of the mean.
For Year 1 and Year 3, averages across fields are shown.For Year 2 and Year 4, values for individual fields are shown for the NGS while averages are shown for the GS.Note that data for the SE field was not available for Year 2 due to measurement issues.The bolded values labelled 4 years average are the average of values shown for each year for Winter, Thaw and GS emissions (for Years 2 and 4 all fields were averaged to represent each year's average).Annual cumulative emissions shown for 4 years are the sum of GS and NGS emissions and percent values are derived from the total for the 4 years value.a November to April for years without a thaw event (Year 1 and Year 3), November to March for Year 2 and November to February for Year 4, representing 50%, 42%, and 33% of the year's duration.b Shown only for years with thaw events: April for Year 2 (DOY 92-121), March-April for Year 4 (DOY 60-120), representing 8% and 17% of the year's duration.c Value for all fields combined.To calculate GS totals: gaps in data for August 2019 were filled with the average of monthly flux measured in August 2020, 2021, and 2022; gaps in data for September 2019 were filled with the average of monthly flux measured in September 2021 and 2022; gaps in data for 2 October 2019 were filled with the average of monthly flux measured in October 2020 and 2021; gaps in data for August 2021 were filled with the average of monthly flux measured in August 2020 and 2022.

Figure 6 .
Figure 6.Relationship of fall soil nitrate concentration (0-15 cm depth) and cumulative N 2 O emissions for November to April over the 2018 to 2022 study period.Labels on each data point indicate the field (e.g., SW denotes the southwest field) and spring year (e.g., "20" shows the spring of the 2019/2020 years) with "AV" used to identify when the average of all fields is displayed.The dashed line shows a linear regression between the soil nitrate levels and the N 2 O emissions where Y = 8.23 X + 189 with n = 9 and r 2 = 0.485 (P = 0.037).

Table 1
Management Applied to the Four Fields Each Year in Spring or Fall Note.For cases when fertilizer was applied in fall, values indicate soil nitrate levels pre-fertilization and values in brackets represent the soil levels plus the fertilizer amount as indicated in Table1.Samples were taken on 11 September 2018; 22 October 2019; 3 September 2020 and 25 October 2021.

Table 3
Cumulative Freezing Degree-Days (FDD) Calculated Using Air Temperature and Soil Temperature at 10 cm Depth <0°C for Each Non-Growing Season From November 1 to April 30 for the Four Study Years (Year 1 Fall 2018/Spring 2019, Year 2 Fall 2019/Spring 2020, Year 3 Fall 2020/Spring 2021 and Year 4 Fall 2020/Spring 2021)

Table 4
NGS is divided into Winter and Thaw periods as per footnote.

Freeze-Thaw N 2 O Emissions Were Small but a Significant Proportion of Annual Totals Measured
Yang et al. (2022)23)) 2017)f 123-938 g N ha1for NGSs with cumulative FDD from 332 to 597 degree-days measured at the Saskatoon site (Table4) were much smaller than values obtained at other cold climate sites subjected to similarly high cumulative FDD, which tended to be >1 kg N ha 1(Wagner-Riddle et al., 2017).By coincidence, the average NGS N 2 O emissions over 4 years in our study (345 g N ha 1 ) is the same value derived for several Prairie sites except one site in eastern Manitoba (Glenlea)(Pelster et al., 2023).For our study, the NGS emissions corresponded to slightly over half (52%) of the annual average emissions of 669 g N ha 1 with 22% of emissions occurring during the Winter period (November to February or Mar) and 30% of emissions at thaw during March to April, while for thePelster et al. (2023)study, the NGS emissions for the Prairie sites corresponded to a lower value of 34.5%.Overall,Pelster et al. (2023)found a ratio of 35.5% for NGS to 64.5% for NGS emissions across Canada and proposed an adjustment factor of 1.55 for emission factors derived solely from GS emission measurements.Based on our average proportion of 52:48 for NGS: GS emissions, this adjustment factor would be equal to 2. However, we caution that such adjustments should be derived from wide-spread experiments across several soil types and climate regions, as conducted byPelster et al. (2023).Based on patterns of simulated NGS N 2 O emissions across US Midwest regions,Yang et al. (2022)proposed that NGS N 2 O emissions are dominantly driven by moisture dynamics at thawing in drier regions and by the intensity of freezing in wetter regions-with 300 mm of precipitation per year as the tipping point between the two scenarios.Our site received less than 300 mm of precipitation per year throughout the study period (TableS2in Supporting Information S1) and experiences a semi-arid climate, in contrast to the other Canadian micrometeorological sites in humid regions where NGS emissions are correlated with cumulative FDD (Wagner-Riddle ).Furthermore, soil nitrate levels are strong positive drivers of NGS N 2 O emissions; thus, agronomic management practices that lower fall soil nitrate levels are recommended a promising strategy to lower NGS emissions.Whereas other micrometeorological research in cold climate but humid sites have demonstrated that soil freezing intensity explains NGS N 2 O emissions (where greater freezing leads to greater thaw N 2 O emissions), this was not the case for our site.Instead, we propose that fall nitrate, temporal dynamics of soil thawing, and soil moisture at thaw are more dominant factors controlling N 2 O production than freezing intensity for cold semi-arid sites.Further experimentation into thaw dynamics throughout freeze-thaw cycles is needed to understand how melting, vaporization, and sublimation processes may interact with N 2 O production pathways and the magnitude of emissions.Regardless, accounting for NGS sources of N 2 O emission in these cold semi-arid sites where GS emissions are generally small is clearly important for more accurate inventory reporting and for developing strategies to mitigate emissions.The use of continuous micrometeorological N 2 O measurements as provided here and expanding their use to include a broader representation of soil types, regions, and cropping systems is needed to improve model predictions, enabling a more complete accounting of N 2 O in national and global GHG budgets. as