Winter wheat responses to enhanced efficiency liquid nitrogen fertilizers in the Canadian Prairies

Abstract To evaluate how enhanced efficiency liquid nitrogen (N) fertilizers affect winter wheat (Triticum aestivum L.) production under irrigated and rain-fed environments, experiments were conducted at two irrigated and five rain-fed sites across the Canadian Prairies from 2013 to 2018 (22 site-years). The N fertilizers included urea ammonium nitrate (UAN) treated with (i) urease inhibitor N-(n-butyl) thiophosphoric triamide (NBPT), (ii) NBPT plus nitrification inhibitor dicyandiamide, and (iii) nitrification inhibitor nitrapyrin (Nitrapyrin), as well as untreated UAN and urea, and polymer-coated urea (PCU). All fertilizers were applied by banding 50% at planting and 50% in-crop in early-spring, except PCU, where PCU was applied at planting and urea was applied in early-spring. Nitrous oxide (N2O) emissions and methane (CH4) uptake were measured at one rain-fed site from 2014 to 2017. NBPT increased grain yield by 1.2%–14% and 2.8%–4% under irrigated and rain-fed environments, respectively, relative to all the other N sources except untreated urea in the rain-fed environment. Total N uptake with NBPT was between 0% and 12% higher than the other N sources across irrigated and rain-fed environments. The results suggested that both grain yield and N use efficiency were optimized when UAN contained a urease inhibitor. All liquid enhanced efficiency fertilizers produced grain protein content greater than 11%, except Nitrapyrin under irrigated environments. Data from three site-years indicated that greenhouse gas emissions were unaffected by N source under rain-fed conditions. Liquid UAN with a urease inhibitor may have the most potential to optimize winter wheat production and N use efficiency in the Canadian Prairies.


Introduction
With climate change expected to increase the frequency of extreme weather events in the Canadian Prairies (Bonsal et al. 2013), it is important to develop nitrogen (N) management options that help to minimize the risk of N loss to the environment under irrigated and rain-fed growing environments (Beres et al. 2018). Liquid N formulations are attractive fertilizers for winter wheat (Triticum aestivum L.) since they can provide uniform application and may be readily assimilated (Fageria et al. 2009). However, they are susceptible to ammonia (NH 3 ) volatilization losses in the presence of humid conditions, warm temperature, alkaline pH, and soil high in carbonates (Sommer et al. 2004). Research has shown that when an N fertilizer strategy mismatches the timing of N supply with crop N demand, not only crop yield declines but nitrous oxide (N 2 O) emissions may increase (Kirkegaard et al. 2014). Even though much of the northern Great Plains are located in a semi-arid climate, the region is still subjected to significant N loss risk. Chantigny et al. (2019) observed that N applied in fall had high vulnerability to loss even in cold and frozen soil.
Their research indicated that 60%-75% of fall broadcasted ammonium N (NH 4 + -N) was lost overwinter in the Lethbridge area via nitrate (NO 3 -N) leaching and denitrification before spring crops were seeded. An ideal winter wheat N management strategy would attain high agronomic productivity and limit environmental consequences associated with applying inorganic N fertilizer. Split-application of N is one important N management strategy to maintain winter wheat grain yield and protein levels and subsequently minimize N losses (Huerfano et al. 2015;Beres et al. 2018). In addition to splitapplication, enhanced efficiency fertilizers (EEFs) may also help improve synchrony between N supply and winter wheat requirements to increase recovery of fertilizer N and limit N losses to the environment.
EEFs are designed to slow down the release of plantavailable N or include additives in their formulation that temporarily reduce urease activity or nitrification rates in soil (AAPECO 2013). These mechanisms may extend the opportunity for crops to utilize N fertilizer and therefore optimize winter wheat grain yields and protein levels. Previous works suggested that EEFs could be coupled with split-application to improve winter wheat agronomic performance. For example, split-applying urea+urease inhibitor N-(n-butyl) thiophosphoric triamide (NBPT) or urea+NBPT and nitrification inhibitor dicyandiamide (DCD) 50% side-banded at planting and 50% broadcast in-crop in spring optimized winter wheat grain yields and protein content relative to urea in the Canadian Prairies (Beres et al. 2018). Research conducted in the Prairies and in Ontario reported that eNtrench and N-Serve (nitrapyrin is the active ingredient of both nitrification inhibitors) were highly effective tools for stabilizing N in urea, urea ammonium nitrate (UAN), and NH 3 to potentially reduce N losses (Degenhardt et al. 2016). The combination of urease and nitrification inhibitors in UAN+Agrotain Plus (containing both NBPT and DCD) reduced N 2 O emissions by 61% compared with urea and 41% compared with UAN in irrigated corn (Zea mays L.) production systems in the Central Great Plains (Halvorson et al. 2014). Other research noted that NBPT alone was efficient in reducing NH 3 volatilization in the Canadian Prairies and performed better in sandy loam soils than in clay loam soils by increasing grain yield, grain N removal, and crop N uptake relative to untreated urea and UAN (Lasisi et al. 2021). Meta-analysis indicated that the benefits of EEFs were more pronounced in irrigated than rain-fed systems (Thapa et al. 2016), and nitrification inhibition effectively enhanced N use efficiency in irrigated compared to rain-fed systems (Li et al. 2018).
Given the recent focus on reducing greenhouse gas emissions, it has become a priority to better understand how N fertilization strategies impact N 2 O and methane (CH 4 ) emissions. While it is well-established that surplus N is subject to loss via N 2 O emissions, excessive N supply may also reduce CH 4 uptake in aerobic soils (Acton and Baggs 2011). Methane uptake can provide some offset for N 2 O emissions from N application, but the extent to which they are simultaneously impacted by EEFs under different soil moisture regimes is uncertain. It is also prudent to consider tradeoffs by scaling N 2 O and CH 4 emissions in carbon dioxide equivalents (CO 2 -eq), per unit of yield or per unit N uptake (Mosier et al. 2006;Van Groenigen et al. 2010). These indices provide a way to aggregate and compare the net impact of different N management decisions that may alter both agronomic performance and N 2 O and CH 4 emissions. This research compared agronomic performance as influenced by liquid UAN fertilizers containing a urease inhibitor, a nitrification inhibitor, or both, with those of untreated UAN and urea, and polymer-coated urea (PCU) in a split-application regime. The objectives of this study were to (i) determine how N source influences grain yield, protein content, and other agronomic performance variables across a range of soil moisture conditions and two moisture regimes; and (ii) determine if N sources integrated with application methods cause differential N 2 O and CH 4 emissions.

Materials and methods
Site description and experimental design From 2013 to 2018, experiments were conducted at locations representing major agroecosystems across the Cana-dian Prairies with a new site established at each location every year (Table 1). Annual precipitation at the sites ranged from 224.2 to 506.4 mm ( Table 2). Monthly precipitation was highest in the spring and summer. Mean annual air temperature ranged from 1.9 to 7.6 • C (Table 2), with monthly average air temperature ranging from −20.5 to 21.9 • C (Fig. 1). "Lethbridge Irr." and "Farming Smarter" were irrigated sites located in Lethbridge, AB, which received 100-230 mm of irrigation water in addition to growing season precipitation-all other sites were rain-fed. Edmonton, St. Albert, and Falher, AB and Brandon, MB were situated on well-drained to moderately well-drained Black Chernozem clay loam soils (Udic Boroll) with 3%-5% organic matter content and a pH range of 7.7-8.2. The "Lethbridge Irr.", "Lethbridge Dry", and "Farming Smarter" sites have Dark Brown Chernozems clay loam soil (Typic Boroll) with 2.7%-3.1% organic matter content and pH of 7-8. Soils at all sites are calcareous ( Table 1). The "Lethbridge Irr.", "Lethbridge Dry", "Farming Smarter", Brandon, and Falher sites were previously planted with canola (Brassica napus L.). Barley (Hordeum vulgare L.) was used as a preceding crop in Edmonton and St. Albert. "AC Flourish", a Canada Western Red Winter milling quality cultivar (Graf et al. 2012), was grown in all experiments. A randomized complete block design with four replicates was employed at all sites. Experimental units (plot sizes) varied by site and were selected based on the equipment used for the experiments (Table 1). Soil nutrient availability was measured at each site before planting using Plant Root Simulator probes (Western Ag Labs, Saskatoon, SK, Canada). The N rates were based on 80% of the recommended rates, which ranged from 62 to 182 kg N ha −1 for rain-fed sites, and from 73 to 196 kg N ha −1 for the irrigated sites.

Nitrogen fertilizer treatments
Six different N sources were tested: (1) untreated granular urea (46-0-0), (2) untreated liquid UAN (28-0-0), (3) NBPT-treated UAN (NBPT; Agrotain Ultra ; Koch Agronomic Services, St. Louis, MO, USA), (4) nitrification inhibitor nitrapyrin-treated UAN (Nitrapyrin; eNtrench ; Corteva Agriscience Canada, Calgary, AB, Canada), (5) NBPT plus DCDtreated UAN (NBPT + DCD; Agrotain Plus ; Koch Agronomic Services, St. Louis, MO, USA), and (6) PCU (Environmentally Smart N, ESN ; Nutrien, Saskatoon, SK, Canada). All N sources were split-applied with 50% of N side-or midrowbanded at planting and 50% applied in-crop in early-spring (Feekes 4) except PCU where PCU was side-banded at planting and untreated urea was broadcast early-spring. Liquid N application was applied with a custom-made 38 cm singlearm sprayer equipped with SJ3 Teejet liquid fertilizer nozzles (Teejet Technologies, Wheaton, IL, USA) or John Deere stream nozzles (John Deere Canada ULC, Grimsby, ON, Canada). The experiments also included a 0 N control, i.e., phosphorus, potassium, and sulfur fertilizers applied at the recommended rates based on soil test results, for N uptake, apparent crop recovery efficiency (RE) and agronomic efficiency (AE) calculations. Winter wheat was seeded with a ConservaPak™ air drill (John Deere Canada ULC, Grimsby, ON, Canada) configured with knife openers spaced 23 or 30 cm apart, a Seedmas-ter™ plot drill (Seedmaster, Emerald Park, SK, Canada) with   Note: "--" represents missing data. * Lethbridge Irr. and Farming Smarter were irrigated, and all the other sites were rain-fed.  Herbicides were applied at each site 24-48 h prior to seeding and in-crop when the average weed growth was the 3-5 leaf stages around mid-October (averaged 32 days after planting) or in the spring (averaged 241 days after planting). Herbicides were selected based on weed species and abundance, and applied at full label rates.

Yield and agronomic variables
Early-season vigor (ESV) was assessed at Feekes 6 and lateseason vigor (LSV) was assessed at Feekes 8. Vigor was rated on a scale of 1-5 with 1 representing poor vigor with light color and 5 indicating vigorous growth with bright green color. Chlorophyll index was determined weekly using a Field Scout CM1000 chlorophyll meter (Spectrum Technologies, Inc., Aurora, IL, USA) at growth stages from Feekes 4 to Feekes 9 following the procedures described in Beres et al. (2018). Heads were counted near physiological maturity from the two 1 m rows marked prior to tillering for plant density in each plot. The number of heads was divided by the number of plants in the same area to calculate the number of heads per plant. Plots were harvested with a plot combine equipped with a straight-cut header, pickup reel, and crop lifter. Grain yield was calculated and corrected to 14% moisture. A 2 kg subsample was retained to measure seed mass (g 1000 kernels −1 ) and grain bulk density (kg hL −1 ). Whole grain protein content was determined from the same subsample using nearinfrared reflectance spectroscopy technology (Foss Decater GrainSpec, Foss Food Technology Inc., Eden Prairie, MN, USA).
Three adjacent 0.5 m row subsections were harvested in each plot near physiological maturity. Winter wheat plants were dried at 60 • C and used to determine biomass. A subsample of straw was retained, ground with a #3 Wiley Mill to pass through a 1 mm screen and analyzed for N content using the Kjeldahl procedure (AACC International 2018). Straw N uptake (kg ha −1 ) was determined by multiplying the straw N content (%) by the straw biomass (kg ha −1 , corrected to 0% relative humidity). Grain N uptake (kg ha −1 ) was determined by multiplying the grain N content (%) by the grain yield (kg ha −1 , corrected to 0% relative humidity). Total N uptake (kg ha −1 ) represents the N in the aboveground biomass and is the sum of straw N uptake and grain N uptake. Apparent crop RE is the percentage of N uptake relative to N applied and was calculated using the modified equation based on that reported in Dobermann (2007): where NF (kg N ha −1 ) is the total N uptake with N fertilization, NC (kg N ha −1 ) is the total N uptake in the 0-N control, and NR is the applied N rate (kg N ha −1 ). AE represents the yield increase (kg) per kg N fertilizer applied and was calculated following the equation in Dobermann (2007): where Y is grain yield (kg ha −1 ) with N fertilization, Y 0 is the grain yield (kg ha −1 ) in the 0-N control, and NR is the applied N rate (kg N ha −1 ).

Greenhouse gas measurements
Soil-to-atmosphere N 2 O and CH 4 emissions were measured from September (planting) to July (harvest) for each treatment at the "Lethbridge Dry" site (Lethbridge, AB, Canada) from 2014 to 2017. Gas samples were collected from the same field as the yield and agronomic factors, on three of the four replicates; the chamber collars were installed on the same day as planting, pulled out during the late-fall and early-spring fertilizer applications, and reinstalled immediately afterwards. This was considered a suitable compromise between optimal sample replication and the time and resource intensity of greenhouse gas collection and analysis. All gas samples were collected between 9:00 am and 12:00 pm local time using vented static chambers with a PVC base collar (30 cm inner diameter, 10 cm height, inserted to 5 cm soil depth, approximate volume 7 L, and one chamber per plot) (Chang et al. 1998). Gas samples (11.3 mL) were drawn from the chamber headspace using a gas-tight syringe at 0, 15, 30, and 60 min after securing the lid to the collar and immediately transferred to a pre-evacuated (−1 atm) vial (5.8 mL) (Labco, Lampeter, Wales, UK). Gas concentrations were determined with a gas chromatograph equipped with electron capture and flame ionization detectors (Varian 3800, Varian Instruments, Palo Alto, CA, USA). The columns used for the N 2 O channel are Porapak N 80/100 mesh (0.50 m × 3.2 mm × 2.2 mm) followed by an Alltech Hayesep D 80/100 mesh (1.83 m × 3.2 mm × 2.2 mm) (Alltech Associates, Inc., Deerfield, IL, USA). The electron capture detector temperature was set at 350 • C. The columns used for the CH 4 channels are Porapak N 80/100 mesh (0.91 m × 3.2 mm × 2.2 mm) followed by Porapak QS 80/100 mesh (1.83 m × 3.2 mm × 2.2 mm). The flame ionization detector was set at 230 • C. The injector and column oven temperatures were kept at 55 • C. The carrier was P10 gas (10% methane, balance argon) for N 2 O and helium for CH 4 . The channel was maintained at a static pressure of 150 kPa.
Gas samples were collected every 1-2 weeks during the fall, spring, and summer, and approximately once a month during the winter. All greenhouse gas fluxes were calculated using changes in gas concentrations over time as determined by the slope of a quadratic or linear regression, the ideal gas law, air temperature, chamber area, and chamber volume (Venterea et al. 2020). The linear regression was used to calculate the flux when the second derivative of the quadratic regression was negative. Fluxes were integrated between days using linear interpolation between sampling events and summed to estimate cumulative emissions. The cumulative emissions ranged from planting to the last greenhouse gas measurement, and therefore extended over 283 days in 2014-2015, 289 days in 2015-2016, and 281 days in 2016-2017. The net N 2 O and CH 4 emissions expressed as carbon dioxide equivalents (kg CO 2 -eq ha −1 ) were calculated by multiplying the cumulative N 2 O and CH 4 emissions by their respective global warming potential relative to CO 2 over a 100-year time horizon (Mosier et al. 2005). The global warming potential values used were 298 for N 2 O and 25 for CH 4 (Solomon et al. 2007). The net CO 2 -eq emissions represent the sum of CO 2 -eq N 2 O emissions and CO 2 -eq CH 4 uptake for each treatment. Yield-scaled CO 2 -eq emissions were calculated by dividing net CO 2 -eq emissions by the corresponding grain yield, grain N, and total N uptake for each treatment (Van Groenigen et al. 2010). For these indices, a lower value indicates lower net N 2 O and CH 4 emissions as CO 2 -eq per unit yield, grain N, or total N uptake, and a reduction in these values can be indicative of lower net N 2 O and CH 4 emissions, higher yield and related components, or a combination of both. The grain yield, grain N, and total N uptake values used for the yield-scaled CO 2 -eq emissions calculations differed slightly from those reported in the yield section because these calculations were only done at one site and only measured in three of the four replicates.

Yield and agronomic variables
Data analyses were performed using SAS 9.4 software (SAS Institute, Cary, NC, USA). The data from irrigated and rainfed sites were analyzed separately. Homogeneity of error variance was tested using the PROC UNIVARIATE procedure and any outlier observations were removed. Data were analyzed using a mixed model with the MIXED procedure. Nitrogen source was treated as a fixed effect, and site (combinations of year and site) and site by treatment (combinations of site and N source) were treated as random effects. Year differences were not explored. Pearson correlation coefficients were calculated using the CORR procedure of SAS to determine the relationship between grain yield and chlorophyll index. Effects were considered significant if P < 0.05. Mean separation tests were performed using Fisher's least significant difference (LSD).
A grouping methodology was used to explore system responses and variability of winter wheat grain yield (Francis and Kannenberg 1978). The means and coefficient of variation (CV) were estimated for each treatment combination across sites and replicates. Means were plotted against CV and used to categorize the biplot data into four quadrants, which included high mean and low variability (Group I), high mean and high variability (Group II), low mean and high variability (Group III), and low mean and low variability (Group IV). Mean-CV biplot provide a relatively simple and general overview of system stability, and supplement results for grain yield, the most important response variable.
A multivariate stability analysis was performed using the additive main effects and multiplicative interaction (AMMI) model to explain the yield variations driven by site-years (the equivalent of environment in the AMMI model). The AMMI model combines the analysis of variance (ANOVA) for genotype and environment with principal component analysis of genotype × environment interactions to characterize the interaction effects (interaction principal components (IPCs)) (Dia et al. 2016). The first stage of the analysis used PROC MEANS to average yield data. The second stage involved a PROC GLM analysis with averaged yield. Both N source (the equivalent of genotype in the AMMI model) and site-year were included in these stages. Residuals were outputted from PROC GLM analysis, and PROC IML converted residuals data to a matrix of residuals, and then subjected the residual matrix to a singular value decomposition. The final steps of IML code estimated the proportion of variability for each AMMI term divided by the total N source × site-year interaction variability. The IPC scores and means derived from preceding AMMI analysis were plotted using PROC TEMPLATE and PROC SGRENDER to generate an AMMI biplot. An IPC score in the AMMI biplot close to zero suggests that N source and siteyear contribute little to the interaction effect, that is, they are stable (Carbonell et al. 2004).
The CO 2 -eq emissions and yield-scaled CO 2 -eq emissions were analyzed using a one-factor mixed model ANOVA with N source as a fixed effect and year as a random effect. These statistics were computed by R Studio Team (2020). Mixed-model ANOVAs were performed using the lme command from the nlme package and the lmer Test library (Bates et al. 2012). Mean separation tests were performed using LSD with the lsmeans command from the emmeans package (Lenth 2022) with a significance level of 0.05. The 0-N control was excluded from the analysis of both agronomic data and CO 2 -eq emissions data. This meant that the analysis could consider a factorial treatment design for N source.

Results and discussion
Responses of grain yield and protein content Under irrigated environments, NBPT optimized grain yield that was similar to untreated UAN and NBPT + DCD, but 6.4%, 8%, and 14% higher than untreated urea, Nitrapyrin, and PCU, respectively (Table 3). Superior total N uptake was observed with NBPT and UAN, which was similar to NBPT + DCD and Nitrapyrin, but 9.6% and 11.8%, respectively, greater than untreated urea and PCU. RE and AE responded similarly to N source, with NBPT having the highest, PCU the lowest, and all other N sources providing intermediate responses. The RE of NBPT was greater than that of Nitrapyrin, PCU, and urea, and the AE of NBPT was greater than that of Nitrapyrin and PCU.
Under rain-fed environments, grain yield increased in the order of Nitrapyrin ≤ PCU ≤ NBPT + DCD ≤ UAN < NBPT < untreated urea (Table 4). Total N uptake was maximized with NBPT, which was 6.2%, 7.1%, and 10.1% higher than untreated UAN, PCU, and Nitrapyrin, respectively, but similar to untreated urea and NBPT + DCD. RE did not differ among N sources, but there was a trend of marginal significance for N source effects (P = 0.068) as treatment responses followed trends and a rank order similar to N uptake. AE was highest with NBPT and urea, which was 31.1% higher than Nitrapyrin; no AE difference was detected among NBPT, NBPT + DCD, PCU, UAN, and urea.
The results from both irrigated and rain-fed environments suggested that combining UAN with a urease inhibitor has more potential to enhance winter wheat grain yield and agronomic performance than with a nitrification inhibitor in liquid formulations used in this study. The positive impact of a urease inhibitor on winter wheat agronomic performance can be attributed to it slowing urea hydrolysis. For untreated urea, it is rapidly hydrolyzed by urease after being applied to the soil, causing soil pH to increase in the surrounding area resulting in NH 3 volatilization (Cantarella et al. 2018). By inhibiting urease activity, it also likely indirectly restricted NH 3 volatilization associated with the high pH caused by urea hydrolysis and offset the potential for NH 3 losses, thereby making more fertilizer N available for crop uptake to support grain yields. The lack of a yield response for NBPT + DCD compared to NBPT in this study may be related to DCD increasing the amount of time NH 4 + remains in soil with lower soil moisture, which may have partly offset the reductions of NH 3 volatilization achieved with the urease inhibitor (Cantarella et al. 2018). Beres et al. (2018) observed that a split-application of NBPT + DCD treated urea usually maximized winter wheat grain yield. The difference in the Note: Different letters within columns after the means indicate significant differences between N sources; NS represents not significant; SE denotes standard error; * and * * denote significance at P < 0.05 and P < 0.01, respectively; and "--" represents missing P-values due to covariance estimate being 0. † RE = recovery efficiency. ‡ AE = agronomic efficiency. § PCU banded at planting and urea applied in early-spring in this treatment. Note: Different letters within columns after the means indicate significant differences between N sources; NS represents not significant; SE denotes standard error; * and * * denote significance at P < 0.05 and P < 0.01, respectively; and "-" represents missing P-values due to covariance estimate being 0. † RE = recovery efficiency. ‡ AE = agronomic efficiency. § PCU banded at planting and urea applied in early-spring in this treatment.
performance of dual inhibitors combined with granular versus liquid N sources could be attributed to the liquid N sources being more susceptible to volatilization losses with the nitrification inhibitor resulting in more NH 3 volatilization when mixed with liquid N fertilizers (Sommer et al. 2004;Soares et al. 2012). Therefore, to some extent, the benefit of liquid EEFs is limited in dry rain-fed regions, which is consistent with a previous meta-analysis (Thapa et al. 2016). This does raise questions about the potential of liquid EEFs to substantially improve winter wheat agronomic performance under rain-fed environments in the Canadian Prairies. Ultimately, moisture limitation may be more impor-tant than N availability under rain-fed environments in this region. An example of the importance of water as a driver of crop performance and yield is illustrated in the biplot in Fig. 2 where the grain yields of winter wheat under irrigated environments were greater than those under rain-fed environments. Under irrigated environments, NBPT, NBPT + DCD, and untreated UAN appeared in Group I, which indicates above average grain yield and yield stability. Conversely, PCU produced lower than average grain yield and yield stability (Group III). Compared to the results under irrigated environments, NBPT, NBPT + DCD, Nitrapyrin, and PCU Fig. 2. Biplot summarizing the effects of N source on mean yield of winter wheat, compared with its respective coefficient of variation (CV). The horizontal and vertical lines represent the average of grain yield and CV, respectively. UAN represents urea ammonium nitrate; NBPT, NBPT + DCD, PCU, and Nitrapyrin represent urease inhibitor NBPT-treated UAN, urease inhibitor NBPT plus nitrification inhibitor DCD-treated UAN, the treatment of PCU banded at planting and urea applied in early-spring, and nitrification inhibitor nitrapyrin-treated UAN, respectively. were re-positioned in the biplot under rain-fed environments. NBPT and NBPT + DCD appeared in Group IV (low mean yield and high yield stability) under rain-fed environments, suggesting that winter wheat benefited less from NBPT and NBPT + DCD under rain-fed environments relative to under irrigated environments. PCU was re-positioned in Group I under rain-fed environments, indicating it performed better in rain-fed environment than under irrigated environments. Nitrapyrin consistently provided lower grain yield under both water management environments, but low yield stability under rain-fed environments. The positions of untreated UAN and urea in the biplot were same under both water management environments. The sensitivity of liquid EEFs to rain-fed sites was high as more rain-fed sites were furthest from the origin (0, 0) than irrigated sites in the AMMI biplot (Fig. 3). Therefore, the yields with liquid EEFs varied more under rain-fed environments than irrigated environments, which is consistent with moisture availability limiting winter wheat performance more than N availability under rain-fed environments in the Canadian Prairies. NBPT is closer to the origin than all the other N sources, suggesting that it resulted in the most stable grain yield response across the irrigated and rainfed environments. On the contrary, Nitrapyrin as well as untreated urea produced the least stable grain yield relative to all the other N sources as they were furthest from the origin.
All N sources produced higher grain protein content than industry preferred levels (≥11%), except Nitrapyrin and untreated urea under irrigated environments (Tables 3 and 4).
Grain N was greatest with NBPT and UAN, which were greater than Nitrapyrin, PCU, and urea under irrigated environments and with NBPT and urea, which were greater than NBPT + DCD, Nitrapyrin, and UAN, under rain-fed environments. Grain protein content was lower with Nitrapyrin than all the other fertilizers under rain-fed environments except NBPT + DCD.
Based on the responses of grain yield and protein content to N treatments, our results suggest that a split-application of liquid UAN with a urease inhibitor will not only improve and stabilize grain yield but will also optimize protein production. Fertilizing winter wheat with untreated UAN under irrigated environments and untreated urea under rainfed environments produced similar grain yield, protein content, RE, and AE to that of NBPT under the same respective environments, suggesting that the split-application of these N sources may be improving synchrony between N supply and crop demand and thus minimizing potential differences among them. Our results are consistent with other studies that have observed split-applying EEFs by side-banding 50% at planting and broadcasting 50% in-crop in spring optimized winter wheat grain yields and protein content in the Canadian Prairies (Beres et al. 2018). In this study, Nitrapyrin treatments accumulated the lowest protein content across rainfed and irrigated environments relative to the other liquid EEFs, which may provide evidence that the nitrification inhibitor alone limited crop N uptake during critical times for protein production. Nitrification inhibitors treated urea have been shown to improve agronomic performance in winter Fig. 3. Additive main effects and multiplicative interaction biplot for grain yield of 6 N source and 22 site-years. Fairr is Farming Smarter irrigated, Farf is Falher rain-fed, Leirr is Lethbridge irrigated, Lerf is Lethbridge rain-fed, Strf is St. Albert rain-fed, Edrf is Edmonton rain-fed, and Brrf is Brandon rain-fed. Numbers 14, 15, 16, 17, and 18 are growing seasons 2013Numbers 14, 15, 16, 17, and 18 are growing seasons -2014Numbers 14, 15, 16, 17, and 18 are growing seasons , 2014Numbers 14, 15, 16, 17, and 18 are growing seasons -2015Numbers 14, 15, 16, 17, and 18 are growing seasons , 2015Numbers 14, 15, 16, 17, and 18 are growing seasons -2016Numbers 14, 15, 16, 17, and 18 are growing seasons , 2016Numbers 14, 15, 16, 17, and 18 are growing seasons -2017Numbers 14, 15, 16, 17, and 18 are growing seasons , and 2017Numbers 14, 15, 16, 17, and 18 are growing seasons -2018. UAN represents urea ammonium nitrate; NBPT, NBPT + DCD, PCU, and Nitrapyrin represent urease inhibitor NBPT-treated UAN, urease inhibitor NBPT plus nitrification inhibitor DCD-treated UAN, the treatment of PCU banded at planting and urea applied in early-spring, and nitrification inhibitor nitrapyrin treated UAN, respectively. wheat in wetter climates or under high irrigation where there is more potential for NO 3 -N leaching (Rao and Popham 1999). Nevertheless, in the context of a liquid UAN source, it did not improve grain yield or protein content in this study, either under irrigated or rain-fed environments. Thus, liquid UAN formulations containing nitrapyrin appear to pose greater economic and agronomic risks when used to fertilize winter wheat in the Canadian Prairies.

Responses of yield components and seedling vigor
No differences in seed mass, heads per plant, ESV, and LSV were detected among N sources across irrigated and rainfed environments (Tables 5 and 6). PCU displayed the lowest heads per plant and ESV at all sites. The responses of heads per plant and ESV to PCU were consistent with that of grain yield, grain N, total N uptake, RE, and AE under irrigated environments.
In this study, PCU is the only EEF in granular form, which was all side-banded at planting at a rate of 80% of the soil test N recommendation. Spring N supply was substituted with untreated urea based on the 80% of the soil test N recommendation. Nitrogen release from PCU during the fall was likely incompatible with winter wheat metabolic needs early in the growth cycle (Azeem et al. 2014), which in turn delayed the development of some crop traits (Baresel et al. 2008). Thus, heads per plant and seedling vigor with PCU were inferior to the other N sources. In the presence of soil water, N release from PCU is positively related to temperature (Cahill et al. 2010). In spring, the co-occurrence of warm temperature and high soil moisture under irrigated environments promoted N release from PCU and spring broadcasted urea. The amount of N released in a short period likely exceeded winter wheat demand and tended to leach under irrigated environments, and subsequently left less available N for winter wheat relative to other N sources. The lowest grain yield, grain N, total N uptake, RE, AE, heads per plant, ESV, and LSV observed in PCU under irrigated environments could be attributed to the distinctive N release characteristics of PCU. Beres et al. (2018) reported that winter wheat cold resistance could be improved by sufficient N availability at planting. When split-applying the 80% of recommended N, it appears that the proportion of PCU side-banded at planting is insufficient for winter wheat plant development before spring. The spring applied balance in the urea form did not fully compensate for the negative influence of PCU on plant development in early stages of winter wheat growth.  Note: NS represents not significant; * and * * denote significance at P < 0.05 and P < 0.01, respectively; SE denotes standard error; "-" represents missing P-values due to covariance estimate being 0. † ESV early-season vigor (1 = poor vigor; 5 = high vigor). ‡ LSV late-season vigor. § PCU banded at planting and urea applied in earlyspring in this treatment.
The apparent asynchrony between fertilizer N availability and winter wheat uptake for PCU is supported by the chlorophyll index as it was correlated with winter wheat grain yield at Feekes 4 (P = 0.004 * * ), Feekes 5 (P = 0.013 * ), Feekes 6 (P = 0.010 * * ), Feekes 7 (P = 0.031 * * ), and Feekes 9 (P < 0.001 * * ) under the irrigated environments and at Feekes 7 (P < 0.001 * * ) and Feekes 9 (P < 0.001 * * ) under rain-fed environments (data not shown). Feekes 6 to Feekes 9 are critical growth stages for winter wheat dry matter accumulation. The chlorophyll index measurements were lowest for PCU from Feekes 6 to Feekes 9 under irrigated environments and at Feekes 7 and Feekes 8 under rain-fed environments (Fig. 4), with trends toward significance at Feekes 7 (P = 0.066) and Feekes 8 (P = 0.079) under irrigated environments. This suggests some N deficiency in the PCU treatment, in particular under irrigated environments, as chlorophyll index measurements are strongly related to leaf N content (Sage et al. 1987). The polymer coating characteristics of PCU result in N release Fig. 4. Mean winter chlorophyll index in response to N sources under (A) irrigated and (B) rain-fed environments. UAN represents urea ammonium nitrate; NBPT, NBPT + DCD, PCU, and Nitrapyrin represent urease inhibitor NBPT-treated UAN, urease inhibitor NBPT plus nitrification inhibitor DCD treated UAN, the treatment of PCU banded at planting and urea applied in early-spring, and nitrification inhibitor nitrapyrin-treated UAN, respectively. Different letters on the top of the columns at Feekes 9 in graph (B) indicate significant differences at P < 0.05.

Responses of net CO 2 -equivalent emissions
The N 2 O and CH 4 emissions, expressed as CO 2 -eq N 2 O emissions, CO 2 -eq CH 4 uptake, net CO 2 -eq emissions, net CO 2eq/grain yield, net CO 2 -eq/grain N, and net CO 2 -eq/total N uptake did not vary among N sources (Table 7). The CO 2eq N 2 O emissions, net CO 2 -eq emissions, net CO 2 -eq/grain yield, net CO 2 -eq/grain N, and net CO 2 -eq/total N uptake were highest with UAN and lowest with Nitrapyrin. All the CO 2eq weighted agronomic metrics varied by year, except CO 2eq CH 4 uptake. Winter wheat grown under rain-fed conditions in the Canadian Prairies is subject to variable precipitation. Recent research conducted by Montoya et al. (2021) suggested that the efficacy of NBPT was compromised when significant rainfall or irrigation events occurred after fertilization. High precipitation or less solar radiation has also led to higher yield-scaled CO 2 -eq emissions in a winter wheat system (Wójcik-Gront 2018). The variations of precipitation among years (Fig. 1) explain the influence of year on CO 2 -eq N 2 O emissions, net CO 2 -eq emissions, net CO 2 -eq/grain yield, net CO 2 -eq/grain N, and net CO 2 -eq/total N uptake.
There was an expectation that if N management increased yields or total N uptake, or decreased greenhouse gas emissions, yield-scaled net CO 2 -eq emissions would decrease in response (Van Groenigen et al. 2010). Grain yield and total N uptake were generally highest when N fertilizer included a urease inhibitor, but increases were incremental and did not offset variation in net CO 2 -eq emissions, leading to no differences in yield-scaled net CO 2 -eq emissions among N sources. Previous works suggested that increasing total N uptake could reduce N 2 O emissions (Van Groenigen et al. 2010), which could have an important impact given that N 2 O makes the largest contribution to net CO 2 -eq emissions. Nitrous oxide emissions have been shown to be less respon- Table 7. Responses of CH 4 and N 2 O emissions and their net emissions expressed in CO 2 -equivalents (CO 2 -eq) and yield-scaled CO 2 -eq emissions to N sources and year by N source interactions in a rain-fed winter wheat production system in Lethbridge, AB, from 2014-2017. The value in parentheses immediately after each mean is standard error of the mean; NS represents not significant; * * denotes significance at P < 0.01. † CO 2 -eq of CH 4 is CO 2 -equivalents converted from cumulative methane; CO 2 -eq multiplicative value for methane was 35, which assumes a 100-year time frame. ‡ CO 2 -eq of N 2 O is CO 2 -equivalents converted from cumulative nitrous oxide; CO 2 -eq multiplicative value for N 2 O was 295, which assumes a 100-year time frame. § Net CO 2 -eq is the sum of CO 2 -eq of CH 4 and CO 2 -eq of N 2 O. || CO 2 -eqs used here are net CO 2 -eq. ¶ PCU banded at planting and urea applied in early-spring in this treatment.
sive to EEFs in moisture limited production systems (Akiyama et al. 2010). Other research also noted that the reduction of N 2 O emissions through the use of EEFs was limited in rain-fed regions due to the episodic nature of N 2 O emissions induced by rainfall events (Parkin and Hatfield 2014). Thus, it is plausible that the impacts of EEFs on N 2 O production were limited by low soil moisture in this study (data from three site-years), contributing to no differences in N 2 O emissions among N sources even though N uptake improved with some EEFs. A reduction of N 2 O emissions with EEFs relative to uncoated fertilizer was observed in spring wheat (unpublished paper). Our results suggested that the growth habit of winter wheat made it more resilient to EEFs effects and to climate smart approach to reduce N 2 O emissions. Van Groeni-gen et al. (2010) indicated that increased N 2 O emissions were associated with increased fertilizer rates. The lack of difference in yield-scaled net CO 2 -eq emissions among N sources could also be attributed to the moderate N application rates (80% of the soil test recommendation) in our study.

Conclusions
Our results suggest that fertilizing winter wheat with liquid UAN containing a urease inhibitor is a consistent top-performing option for winter wheat producers in the Canadian Prairies. Moreover, all EEF treatments in liquid forms produced protein content greater than the industry standard (≥11%), except Nitrapyrin under irrigated environ-ments. Therefore, in addition to EEFs in granular forms, producers in the Canadian Prairies also have an option to use EEFs in liquid forms to optimize winter wheat grain and protein content. There was no detectable improvement in grain yields when applying UAN combined with dual urease and nitrification inhibitors in both irrigated and rain-fed winter wheat production systems. Data from three site-years indicated that greenhouse gas emissions was not reduced with the application of EEFs in rain-fed environment. Asynchronization between N supply and key winter wheat developmental stages was suspected for PCU (PCU applied at planting + urea applied early-spring), which did not perform as well as expected, particularly in irrigated environments. Irrespective of the application method, in the context of liquid sources, UAN treated with a urease inhibitor may have the most potential to optimize winter wheat production and N use efficiency in the Canadian Prairies.