Nitrous Oxide Consumption Potential in a Semi-Arid Agricultural System: Effects of Conservation Soil Management and Nitrogen Timing on nosZ Mediated N2O Consumption

Agricultural soils account for less than 10% of the total greenhouse gas (GHG) emissions in the United States but about 75% of nitrous oxide (N2O) emissions. Soil conservation practices, such as no-tillage, have the potential to mitigate GHG emissions. We examined the short-term consequences of no-tillage with a winter wheat cover crop (NTW) and no-tillage winter fallow (NT) on N2O emissions, N2O reducing bacterial populations, and overall soil bacterial abundance during the summer growing season in the southern Great Plains, United States. Conservation practices were coupled with nitrogen (N) fertilizer application timing (100% pre-plant, 100% mid-season, 40% pre-plant 60% mid-season, 100% pre-plant with N stabilizer). In addition, N2O emissions were measured to determine any functional effects of altering N fertilizer timing and changing bacterial populations. The combination of N treatment and conservation practice affected nosZ clade II abundance in the second year of the study. Diversity of nosZ clade II was evaluated to determine effects on non-typical N2O reducers which were highly abundant in this study. No nosZ clade II diversity effects were determined, although some clustering of conservation system and N treatments was observed in the second year. Nitrogen treatment affected N2O-N emissions during the summer of both years, likely related to overall increased microbial activity and N fertilizer application. Negative fluxes (consumption) of N2O-N were observed in every treatment and tillage combination and were most pronounced in the control (0 kg N ha−1). Negative fluxes are likely due to a combination of low inorganic-N concentrations at various points during the year and a robust clade II population driving N2O consumption. Altering conservation system and the timing of N fertilizer application affects the microbial community and will likely continue to select for unique communities as the system matures. This will also likely further impact N2O emissions from the system and may increase the rate and frequency of N2O consumption.


INTRODUCTION
Nitrous oxide (N 2 O) is a long-lived atmospheric trace gas that has increased its concentration by 20% since preindustrial time (Butterbach-Bahl et al., 2013), has a global warming potential (GWP) approximately 300 times that of CO 2 (IPCC, 2007), and plays an important role in stratospheric ozone depletion (Griffis et al., 2017). Soil is the major global source of nitrous oxide (N 2 O), accounting for about 78% of N 2 O emissions in the United States in 2018 (EPA, 2020) where the predominant pathways for production are denitrification and nitrification (Bremner, 1997;Barnard et al., 2005;Mørkved et al., 2007;Reay et al., 2012). Denitrification most often occurs in anaerobic environments (Knowles, 1982;Linn and Doran, 1984) while nitrification is an aerobic process. How N 2 O is produced is thus often dictated by soil moisture content (Linn and Doran, 1984;Fentabil et al., 2016), with soil texture and organic matter (OM) content impacting soil water holding capacity, and consequently N 2 O production (Firestone and Davidson, 1989;Bremner, 1997). Increased soil moisture and O 2 content affect nitrification and denitrification processes differently. With denitrification, the genes and enzymes responsible are typically O 2 sensitive and thus increased soil moisture and reduced O 2 would enhance this process and N 2 O production (Betlach and Tiedje, 1981;Lloyd, 1993). For nitrification where reduced O 2 conditions are present, the oxidation of NH 3 to NO can be the terminus of the nitrification process, resulting in a potential buildup of NO for soil microbes (Caranto and Lancaster, 2017). This NO may then be reduced to N 2 O by those microbes, or through release to the environment and subsequent use by denitrifiers. With the oxygen sensitive nature of the enzymes responsible for N 2 O reduction, production of N 2 O from NO produced through the first step of nitrification in a mostly aerobic soil usually results in N 2 O as the final product (Lloyd, 1993).
However, soil carbon (C) content is a strong control over N 2 O production, compared to soil water, in semi-arid lands which cover approximately 35% of the terrestrial surface (McLain and Martens, 2006;Barton et al., 2008). In semi-arid lands, additional C inputs are derived from conservation practices such as cover cropping and no-tillage, which are most often implemented to mitigate eolian erosion, and increase soil health factors. These practices have been shown to substantially decrease eolian losses (Zobeck and Van Pelt, 2011), but still only about 38% of cropped area used conservation tillage and 6% used cover crops in Texas in 2012 (NASS, 2012). Studies aimed at understanding the impact these changes have on crop productivity and carbon dioxide emissions in semi-arid areas have been conducted (Keeling et al., 1989;Lewis et al., 2018;McDonald et al., 2019;McDonald et al., 2020). In more temperate climates, conservation tillage practices have been reported to not affect N 2 O emissions, while in dry climates conservation practices have been reported to generally decrease N 2 O emissions (Kessel et al., 2013). However, the effect of these conservation systems on N 2 O reduction and reduction potential is less understood. Conservation practices can increase soil respiration, potentially leading to anaerobic microsites and influencing the production and consumption of N 2 O and other greenhouse gases (Malhi et al., 2006;McLain and Martens, 2006;Barton et al., 2008;Halvorson et al., 2008;Smith et al., 2008). While anaerobic conditions are likely to increase with cover cropping and no-tillage, increased soil respiration, as brought about by the introduction of inorganic fertilizer and increased carbon inputs will also increase the growth and activity of denitrifiers in an otherwise "aerobic" agricultural soil (Wu et al., 2017). The increased potential for denitrification to occur under no-tillage is further supported by a meta-analysis regarding no-tillage effects on the abundance and activity of denitrifying communities (Wang and Zou, 2020). However, it is also under these conditions that the potential for N 2 O mitigation can be derived through increasing the abundance and activity of clade II N 2 O reducing bacteria in anaerobic microsites.
The bacteria capable of N 2 O reduction employ the nitrous oxide reductase (N 2 OR) enzyme produced from two clades of the nosZ gene, clade I and clade II, both of which have been shown to mediate N 2 O consumption, especially clade II . Clade II bacteria tend to be more abundant, diverse, and less likely to be associated with other denitrification genes (Jones et al., 2013;Jones et al., 2014;Higgins et al., 2016). These bacteria are hypothesized to use the reduction of N 2 O to nitrous gas (N 2 ) as a bet-hedging strategy to survive brief periods of anoxic conditions such as those present in agricultural soil under irrigation, or where anaerobic microsites are expected (Lycus et al., 2018). In addition to environmental selection via fluctuations in soil O 2 , low soil inorganic-N concentrations have been shown to increase N 2 O consumption in soil (Ryden, 1981;Minami, 1997;Butterbach-Bahl et al., 1998;Rosenkranz et al., 2006;Kroeze et al., 2007). With N fertilizer application, an increase in soil N 2 O emissions is to be expected due to overall increased N cycling as well as through competition for electrons between N 2 OR and NO 3 − reducing enzymes (Barnard et al., 2005;Mania et al., 2014;Shelton et al., 2017). Thus, best management practices for mitigating N 2 O emissions should include evaluation of N fertilizer application practices in addition to soil conservation strategies. Selecting for a system that can increase the potential for anaerobic microsites where N 2 O reduction can occur while also reducing the amount of labile N for microbial use is challenging and imperative for economic and environmental sustainability in existing and developing semi-arid agricultural soil. Previous reviews have discussed the importance of N 2 O consumption in soils, related to global greenhouse gas accounting, where a large potential for N 2 O consumption and understanding of the practices that enhance this capability could have significant impacts on reducing the effect of agricultural N 2 O emissions (Chapuis-Lardy et al., 2007;Hallin et al., 2018).
This research aimed to quantify the effects of implementing conservation practices coupled with N fertilizer management (timing of fertilizer application) on the abundance of N 2 O reducing genes, the diversity of the clade II population, and N 2 O emissions over a 2-year period in continuous cotton production on the Southern High Plains of Texas (SHP, MLRA 77C). By examining these systems shortly after implementation, we can evaluate the immediate effects of these conservation practices on N 2 O flux, while also providing valuable information for other factors that are considered with suggested changes in agronomic production. This study involved 15 unique treatment combinations of conservation practices and timing of fertilizer application for evaluation of N 2 O reducing potential, soil N resources, and in-situ N 2 O production, and consumption. In addition, the diversity of clade II N 2 O reducers was evaluated in a subset of treatment combinations that are most commonly implemented in the study area to determine if implemented conservation practices and N fertilizer management selected for unique N 2 O reducing communities. It was hypothesized that N 2 O consumption would be greatest where N fertilizer application was split or removed in a no-tillage system with a winter wheat cover crop and would be driven by an increased N 2 O reducing bacteria population. It was also hypothesized that commonly used treatment combinations would result in unique clade II communities due to their differential impacts on soil C and N inputs.

Study Area and Cropping System
This study was conducted at the Texas A&M AgriLife Research and Extension Center in Lubbock, Texas (33.687°, 101.827°). The 30-years (1991-2020) temperature and rainfall averages for this area were 16.1°C and 481 mm, respectively (Noaa, 2021). The soil was an Acuff loam described as fine-loamy, mixed, superactive, thermic Aridic Paleustolls (U.S. Department of Agriculture, 2016). The study design was as a split plot with conservation practice as the main plot and N fertilizer application as the split plot arranged in randomized complete blocks. Main plot conservation treatments included the following: no-till with a winter wheat cover crop (NTW), no-till winter fallow (NT), and conventional tillage winter fallow (CT). Split plot N fertilizer application timings were 1) no-added N (control); 2) 100% of N applied in a pre-plant application (PP); 3) 100% of N applied mid-season (MS) at the first reproductive growth of the cotton crop; 4) 40% of N applied PP and 60% MS applied (SPLIT); and 5) 100% of N applied PP with a N stabilizer product (STB). The stabilizer product used was Limus ® Nitrogen Management (N-butyl-thiophosphoric triamide and N-Propylthiophosphoric triamide, BASF Corporation, United States) a dual action urease inhibitor. Tillage main plots were randomly assigned to four rows (1-m row spacing) within each of the three blocks (replicates), and the N treatments were randomly arranged within each main plot, with each of the 5 N treatments were replicated within each tillage system. There were 45 plots measuring 15 m in length. Prior to seeding the cover crop in fall 2015, this area was under conventional tillage, winter fallow management for at least the last 60 years.
Nitrogen fertilizer was applied via knife injection using a coulter fertilizer applicator at a total rate of 168 kg N ha −1 placed 10-15 cm from the cotton row, as urea ammonium nitrate (UAN-32, 32-0-0). Preplant N treatments were applied on May 10, 2016 and May 11, 2017, and MS applications on July 13, 2016 and July 20, 2017. Wheat (TAM 304) was planted on January 25, 2016, and November 22, 2016 at a seeding rate of 67 kg ha −1 (19 cm row spacing). The planting on January 25, 2016 was a re-plant after a failed stand due to low soil moisture and precipitation at planting in November 2015. Glyphosate [Nphosphonomethyl glycine] at 2.2 kg active ingredient (a.i). ha −1 in 2016 and at 3.5 kg a.i. ha −1 in 2017 was used to chemically terminate the wheat cover crop on April 13, 2016 and April 20, 2017. Cotton (Delta-Pine 1,321) was planted on May 26, 2016 and June 6, 2017 at a rate of 123,553 seeds ha −1 and harvested on November 14, 2016 and November 15, 2017. Furrow irrigation (152 mm) occurred on July 1, 2016, July 27, 2016, August 13, 2016, June 6, 2017, and July 30, 2017. The full field management procedure was reported previously (McDonald et al., 2020). Climate data for the area was collected from the National Oceanic and Atmospheric Administration weather station 2 km south of the research site at the Lubbock International Airport, TX, United States ( Figure 1).

Soil Analysis
Soil samples were collected prior to N fertilizer application on April 8, 2016, and prior to the in-season mid-season application of N fertilizer in each year (July 1, 2016 and July 19, 2017) using a 5.1 cm diameter Giddings probe (Giddings Machine Company, Windsor, CO). Pre-season soil analysis was reported previously (Supplementary Table S1; McDonald et al., 2019). In-season soil samples were collected to a depth of 60 cm and subdivided into three increments: 0-15, 15-30, and 30-60 cm. A representative soil sample for each plot was collected by compositing two cores from each plot. Data from the 15-30 cm and 30-60 cm depth are not presented here. Sub-samples (50 ml) of the in-season soil samples (0-15 cm depth) were stored at −80°C until DNA extraction for microbial analysis. The remainder of the inseason samples were mechanically ground to pass a 2-mm mesh screen after drying at 60°C for 7 days. Soil samples were then stored at room temperature until nutrient analysis.
The samples collected in-season were analyzed for ammonium (NH 4 + -N) and NO 3 − -N by extracting with 2 M KCl using a 1:5 soil to extractant ratio (5 g soil and 25 ml 2 M KCl). Extracted samples were then analyzed for NH 4 + -N by the Berthelot reaction involving salicylate and NO 3 − by cadmium reduction prior to analysis using flow injection spectrometry [FIAlab 2600, FIAlab Instruments Inc., Bellevue, WA; Keeney and Nelson (1982)].

N 2 O Flux Measurements
Full details of the gas sampling procedure are outlined in McDonald et al. (2019). Gas samples were collected monthly as well as 1, 3, and 7 days post-fertilizer application (weather permitting) in both years (13 samplings from April 2016-March 2017; 12 samplings from April 2017-December 2017). A Gasmet DX-4040 portable FTIR (Fourier Transform Infrared) multi-gas analyzer (Gasmet Technologies, Helsinki, Finland) integrated with a 20 cm diameter Li-Cor survey gas chamber (Li-8100-103, Li-Cor Biosciences, Lincoln, NE United States) was used for in-situ gas flux analysis. A PVC collar (19.5 cm diameter, 11 cm height) was placed between the 2 nd and 3 rd row of each plot at least 10 h prior to sampling at a depth of at least 3 cm below the soil surface. During sampling, the Li-Cor chamber was deployed on each collar for 8 min, with a 20 s sampling time to yield 24 samples per plot. Linear regression of N 2 O concentration over the deployed time was conducted and fluxes with an r 2 of >0.7 were considered significant. Slope of the trendline was then calculated and used to determine the soil gas flux with the ideal gas law (McDonald et al., 2019). The minimal detectable concentration of N 2 O for the Gasmet DX4040 is 7 ppb (J. Cornish, Gasmet Technologies, personal communication, November 8, 2016). The data was sorted into seasons according to the major cotton growing periods of the year where Spring was April through May, Summer was June through September, and Fall/ Winter was October through March. These seasonal determinations coincide with pre-plant field operations for the spring, the major growing season in the summer, and the harvest and post-harvest season for the Fall/Winter on the SHP, respectively. Cumulative fluxes were determined from cover crop termination (April 13, 2016, April 20, 2017) through cotton harvest (November 14, 2016 andNovember 15, 2017). Calculation of cumulative emissions was determined by averaging the two most recent daily flux rates and extrapolating over the time between the two flux rate measurements. Fluxes of N 2 O for this study are considered baseline rates, due to complications in gas measurement following major rainfall and furrow irrigation events related to field access and equipment capabilities. However, these moisture events are infrequent on the SHP (29 days with >10 mm across 2016 and 2017, Menne et al., 2012a;Menne et al., 2012b) and thus "dry" measurements would represent the most common soil flux rate. Cumulative emission calculations estimated the total baseline N 2 O flux during the crop growing season to determine potential treatment differences in this semi-arid agricultural system and likely underestimate total N 2 O flux across the year due to the inability to measure post-wetting emission events. Winter fluxes of N 2 O were measured between the first and second year of the study and were determined to be low, or negative and thus likely do not significantly contribute to total N 2 O emissions from the treatments evaluated here.

Microbial Analyses
In addition to in-season soil chemical measurements, microbial analyses were performed for the 0-15 cm depth and included: qPCR of the 16S, nosZ clade I, and nosZ clade II genes; sequencing of the nosZ clade II genes. DNA from soil samples was extracted using DNeasy PowerSoil DNA isolation kits (Qiagen, Germantown, MD, United States) according to manufacturer protocol with a 5 min incubation at 2-8°C following solution C2 addition. A NanoDrop spectrometer (ND-1000, NanoDrop Technologies, LLC, DE, United States) was used to assess DNA purity after extraction. Quantitative polymerase chain reaction (qPCR) was conducted on extracted DNA to determine overall bacterial abundance as well as the abundance of both clades of the nosZ gene. Extracted DNA was quality checked for qPCR inhibitors in a process similar to Hartman et al. (2005) where a spiked sample with a threshold cycle (C t ) value within three standard deviations of the quality control C t mean was determined to not contain inhibitors. Quality control was conducted by spiking a qPCR assay with extracted DNA from the collected soil samples. The assay chosen quantified the abundance of Vibrio alginolyticus with gyrB as the gene target (Zhou et al., 2007), and no added DNA from collected soil samples inhibited the reaction. Quality control qPCR was conducted on a Bio-Rad CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, Inc., Hercules, CA, United States). Non quality control qPCR was conducted using an Eppendorf Mastercycler epgradient realplex 2 (Eppendorf North America, Hauppauge, NY, United States) with the programs outlined in Table 1. Each program begins with a denaturing step at 95°C with the 16S and nosZ clade I running for 15 min and the nosZ clade II running for 5 min. Both the 16S and clade II programs run for 40 cycles while the clade I program is a touchdown program from 67 to 62°C where it then runs for 40 cycles at 62°C. Sequences of nosZ Clade II were amplified with the primers listed in Table 1 and determined with a Pac-Bio Sequel (MR DNA, Shallowater, TX) due to the length of the nosZ clade II gene (∼700 bp). Sequence depth was about 5,000 sequences per sample. Sequences were trimmed with cutadapt (Martin, 2011), and denoised and dereplicated using DADA2 in R (Callahan et al., 2016;Callahan et al., 2019). Sequences were then uploaded to qiime2 (Bolyen et al., 2019) and low abundance sequences were removed prior to analysis (minimum five sequence occurrence and present in at least two samples). Downstream analysis was conducted in qiime2 including taxonomic classification using a Naïve Bayes classifier trained on about 5,000 nosZ sequences from both the National Center for Biotechnology Information (NCBI) database and the FunGene repository (Fish et al., 2013). Reference sequence taxonomy was downloaded from NCBI via the Entrez Direct module (Kans, 2020) and annotated to emulate the 16S Greengenes taxonomy file. Bray-Curtis dissimilarity PCoA plots were constructed in R with ggplot2 (Wickham, 2016). The eight most abundant amplicon sequence variants (ASVs) were chosen for further analysis where each ASV individually represented at least 2% of the total number of sequences (27,408 total sequences) and together amounted to approximately 40% of the total sequence count. These eight abundant ASVs were chosen for Basic Local Alignment Search Tool (BLAST).

Statistical Analyses
Data was analyzed using Proc GLIMMIX at a significance level of α 0.1 for soil N concentrations, N 2 O emissions, and microbial analysis using SAS version 9.4 (SAS Institute Inc., Cary, NC). Statistical analysis of N 2 O flux rate determined a year interaction effect with N treatment (p 0.015), so data was analyzed within year. In addition, an interaction of season (fall/winter, spring, and summer) and N treatment was determined for N 2 O flux rate in 2016 (p 0.003), so data was analyzed within season for both years of the study. Analysis of cumulative emissions was conducted for emissions occurring between cover crop termination (April 13, 2016, April 20, 2017) through cotton harvest (November 14, 2016 andNovember 15, 2017) for each year of the study due to a significant year interaction with N treatment (p 0.025). Soil inorganic N analysis was conducted for the 0-15 cm depth, and a year interaction with N treatment was determined (p < 0.001) so NO 3 − -N and NH 4 + -N were analyzed within year. For all analysis of variances, main-plot treatments (NTW, NT, CT) as well as split-plot treatments (control, PP, MS, SPLIT, STB) were treated as fixed effects and replication and replication by conservation system was treated as a random effect. Fisher's protected LSD was used to separate means of significant effects at α 0.1, unless otherwise stated. Correlation analysis was conducted using Pearson's correlation, proc CORR, and regression analysis, proc RSREG and proc REG (SAS, 2013) for determination of any correlation or relationship between soil chemical (N concentrations), biological (microbial abundance and diversity), cover crop biomass, and N 2 O-N flux rates and cumulative emissions. Bray-Curtis dissimilarity was determined in qiime2 (Bolyen et al., 2019); principal coordinates analysis plots were conducted in R with ggplot2 (Wickham, 2016); PERMANOVA and PERMDISP were conducted in R with the vegan package (Oksanen et al., 2019). Average sequence abundance within treatment for the eight most abundant features was calculated across both years due to no significant interactions of sequence percentage within year (p < 0.05).

Climatic Conditions
The climatic conditions collected in 2016 were more typical for the SHP, with average temperatures of 29°C and 25°C in July and August, respectively, and cooling off to an average temperature of   Table 2). Main plot and split-  Table S2). Many of the samples contained NH 4 + -N concentration lower than the detectable limit, which may have led to the lack of differences present within either year (analysis included a zero standard and thus zero values for NH 4 + were treated as zero in statistical analysis). However, there was an interaction effect on NH 4 + -N concentrations in 2017 (p 0.078; Table 2). Total inorganic N (N inorg ) was calculated as the sum of NO 3 − -N and NH 4 + -N. No differences in N inorg were determined for main plot, or the interaction in 2016 (Supplementary Table S2). In both 2016 and 2017, both N treatment significantly affected N inorg (p 0.005, p 0.003, respectively), with the PP and STB treatments having greater N inorg concentrations than the MS treatment and the control in both years. The interaction of conservation system and N treatment affected N inorg concentrations in 2017 (p 0.013; Table 2).

Microbial Abundance
Abundance of 16S genes in the 0-15 cm depth of soil was not affected by year interactions with conservation system, N treatment, or the interaction of conservation system and N treatment, and thus years were combined for further analysis. Conservation system (p 0.932), N treatment (p 0.608), and interaction effects (p 0.917) were not significant when years were averaged. The abundance of 16S gene copies was also analyzed by year due to significant year effects determined for other measured parameters. There was no difference in 16S gene abundance due to conservation system, N treatment, or their interaction in 2016 (Supplementary Table S3). In 2017, there were also no conservation system or interaction effects (Supplementary Table S3), although N treatment did affect 16S abundance (p 0.004). The control treatment (5.43*10 8 ) was determined to have a greater 16S abundance than the SPLIT (4.21*10 8 ) and PP treatments (3.98*10 8 ) in 2017, as well as the MS treatment (5.12*10 8 ) having a greater abundance than the PP treatment. Clade I abundance was not affected by N treatment, conservation system, or the interaction of N treatment and conservation system (Supplementary Table S3). Clade II was affected by the interaction of conservation system and N treatment in 2017 (p 0.081; Table 3), but not conservation system or N treatment (Supplementary Table S3). With no main plot or split plot effects determined within each year, except for the interaction effect on clade II abundance in 2017, N 2 O reducing population abundance was summarized across treatments. Clade I abundance in 2016 and 2017 was about 3.950 × 10 5 and 1.12 × 10 5 gene copies g soil −1 , respectively. Clade II abundance in 2016 was about 1.68 × 10 8 and 6.88 × 10 7 gene copies g soil −1 in 2017 (Figure 2)

Microbial Diversity
Due to the much larger abundance of clade II compared to clade I organisms present, sequence analysis of nosZ clade II was conducted on a subset of the in-season samples for 2016 and 2017. The subset included the PP treatment and the control from the CT and NTW systems, which were selected for the greatest potential differences and most common agronomic practices. All three replicates of each combination were sequenced for a total of 24 samples. It was determined that alpha diversity was not affected by conservation system or N treatment (data not shown).
No differences in distance within groups were determined to be significant, so PERMANOVA was conducted to evaluate Bray-Curtis dissimilarity. It was determined that no factors (Year, Tillage, Treatment) or their interactions affected dissimilarity of the microbial communities (α 0.05; Table 4). Furthermore, dissimilarity was analyzed with principal coordinates analysis and no distinct clustering occurred with treatment and tillage combinations over the 2 years (Figure 3), clustering occurs with similar values of dissimilarity, meaning a treatment with similar dissimilarity (across replication) compared to the rest of the treatments would be more tightly clustered. The principal coordinates axes combined to explain 35% of the variability within the data (PC1: 22%, PC2: 13%; Figure 3).
Classification of ASVs with the NCBI and FunGene databases resulted in few sequences having deeper classification than domain. In general, there is less taxonomic information for functional genes such as nosZ clade II, leading to many of the samples being classified as "environmental samples," pointing to

Factor
Bray-curtis dissimilarity Unweighted unifrac distance  Table S5) to provide some insight into the microbial identity ( Table 5). No year, treatment, tillage, or interaction effects were determined for the treatment abundance of any ASV (α 0.05).

Nitrous Oxide Flux Rates
Nitrous oxide emissions were analyzed within season for each year of the study. No differences were found between conservation systems, N treatments, or their interactions for the Spring, and Fall seasons in both years of the study (Supplementary Table S6). However, N 2 O-N flux was affected by N treatment in the Summer of 2016 (p 0.013) and 2017 (p 0.076) with all the N treatments having a greater flux than the control in Summer 2016 ( Figure 4A) and all N treatments except the PP treatment having a greater flux than the control in Summer 2017 ( Figure 4B). Negative fluxes and no N 2 O flux were determined in the MS treatment and the control, respectively, in the spring of 2016 ( Figure 4A). Nitrous oxide consumption was also recorded in the Fall/Winter of both years for all treatments with the exception of the PP treatment in 2017 ( Figures 4A,B). No correlation between 16S, clade I, or clade II abundance, and summer N 2 O flux rate was determined (α 0.05).

Cumulative Emissions
Nitrogen treatment impacted average growing season cumulative emissions in 2016 with the N treatments producing a greater average cumulative flux than the control (p 0.027; Figure 5). The control had net negative cumulative emissions of N 2 O-N in both years of the study ( Figure 5). Although not statistically different, cumulative emissions were lower for the STB treatment in 2016. No conservation system or conservation system and N treatment interaction effects were determined in either year of the study (Supplementary Table S7). In 2017, N treatment did not affect cumulative N 2 O-N emissions. Emissions during the second year of the study were greatly increased in the STB treatment compared to the first year while the rest of the N treatments were slightly reduced.

DISCUSSION
Results of this study indicate that altering the timing of N fertilizer application will affect the microbial community, which may alter N 2 O emissions. In the second year of N treatment implementation, bacterial abundance was reduced for the PP treatment compared to treatments with no N applied prior to sampling, the control and MS treatment. In addition, the SPLIT treatment had reduced bacterial abundance compared to the control. Both of these observations indicate potentially deleterious effects of high rates of N fertilizer application on overall bacterial abundance on the SHP, although other research suggests the deleterious effect of N fertilizer application may be less pronounced in agricultural systems where application is more common (Geisseler and Scow, 2014). The reduction in the overall bacteria population due to N treatment did not translate to the N 2 O reducing community, but  the interaction of N treatment and conservation system did affect clade II abundance in the second year of the study (Table 3). In addition, it was clear that clade II greatly outnumbered clade I, indicating that N 2 O reduction potential lies mostly within the more diverse, abundant, and efficient form of N 2 O reductase (Sanford et al., 2012). Previous research supports greater clade II abundance in soil (Sanford et al., 2012;Jones et al., 2013;Domeignoz-Horta et al., 2015), with clade II being associated with a lifestyle strategy involving the survival of anoxic conditions in a more energetically favorable way (Lycus et al., 2018). In agricultural soil, oxygen concentrations may change rapidly due to precipitation or irrigation events, as well as pore space O 2 concentrations changing due to soil respiration so a more efficient survival mechanism such as clade II activation is helpful. Although there are no clear patterns across the entire study related to conservation system and N treatment, the most apparent differences in clade II abundance were within the NT system in 2017 where nutrient stratification and low soil C would both be expected. Where N fertilizer was applied before the season (PP, SPLIT, STB) clade II abundance was reduced and where N was not applied for at least one full year, clade II abundance was increased (MS, control). Previous research indicates that N 2 O consumption can be enhanced during periods of low soil inorganic N (Butterbach-Bahl et al., 1998;Rosenkranz et al., 2006;Kroeze et al., 2007) and thus would be expected for the control and MS treatment. Nitrous oxide consumption would thus likely be driven by clade II microbes in N limited environments. Nitrate can act as a proximal control over N 2 O consumption where high concentrations can negatively affect the production of the N 2 OR enzyme through competition for electrons (Highton et al., 2020). This was observed in the PP and STB treatments in 2017, where increased soil NO 3 − -N due to pre-season application acted as a distal control, reducing the N 2 O consuming population at the time of sampling ( Table 2). A similar but less distinct pattern was present for the CT system, while in the NTW system there was no apparent pattern ( Table 2). The lack of a pattern in the NTW system could be due to several biological and physical phenomena including overall reduced soil inorganic N prior to the start of the season due to wheat cover crop use of residual soil N (Lyons et al., 2017). This would likely enhance the N 2 O reducing population over the winter and may conceal any effects on clade II abundance in-season and will likely increase as the system matures and selects for a specific microbial community. The NTW system also encourages water infiltration, soil aeration, and soil C resources, all of which would have differential effects on anaerobic microbial processes further complicating any patterns within that system. Wang et al. (2021) determined greater clade II abundance with long term conservation tillage practices likely leading to greater moderation of N 2 O emission within those systems, supporting previous determinations of conservation practice effects on N 2 O emissions in systems in place for ≥10 years (Kessel et al., 2013). The increased population of Frontiers in Environmental Science | www.frontiersin.org November 2021 | Volume 9 | Article 702806 clade II organisms with conservation tillage was not determined in this study, nor were N 2 O reductions for conservation practices determined. However, this study comprised the first 2 years of conservation practice implementation for the research site and it is likely that as these systems mature similar patterns will emerge. Abundance of nosZ clade I genes was correlated with the amount of aboveground wheat cover crop biomass (data not shown) produced in the NTW system during this study (p 0.005;McDonald et al., 2019), where clade I abundance linearly increased with increasing wheat biomass (p 0.0025, r 2 0.56). However, there was no relationship between nosZ clade II abundance and wheat biomass (p 0.100). The correlation between clade I abundance and wheat biomass was expected, as greater wheat residue would increase conditions favorable for denitrification, selecting for more typical, complete, denitrifiers such as those associated with nosZ clade I.
To further elucidate any patterns in potential N 2 O consumption, sequence analysis was conducted for a subset of the conservation system and N treatments within both years of the study. The subset included the PP treatment and the control from the CT and NTW systems, which were selected for the greatest potential differences and most common agronomic practices. Although no year, N treatment, conservation system, or interaction effects were determined for Bray-Curtis dissimilarity ( Supplementary Table S5), there is clearly some clustering of treatment and tillage combinations in the second year of the study (Figure 3), indicating potential development of unique N 2 O reducing communities as also indicated with the significant conservation system and N treatment interaction for clade II abundance in the second year. In a previous study, homologs of nosZ were found in 12% of sequenced bacterial genomes (Graf et al., 2014), and while no significant classifications could be made from nosZ sequences alone in our study, the individual query of the most abundant features allowed for some taxonomic evaluations to be made from fully sequenced soil microbial genomes. These classifications included Gemmatimonadetes which are the most abundant N 2 O reducers in soil (Jones et al., 2013) and a common soil bacteria. The association of such common soil bacteria with N 2 O reduction speaks to the ubiquitous and environmentally relevant nature of N 2 O reduction in soil, and further supports potential for the soil to act as an N 2 O consumer even in semi-arid agricultural systems.
With the potential for N 2 O consumption being observed in this study, N 2 O emissions were measured throughout 2016 and 2017 ( Figures 4A,B). Negative fluxes of N 2 O were determined during the fall/winter of both years of the study and are likely due to low levels of NO 3 − -N present in the soil (Ryden, 1981;Minami, 1997;Butterbach-Bahl et al., 1998;Rosenkranz et al., 2006;Kroeze et al., 2007) and an abundant clade II population. In the spring of 2016 ( Figure 4A) N 2 O consumption was also determined for the MS treatment which had not yet received N fertilizer during the study period. Treatments with N fertilizer application increased N 2 O emissions in Spring 2016. Increased emissions of N 2 O are often associated with greater levels of NO 3 − -N (Minami, 1997;Butterbach-Bahl et al., 1998;Kroeze et al., 2007;Mania et al., 2014;Mania et al.,2016) which can negatively impact the formation of the N 2 O reducing enzymes (Highton et al., 2020). Similar assumptions can be made regarding the negative and zero fluxes of N 2 O-N in the fall/winter of 2016 and 2017 ( Figures 4A,B), where low levels of NO 3 − -N would be present due to plant and microbial use of available soil N throughout the growing season. There is a significant increase in N 2 O-N emissions during the summer season ( Figures 4A,B) which can be attributed to several factors including: increased temperature and moisture, increased plant and microbial activity, and the application of N fertilizer (Ryden, 1981;Dobbie et al., 1999;Barnard et al., 2005;McSwiney and Robertson, 2005;Shelton et al., 2017). However, where N was not applied in the control, net consumption of N 2 O-N was determined, further supporting the association of N 2 O consumption in low inorganic N environments and providing evidence of an active N 2 O reducing community where clade II organisms play a significant role. After 1 year of treatment implementation, N 2 O-N flux rates were reduced during the spring ( Figure 4B), which may be attributed to reduced NO 3 − -N concentrations, and thus increased N 2 OR activity. Monthly measurements were used for evaluation of treatment differences across the cotton growing season. Recent studies have reported the potential for under or over estimation of total N 2 O emissions with less frequent measurements (Su et al., 2021), however our reported seasonal average emissions compare favorably with N 2 O fluxes measured under similar climatic conditions (Ryden, 1981;Shelton et al., 2017;Domeignoz-Horta et al., 2018), although they were lower compared to emissions from wetter climates (Chantigny et al., 2010;Domeignoz-Horta et al., 2018). When the rate of N 2 O consumption was compared, it was similar to previously reported rates in varied study areas indicating a functionally ubiquitous N 2 O consuming population in soil regardless of environment.
Cumulative emissions were calculated based upon monthly measurements of N 2 O emissions and were positive for all treatments with N fertilizer application ( Figure 5). However, the control resulted in net negative emissions over the 2-year study. The negative emissions recorded are likely the result of inorganic N loss (plant uptake, microbial use, leaching, etc.) from the system without significant replacement (Butterbach-Bahl et al., 1998;Rosenkranz et al., 2006;Kroeze et al., 2007) combined with selection for a N 2 O consuming population by both the soil chemical and physical characteristics. The calculation of cumulative emissions was conducted to estimate treatment effects on yearly N 2 O emissions to determine potential best practices based on the data available. The total number of treatments evaluated for N 2 O fluxes included 15 unique combinations of conservation systems and N treatments which required various field operations throughout the growing season to maintain production-level field conditions and thus installation of long-term chambers and stationary gas analysis was not feasible for a study of this size and scale. However, cumulative emissions produced from this study were comparable to those from other studies in semi-arid and sub-humid regions (Dong et al., 2018) although total emissions were potentially underestimated due to the inability to measure emissions following precipitation or irrigation events.
Although no definitive relationships can be observed between recorded N 2 O emissions and the abundance of N 2 O reducing Frontiers in Environmental Science | www.frontiersin.org November 2021 | Volume 9 | Article 702806 genes, it is clear that the treatments implemented here affect the pattern of N 2 O emissions through alterations to the soil biological and chemical composition even within the first few years of implementation. It will thus be important to continue this research as the system matures to observe these expected changes and better elucidate how conservation system and N timing affect N 2 O emissions in semi-arid agricultural systems. Understanding the mechanisms behind these changes in emissions, specifically the consumption of N 2 O, will aid in the choosing of best management practices for reducing N 2 O emissions in expanding semi-arid areas and potentially provide practices suited for net N 2 O consumption.

CONCLUSION
Changes in N fertilizer management can alter the microbial community and change the rate at which N 2 O gas is produced or consumed within semi-arid agricultural soils. The microbial community measured in this study contained an abundant nosZ clade II N 2 O reducing population that is likely the driver for N 2 O consumption. The mechanisms behind the population shift are still being determined, but after 2 years of treatment implementation distinct communities appear to be forming which may further alter N 2 O consumption and production. Based on the results from this study, it is likely that a N treatment and conservation practice best suited for mitigating N 2 O will emerge as the system matures.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: NCBI [accession: PRJNA612879].

AUTHOR CONTRIBUTIONS
MM: data acquisition, data interpretation, and writing of the article. KL: conception and design, data interpretation, and writing of the article. PD, TB, and JR: writing of the article. TG: data interpretation and writing of the article. All authors contributed to the article and approved the submitted version.

FUNDING
Funding for this research was provided through the Texas A&M AgriLife Research Air Quality Initiative.