Assessing Seasonal Methane and Nitrous Oxide Emissions from Furrow-Irrigated Rice with Cover Crops

: Improved irrigation management is identiﬁed as a potential mitigation option for methane (CH 4 ) emissions from rice ( Oryza sativa ). Furrow-irrigated rice (FR), an alternative method to grow rice, is increasingly adopted in the Mid-South U.S. However, FR may provide a potential risk to yield performance and higher emissions of nitrous oxide (N 2 O). This study quantiﬁed the grain yields, CH 4 and N 2 O emissions from three different water management practices in rice: multiple-inlet rice irrigation (MIRI), FR, and FR with cereal rye ( Secale cereale ) and barley ( Hordeum vulgare ) as preceding winter cover crops (FRCC). CH 4 and N 2 O ﬂuxes were measured from May to September 2019 using a static chamber technique. Grain yield from FR (11.8 Mg ha − 1 ) and MIRI (12.0 Mg ha − 1 ) was similar, and signiﬁcantly higher than FRCC (8.5 Mg ha − 1 ). FR and FRCC drastically reduced CH 4 emissions compared to MIRI. Total seasonal CH 4 emissions decreased in the order of 44 > 11 > 3 kg CH 4 -C ha − 1 from MIRI, FR, and FRCC, respectively. Cumulative seasonal N 2 O emissions were low from MIRI (0.1 kg N 2 O-N ha − 1 ) but signiﬁcantly higher from FR (4.4 kg N 2 O-N ha − 1 ) and FRCC (3.0 kg N 2 O-N ha − 1 ). However, there was no net difference in global warming potential among FR, FRCC and MIRI. These results suggest that the increased N 2 O ﬂux from furrow-irrigated rice may not greatly detract from the potential beneﬁts that furrow-irrigation offers rice producers. and


Introduction
One of the greatest challenges of agriculture in the 21st century is to meet the growing food demand while simultaneously abating its environmental impacts. Food production needs to increase by 2050 to meet global food demand while agriculture must concurrently address climate change, biodiversity loss, and soil and water degradation [1]. Rice is a major staple food for almost half of the world's population and its demand is expected to grow through 2025 with an increase in population [2]. Yet, traditional rice cultivation practices in flooded paddy have the highest global warming potential (GWP) compared to other cereal crops primarily due to high methane (CH 4 ) emissions [3], accounting for about 11% of global anthropogenic CH 4 emissions [4]. Furthermore, flooded rice systems also face major environmental issues such as higher water use. About 75% of global rice is produced in irrigated lowlands [5], where the fields are mostly continuously flooded throughout the growing season. Irrigated rice uses 34-43% of total world's irrigation water [5]. Thus, its sustainability is highly threatened by global water scarcity [6]. Therefore, recent research is focused on alternative water management practices to concurrently increase irrigation water productivity and reduce greenhouse gas (GHG) emissions from rice without altering the grain yield.
Furrow-irrigated rice, also known as row rice, is being increasingly adopted by farmers in the Mid-South U.S. In this system, rice is grown on relatively short raised beds and

Study Site and Crop Management
Field studies were conducted at a commercial farm (35 • 48 53 N, 89 • 59 50 W) located in Burdette, Arkansas. The soil at the study site was Sharkey silty clay (very fine, montmorillonitic, nonacid, thermic Vertic Haplaquepts) that had been under a long-term rice and soybean rotation. Prior to this study, all fields were cropped to soybean, and the residues from the preceding soybean crop were left on the ground after harvest. No intentional effort was made to capture rainfall within the fields, but it should be noted that rainfall capture is inherent to any system that creates freeboard.
The study was conducted from May to September 2019 in a complete block design replicated five times. Three fields were used with different irrigation treatments: contin- uously flooded with multiple-inlet rice irrigation (MIRI), furrow-irrigated rice (FR), and furrow-irrigated rice with winter cover crops (FRCC). Each field was approximately 30 ha and had been previously precision graded to 0.1% slope. In all fields, irrigation water was delivered using 10 mil (thickness) by 38 cm (diameter) lay-flat, plastic tubing (Delta Plastics, Little Rock, AR, USA).
The MIRI field served as the conventional control and consisted of seven paddies separated by six straight earthen levees. Plastic tubing was installed perpendicular to the levees and multiple holes were installed in each paddy to allow irrigation water to be supplied simultaneously to all paddies. After flood initiation, a flood depth ranging from approximately 2 to 10 cm was maintained until about two weeks prior to harvest, at which time the field was drained. The FR and FRCC fields consisted of raised beds that were approximately 10 cm tall and spaced 91 cm apart (center to center). The beds were used to grow soybeans in 2018. Immediately after the soybean crop was harvested in August 2018, a 50:50 mixture of annual rye and barley was drilled into the beds of the FRCC field. The cover crop was terminated at the time of rice planting on 23 April 2019 using glyphosate herbicide. A disk-bedder was used to "freshen" the beds of the FR field prior to rice planting in 2019 while a modified field cultivator was used to clean the furrows of the FRCC field prior to the first irrigation. Plastic tubing was installed across the top of each field and one hole punched at every furrow to allow irrigation water to flow down each furrow.
Hybrid XP753 (RiceTec, Inc., Alvin, TX, USA) were drill seeded at the rate of 27 kg ha −1 on 23-24 April 2019 in all treatments. No irrigation water was applied following seeding because rainfall was sufficient for crop establishment. On 6 June, that is, 43-44 days after sowing, irrigation was commenced on all fields. Approximately three weeks after irrigation initiation, field outlets were blocked to create 5 to 10 cm standing floods across the entire MIRI field and the lower portions of FR and FRCC fields. The MIRI treatment was kept flooded through the growing season with the supply of irrigation at approximately a 5-day interval. In FR and FRCC, irrigation water was supplied at approximately a 3-day interval. Irrigation was halted on all fields on 29 August, resulting in an 84-d irrigation period. The combination of the 0.1% field slope and lack of levees on the FR and FRCC fields resulted in different management zones where the upper portion (~30%) of the fields was moist but not flooded (i.e., aerobic) and the lower portion (~30%) of the fields was maintained with a standing flood. Thus, the FR and FRCC fields were divided into top and bottom sections for greenhouse gas flux measurement as discussed below. The top sections (FR Top, FRCC Top) were located 76 m from the uppermost field edge while the bottom sections (FR Bottom, FRCC Bottom) were located 76 m from lower (i.e., downslope) side of each field.
Urea fertilizer was aerially broadcast over the whole field in a two-way (MIRI) and three-way split application (FR and FRCC fields) as shown in Table 1. First application of fertilizer was applied a few days prior to the permanent flood establishment or prior irrigation for all fields. For FRCC, an additional 39 kg N ha −1 was applied in the top portion of the field due to the poor growth of rice plants.

Soil Sampling and Analysis
Five composite soil samples (0-25 cm) for physico-chemical analysis were randomly collected in April 2019 from top and bottom sections of each field. The soil samples were air-dried and ground to pass a 2 mm sieve. The soil was analyzed by the University of Missouri Soil and Plant Testing Laboratory, Columbia, Missouri, USA for texture and various chemical properties (pH, electrical conductivity, cation exchange capacity, total C and total N).
Soil core samples (diameter, 22 mm) from the 0-25 cm depth were also collected for mineral N (NH 4 + , NO 3 − ) analysis on 11 July and 2 and 15 August. Soil-exchangeable N was determined by extracting 10 g of fresh soil with 80 mL of 2 M KCl. The extracts were stored at −20 • C and later thawed and analyzed for mineral N content. Soil exchangeable NH 4 -N concentration was determined using the salicylate colorimetric method while soil NO 3 -N concentration was measured using the vanadium reduction method [21].

Measurements and Calculation of Greenhouse Gas Fluxes
Greenhouse gas fluxes were measured using static flux chamber technique. The chamber consisted of a PVC base (collars), extensions of various heights (15.3-91.4 cm) to accommodate growing rice plants and a chamber lid. The chamber lid was equipped with a vent tube, fan and thermocouple wire for measurement of air temperature. Each PVC base (diameter, 29.5 cm and height, 22.9 cm) had four holes (diameter, 11 cm) at the bottom to facilitate water movement and root interaction. Additionally, two holes (diameter, 2.86 cm) were drilled into the top parts of the collars to allow free movement of water inside collars. The holes were plugged with rubber stoppers during gas sampling. The bases were permanently installed to an approximate depth of 15 cm into the soil when the rice was at the 3-4 leaf stage. Permanent boardwalks were installed to access each collar so as to minimize soil and plant disturbance during gas sampling.
Three gas samples (25 mL) were drawn from chamber headspace at equal time intervals within 1 h of chamber closure. Headspace gas was mixed for 1 min using a 12-V fan installed inside the chamber lid (Allied Electronics, Forth Worth, TX, USA) before gas sampling when the chamber height reached >30 cm. Also, ambient gas samples were collected from each plot at 0 min. The gas samples were transferred in pre-evacuated 12.5 mL vials. The glass vials were sealed with rubber septa and silicon to avoid gas leakage. Gas samples were generally collected from 9:00 to 13:00. The sequence of gas sampling was randomized to avoid possible bias with temperature difference during gas measurements. Gas samplings were performed at weekly intervals. Gas concentrations of CH 4 and N 2 O were determined on a GC-2014 gas chromatograph (Shimadzu Scientific, Inst., Columbia, MD, USA) connected with an autosampler (XYZTEK, Sacramento, CA, USA) configured and calibrated as described by Adviento-Borbe et al. [22]. Fluxes of CH 4 and N 2 O were calculated from the linear increase of gas concentration over time. CH 4 and N 2 O fluxes below the detection limit of GC (0.203 ppm for CH 4 and 0.010 ppm for N 2 O) were treated as zero fluxes.

Yield Component Analysis
A 1-m 2 area from the middle of each replicate treatment plot was manually harvested at maturity to estimate grain yield and yield components. Approximately 2 kg of fresh biomass was subsampled by hand. Rice biomass from subsamples were analyzed for grain yield and yield components such as plant height, total panicle count, tiller number and 1000 grain weight. Plant heights were noted for 10 randomly selected plants from the subsamples. Grain yield, tiller, and panicle numbers were then calculated on area basis after upscaling the measured variables from 2 kg subsample to total fresh biomass from 1 m 2 measured in the field. Grains were manually threshed from the panicles and cleaned with a seed blower (SeedBuro, Des Plaines, IL, USA), oven-dried at 60 • C until constant weight and adjusted to 14% moisture for grain yield. Also, the weight of the straw was noted after oven drying at 60 • C. The harvest index was then determined as the ratio of grains to the total aboveground biomass.

Environmental Variables
Air temperature and precipitation data were obtained from a nearby NOAA weather station located 20 km from the study site. Air temperature during chamber closure was recorded using a thermocouple wire attached to each chamber. Soil temperature at 5 and 10 cm soil depths was recorded using a digital thermometer (Thermo Fisher Scientific, Pittsburgh, PA, USA). Floodwater depth was measured manually at every gas sampling event. Additionally, soil moisture at 15, 30 and 45 cm depth were recorded at daily intervals using soil moisture sensors (Irrometer, Riverside, CA, USA).

Data Analysis
Cumulative seasonal fluxes were calculated by linear interpolation between sampling dates. Cumulative fluxes were calculated for each collar and then averaged for each water treatment (n = 5). Global warming potential (GWP) of CH 4 and N 2 O was calculated in mass of CO 2 equivalent (kg CO 2 eq ha −1 ) over 100-yr time horizon. A radiative forcing potential relative to CO 2 of 28 for CH 4 and 265 for N 2 O [23] was used. Yield-scaled global warming potential (GWP Y ) was calculated by taking the ratio of GWP and corresponding grain yield for each treatment.
The differences in mean cumulative CH 4 , N 2 O, and GWP among treatments were computed using R version 3.6.1 for completely randomized design at p-level < 0.05 [24]. Data that failed normality test (Shapiro Wilk test) and homogeneous variance were logtransformed. Greenhouse gas emissions, grain yield, yield components, GWP and GWP Y due to main effects such as irrigation water treatments and location (top and bottom) were analyzed using R with function "lm". Pairwise comparison in between the treatments was performed using package "lsmeans" in R.

Characteristic of Soil and Soil Mineral Nitrogen
Soil texture differed among treatment plots. The soil texture was loam, clay loam and clay from FR, FRCC and MIRI, respectively ( Table 2). Other soil properties at 0-25 cm depth did not differ among the treatment plots with average total organic carbon, 11 g kg −1 , total nitrogen, 1.1 g kg −1 , soil pH w 7.1, electrical conductivity, 0.18 ds m −1 and cation exchange capacity, 26.7 cmol c kg −1 ( Table 2). Table 2. Soil characteristics (0-25 cm) from furrow-irrigated rice (FR), furrow-irrigated rice with winter cover crops (FRCC) and multiple-inlet rice irrigation (MIRI) fields studied in northeast Arkansas.

Treatment (Field Location)
Texture Soil exchangeable nitrate concentration was negligible or below the detection limit across water treatments in all soil sampling occasions. Also, soil exchangeable ammonium was low (<1 mg N kg −1 dry soil) and similar in all treatments during soil sampling occasions ( Figure 1). Soil exchangeable nitrate concentration was negligible or below the detection limit across water treatments in all soil sampling occasions. Also, soil exchangeable ammonium was low (<1 mg N kg -1 dry soil) and similar in all treatments during soil sampling occasions ( Figure 1).

Environmental Conditions
The mean daily air temperature ranged from 21 to 31 °C during the study period. (Figure 2a). Total annual precipitation was 1218 mm and only around 25% occurred during the study period. The soil temperature during gas measurements ranged between 22 to 33 °C at 0-5 cm depth across all treatments (Figure 3a). Soil temperatures at 0-5 cm depth were similar across all treatments, except for FRCC in top sections where higher

Environmental Conditions
The mean daily air temperature ranged from 21 to 31 • C during the study period. (Figure 2a). Total annual precipitation was 1218 mm and only around 25% occurred during the study period. The soil temperature during gas measurements ranged between 22 to 33 • C at 0-5 cm depth across all treatments (Figure 3a). Soil temperatures at 0-5 cm depth were similar across all treatments, except for FRCC in top sections where higher temperatures were observed mostly throughout the measurement period. Soil temperatures at the 0-15 cm soil depth ranged from 18-29 • C across all treatments with no differences among treatments. Soil water contents at 15 cm soil depth in top field sections were relatively constant at 0 kPa for FRCC but they fluctuated between 0 kPa to −80 kPa in FR ( Figure 2b). In bottom field sections, soil water contents were higher in FR as compared to FRCC (Figure 2c). Soil water contents were similar for FR and FRCC after mid-July. Soil water contents were 0 kPa at the 30 to 45 cm soil depths for all treatments.
During gas sampling, standing floodwater was never observed within the upper sections of the row rice fields except for two occasions for FRCC (Figure 3b). In the bottom sections of the furrow-irrigated fields, floodwater depths fluctuated between 0 to 9 cm above the soil surface. The MIRI field was always flooded from June to September with floodwater depth ranging from approximately 2 to 10 cm (Figure 3b).

Grain Yield and Yield Components
There was a significant effect of water management on rice grain yield (p = 0.03). The average grain yield from FRCC was 8.5 Mg ha −1 , which was significantly lower than MIRI and FR by almost 29% (Figure 4). For FR, the biomass yield from the top field section was significantly (p = 0.004) lower than the bottom section. A similar trend was obtained for tiller number and panicle number (Table 3). However, there was no significant effect of irrigation water management on plant height, harvest index, and 1000-seed grain weight. temperatures were observed mostly throughout the measurement period. Soil temperatures at the 0-15 cm soil depth ranged from 18-29°C across all treatments with no differences among treatments. Soil water contents at 15 cm soil depth in top field sections were relatively constant at 0 kPa for FRCC but they fluctuated between 0 kPa to −80 kPa in FR (Figure 2b). In bottom field sections, soil water contents were higher in FR as compared to FRCC (Figure 2c). Soil water contents were similar for FR and FRCC after mid-July. Soil water contents were 0 kPa at the 30 to 45 cm soil depths for all treatments. During gas sampling, standing floodwater was never observed within the upper sections of the row rice fields except for two occasions for FRCC (Figure 3b). In the bottom sections of the furrow-irrigated fields, floodwater depths fluctuated between 0 to 9 cm above the soil surface. The MIRI field was always flooded from June to September with floodwater depth ranging from approximately 2 to 10 cm (Figure 3b).

Grain Yield and Yield Components
There was a significant effect of water management on rice grain yield (p = 0.03). The average grain yield from FRCC was 8.5 Mg ha −1 , which was significantly lower than MIRI and FR by almost 29% (Figure 4). For FR, the biomass yield from the top field section was significantly (p = 0.004) lower than the bottom section. A similar trend was obtained for tiller number and panicle number (Table 3). However, there was no significant effect of irrigation water management on plant height, harvest index, and 1000-seed grain weight.

Seasonal Dynamics of Methane and Nitrous Oxide Emissions
Methane fluxes were not detected or were close to zero for all treatments during the initial period of crop development (i.e., 70 days after seeding) (Figure 5a). CH4 emissions in MIRI increased after two weeks of flooding in July and continued to increase, reaching a peak emission of 1174 g CH4-C ha −1 day −1 in mid-July (Figure 5a). Thereafter, CH4 fluxes declined to near zero a few days following removal of the flood from the field. In FR and

Seasonal Dynamics of Methane and Nitrous Oxide Emissions
Methane fluxes were not detected or were close to zero for all treatments during the initial period of crop development (i.e., 70 days after seeding) (Figure 5a). CH 4 emissions in MIRI increased after two weeks of flooding in July and continued to increase, reaching a peak emission of 1174 g CH 4 -C ha −1 day −1 in mid-July (Figure 5a). Thereafter, CH 4 fluxes declined to near zero a few days following removal of the flood from the field. In FR and FRCC, CH 4 emissions were significantly higher from the bottom versus the top section. CH 4 emissions from the FR bottom section were detected at the same time as MIRI but were later in July for the bottom section of FRCC (Figure 5a). Peak emissions of 640 and 390 g CH 4 -C ha −1 day −1 were observed from FR bottom and FRCC bottom, respectively, by the end of July after which time the fluxes began to decline. With the exception of two events when FR emissions were higher following reflooding of the field, the CH 4 fluxes fluctuated between −22 and 19 g CH 4 -C ha −1 day −1 for the top sections of the FR and FRCC treatments.

Cumulative Emissions, GWP, and Yield Scaled GWP
Cumulative seasonal CH4 emissions were significantly affected by irrigation treatment (p < 0.001), with the highest emissions from MIRI (Table 4). Seasonal CH4 Daily fluxes of N 2 O were low and fluctuated around zero fluxes in all field treatments except for a short period after N fertilization in FR and FRCC (Figure 5b). Nitrogen fertilizerinduced peak emissions were not observed from MIRI where daily emissions ranged from −3 to 10 g N 2 O-N ha −1 day −1 . In the FR and FRCC treatments, the largest N fertilizerinduced peak emissions ranged from 37 to 520 g N 2 O-N ha −1 day −1 and were observed one week after the second N application at pre-flood + nine days (Table 1). Among the different treatments, the highest N fertilizer-induced peak was observed from FR top and FRCC bottom and lowest from FRCC top. FRCC bottom emissions continued until the first week of July, that is, almost a week after the third N application. Outside of these periods, daily N 2 O fluxes from FR and FRCC were low and ranged from −6 to 21 g N 2 O-N ha −1 day −1 .

Cumulative Emissions, GWP, and Yield Scaled GWP
Cumulative seasonal CH 4 emissions were significantly affected by irrigation treatment (p < 0.001), with the highest emissions from MIRI (Table 4). Seasonal CH 4 emissions from the upper field sections were low for both FR and FRCC. The cumulative seasonal N 2 O emissions were lowest from MIRI and highest from FRCC bottom (Table 4). Combining the emissions from both the top and bottom field sections, FR and FRCC reduced CH 4 emissions by 74% and 93%, respectively, relative to MIRI. In contrast, both FR and FRCC exhibited increased N 2 O emissions as compared to MIRI (p < 0.001). Table 4. Cumulative CH 4 and N 2 O emissions, total global warming potential (GWP) and yield scaled GWP (GWP Y ) of rice from top and bottom sections of furrow rice (FR), furrow rice with cover crops (FRCC) and multiple inlet rice irrigation (MIRI) system of rice cultivation during cropping season of 2019. Data shown are average ± standard error (n = 5). Different letters denote statistical difference (p < 0.05) among treatments.

CH 4 Emissions (kg CH 4 -C ha
Total GWP (kg CO 2 eq. ha −1 Season −1 ) GWP Y (kg CO 2 eq. Mg −1 Season −1 ) CH 4 emissions dominated the global warming potential (GWP) calculated for MIRI, contributing about 96% of its total GWP. In contrast, N 2 O emissions dominated the GWP in row rice with (FRCC) and without (FR) cover crop, contributing about 94 and 71% for FRCC and FR, respectively. In terms of GWP, decreases in CH 4 emissions were offset by increases in N 2 O for both FR and FRCC. Overall, there were no significant differences in GWP. As a result of relatively similar grain yields and GWP emissions, there was no significant difference in yield scaled GWP (GWP Y ) across field treatments.

Discussion
This field experiment was designed to compare the yield-scaled GHG emissions from furrow-irrigated and continuously flooded rice systems on a commercial farm in eastern Arkansas. Though the three fields were located adjacent to each other, there were differences in soil texture, which may have a confounding effect on measured greenhouse gas fluxes [13,25]. In flooded rice systems, the interactions among water, fertilizer N, and GHG cycle are complex with many processes occurring at different levels. These differences complicate the underlying mechanisms or drivers contributing to net effects on CH 4 and N 2 O emissions [26,27]. It has been determined that the amounts of CH 4 and N 2 O emitted in paddy rice are mainly influenced by processes such as gas production, oxidation and transport from the soil to the environment. At the field level, other soil parameters such as soil texture, in addition to water soluble organic C, mineralizable N, irrigation, and N fertilization may increase or decrease net GHG emissions depending on dominant processes occurring in the soil [28][29][30]. However, at the ecosystem level, water and fertilizer N addition increase plant growth, which both increases C substrate supply for methanoptrophs and denitrifiers/nitrifiers and promotes larger aerenchyma cells for transport of CH 4 or N 2 O gas from the soil to the atmosphere. While various sets of effects dominate the overall GHG efflux in a flooded rice system, it is not the intent of this paper to study the effects of soil texture on CH 4 and N 2 O emissions. Our goal was to determine at the ecosystem level the net effect of row rice cropping practice on CH 4 and N 2 O emissions.

Grain Yield
A relatively poor and delayed rice stand in the FRCC field may have been due to the delayed termination of cover crops [31]. The time of cover crop termination is one of the biggest challenges for their successful field implementation. Termination of cover crops two weeks prior to crop planting is suggested to avoid the adverse effect of cover crop biomass on planting operations, physical impedance for crop emergence, shading and translocating the herbicide to nontarget plants [31,32]. In this study, the presence of cover crop biomass affected the perfomance of the air seeder during planting such that some rice seeds did not have adequate contact with soil, resulting in reduced germination. In addition, minimal soil tillage in FRCC fields in late spring may have influenced rice emergence [33]. Despite the reduced stand establishment, the rice grain yield from FRCC was similar to the national average of 8.2 Mg ha −1 in 2019 [34].
Row rice has been reported to have reduced grain yields mostly due to yield declines in the upper, nonflooded portions of fields [7]. While lower yields in the upper portion of fields were observed in the current study, the average yield for FR was similar to that of MIRI. The moisture content at 15-cm soil depth in FR remained wetter than the threshold limit of −20 kPa reported to influence grain yield [35]. Fertilizer N management is also a major issue in row rice production. The recommended N fertilizer rate for hybrid XP753 is 168 kg N ha −1 for flooded rice production in Arkansas [36]. The MIRI, FR and FRCC fields actually received 22, 47, and 86 kg N ha −1 , respectively, beyond this rate ( Table 1). The rough rice yields obtained from MIRI and FR were higher than the 2019 national average yield of 9.2 Mg ha −1 [34], but similar yields were reported from other Arkansas studies [36][37][38][39]. Moreover, the average grain yield for XP753 under continuously flooded regime was 11.6 Mg ha −1 [40].

Methane Emissions
Soil oxygen content is one of the most important factors influencing CH 4 production. The increase in soil oxygen content not only inhibits CH 4 but also reduces CH 4 emissions by stimulating CH 4 oxidation [41,42]. Therefore, water management practices such as alternate wetting and drying (AWD) that introduce oxygen to create aerobic soil conditions during the growing season are considered as a means to reduce rice CH 4 emissions [12,13]. Based on the water depth data of the present study, between two and four partial drying events were observed in the bottom of the FR and FRCC fields, respectively. These occurred exclusively during the vegetative stages of rice growth. The reduction in total cumulative CH 4 emissions of 50% from the bottom of FR as compared to MIRI was close to that reported by Balaine et al. [43] and global analyses by Jiang et al. [12] with similar AWD drying events. Similarly, the reduction of 93% from the bottom of the FRCC field as compared to MIRI was close to the 83% reduction reported by Linquist et al. [13] for multiple AWD dry-down events analyzed from the U.S. The upper field sections reduced CH 4 emissions by more than 94% as compared to bottom sections of both the FR and FRCC fields owing to the aerobic conditions created by nonflooded conditions during gas measurement period.
The cumulative seasonal CH 4 emission from MIRI was 44 kg CH 4 -C ha −1 , which is lower than the average seasonal CH 4 emissions reported for the Mid-South (146 kg CH 4 -C ha −1 ) by Linquist et al. [13]. However, cumulative CH 4 emissions ranging from 15 to 77 kg CH 4 -C ha −1 were reported from different studies conducted in Arkansas fields managed using a continuous flood [22,39,44].

Nitrous Oxide Emissions
Nitrous oxide is produced in soil by two major pathways: nitrification and denitrification. Both of these processes are strongly affected by soil water content and oxygen availability. N 2 O emissions are generally low in flooded rice fields as most of the nitrogen is lost as N 2 rather than N 2 O [27]. An introduction of aerobic conditions has often been reported to increase N 2 O emissions in flooded rice fields [12,45,46]. In this study, N 2 O emissions were higher in both FR and FRCC as compared to MIRI. The total N 2 O emissions were not different between FR and FRCC; however, there was a contrasting emissions pattern in top and bottom locations. Higher N 2 O emissions were observed in FR top and FRCC bottom as soil water condition varied in these treatments as compared to more constant soil water condition in FRCC top and FR bottom. Soil water-filled pore space above 60% was considered optimum for maximum N 2 O emissions [47]. Though soil porosity was not measured, it is likely that soil moisture conditions in FR and FRCC were sufficient to cause relatively high N 2 O emissions. The seasonal emissions from FR (4.4 kg N 2 O-N ha −1 ) were greater than the average seasonal N 2 O emissions reported for wheat and maize globally 1.44 and 3.01 kg N 2 O-N ha −1 , respectively [3]. However, it should be noted that higher seasonal emissions up to 4.6-4.9 kg N 2 O-N ha −1 have been reported for irrigated maize in the U.S. [48].
Higher N 2 O emissions from both FR and FRCC as compared to MIRI could also be due to higher application of fertilizer. N 2 O emissions peaked after each N fertilization event in FR and FRCC. The higher N fertilizer-induced N 2 O emissions in FR might be due to favorable soil water and warm conditions during fertilization. The pronounced effect of N fertilizer on N 2 O emissions was visible after the second N fertilization events, which coincided with rainfall. Cumulative rain of 36 mm was recorded for five days prior to GHG measurenents as per onsite manual rain-guage reading. Several studies have reported increases in N 2 O emissions following rainfall events coupled with the presence of high nitrate in the soil [48,49]. Notably, these N fertilizer-induced N 2 O emissions decreased in the later growing season despite favorable water content and warm weather due to the low amount of mineral nitrogen present in the soil (Figure 1). Similarly, LaHue et al. [50] and Balaine et al. [43] reported low N 2 O emissions with AWD in later growing season due to low mineral nitrogen present at the time of dry down.

GWP and Future Directions
The GWP from MIRI (1679 kg CO 2 eq ha −1 ) was within the range of GWP (573-3371 kg CO 2 eq ha −1 ) reported for flooded rice in Arkansas [44,45]. For the MIRI treatment, CH 4 dominated the total GWP in accordance with the previous results from flooded rice [12,[43][44][45]. In contrast, N 2 O emissions contributed most, that is, 71 and 94% to the total GWP for FR and FRCC, respectively. Despite the substantial decrease in CH 4 emissions in both FR and FRCC, GWP and GWP Y were similar to MIRI as higher N 2 O emissions from both these treatments offset decreases in CH 4 emissions. Therefore, improved N fertilizer management options are needed to reduce N 2 O emissions from FR. Future studies should focus on irrigation water and nitrogen management options that may potentially decrease N 2 O emissions [51,52] from rice. In this study, gas samples were not collected in the spring immediately after the termination of cover crops. The addition of residues from winter cover-crops termination may affect both CH 4 and N 2 O emissions [16,17] that need to be considered in the future studies.

Conclusions
This study is the first to present greenhouse gas emissions from commercial furrowirrigated rice fields in Arkansas, United States. Our results suggest that furrow irrigation is an effective means to reduce CH 4 emissions from traditionally flooded rice fields. However, furrow-irrigated rice also increased N 2 O emissions, which prevented an overall decrease in total GWP. Rice grain yield was not different in FR and MIRI. However, reduced rice stand establishment in FRCC reduced grain yield. Although our results were relatively benign and further research is required to optimize irrigation and N management, these suggest that growing lowland rice using furrow irrigation has potential to preserve grain yields while not increasing net GWP relative to continuously flooded rice. Taken together, these results suggest that increased N 2 O flux from furrow irrigated rice may not greatly detract from the potential benefits (e.g., improved flexibility for adapting to changing commodity markets and weather patterns; reduced soil disturbance and labor associated with rice levees) that furrow irrigation offers rice producers.