This field study was conducted from April to September at the Rice Research and Extension Center (RREC, 34.46°N,−91.46°W) near Stuttgart, Arkansas throughout the 2022 growing season. The soil series mapped throughout the study area was classified as DeWitt silt loam (fine, smectitic, thermic Typic Albaqualf), which is a poorly drained and slowly permeable soil formed from alluvial terrace deposits of the Lower Mississippi River during the Quaternary period (USDA-NRCS, 2014). The field has been under cultivation for the past 15 years, but, since 2014, has been managed exclusively in a furrow-irrigated rice-rice rotation. The study area encompassed a portion of the southwest corner of a larger 16-ha rice field. The study area consisted of 12 raised beds, 30 cm wide and 15 cm tall, with a furrow-to-furrow distance of 76 cm. The whole study area was 54-m long and 20-m wide. The study area was entirely prepared with a reduced tillage practice (i.e., stale-seedbed), hereafter referred to as no-tillage (NT), during the spring prior to planting each year's rice crop. The south-north-oriented slope within the study area was ~0.2%.

With a 30-year (1991 to 2020) mean annual precipitation and air temperature of 128.8 cm and 17.3°C, respectively, the climate in the region is classified as Humid Subtropical or Cfa by the Koppen Classification System (Arnfield, 2016). Archived meteorological data, specifically air temperature (°C) and precipitation (cm), for April to September from the past 10 years (i.e., 2012 to 2021) and for the study year (i.e., 2022) were also retrieved from a weather station located at the RREC just a couple of kilometers away from the study area (SRCC, 2023). Additionally, monthly 30-years normal (1991 to 2020) data were retrieved from April to September (NCEI, 2023). Monthly archived meteorological data, 30-years normal data, and 2022 data were summarized and the 6-month average for air temperature and 6-month total for precipitation were calculated. A relative percentage difference (RPD) was also calculated by subtracting the 10-years monthly, 30-years monthly normal, and 6-month values from the 2022 monthly and 6-month values and dividing the difference by their average.

A split-plot design was established with fertilizer source as the whole-plot factor in a randomized complete block design (RCBD) with time (i.e., sample date) as the split-plot factor. The six fertilizer treatments (i.e., TSP, ECST, CPST, DAP, ESN, and CT) were randomized within each of three blocks, for a total of 18 plots (Figure 1). Each plot had an area of ~ 2.3 m² and had an empty raised bed between adjacent treatments to avoid cross-contamination, except the two west-most, external plots that were positioned on adjacent raised beds out of necessity from limited available space in the field (Figure 1). The first block was located 15 m downslope from the upper crown of the field, while the second and the third blocks were 15 m downslope from the first and second blocks, respectively, for a total study area length of 54 m (Figure 1).

Fertilizer sources were characterized by different nutrient grades [i.e., TSP fertilizer grade: 0–46–0; ECST fertilizer grade: 5–37–0 (Anderson et al., 2021); CPST fertilizer grade: 6–27–0.1; DAP fertilizer grade: 18–46–0; and ESN fertilizer grade: 44–0-0]. Electrochemically precipitated struvite was created from synthetic wastewater of known P and N concentrations by the Chemical Engineering Department at the University of Arkansas as detailed in Della Lunga et al. (2023b), while CPST, under the tradename of Crystal Green, created via chemical precipitation from a municipal wastewater source near Atlanta, GA, was purchased from Ostara Nutrient Recovery Technologies (Vancouver, Canada). Environmentally Smart N represented a slow-release, polymer-coated urea product (Agrium Wholesales, Denver, CO), where the release of N is regulated by soil temperature (Nutrient, 2023).

On 14 September, 2021, 2 weeks after the 2021 harvest, rice crop residues throughout the entire 16-ha field were burned. On 15 September, 2021, a furrow-runner implement (Perkins Sales Inc., Bernie, MO) was used in the field to clean and open the furrows. On 29 September, 2021, a cover crop mix, consisting of 50% annual rye (Lolium multiflorum), 25% Australian winter peas (Pisum sativum subsp. arvense), 12.5% crimson clover (Trifolium incarnatum), and 12.5% radish (Raphanus sativus), was seeded at a rate of 28.0 kg seed ha⁻¹. On 29 April, 2022, the entire field was seeded with the hybrid rice cultivar “FP7521” (RiceTec, Alvin, TX) at a rate of 23.5 kg seed ha⁻¹. A larger portion of the field that included the study area was left untreated until 13 May, 2022 when P from the fertilizer treatments and zinc (Zn) as zinc sulfate were manually broadcast in each plot at equivalent rates of 14.7 and 11.2 kg ha⁻¹, respectively, except in the CT replicates that received Zn, but no P addition at any time. The plots that were going to be fertilized with ESN, received DAP as their P source to have balanced P additions among all fertilized plots. Due to the greater N concentration of DAP that provided 14.5 kg ha⁻¹ of N, plots that were going to be fertilized with TSP, ECST, and CPST and the control plots received a N application as N-(n-butyl) thiophosphoric triamide (NBPT)-coated urea (fertilizer grade: 46–0–0), hereafter referred to as coated urea, to balance the N content among all the fertilizer source treatments. Except in plots treated with ESN, all other plots were fertilized with coated urea as the N source. Plots did not receive any potassium (K) additions based on soil-test results and the University of Arkansas recommendations for rice production systems (UA-DA-CES, 2019).

On 2 June, 2022, the ESN plots were manually broadcast with the equivalent of 140 kg N ha⁻¹ as ESN as all upfront application, while the TSP, DAP, ECST, CPST, and CT plots received the equivalent of 37.0 kg N ha⁻¹ as coated urea as the first N application of a three-way split application recommended by the University of Arkansas for furrow-irrigated rice on a silt loam (Hardke, 2020b). On 9 and 16 June, 2022, all plots, except for the ESN plots, received the second and third split-N applications, respectively, as coated urea at the equivalent rate of 51.6 kg N ha⁻¹. Therefore, all the treatments received a total of 154.5 kg N ha⁻¹ according to University of Arkansas recommendations (UA-DA-CES, 2019).

The study area was furrow-irrigated approximately once a week with a 30-cm-diameter, polyvinyl chloride (PVC) poly pipe laid perpendicular to the slope at the upper end of the field. Once activated, irrigation was left on for more than 10 h in order to have the wetting front reach the end of the field, usually occurring within 4 h, and to allow enough time for the water to infiltrate laterally and vertically in the raised beds. A tail-water pump at the lower end of the field returned tailwater to the crown of the field when ~ 30 cm of ponded water was reached in the lower third of the entire field (Kandpal, 2018). Irrigation applications stopped on 16 June, 2022 for about a week to avoid straight head, as recommended by the University of Arkansas guidelines for rice production (UA-DA-CES, 2019). The irrigation system was supplied by a rain-fed reservoir located adjacent to the northwest corner of the field. At the end of the season, a total of 44 cm of water had been applied as irrigation.

Pesticides and herbicides were applied as-needed throughout the growing season according to University of Arkansas recommendations (UA-DA-CES, 2019). On April 30, a solution composed of residual herbicide Command (2-[2-chlorophenyl)methyl]-4,4-dimethyl-3-isoxazolidinone, 0.9 L ha⁻¹), Facet (3,7-dichloro-8-quinolinecarboxylic acid, 1.6 L ha⁻¹) and League (2-chloro-N-[[(4,6-dimethoxy-2-pyrimidinyl)-amino]carbonyl]imidazo[1,2-a]pyridine-3-sulfonamide, S-[(4-chlorophenyl)methyl] diethylcarbamothioate, 0.23 L ha⁻¹) was applied in the entire study area. On May 19, a solution composed of Clincher (cyhalofop: 2-[4-(4-cyano-2-fluorophenoxy) phenoxy] propanoic acid, butyl ester (R), 2.2 L ha⁻¹) and Zurax (3,7-dichloro-8-quinolinecarboxylic acid, 1.5 L ha⁻¹) and on June 5 a solution composed by Preface (0.3 L ha⁻¹), Gambit [Halosulfuron-methyl, methyl 3-chloro-5-(4,6-dimethoxypyrimidin-2-ylcarbamoylsulfamoyl)-1-methylpyrazole-4, 0.1 L ha⁻¹], and Rice One [N-(1-ethylpropyl)-3,4-dimethyl-2,6-dinitrobenzenamine, 2.2 L ha⁻¹] was applied to the entire study area.

Prior to individual plot and treatment establishment, five random sub-samples were collected from the top 10 cm of the raised beds in each of three predetermined blocks on 22 March, 2022 using a 2-cm diameter push probe and combined for one sample per block for initial particle-size and chemical analyses. An additional set of three soil samples, one in each block, were collected from the top 10 cm of the raised beds using a slide hammer and 4.8-cm-diameter core chamber for bulk density determination. Soil samples were oven-dried at 70°C for at least 48 h, ground, and sieved through a 2-mm mesh screen. Sand, silt, and clay percentages were determined using a modified 12-h hydrometer method (Gee and Or, 2002). Extractable soil nutrient (i.e., P, K, Ca, Mg, S, Na, Fe, Mn, and Zn) concentrations were determined with a Mehlich-3 extraction (i.e., 1:10 soil mass:extraction solution volume ratio) followed by analysis with inductively coupled, argon-plasma spectrophotometry (ICAPS; Tucker, 1992). Soil electrical conductivity (EC) and pH were measured in a 1:2 soil mass:water volume suspension with a potentiometer. Soil organic matter (SOM) and TC and TN concentrations were determined by weight-loss-on-ignition following combustion at 360^oC for 2 h and high-temperature combustion in a VarioMax CN analyzer (Elementar Americas Inc., Mt. Laurel, NJ), respectively. No effervesce upon soil treatment with dilute hydrochloric acid was observed, therefore all measured TC was considered organic C. Measured soil nutrient, TC, TN, and SOM concentrations were converted to contents (kg ha⁻¹) using the measured mean bulk density and the 10-cm soil depth. Soil C:N, N:P, and C:P ratios were calculated from soil contents.

Soil samples were collected from the top 10 cm of the raised bed within each plot at the end of the growing season on August 27, 2022 right before the drying period that precedes harvest. Following the same procedure described above, soil samples were prepared and analyzed for chemical properties to evaluate the soil chemical property change from beginning to end of the season (i.e., end minus beginning of the season).

Following procedures detailed in previous studies (Rogers et al., 2014; Smartt et al., 2016; Rector et al., 2018; Della Lunga et al., 2023a), on May 6, 2022, 18, 30-cm diameter PVC base collars were installed on top of the raised bed in each plot to a depth of 12 cm to allow free water movement through four, 12.5-mm-diameter holes drilled in each collar. Each base collar contained four plants as a portion of two adjacent rice rows. Gas sampling occurred approximately weekly between rice planting and harvest (i.e., 14, 21, 28, 35, 43, 49, 56, 63, 70, 77, 84, 91, 98, 105, 112, and 128 days after planting), for a total of 16 sampling dates.

Gas sample collection occurred on each sampling date between the same temporal window (i.e., 0800 h to 0900 h) following procedures used in numerous previous studies (Rogers et al., 2014; Smartt et al., 2016; Rector et al., 2018; Della Lunga et al., 2021a,c, 2023a). Gas sampling performed during the temporal window between 0800 and 1000 h represents the best time window to estimate daily GHG fluxes in rice production systems under flood- (Rogers et al., 2014; Smartt et al., 2016; Rector et al., 2018) and furrow-irrigated conditions (Della Lunga et al., 2021c). During the 1-h sampling period, gas samples were collected at 20-min intervals (i.e., 0 min, 20 min, 40 min, and 60 min) using the vented, non-flow-through, chamber method (Parkin and Venterea, 2010; Della Lunga et al., 2021a, 2023a). Additional details of the gas sampling procedures have been described previously in Della Lunga et al. (2021a, 2023a).

Gas samples were analyzed within 24 h of sample collection using a Shimadzu GC-2014 ATFSPL 115V gas chromatograph (GC; Shimadzu North America/Shimadzu Scientific Instruments Inc., Columbia, MD) equipped with a flame ionization detector (FID) for CH₄ and CO₂ and an electron capture detector (ECD) for N₂O. The slope of the best-fit linear regression line between the concentrations over the 20-min sampling intervals was used to calculate gas fluxes (μL m⁻² min⁻¹) on a plot-by-plot basis (Parkin and Venterea, 2010; Rogers et al., 2014; Smartt et al., 2016; Della Lunga et al., 2021a). Negative slopes were not retained nor considered in this study following protocols and methodologies developed and published in previous studies (Parkin and Venterea, 2010; Rogers et al., 2014; Smartt et al., 2016; Della Lunga et al., 2021a).

Seasonal emissions (kg ha⁻¹ season⁻¹) were calculated through linear interpolation between measured fluxes on sequential sampling dates on a plot-by-plot basis. Global warming potential (GWP) values were also calculated on a plot-by-plot basis using conversion factors to obtain CO₂ equivalents of 265 and 28 for N₂O and CH₄, respectively, according to the Intergovernmental Panel on Climate Change (IPCC) 6th assessment (IPCC, 2021). The 6th IPCC assessment adopted conversion factors for CH₄ and N₂O based on most recent studies and publications on GHGs (IPCC, 2021). Global warming potential was calculated including only CH₄ and N₂O due to the large, expected CO₂ response. Seasonal emissions (kg ha⁻¹ season⁻¹) for each GHG were divided by the measured chamber yield (Mg ha⁻¹ season⁻¹; described below) to calculate emissions intensity [EI; kg GHG (Mg yield)⁻¹]. The same procedure was used to calculate GWP emissions intensity [kg CO₂eq. (Mg yield)⁻¹]. The difference between raised beds and furrows with regard to GHG fluxes and emissions was not considered in this study, but future research should evaluate GHG emissions directly from both furrows and atop the raised beds to better estimate seasonal emissions at the field scale.

At the end of the growing season, on 11 September, 2022, plants were cut from inside the base collars at 2 cm from the soil surface, dried for 7 days at 55^oC, then weighed to obtain vegetative dry matter on a plot-by-plot basis. Rice grains were manually separated from the panicles and weighed to determine grain yield, which was corrected to 12% moisture content for reporting purposes. A subsample of vegetative tissue and grain from each plot was ground and sieved through a 1-mm mesh screen and analyzed for total N concentration by high-temperature combustion (VarioMax CN analyzer, Elementar Americas Inc., Mt. Laurel, NJ) and for total P, K, Mg, and Zn concentrations by ICAPS after strong acid digestion as described in previous studies (Tucker, 1992; Nelson and Sommers, 1996; Rector et al., 2018; Della Lunga et al., 2021a). Measured dry matter and nutrient concentrations were used to calculate nutrient uptakes on a plot-by-plot basis. Dry matter and nutrient uptakes were converted to Mg ha⁻¹ and kg ha⁻¹ respectively, for reporting purposes.

Based on a split-plot design with fertilizer treatment as the whole-plot factor in a RCBD arrangement and time as the split-plot factor, a two-factor ANOVA was conducted using the PROC GLIMMIX procedure in SAS (version 9.4, SAS Institute, Inc., Cary, NC) and using a gamma data distribution to evaluate the effects of time as days after planting, fertilizer source, and their interaction on N₂O, CH₄, and CO₂ fluxes. Based on the RCBD design of the whole-plot factor, a one-factor ANOVA was conducted using the PROC GLIMMIX procedure in SAS, using a gamma data distribution, to evaluate the effect of fertilizer source on seasonal GHG emissions and emissions intensities, GWP and GWP intensities, vegetative dry matter, grain yield, total aboveground dry matter (vegetative + grain dry matter), vegetative nutrient concentrations and uptakes, grain nutrient concentrations and uptakes, and total aboveground nutrient uptakes. A one-factor ANOVA was conducted using the PROC GLIMMIX procedure in SAS, using a normal data distribution, to evaluate the effect of fertilizer source on the soil property change, calculated as the end minus the beginning of the season. Externally studentized residuals were evaluated for each ANOVA performed and, as a result, no outliers were identified, thus no data points were removed for any tested variable. Homoskedasticity, normality, and independence of the residuals were visually checked. Significant treatment effects were judged at P < 0.05. When appropriate, means were separated with the least-significant-difference procedure at the alpha value of 0.05 (P < 0.05).

Meteorological parameters in 2022 depicted a growing season where air temperature was similar to, while precipitation was substantially divergent from the calculated 10-year average (2012–2021) and the 30-years normals for the study (Figure 2). From April to September 2022, monthly average air temperature had a RPD from the 10-years monthly average and the 30-years monthly normals < 7% (absolute value), except for May where the RPD was > 24% (absolute value). In April, May, and August, the air temperature in 2022 was, on average, lower than the respective 10-years monthly average and the 30-years monthly normals, while the opposite trend occurred in July (Figure 2). Averaged across the 6 months encompassing the 2022 growing season, the RPD for air temperature was < 6% (absolute value), highlighting how the 2022 year could be considered representative of the meteorological conditions reported within the region in the last 10 to 30 years (Figure 2).

In contrast to air temperature, precipitation during the 2022 growing season was lower than the 10-years monthly averages or 30-years monthly normals (Figure 2). Averaged across the 6 months encompassing the 2022 growing season, rainfall in 2022 was less than half of the 10-years average and the 30-years normals (Figure 2). However, when the amount of applied irrigation in 2022 (i.e., 44 cm) is added to the 6-month precipitation, the total amount of water that the field received was much closer to the 10-years average and 30-years normals (Figure 2). Irrigation is commonly used in agricultural settings to limit and reduce plant moisture stress and is applied in consideration of rainfall additions. Therefore, considering the combination of air temperature, precipitation, and applied irrigation, the 2022 growing season was reasonably representative of the typical regional meteorological conditions (NCEI, 2023) without having to repeat the field study during the subsequent growing season.

Soil physical and chemical properties were assessed at the beginning of the growing season, before any fertilizer treatment was applied. Throughout the entire study area, a textural class of silt loam was confirmed for the top 10 cm, with average sand, silt, and clay of 16%, 72%, and 12%, respectively (Brye et al., 2023). As described in Brye et al. (2023), all near-surface soil properties, except for pH, were within optimal ranges for furrow-irrigated rice production in Arkansas (Table 1). Soil K (131 to 175 mg kg⁻¹), Ca (< 400 mg kg⁻¹), Mg (< 30 mg kg⁻¹), S (> 10 mg kg⁻¹), Mn (> 40 mg kg⁻¹) and Zn (> 4.1 mg kg⁻¹) concentrations were within optimal soil-test ranges for rice cultivation according to University of Arkansas recommendations (Norman et al., 2013; UA-DA-CES, 2019; Table 1).

In contrast to many extractable soil nutrients, the mean measured soil pH (i.e., 5.3) was close to the lower limit (i.e., 5 to 6.5) for optimal rice cultivation (Havlin et al., 2014). Although the soil pH under flooded conditions tends to migrate close to neutrality due to the presence of the flood and resulting chemical interactions, in a furrow-irrigated setting, acidic soil conditions can persist and limit nutrient availability, specifically P, where P remains bound to Al and/or Fe oxides (Stevenson and Cole, 1999; Havlin et al., 2014). The extractable Fe content in the study area was lower than what was measured at the down-slope end of the same field in 2018 and 2019, where the presence of ponded water and reducing conditions likely resulted in the dissolution of some of the non-readily-available P (Della Lunga et al., 2021a). However, soil pH measured in the current study was within the range of values measured at the beginning of the 2018 and 2019 growing seasons in the same furrow-irrigated field (Della Lunga et al., 2021a,b), suggesting that soil-fertility-related parameters associated with furrow-irrigated rice represent a substantial deviation from under flood-irrigated rice conditions.

Despite the low soil pH status, extractable soil P was within the upper limit of the medium soil-test level (17 mg kg⁻¹ to 25 mg kg⁻¹) according to the University of Arkansas soil-test evaluations, which, in conjunction with the low soil pH, indicated a plant response from fertilizer-P applications was expected, but not definite (Table 1; UA-DA-CES, 2019). The near optimum soil-P level in the top 10 cm throughout the study area at the beginning of the growing season could have represented a limiting factor in the evaluation of fertilizer treatments. Furthermore, as rice roots developed, soil P from below the top 10 cm was easily accessible to provide an additional source of plant-available P that was not assessed at the beginning of the study. However, assessing the near-surface, soil-test P levels is common practice when deciding on a fertilizer-P recommendation.

Soil TN and TC were within the range reported by Roberts et al. (2012) from silt-loam soil series across the state of Arkansas in the top 15 cm (Table 1). Soil organic matter exceeded what is considered a minimum desirable level (2.5%) for crop production in Arkansas (Espinoza et al., 2018). The calculated C:N and both N:P and C:P ratios indicated that, upon fertilizer-N and -P application, rapid mineralization of inorganic N and microbial P immobilization were expected, respectively, to facilitate the evaluation of fertilizer sources in this study (Li et al., 2016; Sheng et al., 2022; Table 1).

Seasonal emissions in the current study need to be considered an overestimation of the actual net production of GHGs from the field to the atmosphere, due to the interpolation method and to the limits and methodological biases of the vented, non-flow-through, chamber approach as detailed by Parkin and Venterea (2010) and Wood et al. (2013). Interpolation techniques assume a constant relationship, often linear, between the interpolated values, which may not accurately represent the flux pattern between measurement dates due to heterogeneity in temporal fluxes (Wood et al., 2013). Regardless of the type of model chosen for flux interpolation, an interpolation bias is always assumed (Seok-In et al., 2013; Wood et al., 2013). The lack of fixed temporal and spatial patterns in GHG fluxes renders the choice of location in a field, time of day of the measurements, and duration of sampling as relevant parameters to capture average or representative fluxes rather than flux maxima or minima (Wood et al., 2013). As hypothesized, seasonal CH₄ and CO₂ emissions did not differ (P > 0.05), while N₂O differed (P = 0.02) among fertilizer treatments (Table 3). Due to the different techniques used to measure and/or calculate fluxes and seasonal emissions and due to the lack of the temporal component associated with emissions, fluxes and emissions depict results that need to be evaluated separately.

Seasonal CH₄ emissions ranged from 0.62 kg ha⁻¹ season⁻¹ from CPST to 1.34 kg ha⁻¹ season⁻¹ from ESN, with an overall mean of 1.0 kg ha⁻¹ season⁻¹, while seasonal CO₂ emissions ranged from 20,547 kg ha⁻¹ season⁻¹ from CT to 28,922 kg ha⁻¹ season⁻¹ from TSP, with an overall mean of 23,977 kg ha⁻¹ season⁻¹ (Table 3). Soil macronutrients close to the optimal levels for rice production and the equal amount of fertilizer-P and -N applications across treatments throughout the growing season created a similar amount of substrate for microbial and phytological processes, resulting in no seasonal CH₄ or CO₂ emissions differences among fertilizer treatments (Table 3). While the generally visibly dry-soil conditions in the study area likely limited methanogenic activity, the amount of nutrients provided by the topsoil and fertilizers likely enhanced overall soil respiration, potentially resulting in relatively large CO₂ emissions (Paustian et al., 1998; Della Lunga et al., 2023a; Table 3). Additionally, the relationship between CH₄ emissions in rice fields and soil-P levels is strictly related to the rhizosphere activity (Lu et al., 1999), where methanogens are often concentrated. The methanogenic microorganisms associated with plant roots in rice production systems may have adapted to low-P conditions, partially explaining the inhibitory effect on CH₄ production at larger soil-P levels (Conrad et al., 2000), as describes the field conditions of the current study. Soil-P conditions close to optimal levels may have caused decreased root exudation of organic C, specifically organic acids, providing less C substrate for methanogenic activity (Adhya et al., 1998). The availability of P in different ecosystems has often been correlated to the rate of microbial metabolic activity, without affecting abundance and composition of the microorganisms (Hui et al., 2019). Although characterized by different solubilities, all of the fertilizer-P treatments provided equal total nutrient amounts, partially explaining the relatively large CO₂ emissions in the current study and the lack of differences among fertilizer-P treatments.

Seasonal CH₄ and CO₂ emissions reported in the current study were numerically lower and greater, respectively, than results reported by Della Lunga et al. (2023a) from the up-slope portion of a furrow-irrigated field on a silt-loam soil under NT management in east-central Arkansas. However, CH₄ emissions were similar to those reported by Karki et al. (2021) from the upper portion of a furrow-irrigated rice field on a silty-clay soil in northeast Arkansas. Contrary to that hypothesized, N₂O emissions were greater from CT (i.e., 5.97 kg ha⁻¹ season⁻¹), which did not differ from ECST, and were lowest from ESN (1.50 kg ha⁻¹ season⁻¹), which did not differ from TSP, CPST, ECST and DAP (Table 3). The slower release of N from ESN compared to coated urea likely limited the amount of substrate available for nitrification-denitrification processes, while, in the CT treatment, the lack of fertilizer-P application likely limited plant growth and plant nutrient uptake, leaving a greater amount of N in the soil available for microbial processes and N₂O production (Table 3). Nitrous oxide emissions in the current study were at least two times lower than previously reported from furrow-irrigated rice fields on a silt-loam soil in east-central (Slayden et al., 2021) and on a silty-clay soil in northeast Arkansas (Karki et al., 2021).

As hypothesized, rice yield and CO₂ and CH₄ EI did not differ (P > 0.05) among fertilizer treatments (Table 3). Rice yields ranged from 5.53 Mg ha⁻¹ from ECST to 9.85 Mg ha⁻¹ from CPST, with an overall mean of 8.33 Mg ha⁻¹, which was nearly identical to the 8.31 Mg ha⁻¹ state average rice yield in 2022 across all rice production systems in Arkansas (USDA-NASS, 2023). The lack of a significant difference between ECST and CPST was related to a large degree of variability among replicates that resulted in an overall large confidence interval (Table 3).

Field studies are often exposed to environmental conditions that depict spatial variability, even within short distances, resulting in differential plant responses among replicates of the same treatments, as in the current study. The lack of significant differences in yield among fertilizer-P treatments indicated a similar plant response between all the plots, which was likely related to the near-optimum initial soil nutrient status throughout the study area and the equal total amounts of nutrients provided by the different fertilizers. However, the similar yields among treatments reinforced the validity of the slow-release fertilizers and emphasized the role of fertilizer solubility when significant differences among fertilizer-P treatments occurred, such as for N₂O emissions (Table 3).

Similar to rice yield, CH₄ and CO₂ emissions intensities (EI-CH₄ and EI-CO₂, respectively) were unaffected by fertilizer treatment. Methane EI ranged from 0.06 kg CH₄ (Mg yield)⁻¹ from CPST to 0.19 kg CH₄ (Mg yield)⁻¹ from ECST, with an overall mean of 0.13 kg CH₄ (Mg yield)⁻¹, while EI-CO₂ ranged from 2,344 kg CO₂ (Mg yield)⁻¹ from CT to 4,434 kg CO₂ (Mg yield)⁻¹ from ECST, with an overall mean of 3182 kg CO₂ (Mg yield)⁻¹ (Table 3). Estimated EIs in the current study were 900% and 30% lower for CH₄ and CO₂, respectively, than those reported by Della Lunga et al. (2023a) from the NT/up-slope tillage treatment-site position combination in a furrow-irrigated rice field on a silt-loam soil in east-central Arkansas.

In contrast to CH₄ and CO₂, but similar to seasonal N₂O emissions, EI-N₂O differed (P = 0.04) among fertilizer treatments (Table 3). As for seasonal emissions, and contrary to that hypothesized, EI-N₂O was greater from CT [0.76 kg N₂O (Mg yield)⁻¹], which did not differ from ECST, and was lowest from ESN [0.19 kg N₂O (Mg yield)⁻¹], which equaled values from TSP, CPST, and DAP (Table 3). The significantly greater N₂O emissions from CT and lower emissions from ESN, coupled with the numerically similar yields among treatments, were responsible for the EI-N₂O results (Table 3), where the statistical differences were likely due to similar reason as for seasonal N₂O emissions (Table 3).

Global warming potential was calculated using only seasonal CH₄ and N₂O emissions, excluding CO₂ from the GWP calculation. As reported in previous studies, the magnitude and variance of CO₂ emissions can mathematically absorb and mask trends and differences in CH₄ and N₂O contributions to GWP [Intergovernmental Panel on Climate Change (IPCC), 2014; Rector et al., 2018; Della Lunga et al., 2023a]. Additionally, due to the difficulty in controlling soil respiration, mitigation approaches primarily focus on CH₄ and N₂O regulation [Paustian et al., 1998; Intergovernmental Panel on Climate Change (IPCC), 2014].

Global warming potential and GWP intensity differed (P < 0.05) among fertilizer treatments (Table 3). Contrary to that hypothesized, GWP was greatest from CT (1612 kg CO₂ eq. ha⁻¹ season⁻¹, standard error = 415), which did not differ from ECST, and was lowest from ESN (436 kg CO₂ eq. ha⁻¹ season⁻¹, standard error = 112), which did not differ from TSP, ECST, CPST, and DAP (Table 3).

The difference in GWP estimates highlighted in the current study were mainly related to N₂O's contribution to GWP (GWP-N₂O; Table 4). With factors from the 6th IPCC assessment, GWP-N₂O across fertilizer treatments was not lower than 90%, highlighting how upland rice cultivation should be mainly managed for N-loss reductions (Table 4). The GWP from CT was between 3.6 times numerically or significantly greater than from the other fertilized treatments (Tables 3, 4). Global warming potential estimates from the current study were substantially lower than those reported from flooded rice paddies in South Korea (9725 kg CO₂ eq. ha⁻¹ season⁻¹), water-saving conditions (2114 kg CO₂ eq. ha⁻¹ season⁻¹; Ahn et al., 2014), continuous irrigation (4380 kg CO₂ eq. ha⁻¹ season⁻¹), and intermittent irrigation (3352 kg CO₂ eq. ha⁻¹ season⁻¹; Ali et al., 2013), and from NT paddies in Spain under flooded conditions (8477 kg CO₂ eq. ha⁻¹ season⁻¹) and sprinkler irrigation (1784 kg CO₂ eq. ha⁻¹ season⁻¹; Fangueiro et al., 2017). However, GWP estimates from the current study were similar to those from rice paddies in US under alternate-wet-and-dry irrigation (1671 kg CO₂ eq. ha⁻¹ season⁻¹; Linquist et al., 2015) and from cotton (Gossypium) fields in China under furrow irrigation (410 kg CO₂ eq. ha⁻¹ season⁻¹; Wu et al., 2014). These results highlight how water and irrigation management represent an effective tool to mitigate GHGs and GWP across upland and lowland crops and climatic regions.

Similar to GWP, but contrary to that hypothesized, GWP intensity was greatest from CT for the 6th [204.2 kg CO₂ eq. (Mg yield)⁻¹] IPCC assessment and did not differ from ECST (Table 3). However, in contrast to GWP, GWP intensity was lowest from CPST from the 6th [52.1 kg CO₂ eq. (Mg yield)⁻¹] IPCC assessment and did not differ from TSP, DAP, and ESN (Table 3). The combination of numerically greater yield and lower seasonal N₂O emissions from CPST and ESN resulted in a relatively low GWP intensity, suggesting that slow-release fertilizers could be an effective mitigation tool to reduce GHG emissions (Table 3). Although ECST is generally considered a slow-release fertilizer in the absence of substantial evidence to the contrary, the crystalline-flaky constitution of ECST could have affected ECST's solubility and soil behavior, due to the greater surface area of the ECST crystalline flakes compared to the pelletized fertilizers that can enhance dissolution processes, suggesting that further studies should evaluate the behavior of ECST mass-produced in pelletized form. Global warming potential intensities from the current study were similar to those reported in previous research from furrow-irrigated rice fields in Arkansas (Karki et al., 2021; Slayden et al., 2021; Della Lunga et al., 2023a).

Plant nutrient concentrations and uptakes reflect the nutrient status at the end of the growing season at harvest time and are intended to represent the culmination of uptake and translocation processes (Masclaux-Daubresse et al., 2010). However, the portion of macro- and micronutrients coming from the fertilizer applications and the portion coming from the pedosphere was not independently determined, meaning that plant nutrient concentrations and uptakes reported in the current study have been supplied collectively by the native soil's nutrient pool and by the inorganic fertilizer additions and cannot be used to determine fertilizer-nutrient-use efficiency (Singh et al., 2021).

As hypothesized, all aboveground plant properties measured in the study were unaffected (P > 0.05) by fertilizer treatment. The near-optimum initial soil nutrient levels, specifically soil P, in the top 10 cm, coupled with the plant roots' ability to secure nutrients from below the top 10 cm, which was not assessed early in the growing season, likely provided sufficient nutrient supplies and buffered the rice plants from any nutrient shortfall to minimize significant differences in plant response among the various fertilizer treatments. Although the initial, near-optimum soil-test P status of the study area could have masked the effects of the fertilizer treatments, the current project still provided useful information and guidance for agronomic and environmental management practices.

Averaged across fertilizers, aboveground tissue N, P, K, Mg, and Zn concentrations were 0.92, 0.07, 1.97, 0.18%, and 188 mg kg⁻¹, respectively. Plant tissue concentrations in the current study were within or above the optimal levels of N (0.65% to 0.8%), K (1.55% to 2.0%), Mg (0.25% to 0.3%), and Zn (25 mg kg⁻¹ to 50 mg kg⁻¹) in rice vegetative tissue at maturity (Linquist, 2020). However, plant tissue P concentration was below the optimal P level for rice tissue at maturity (0.1 to 0.15%) but was above the critical level for deficiency (< 0.06%; Linquist, 2020). Vegetative dry matter (VDM) across fertilizer treatments averaged 11.2 Mg ha⁻¹ and, together with measured tissue concentrations, N, P, K, Mg and Zn uptakes averaged 99.8 kg ha⁻¹, 8.1 kg ha⁻¹, 216 kg ha⁻¹, 19.4 kg ha⁻¹, and 2.11 kg ha⁻¹, respectively. Rice grain nutrient concentrations averaged 1.54% N, 0.24% P, 0.28% K, 0.10% Mg, and 37.5 mg Zn kg⁻¹ across fertilizer treatments. Measured rice yields, coupled with measured grain concentrations, resulted in mean nutrient uptakes of 125 kg ha⁻¹, 20.2 kg ha⁻¹, 23.4 kg ha⁻¹, 8.1 kg ha⁻¹, and 0.31 kg ha⁻¹ for N, P, K, Mg, and Zn, respectively. Combining VDM and grain uptakes, total aboveground nutrient uptakes across all treatments averaged 224 kg ha⁻¹, 28.3 kg ha⁻¹, 240 kg ha⁻¹, 27.5 kg ha⁻¹, and 2.42 kg ha⁻¹ for N, P, K, Mg, and Zn, respectively. Although to date, to the authors' knowledge, no studies on furrow-irrigated rice have reported optimal nutrient concentration in vegetative and grain tissues, the macro- and micro nutrient uptakes reported in the current study were, in all instances, near the upper limit of agronomic evaluations for upland rice on silt-loam soils in Arkansas (Chlapecka, 2021).

The change in soil properties throughout the 2022 rice growing season was calculated as end-of-season minus beginning-of-season properties on a plot-by-plot basis. A positive difference in soil properties indicated an accumulation process, while a negative difference indicated a removal (i.e., plant uptake/consumption) process and/or chemical alteration. The accumulation and removal processes occurred as a combination of biotic and abiotic mechanisms. However, the specific processes that resulted in accumulation or removal were not determined.

Except for soil pH, and contrary to that hypothesized, changes in all other measured soil properties in the top 10 cm throughout the 2022 growing season did not differ (P > 0.05) among fertilizer treatments (Table 5). However, soil pH had a positive difference, thus soil pH increased (P < 0.01), over time in all treatments, indicating an accumulation of hydroxyl radicals (OH⁻) or a consumption of hydrogen ions (H⁺) across all fertilizer treatments (Table 5). Soil acidification is mainly due to the export of hydrogen ions by rice roots in order to balance the intake of cations from the soil solution, a balance that is related to the form of N taken up (Nye, 1986). The largest soil pH increase occurred from ESN (0.64 units), which did not differ from CPST and CT, while the lowest soil pH increase occurred from ECST (0.09 units), which did not differ from TSP and DAP (Table 5). In all fertilizer treatments, soil conditions remained acidic throughout the growing season, a condition substantially divergent from what has been reported under flooded-soil conditions, where soil pH tends to increase and achieve near-neutral conditions [Intergovernmental Panel on Climate Change (IPCC), 2014]. Increasing soil pH can result in a greater availability of acetate and rhizodeposition that provide substrate for methanogenic activity (Ye et al., 2013). The large soil pH increase from ESN achieved values closer to the optimum pH range for methanogen activity (i.e., 6.5 to 7.5), likely explaining the numerically greater CH₄ emissions from ESN compared to the other fertilizer treatments (Tables 3, 5).

Among the soil properties considered in this study, BD and extractable soil K, Ca, and Fe, and TN, TC, and SOM contents decreased over the growing season in all treatments, by means of −0.06 g cm⁻³ and −63.1 kg ha⁻¹, −38.2 kg ha⁻¹, −156 kg ha⁻¹, −204 kg ha⁻¹, −1437 kg ha⁻¹, and −1690 kg ha⁻¹, respectively (Table 5). In contrast, extractable soil Na and Mn increased over the growing season in all treatments (i.e., 59.5 kg ha⁻¹ and 119 kg ha⁻¹, respectively; Table 5). Soil EC and extractable soil P, Mg, S, and Zn experienced increases and decreases over time across treatments, with overall means of 0.018 ds m⁻¹ and −1.97 kg ha⁻¹, −2.2 kg ha⁻¹, 2.9 kg ha⁻¹, and 3.1 kg ha⁻¹, respectively (Table 5). Soil properties changes over time were significantly different than a change of zero (P < 0.05) in all treatments for extractable soil Fe and Mn and TN, but did not differ from a change of zero (P > 0.05) in all treatments for soil EC and extractable soil P, Mg, and Zn. Additionally, soil BD differences did not differ from zero (P > 0.05) in all treatments, except for DAP where soil BD decreased over time, while extractable soil S differences also did not differ from zero (P > 0.05) in all treatments, except for TSP where extractable soil S increased over time (Table 5). Extractable soil Ca differences did not differ from zero (P > 0.05) in DAP, CPST, and ESN and soil pH differences did not differ from zero (P > 0.05) in DAP, ECST, and TSP (Table 5). Soil OM differences did not differ from zero (P > 0.05) in CPST, ESN, and CT and soil Na differences did not differ from zero (P > 0.05) in DAP, ESN, and TSP (Table 5). Soil TC differences did not differ from zero (P > 0.05) in CPST, ESN, and TS and soil K differences did not differ from zero (P > 0.05) in DAP and ECST (Table 5).

While the decreasing temporal trends measured for extractable soil P, K, Ca, and Mg can be attributed to plant uptake processes (Stevenson and Cole, 1999; Julia et al., 2016; Dhillon et al., 2019), similar temporal trends for extractable soil Fe and TN, TC, and SOM were indicative of predominantly aerobic soil conditions, where redox-active minerals are oxidized, inorganic minerals are mineralized, and organic material is rapidly decomposed (Table 5; Ponnapempura, 1972). The accumulation of soil Mn likely was related to the acidic conditions that can solubilize Mn into soil solution even in aerobic conditions (Goth and Patrick, 1972). The lack of differences in soil properties changes among fertilizer treatments was likely due to the large initial soil nutrient levels at the beginning of the growing season and to the analysis being limited to only one growing season. The wet and dry cycles that commonly occur in the up-slope portion of furrow-irrigated rice fields (Della Lunga et al., 2020b) likely contributed to the accumulation of extractable Na in the topsoil as evaporation water losses left salts, particularly Na, behind to be concentrated in the near-surface soil (Table 5). Changes in soil properties over the course of the growing season can provide useful information to develop management practices tailored to furrow irrigation settings. However, further studies should evaluate the long-term effects of fertilizer-P and -N sources on agronomic and environmental soil parameters.