The study was conducted on the Research Field of the Department of Soil Science at Bangladesh Agricultural University (BAU) farm, Mymensingh (24°71.60′N, 90°42.51′E), Bangladesh, that has been under CT for a long time with negligible aboveground crop residue incorporation (see also Uddin et al. (2021)). Since 2018, an annual rice–rice–rice pattern, a common cropping sequence on the floodplain soils of Bangladesh, has been followed. The farm site, in the Old Brahmaputra Floodplain, has a subtropical monsoon climate with a mean annual temperature of 26°C, average annual rainfall of 1,800 mm, and relative humidity of 85% (Weather Yard, BAU). The mean monthly rainfall and temperature during the rice cropping season were 127 mm and 25.4°C in 2019, respectively, and 171 mm and 26.8°C in 2020, respectively (Uddin et al., 2021). It is notable that the temperature and rainfall in this subtropical region has been relatively stable over the years. The field site with non-calcareous dark gray floodplain soil (Aeric Haplaquept in the U.S. Soil Taxonomy) was moderately drained with silt loam texture and had near neutral pH (6.5). At the onset of the experiment, soil organic carbon (0–15 cm depth) was 11.3 g kg⁻¹, total N was 1.2 g kg⁻¹, available P was 3.2 mg kg⁻¹, exchangeable K was 0.04 g kg⁻¹, and available S was 10.5 mg kg⁻¹ (Uddin et al., 2021).

An annual sequence of three rice crops (Oryza sativa L), called hereafter Boro rice, transplanted Aus rice, and transplanted Aman rice, has been cultivated since 2018. The first crop of the sequence, Boro rice, was grown from January to May (mid-winter to pre-monsoon season), followed by transplanted Aus rice as a rain-fed crop from June to August (monsoon), and then transplanted Aman rice from August to December (late monsoon to winter). In the last week of January, land preparation began for transplanting Boro rice seedlings, followed by transplanting of Aus rice in the last week of May and transplanting of Aman rice seedlings in late August. Three seedlings were transplanted per hill with a 20 cm × 20 cm spacing. The variety for Boro rice was BRRI dhan63. The recommended dose (RD) of fertilizers for this soil was 144 kg N, 21 kg P, 60 kg K, 8 kg S, and 1.5 kg Zn ha⁻¹ for Boro rice; 72 kg N, 7 kg P, 40 kg K, and 3 kg S ha⁻¹ for Aus rice; and 90 kg N, 8.5 kg P, 50 kg K, 4 kg S, and 1 kg Zn ha⁻¹ for Aman rice. N, P, K, S, and Zn were applied as urea, triple super phosphate, muriate of potash, gypsum, and zinc sulfate, respectively. Urea was applied in three equal splits at early tillering, active tillering, and panicle initiation stages on days 10, 35, and 55, respectively, after transplanting as a conventional application practice. Glyphosate (Round up®; ACI Bangladesh Ltd.), a non-selective herbicide, was sprayed over the field at a rate of 1.85 kg ha⁻¹ 3 days before the final land preparation. In addition, Pretilachlor (Superhit®, post emergence herbicide) was used at a rate of 450 g ha⁻¹ 7 days after transplanting the rice seedlings. Brifer 5G and Cidial 5G (ACI Bangladesh Ltd.) were applied as required to control rice insects. The rice fields were irrigated a day before the final land preparation and then when necessary to maintain standing water at about 3 cm above the soil surface throughout the growing season.

In July 2018, the experiment was initiated with two soil disturbance levels (strip tillage, ST and conventional tillage, CT), two crop residue retention levels [low residue (LR): 15% and high residue (HR): 40%, by height], and three N fertilization rates [108 (N1), 144 (N2), and 180 (N3) kg N ha⁻¹], with 144 kg N ha⁻¹ being the recommended N fertilizer rate (FRG, 2018). In the ST system, soil was not puddled and left undisturbed except for a rotary tillage of 3 cm furrows for seeding or transplanting separated by 20 cm of undisturbed soil between rows using a versatile multi-crop planter (Haque et al., 2016). In the ST system, 15 and 40% residues of the previous crop were left standing, whereas in the CT system, the same amounts of residue were incorporated into the soil by repeated puddling using a rotary tiller. The 15% retention was comparable to present farmers’ practices. The experiment was laid out in a split-plot design, established with three replications for each treatment combination (Supplementary Figure S1). Rice soil is managed to develop a plough pan-limiting water percolation and increasing water retention capacity. The conversion of CT to ST was found to increase the water percolation rate due to lack of puddling. Therefore, the required irrigation interval was lower in ST than in CT to maintain a similar amount of standing water or water content in both treatments. For the Boro season, the water table depth is generally between 1 and 2 m below the ground level. Tillage treatment was assigned to the main plots, residue to the sub-plots, and fertilizer on top of the residue plots (sub-sub plot). The size of each sub-sub plot was 10 m × 4.2 m with a separating bund between the plots. The total number of plots was 36, including two tillage × two residue levels × three N rates × three replications.

The GHG field measurements were conducted in Boro rice fields during the fifth crop after the initiation of the experiment in 2018 via a static chamber method (Hutchinson and Mosier, 1981). The observation period was started from the first split application of urea under continuous flooded condition and continued until the emissions declined to background levels (ambient concentration). Soda glass chambers (40 cm × 40 cm wide and 50 cm high) with stainless steel collars were placed on top of the rows, covering eight rice plants, to a depth of 10 cm (Zaman et al., 2021). Each collar had a neoprene seal which ensured an air-tight connection between the chamber lid and the frame. Urea was applied by the broadcast application method inside the pre-installed collars of the gas-collecting chambers. At each sampling event, the lids were placed on collars and gas samples were collected through the air tight rubber septa using a 20-ml polypropylene syringe equipped with a 25-gauge Luer lock needle at 0, 30, and 60 min. A 16-ml gas sample was collected from the headspace and injected into a pre-evacuated 12-ml vial (Labco Wycom Ltd.) to over pressurize the vials. The gas samples were collected on days 0, 1, 3, 5, 7, 10, 15, and 21 and repeated three times at 0, 30, and 60 min after chamber installation during the day between 10:00 a.m. and 4:00 p.m. after each split application of urea, deemed as the most representative time of a 24-h cycle (de Lima et al., 2018). The sample vials were stored for up to 7 days before analysis on a Varian 3,800 gas chromatograph (CP-3800, Varian, Inc., Switzerland) equipped with an electron capture detector (N₂O) using argon (Ar) as the carrier gas, a thermal conductivity detector (CO₂) using N₂ as the carrier gas, and a flame ionization detector (CH₄) using N₂ as the carrier gas.

Methane and N₂O emissions at each sampling time were calculated from the change in headspace concentration based on a linear regression of measurements over 60 min Eq. 1 F l u x = d G a s d t ∗ 10 x ∗ V c h a m b e r ∗ p ∗ 100 ∗ M W R ∗ T ∗ 10 y ∗ 1 A , (1) where dGas in ppb is the concentration change over time, dt is the difference in time, 10^x is recalculation (here 10^–9), V _chamber is the volume of the chamber used, p is the atmospheric pressure in Pa (100 is to convert Pa to hPa), MW is the molecular weight of CH₄-C or N₂O-N, R is the gas constant 8.314 J mol⁻¹ K⁻¹, T is the temperature in Kelvin, 10^y is recalculation (here 10⁶ (µg)), and A is the area of the chamber. Emissions for CT and ST were averaged across the residue and N-input rates and for LR and HR across CT and ST systems and N input rates. The cumulative CH₄ and N₂O emissions were calculated by summing up all daily fluxes for the entire experimental period by linear interpolation between the sample points (Zhang et al., 2013). Seasonal cumulative CH₄ and N₂O emissions were estimated following the method proposed by Mosier et al. (2006).

The N₂O EFs were calculated based on the method proposed by Huang et al. (2017) Eq. 2, assuming that N₂O emissions from a true 0 N control treatment was negligible. E F =   N 2 O   e m i s s i o n s   f r o m   f e r t i l i z e r   N   t r e a t m e n t A p p l i e d   f e r t i l i z e r   N , (2) where EF = emission factor, N ₂ O emissions from fertilizer N treatment are in kg N ha⁻¹ season⁻¹, and applied fertilizer N is in kg N ha⁻¹ season⁻¹.

The yield-scaled GHG gas emissions (kg per t of grain) were estimated by dividing the cumulative CH₄ or N₂O emissions (kg CH₄-C or N₂O-N ha⁻¹) by the grain yield (t ha⁻¹). For calculating the area-based net global warming potential (GWP), the combined seasonal cumulative emissions of CH₄ and N₂O were converted to their CO₂ equivalent (Ahmed et al., 2009; Hou et al., 2012). The GWP (over 100 years) conversion parameters used for CH₄ and N₂O were 34 and 298 kg ha⁻¹ CO₂ equivalents, respectively [IPCC (Intergovernmental Panel on Climate Change), 2013; Wang et al., 2015]. The yield-scaled global warming potential (GWPY), a metric that assesses the GWP per unit of yield, was calculated following the method proposed by van Groenigen et al. (2010).

Composite soil samples were collected from each replicated plot at 0–15 cm depth with an auger at 4 days after the second split application of urea (active tillering stage) which corresponded very well to the maximum peak of N₂O emissions. Soil samples were collected from several spots in a plot adjacent to each GHG gas sampling chamber and stored in sealable plastic bags in a cold room at 4°C. The soil pH was measured in the field during GHG sampling using a portable pH meter (Direct Soil pH Portable Meter, HI12923; Hanna Instruments). A portion of the field-moist soil was processed after sieving through a 2-mm mesh to remove visible plant roots and litters, and analyzed for soil ammonium (NH₄ ⁺) and nitrate (NO₃ ⁻) contents using the colorimetric method (Keeny and Nelson, 1982). The other portion of the soil was air-dried under shade at room temperature (∼25°C) for two weeks and processed (2- mm sieved) for the analysis of soil organic carbon (SOC) by the wet oxidation method (Walkley, 1947) and total N by the Kjeldahl method (Fawcett, 1954).

At harvesting, the grain yield in t ha⁻¹ was determined from 4 m² areas of each replicated plot. Grain yields were adjusted to 14% moisture.

A split-split plot three-way analysis of variance was performed using tillage, residue level, and N application rate as fixed variables, where each of the three factors was considered as a main factor. The distribution of data for normality was checked before the analysis of variance. Data were statistically analyzed to ascertain the significant differences for the main effects and interactions among tillage, residue level, and N application rate treatments. To separate differences among the means, post hoc tests were performed using the Tukey-Kramer multiple comparison test. All statistical analyses were considered significant at p ≤ 0.05, unless otherwise mentioned, and were performed on Statistics 10 and Jamovi1.0.0.0. (R Package).

The N₂O emission peaks increased with the growth stages of rice plants reaching the highest emission peaks at the panicle initiation stage (50–55 days after transplanting, DAT) and the lowest at the early tillering stage (15 DAT) (Figure 1). The highest peak was found on day three after urea application in all three stages of plant growth and amounted to about 10% of total N₂O emissions. Surprisingly, the emission peaks were similar among the treatments at the early tillering stage but appeared to be different at latter stages. For example, the N₂O emission peak was 14, 52, and 77 g N ha⁻¹ d⁻¹ in ST, and 14, 44, and 64 g N ha⁻¹ d⁻¹ in CT at the early tillering, active tillering, and panicle initiation (PI) stages, respectively (Figure 1A).

Like N₂O, the CH₄ emission peaks were lower at the early tillering stage than at the active tillering and PI stages (Figure 2). The CH₄ emission peaks were as high as 6, 15, and 12 kg C ha⁻¹ d⁻¹ in CT and 4, 12, and 10 kg C ha⁻¹ d⁻¹ in ST at early tillering, active tillering, and PI stages, respectively. The highest CH₄ emission peak accounted for 10–15% of the total emissions in all treatments.

The N₂O emissions from the rice paddy fields were significantly influenced by tillage, residue levels, and N fertilization rates (Table 1). The mean (±SE) N₂O emission rates were significantly higher in ST (15.7 ± 0.69 g N ha⁻¹ d⁻¹) than in CT (13.5 ± 0.70 g N ha⁻¹ d⁻¹) (Table 1). Likewise, the cumulative N₂O emissions were 16% higher in ST (957 g N ha⁻¹) than in CT (822 g N ha⁻¹). The mean N₂O emission rates were 18% higher in HR than in LR. The N₂O emission rates linearly increased with the fertilizer N rate, from 12.7 ± 0.61 to 14.6 ± 0.62 and 16.9 ± 0.91 mg N ha⁻¹ d⁻¹ in N1, N2, and N3, respectively, irrespective of the tillage and residue level. The mean N₂O emission rates increased significantly by about 15% with each increase in the N rate. The interaction effects of tillage × residue × N rate were significant, resulting in the highest N₂O emissions in ST with HR coupled with N3 (21.3 g N ha⁻¹ d⁻¹), whereas the lowest N₂O emissions were found in the combination of CT, LR, and N1 (9.63 g N ha⁻¹ d⁻¹) (Table 1). The highest cumulative N₂O emissions were also observed in ST coupled with HR and N3 which were 117% higher than the lowest emissions in CT with LR and N1 and 80% higher than the conventional management practices (CT with LR in N2). The cumulative N₂O emissions were 30% higher in ST with HR coupled with N2 than the conventional management practices (CT with LR in N2).

The mean CH₄ emission rates were significantly influenced by tillage systems, residue levels, and N fertilization rates (Table 1). The mean CH₄ emission rates were significantly higher in CT than in ST by 32%. The cumulative CH₄ emissions during the measurement period were also higher in CT than in ST (Table 1). High crop residue levels significantly increased the mean CH₄ emission rates, which were 5.76 and 5.33 kg C ha⁻¹ d⁻¹ in HR and LR, respectively. Similarly, cumulative CH₄ emissions were higher in HR than in LR. Mean CH₄ emission rates were 21 and 11% higher in N1 and N3 than in N2, respectively (Table 1). The recommended N application rate (N2) in ST plus LR or HR reduced the cumulative CH₄ emissions by 10–18% over the current tillage and residue management practices (i.e., N2 in CT with LR).

The total yield-scaled N₂O emissions over the experimental duration were not significantly different in CT and ST, with a mean value of 0.27 and 0.29 kg N t⁻¹, respectively (Table 2). Like tillage, the crop residue levels had no significant effect on yield-scaled N₂O emissions. However, the yield-scaled N₂O emissions were significantly lower in N3 than in N1 but equal to N2 since the latter two were similar to each other. The interaction effects of tillage, residue level, and N fertilization rate were non-significant.

The mean N₂O EFs were significantly higher in ST (0.0062 ± 0.0003) than in CT (0.0051 ± 0.0002) (Table 2). Similar to the tillage effect, the N₂O EF was significantly higher in HR (0.0062 ± 0.0002) than in LR (0.0051 ± 0.0002). The N₂O EF was significantly higher in N1 than in N2 and N3 where the latter two were alike (Table 2). The interaction effects of tillage x N fertilization rate, residue level × N fertilization rate, and tillage × residue level × N fertilization rate were significant. The highest N₂O EF was found in ST with HR coupled with N1, while the lowest was found in CT with LR in all N fertilization rates, which is equal to ST with LR in N3.

The ST significantly reduced yield-scaled CH₄ emissions by 23% over the CT (Table 2). Yield-scaled CH₄ emissions between high and low residue levels were similar to each other, with the mean values of 46.7 ± 2.35, and 47.4 ± 2.01 kg C t⁻¹, respectively. The lowest rate of N fertilizer (N1) increased yield-scaled CH₄ emissions by 33 and 27% over N2 and N3, but the higher N rates were similar to each other. The interaction effects of tillage × N rate were significant, resulting in the highest yield-scaled CH₄ emissions in CT coupled with N1 in either residue level. The lowest yield-scaled CH₄ emissions were in ST coupled with N3 in both residue levels which was equal to the emissions in ST with N2 at both residue levels.

The conversion of CT to ST significantly reduced the seasonal GWP from the combined emissions of CH₄ and N₂O from the wetland rice field. The mean GWP was 24% higher in CT (293 kg CO_2-e ha⁻¹) than in ST (224 kg CO_2-e ha⁻¹) (Table 3; Figure 3). Crop residue levels significantly increased the GWP of combined CH₄ and N₂O emissions in rice by 21%, with a mean GWP of 269 and 248 kg CO₂-e ha⁻¹ in HR and LR, respectively. Interestingly, the optimum fertilizer N rate had a significantly lower GWP (N2; 233 kg CO₂-e ha⁻¹) than the higher (N3; 282 kg CO₂-e ha⁻¹) and the lower N rates (N1; 260 kg CO₂-e ha⁻¹) (Table 3). The interaction effects of tillage × residue level × fertilizer N rate were significant so that the highest GWP from combined CH₄ and N₂O emissions was in CT with HR coupled with N1 and lowest in ST with either residue level coupled with N2 (Table 3). The ST with LR in N2 reduced the GWP from combined CH₄ and N₂O emissions by 39% over the GWP in CT with HR in N1. The yield-scaled GWP was significantly higher in CT with N1 in either residue level than in ST with N1and ST with N2 where the latter two were similar to each other in either residue level (Table 3).

After four crops of continued CA practices, the Boro rice yield was significantly higher in CT than in ST at a mean yield of 7.51 ± 0.19 and 7.01 ± 0.22 t ha⁻¹, respectively (Table 2). The mean rice yields at the two residue levels were not significantly different. With each increment in the fertilizer N rate, the rice yield increased significantly. The interaction effects of tillage, residue level, and N fertilizer rate were non-significant.

After 2 years of CA practices, the soil pH was slightly higher in ST (6.66 ± 0.03) than in CT (6.61 ± 0.02) but not significantly different between HR and LR (Table 4). The N fertilization rate significantly increased the soil pH, from 6.52 in N1 to 6.75 in N3, independent of tillage and residue levels (Table 4). The interaction effects of the tillage and N fertilization rate caused a significantly higher pH in ST with N3 in either residue. The tillage × residue level × N fertilization interaction showed a significantly higher TN in ST with either residue level in combination with N2 or N3 than any other treatment combinations that included N1. The soil NH₄ ⁺ content was significantly higher in ST than in CT (ca. 9.9 ± 0.46 and 8.4 ± 0.41 mg N kg⁻¹ in ST and CT, irrespective of residue and fertilizer rate), but the values were similar in HR to LR (Table 4). The N fertilization rate significantly increased the NH₄ ⁺ content in N3 compared to N1 and N2. By contrast, the soil NO₃ ⁻ —N content was similar in ST to CT and in HR to LR. The NO₃ ⁻–N content was lower in N1 than N2 and N3 where the latter two were similar to each other. After five crops, the SOC was unchanged by tillage and residue levels, but the suboptimal N fertilization rate (N1) caused a lower SOC than under N2 and N3.

The adoption of minimum soil disturbance (ST) and HR retention with the recommended fertilizer N rate reduced the GWP from CH₄ and N₂O emissions by 10% relative to the present farmers’ practice (CT with LR) at the same N fertilizer rate. However, if ST is combined with LR and the recommended fertilizer rate (N2), it reduced the GWP by 39% relative to LR in combination with the recommended fertilizer rate under CT. Equally, if we consider yield-scaled GWP for the recommended N rate, ST with HR, which represents a CA cropping system, reduced the GWP from combined CH₄ and N₂O emissions by 9% relative to present farmers’ practices (CT with LR). Overall, the GWP was lower for ST with either LR or HR than CT with LR under the recommended N fertilizer rate.

While CA had opposite effects on CH₄ and N₂O emissions, the net effect was to decrease the GWP since CH₄ emissions dominate the total GWP in paddy rice, accounting for around 96% of the total GWP. Since rice is usually grown under flooded conditions, CH₄ emissions make a major contribution to the GWP (Linquist B. et al., 2012). Contributions of N₂O emissions to the GWP of GHGs in rice were reported in previous studies to be low compared to CH₄ emissions (Ly et al., 2013; Vu et al., 2015; Alam et al., 2016, 2019a), but emissions increased after N fertilizer application (Zou et al., 2005; Pandey et al., 2014).

If we consider the total seasonal N₂O emissions, they were relatively large, ranging from 0.59 kg N ha⁻¹ to 1.30 kg N ha⁻¹. These results were comparable with seasonal N₂O emissions from wheat (ranged 0.32 kg N ha⁻¹–1.20 kg N ha⁻¹) under the same management practices and the same agroclimatic conditions (Jahangir et al., 2021b). This suggests that paddy rice is also a presently insufficiently recognized source of N₂O even though it is relatively less important than CH₄ under these conditions. Hence, management interventions such as optimum N fertilizer rates that decrease the seasonal N₂O emissions from rice paddy fields also need further attention as a means to lower the GWP of rice paddy.

In our study, the effect of tillage, residue level, and N rate on mineral N concentrations was non-significant, whereas both ST and higher N rates resulted in higher NH₄ ⁺ concentrations. Overall, the CA (ST with HR) increased soil TN, nitrate, and pH, whereas ST alone increased NH₄ ⁺ availability. In this study, in the fifth crop of the rotation after conversion of CT to ST, the irrigated rice yield was 6.7% higher in CT than in ST. The rice yield in the first few years after the conversion of CT to ST was higher in CT than in ST (Li et al., 2015) or equal between CT and ST (Haque et al., 2016). The decrease in the yield due to conversion of CT to ST in the present study can be attributed to the transitional effect on soil physicochemical and biological properties. The ST can increase the soil microbial growth that can stimulate the mineralization–immobilization turnover (MIT) in later years. The mechanism for the temporal yield recovery under ST can be attributed to an increased SOM and nutrient availability, biological efficiency, and improved soil fertility properties (Islam and Weil, 2000; Samal et al., 2017). An increase in the rice yield with the increase in the N application rate indicates that the soil was deficient in N content and thus responded to the added N fertilizer. However, ST with HR and the recommended fertilizer N rate significantly increased soil TN over the conventional management (e.g., CT with LR) at the same N rate. The SOC was similar for all treatment combinations, which could be because of the short time period (<2 years) after the switch from CT to ST. In addition, SOC consumption by denitrification in some treatments or by CH₄ production in other treatments can minimize the differences as higher N₂O production corresponded to a lower CH₄ production. Nitrate was easily denitrified under anaerobic conditions and the processes of denitrification consume electrons and H₂, competing with CH₄ producers (Kluber and Conrad, 1998).