Intercontinental comparison of greenhouse gas emissions from irrigated rice fields under feasible water management practices: Brazil and Japan

ABSTRACT

Flooded rice fields are a significant anthropogenic source of methane (CH₄) and nitrous oxide (N₂O) from agriculture in Asia, Latin America, and the Caribbean regions. In this work, we comparatively assessed the potential of intermittent irrigation and continuous rice flooding for reducing soil CH₄ and N₂O emissions, partial global warming potential (pGWP), and its yield-scaled version (YpGWP) in northwestern Japan and southern Brazil. Seasonal CH₄ emissions under continuous flooded soils were slight higher in Japan (738 ± 87 kg ha⁻¹) than in Brazil (623 ± 197 kg ha⁻¹), and they were probably related to the higher level of soil organic C and the longer period under flooding in the seedling transplanting system in the Japanese site. Intermittent irrigation had similar efficiency in decreasing soil CH₄ emissions in both study areas, with the maximum mitigation potential of 71% in northwestern Japan and of 62% in southern Brazil. No significant difference in seasonal soil N₂O emissions (−0.17 ± 0.05 to −0.24 ± 0.06 kg N₂O ha⁻¹ in Japan and 0.32 ± 0.08–1.16 ± 0.40 kg ha⁻¹ in Brazil) or rice yield (7328–8064 kg ha⁻¹ in Japan and 9391–10,231 kg ha⁻¹ in Brazil) between irrigation systems was observed in either area. The potential of intermittent irrigation for reducing pGWP was around three times higher than that of continuous flooding in both sites. Thus, a reduction by 47–63% and 62–80% in yield-scaled pGWP was observed in southern Brazil and northwestern Japan, respectively. Like the well-established labor-intensive rice transplanting systems used in Asia, the introduction of feasible irrigation suppression systems in mechanized direct seeding rice fields in southern Brazil and other countries of Latin America and the Caribbean region is an effective choice for reducing greenhouse gas emissions with no adverse impact on rice yield.

KEY WORDS:

• Methane
• nitrous oxide
• water regime
• paddy field
• global warming potential

1. Introduction

Flood irrigated rice spans an area of approximately 80 million ha and accounts for about 75% of the world’s rice production (IRRI 2013). Because of the anoxic soil environment, irrigated rice lowland is a significant anthropogenic source of methane (CH₄) and nitrous oxide (N₂O), two major greenhouse gases (GHGs) with a global warming potential (GWP) 25 and 298 times, respectively, greater than that of CO₂ on a one hundred-year time horizon (IPCC 2007).

Although 90% of the world’s irrigated rice is grown in Asia, rice production in Latin America and the Caribbean (LAC) has grown substantially in recent times (at an annual rate of 2.8% over the period 2000–2010 according to GRiSP, 2013). Brazil, which is currently the greatest producer among LAC countries, accounts for nearly one-half of the regional rice production (GRiSP 2013). Brazilian rice comes mainly from the subtropic region (southern Brazil). Thus, the Rio Grande do Sul state is a rice granary with a cultivated area of more than 1 million ha (CONAB 2015) that accounts not only for nearly 70% of the national production (IBGE 2014) but also for 18% of the overall CH₄ emissions from the national agricultural sector (SEEG 2014). The need of both Asia and LAC to increase rice production in order to fulfill the demands of an increasing population can be met more sustainably by intensifying rice production based on environmental friendly agricultural practices than by expanding the cultivated area.

Studies on soil CH₄ mitigation in rice fields are abundant in Asia but scant in the LAC region. In Japan, the most effective way of alleviating soil CH₄ emissions involves not only nitrogen fertilization and organic matter application but also the use of effective water management practices such as midseason drainage (MD) and intermittent irrigation to prevent the development of soil reductive conditions (Nishimura et al. 2004; Zou et al. 2005; Minamikawa et al. 2006). In Brazil, research in this field is fairly recent and only a few studies have dealt with the potential of no-till and anticipated tillage (Bayer et al. 2014; Bayer et al. 2015), and water management (Moterle et al. 2013; Zschornack et al. 2016) for mitigating soil CH₄ emissions in subtropical lowland rice fields. In both world regions, intermittent water management has been found to produce soil N₂O emissions similar to or even higher than those of continuous flooded rice fields (Cai et al. 1997; Yan et al. 2000; Zschornack et al. 2016); however, a positive GHG balance was invariably observed (Zschornack et al. 2016; Hou et al. 2012; Hadi et al. 2010).

Partly because of advanced research, the most widely used water management practices in labor-intensive rice transplanting programs in Japan are MD and intermittent irrigation (Minamikawa et al. 2006). In contrast, mechanized direct seeding rice in Brazil and LAC region is produced mainly under continuous flooding (CF) because of the uncertainty in the timing and span of water-saving regimes as regards to rice yields (Chirinda et al. 2017). Therefore, Asian farmers are currently exploiting the whole potential of effective water management practices for decreasing CH₄ emissions, whereas Brazilian farmers are still to explore the ability of these practices to increase rice production and food security without increasing GHG emissions. Although adopting feasible water management systems to mitigate CH₄ emissions from rice fields might benefit the whole LAC region, the actual effectiveness of this practice should be better assessed on a more local level.

The scarcity of studies in this direction in Brazil and the LAC region led us to comparatively assess the potential of feasible water management systems for decreasing GHG emissions in southern Brazil and northwestern Japan. Our starting hypothesis was that the potential of intermittent irrigation for decreasing soil CH₄ emissions would be similar in both world regions and that the magnitude of the effect would vary with soil and climate. In addition, we hypothesized that the GHG mitigation potential of intermittent water management would be partially offset by increased soil N₂O emissions relative to CF and also that properly performed water irrigation suppression would have no negative impact on rice yield.

2. Material and methods

2.1. Sites and field management practices

2.1.1. Southern Brazil

The experiment was conducted in the crop season 2012/13 at the Research Station of the Riograndense Rice Institute (IRGA), Rio Grande do Sul State, southern Brazil. The soil is an Entisol as per the US Soil Taxonomy and the climate is humid subtropical, with warm summers and cool winters. Rainfall is evenly distributed through the year and no definite dry season exists (Peel et al. 2007). Table 1 summarizes the properties of soil and climate in the studied area.

Table 1. Soil and climate characteristics of the studied sites and main characteristics of the experiments.

	Southern Brazil	Northwestern Japan
Site description
Municipality	Cachoeirinha	Nagaoka
State/Province	Rio Grande do Sul	Niigata
Coordinates	29°55′30″S 50°58′21″W	37°54′N 139°3′E
Altitude (m)	6	7
Climate (Koppen–Geiger)	Cfa	Cfa
Air temperature
Annual mean (°C)	20	12.6
Minimum (°C)	10	2.1
Maximum (°C)	29	26.2
Annual rainfall (mm)	1350	2336
Soil classification (USDA taxonomy)	Entisol	Fluvaquents
Grain size distribution
Sand (g kg^−1)	490	430
Silt (g kg^−1)	420	270
Clay (g kg^−1)	100	290
Chemical properties
TOC (g kg^−1)	7.8	21.5
TN (g kg^−1)	0.67	1.95
Fe2O3 (g kg^−1)	10.9	21.7
Experiment characteristics
Irrigation treatment	Crop season CF MI SI	Off-season D ND Crop season CF MD
Experimental design	Randomized complete blocks	Randomized complete blocks
Number of replicates	3	3
Previous crop	Ryegrass	Winter fallow
Rice variety	IRGA 424	Koshihikari
Seeding method	Direct in dry soil	Transplanting
Fertilization at planting
N (kg ha^−1) P2O5 (kg ha^−1) K2O (kg ha^−1)	100 50 90	45 25 25
Topdressing N fertilization (kg ha^−1)	50 50	10 10

TOC: total organic carbon; CF: continuous flooding; MI: moderate intermittent irrigation; SI: severe intermittent irrigation; D: drainage; MD: midsummer drainage; ND: no drainage; TN: total nitrogen, Fe₂O₃: total iron oxides.

Climate data from Climate-data (2014).

The values for Japanese soil are the means of those for drained and undrained areas off-season.

Soil tillage was performed in the autumn of 2012 and involved one plowing disk pass and two leveling disk passes to a depth of 20 cm. Approximately 6 t ha⁻¹ of rice biomass was incorporated into the soil by tillage. Self-reseeding ryegrass was grown as a cover crop without irrigation during the winter months. A glyphosate-based herbicide at a rate of 3 L ha⁻¹ was applied to the ryegrass in the spring, followed by machine chopping of the standing biomass. Rice was drill seeded and grown under different irrigation treatments during the summer.

Shortly after desiccant herbicide and the first nitrogen fertilization were applied, the field was flooded at rice development stage V3 (three leaves; Counce et al. 2000) in all treatments. Three different irrigation treatments were applied during the rice growing season, namely, (a) CF, with which the soil was kept under a 5–10-cm thick water layer throughout the growth season; (b) moderate intermittent irrigation (MI), which involved suppressing water at stages V6–V8 (6–8 leaves); and (c) severe intermittent irrigation (SI), where water was suppressed at stages V6–V8 and V8–V10 (8–10 leaves). Irrigation was suspended 15 days before rice harvest in all treatments.

All cultivation practices followed the technical recommendations for the rice in southern Brazil (Sosbai 2014). Fig. 1A,C shows the daily air temperature and rainfall over the rice growing period as obtained from the National Meteorological Institute (INMET) of Brazil. Table 1 provides additional information about the experimental design, rice variety, and fertilization procedure used.

Figure 1. Seasonal variations in daily rainfall (A,B), daily mean air temperature (C,D), surface water depth (E,F), CH₄ flux (G,H), and N₂O flux (I,J) under different water management systems at Cachoeirinha in southern Brazil and at Nagaoka in northwestern Japan. Grey areas in A–D indicate lack of observation data. Tp: Transplanting; N: nitrogen fertilization; Hv: harvest. Vertical bars in G–J indicate standard deviation (n = 3). Note: Baseline N application in Japan was done 14 days before transplanting.

2.1.2. Northwestern Japan

The Asian experiment was conducted at the Niigata Agricultural Research Institute, Niigata Province, northwestern Japan. Niigata Province is on the coastal plain between the mountains and sea in Japan. It has a humid subtropical climate (Cfa as per Peel et al. 2007). Off-season is characterized by a cold winter with about 2 m of snowfall. The snow melts during late winter and spring, and rice is grown in summer.

Water management treatments were applied off-season and in the rice season. Off-season, the soil was subjected to two different treatments after the snow had melted, namely, drainage (D) and no drainage (ND). Also, two different irrigation treatments were applied in the rice season, namely, CF, with which the soil was kept under a 5–10-cm thick layer of water throughout the growth season; and mid-summer drainage (MD), which involved flooding before rice was transplanted and was followed by drainage around 15–20 days before maximum tillering, and by intermittent irrigation at 4-day intervals for 20 days before harvest (Shiratori et al. 2007), when irrigation was suspended altogether.

In autumn 2013, about 6 t ha⁻¹ rice straw from the previous rice was incorporated into the surface soil layer (10 cm) by using a rotary hoe. In spring (May) 2014, three to four rice seedlings 2.5–2.7 cm high per pit were transplanted to the experimental area after flooding. The area was kept flooded until the seedlings began to grow and then the two irrigation treatments (CF and MD) were applied. Since MD is the irrigation system traditionally used by Japanese farmers, CF was only introduced for comparison of the impact of intermittent irrigation with that in southern Brazil.

The air temperature and rainfall during the rice growing season were measured on a daily basis at the Climate Station of the Niigata Agricultural Research Institute (see Fig. 1B,D).

2.2. Rice grain yield

Rice grain yield was manually assessed in an area of 10 m in southern Brazil and of 4 m² in northwestern Japan. All yield data are expressed on a 13% humidity basis.

2.3. GHG measurement and calculations

GHGs were measured weekly by using the static chamber method (Minamikawa et al. 2012; Bayer et al. 2014) during the rice growing season (November–March in Brazil and May–September in Japan). Acrylic chambers 60 cm long × 30 cm wide × 60 cm high were placed on the soil surface on 60 × 30 × 5 cm PVC bases in northwestern Japan, whereas 60 × 60 × 30 cm aluminum chambers and 60 × 60 × 20 cm bases were used in southern Brazil.

The bases were driven 5 cm deep into the soil before flooding and remained in it throughout the growth season. Each base had an open bottom and sealable channels on the sides for irrigation water to flow freely, all being sealed during air sampling events. Each base covered three rice plant rows in southern Brazil and three rice seedlings in northwestern Japan. Additional 20 or 30 cm extensors were staked on the bases as the rice plants grew in order to accommodate plant growth, and the chamber volume considered in GHG emission calculations.

Each chamber top had a rubber septum sampling port, a stainless steel thermometer, and two or three battery operated fans to circulate and homogenize air inside the chamber (Banker et al. 1995). Chamber closing and air sampling were done from 09:00 to 11:00 am each day and involved four samplings (0, 5, 10, and 20 min) in Brazil and three (0, 15, and 30 min) in Japan. Air was drawn with polypropylene syringes (20 mL) and transferred to glass bottles with double wadded silicone caps (Labco Corp.) to determine CH₄ and N₂O on a gas chromatograph equipped with a flame ionization and electron capture detector.

GHG fluxes were calculated as follows:

(1)

where F is the gas emission rate (g CH₄ m⁻² h⁻¹ or mg N₂O m⁻² h⁻¹), dC/dt is the variation of the concentration of CH₄ or N₂O (mol h⁻¹), is the gas molar mass (g mol⁻¹), P is the atmospheric pressure in the chamber (assumed to be 1 atm), V is the chamber volume (L), R is the ideal gas constant (0.08205 atm L mol⁻¹ K⁻¹), T is the chamber temperature (K), and A the chamber basal area (m²). The flux rate of GHG from air sampling at 9:00–11:00 am was assumed to be equivalent to the mean daily flux rate (Costa et al. 2008). The gas production rate was subsequently converted into kg CH₄ ha⁻¹ day⁻¹ and g N₂O ha⁻¹ day⁻¹. Seasonal emissions were calculated by trapezoidal integration of the daily GHG fluxes throughout the growth season (Bayer et al. 2014).

The partial GWP (pGWP) was calculated by multiplying the seasonal CH₄ and N₂O emissions by their respective radiative forcing potential and combining the two:

(2)

where pGWP is the partial global warming potential (kg CO₂ eq ha⁻¹); CH₄ and N₂O are the seasonal emissions of methane and nitrous oxide, respectively (kg ha⁻¹); and 25 and 298 are the radiative forcing potential of CH₄ and N₂O relative to carbon dioxide (CO₂) on a 100-year horizon, respectively (IPCC 2007).

Yield-scaled pGWP (kg CO₂ eq kg⁻¹ grain yield) was calculated as the ratio of pGWP to rice grain yield:

(3)

where pGWP is the partial global warming potential (kg CO₂ eq ha⁻¹) and Y grain yield (kg ha⁻¹).

2.4. Statistical analyses

Data were visually inspected for normality and constant variance of errors and subjected to appropriated transformations when the assumptions were violated. Analysis of variance was done by using the Mixed Procedure in SAS (SAS Institute Inc., Cary, NC, USA). Fluxes and seasonal soil CH₄ and N₂O emissions, pGWP, yield-scaled GWP (YpGWP), and grain yield were considered dependent variables, while irrigation systems were used as fixed effects and replication as random effect. The dependent variables seasonal soil CH₄ and N₂O emissions, pGWP, YpGWP, and grain yield in each site were analyzed by using GLM Procedure in SAS. All differences were deemed significant at the 95% level (P < 0.05) as per Tukey test.

3. Results

3.1. Soil CH₄ and N₂O fluxes

As can be seen from Table 2, mean soil CH₄ fluxes differed significantly between irrigation systems in northwestern Japan and southern Brazil, intermittent irrigation during rice growth leading to low CH₄ fluxes relative CF in both study areas (see Fig. 1G,H).

Table 2. Results of the ANOVA of fluxes and seasonal CH₄ and N₂O emissions, pGWP, and YpGWP in rice fields under different irrigation systems in southern Brazil (continuous flooding, intermediate intermittent, and severe intermittent irrigation) and northwestern Japan (continuous flooding and midsummer drainage in the rice season, and drainage or no drainage off-season).

Southern Brazil				Northwestern Japan
Irrigation management				Irrigation management				Off-season management
DF	Value	Pr > F		DF	Value	Pr > F		DF	Value	Pr > F
CH4 flux
2	36.76	<.0001		1	292.12	<.0001		1	4.76	0.0303
N2O flux
2	7.65	0.0008		1	4.95	0.0272		1	9.05	0.0030
Cumulative CH4
2	9.85	0.0285		1	97.53	<.0001		1	1.40	0.2811
Cumulative N2O
2	2.49	0.1986		1	0.81	0.4020		1	1.99	0.2077
pGWP
2	92.66	0.0004		1	96.64	<.0001		1	1.34	0.2906
pGWP/Yield
2	32.38	0.0034		1	72.23	0.0001		1	0.45	0.5275

Soil CH₄ fluxes started to increase at 22 after flooding (days after flooding [DAF] in southern Brazil and 32 DAF in northwestern Japan; also, they peaked from 71 to 87 DAF in the former area and from 88 to 123 DAF in the latter (Fig. 1G,H). In general, soil CH₄ emissions were maximal at the vegetative and reproductive stages of rice growth with all treatments, but much lower with intermittent irrigation. Thus, the highest CH₄ flux with CF in northwestern Japan was roughly twice that in southern Brazil (743 ± 220 vs. 395 ± 17 g ha⁻¹ h⁻¹; Fig. 1G,H).

Mean soil CH₄ fluxes under CF in southern Brazil (173 ± 54 g ha⁻¹ h⁻¹) were significantly reduced by moderate and SI (to 125 ± 2 and 68 ± 22 g CH₄ ha⁻¹ h⁻¹, respectively) (Fig. 1G). A similar trend in soil CH₄ fluxes with CF and intermittent irrigation (MD) was observed in northwestern Japan (Fig. 1H). Thus, irrigation with MD was suppressed and a decrease in soil CH₄ fluxes observed around 52 DAF. In contrast, a constant increase in CH₄ fluxes was observed with CF (Fig. 1H). Mean soil CH₄ fluxes were higher with ND-CF (235 ± 28 g ha⁻¹ h⁻¹) and D-CF (228 ± 61 g ha⁻¹ h⁻¹) than with ND-MD (81 ± 11 g ha⁻¹ h⁻¹) and D-MD (48 ± 9 g ha⁻¹ h⁻¹) systems.

Off-season field management (D or ND) in northwestern Japan had a significant influence on soil CH₄ fluxes during the rice growing season (Table 2 and Fig. 1H), the fields where the soil was drained before the rice season exhibiting slightly lower mean soil CH₄ fluxes (138 ± 35 g ha⁻¹ h⁻¹ on average for the irrigation treatments applied in the rice growing season) than those which were not drained (158 ± 19 g ha⁻¹ h⁻¹ under CF and MD).

Mean soil N₂O fluxes were higher with intermittent irrigation than with CF in both study areas. In northwestern Japan, all irrigation treatments resulted in mostly negative soil N₂O fluxes (influxes) that increased from a mean value of −62 ± 7 mg ha⁻¹ h⁻¹ with CF to −45 ± 8 mg ha⁻¹ h⁻¹ with intermittent irrigation; in southern Brazil, mean soil N₂O fluxes increased from 65 ± 20 mg ha⁻¹ h⁻¹ with CF to 186 ± 15 and 386 ± 123 mg ha⁻¹ h⁻¹ under MI and SI, respectively.

3.2. Seasonal soil CH₄ and N₂O emissions

Seasonal soil CH₄ emissions also differed significantly between irrigation systems in both northwestern Japan and southern Brazil (Tables 2 and 3), intermittent irrigation resulting in lower emissions than CF.

Table 3. Seasonal CH₄ and N₂O emissions (±SD values), partial global warming potential (pGWP), grain yield, and yield-scaled pGWP in irrigated rice fields under different irrigation systems in southern Brazil and northwestern Japan.

Site	Irrigation system	Seasonal CH4 (kg ha^−1)	Seasonal N2O (kg ha^−1)	pGWP (kg CO2 eq ha^−1)	Grain yield (kg ha^−1)	YpGWP (kg CO2 eq kg^−1 rice)
Southern Brazil	CF	623 ± 196a	0.32 ± 0.08ns	15,670 ± 4870a	9391a	1.67 ± 0.52a
	MI	466 ± 5ab	0.56 ± 0.94	11,817 ± 405b	10,059a	1.17 ± b
	SI	237 ± 75b	1.16 ± 0.40	6271 ± 6270c	10,231a	0.61 ± 0.17c
Northwestern Japan	ND-CF	738 ± 87a	−0.24 ± 0.06ns	18,378 ± 2184a	8064a	2.27 ± 0.27a
	D-CF	720 ± 194a	−0.17 ± 0.05	17,949 ± 4861a	7414a	2.42 ± 0.66a
	ND-MD	262 ± 35b	−0.23 ± 0.05	6481 ± 884b	7328a	0.88 ± 0.12b
	D-MD	159 ± 30b	–0.09 ± 0.06	3948 ± 761b	8390a	0.47 ± 0.09b

Different letters indicate statistically significant differences between treatments in each country as per Tukey’s test at the 5% level.

ns not significant. CF continuous flooding; MI moderate intermittent irrigation; SI severe intermittent irrigation; ND-CF no drainage off-season and continuous irrigation in the rice season; D-CF drainage off-season and continuous irrigation in the rice season; ND-MD no drainage off-season and midsummer drainage in the rice season; D-MD drainage off-season and midsummer drainage in the rice season. SD standard deviation (n = 3).

In northwestern Japan, seasonal soil CH₄ emissions under CF amounted to 738 ± 87 kg ha⁻¹ but were lower under ND-MD (159 ± 30 kg ha⁻¹) and D-MD (262 ± 35 kg ha⁻¹). No significant differences in soil CH₄ emissions between off-season management systems were observed, however. In southern Brazil, seasonal emissions under CF were 623 ± 196 kg ha⁻¹ (18% lower than in Japan) but lower with MI and SI (466 ± 5 and 237 ± 75 kg ha⁻¹, respectively) (Table 3). Therefore, the maximum potential of intermittent irrigation for decreasing soil CH₄ emissions was 71% in Japan (mean of drained and not drained soil off-season) and 62% in Brazil (SI).

The amount of seasonal N₂O emissions during the rice season was not significantly affected by irrigation system in either study area (Table 3). Thus, emissions in southern Brazil ranged from 0.32 ± 0.08 to 1.16 ± 0.40 kg ha⁻¹, whereas those in northwestern Japan consisted solely of N₂O influxes to soil and ranged from −0.24 ± 0.06 to −0.09 ± 0.06 kg ha⁻¹ (Table 3).

3.3. pGWP and YpGWP

pGWP (kg CO₂ eq ha⁻¹), which represents the combination of soil CH₄ and N₂O emissions accounting for the radiative forcing potential of each gas, was significantly affected by irrigation treatment in both northwestern Japan and southern Brazil (Tables 2 and 3). The mean pGWP values for CF (18,163 ± 2184 kg CO₂ eq ha⁻¹ in Japan and 15,670 ± 4869 kg CO₂ eq ha⁻¹ in Brazil) were roughly three times higher than those for intermittent irritation (5214 ± 822 kg CO₂ eq ha⁻¹ for the former country and 9043 ± 1098 kg CO₂ eq ha⁻¹ for the latter). The reduction in pGWP obtained by adopting intermittent irrigation was thus 71% on average in Japan and 42% in Brazil. There were no significant differences in pGWP between off-season management practices, however (Table 2).

No significant difference in rice grain yield between irrigation treatments was observed in either study area. Yields ranged from 7328 to 8390 kg ha⁻¹ in northwestern Japan and from 9391 to 10,231 kg ha⁻¹ in southern Brazil. Because intermittent irrigation reduced pGWP but water suppression had no effect on rice grain yield, YpGWP was significantly lower with intermittent irrigation than under CF in both Japan and Brazil (Tables 2 and 3). On average for the two off-season management systems in Japan, producing 1 kg of rice under CF caused the emission of 2.34 ± 0.46 kg CO₂ eq to the atmosphere. The corresponding figure for Brazil was 27% lower (1.66 ± 0.52 kg CO₂ eq). With intermittent irrigation, GHG emissions per kilogram of rice were low as 0.48 ± 0.09 kg CO₂ eq in northwestern Japan and 0.61 ± 0.17 kg CO₂ eq in southern Brazil (Table 3).

4. Discussion

4.1. Soil CH₄ fluxes and seasonal emissions

Soil CH₄ fluxes increased with the number of DAF in both northwestern Japan and southern Brazil; fluxes increased as the rice crops developed, with peaks at the vegetative (tillering) and reproductive (panicle initiation) stages. The increase can be ascribed to progress in soil reduction under CF and to increase root exudation (Denier vand der Gon, Neue HAC, Neue HU 1995; Lu et al. 2000). An apparent slight increase in soil CH₄ fluxes was observed in northwestern Japan (Fig. 1H) that led to their peaking later than in southern Brazil (Fig. 1G).

The less marked increase in soil CH₄ fluxes in Japanese soil was probably the result of its high content in iron oxides (Fe₂O₃; Table 1) relative to Brazilian soil since CH₄ production by anaerobic decomposition of organic matter in soil only occurs after most substances with a greater affinity as electron acceptors have been reduced. Such substances (Fe³⁺, Mn⁴⁺, and ) are reduced to Fe²⁺, Mn²⁺, and S²⁻, respectively, prior to CH₄ production (Yu et al. 2007). On the other hand, the fact that soil CH₄ peaks were higher in Japan than in Brazil probably resulted from the higher content in organic C of soil in the former (Table 1) since organic matter is known to be an effective substrate for methanogenic microorganisms (Minamikawa et al. 2006). The differences may also have arisen partly from differences in root exudation among rice varieties (Chirinda et al. 2017), but this is merely a speculative assumption that requires confirmation.

The less substantial increase in soil CH₄ was offset by the highest peaks resulting in higher seasonal soil CH₄ emissions under CF in northwestern Japan (738 ± 87 kg ha⁻¹) than in southern Brazil (623±kg ha⁻¹). Since the mean air temperature in the studied period was slightly lower at the Japanese site (22.1°C) than in southern Brazil (24.5°C), the increased CH₄ emissions in northwestern Japan may again have resulted from its increased content in soil organic C, which acts as an energy source for methanogenic microorganisms (Itoh et al. 2011; Silva et al. 2011; Ali et al. 2013). However, differences in seeding methods may also have contributed to the higher seasonal soil CH₄ emissions observed in northwestern Japan. Unlike transplanting rice in a previously flooded field in Japan, direct seeding in dry soil and flooding fields only after rice has emerged in southern Brazil results in 20–40 fewer days of soil saturation and leads to less favorable conditions for methanogenesis. Previous studies in Asia suggest that soil CH₄ emissions associated with direct seeding may be 8-92% lower than those observed from transplanted rice (DeAngelo et al. 2006; Kumar and Ladha 2011).

In general, while soil CH₄ production increased continuously with time under CF, water suppression in intermittent irrigation arrested the emission reduction process in soil and decreased soil CH₄ fluxes. Suppressing irrigation facilitates diffusion of O₂ from the atmosphere into the soil, thereby promoting re-oxidation of the soil and hindering CH₄ production (Zhang et al. 2012; Kim et al. 2014). In this study, soil CH₄ fluxes resumed during intermittent irrigation periods and increased when the soil was re-flooded. This deleterious effect of intermittent irrigation may also have arisen from population changes in methanogenic and methanotrophic microorganisms in the soil as previously observed by Ma and Lu (2011). According to these authors, soil drainage suppresses growth of methanogenic microorganisms, and emission reduction during draining events can be ascribed in part to methanotrophic microorganisms remaining active thanks to O₂ availability and CH₄ trapping in the soil.

Repeated drainage in northwestern Japan (every 4 days from tillering to 20 days before harvest) decreased seasonal soil CH₄ emissions by 71% (mean of the two off-season treatments), which exceeds the reduction observed with a single water suppression event (6–8 leaves) in the MI treatment, 40%, and a little higher than reduction observed with repeated water suppression (6–8 leaves and 8–10 leaves) in the SI treatment (SI) in southern Brazil (62%). The magnitude of the reduction of seasonal soil CH₄ emissions under intermittent irrigation is related to the number and duration of the water suppression or drainage events (Zhang et al. 2012; Ma et al. 2013; Zschornack et al. 2016). Thus, water suppression should last long enough to promote soil re-oxidation; also, the more suppression events during the rice growing season, the more efficient intermittent irrigation in decreasing soil CH₄ emissions (Towprayoon et al. 2005). Since soil reduction is quite a slow process, the decrease in seasonal soil CH₄ emissions is usually more marked with repeated events of water suppression or drainage than with a single one (Zhang et al. 2012).

In addition to soil and climate, the efficiency with which soil CH₄ emissions are reduced in different parts of the world is influenced by the design of the intermittent irrigation system (particularly by frequency and duration of water suppression or drainage events). Based on the scant available literature on the topic, the potential of intermittent irrigation systems for decreasing soil CH₄ emissions exceeds that of CF by 40–85% in southern Brazil (Moterle et al. 2013; Zschornack et al. 2016). This range is similar to the values of 12–88% derived from massive literature from Japan and other countries on the Asian Continent (Malyan et al. 2016).

Although no significant difference in seasonal soil CH₄ emissions between off-season treatments (D or ND) was observed, the low emission levels from drained soil (720 kg ha⁻¹ under CF and 159 kg ha⁻¹ under MD) relative to undrained soil (738 kg ha⁻¹ under CF and 262 kg ha⁻¹ under MD) apparently follow the same logic as the adoption of intermittent irrigation in the rice growing season. According to Shiratori et al. (2007), draining fields off-season causes stronger decomposition of organic compounds in soil owing to the prevalence of aerobic conditions; as a result, it decreases the amount of C available as a substrate for methanogenic microorganisms under anaerobic conditions in the rice growing season. Also, off-season soil drainage promotes re-oxidation of iron oxides, which remain in their reduced forms by effect of soil saturation in the absence of drainage. In this way, after flooding in the rice season, undrained soil may contain smaller amounts of reducible iron Fe than drained soil and hence lead to earlier soil CH₄ emissions.

4.2. Soil N₂O fluxes and seasonal emissions

As can be seen from Fig. 1I,J, top-dress N fertilization had no effect on soil N₂O fluxes under continuously flooding (Fig. 1I,J). This result is consistent with those of other studies (Bayer et al. 2014, 2015; Liu et al. 2010) and may be ascribed to sequential reduction of N₂O to N₂ under low E_h soil conditions (Reddy and DeLaune 2008) and also to N absorption by rice plants.

Since the formation of CH₄ and N₂O in soil requires contrasting redox potential conditions, reduced soil CH₄ emissions in periods of water suppression are in contrast with increased N₂O fluxes under intermittent irrigation (Fig. 1 and Table 3). This is a well-documented trade-off (Zschornack et al. 2016; Hou et al. 2012; Johnson-Beebout et al. 2009; Liu et al. 2010; Zou et al. 2004). In this study, seasonal soil N₂O emissions in intermittent systems in southern Brazil were three times higher than in flooded soils. However, the difference was not significant at Tukey’s 0.05 level owing to the high variability in the data among replicates. In northwestern Japan, seasonal soil N₂O influxes were lower than in southern Brazil, partly as a result of the also lower N application rates used in the former area (45 vs. 150 kg ha⁻¹). The low soil N availability (Chapuis-Lardy et al. 2007) and the number of days allocated between irrigation periods in Japan proved insufficient to increase E_h (Towprayoon et al. 2005) may be related to soil N₂O influxes.

4.3. pGWP, rice yield, and YpGWP

The increased pGWP values obtained with CF in both study areas can be ascribed to the major contribution of soil CH₄ emissions during the rice growing season (Table 3). By contrast, the contribution of N₂O emissions to pGWP was very small in southern Brazil (0.6–6.0% pGWP) and negative in northwestern Japan, even under intermittent irrigation. These results suggest that efforts should focus on reducing CH₄ emissions during rice growth. Itoh et al. (2011) previously found the contribution of N₂O to pGWP at nine sites with different irrigation systems in Japan to be very small (<10%).

The similarity of yields under CF and intermittent irrigation in both study areas reinforce previous results of Hou et al. (2012) and Kim et al. (2014) and suggest that water supply from both systems fulfills the requirements of rice crops. This is especially so in Asia, where some vagueness in timing and span remains but appropriate water management was previously found to result in similar or even higher rice yields (Minamikawa et al. 2006). In contrast, these results are very important for Brazil and other LAC countries where CF is the prevalent water management practice. In this region, farmers have been largely reluctant to adopt intermittent irrigation despite its emission mitigation benefits (Mushtaq et al. 2009). One of the main challenges of intermittent irrigation is avoiding the yield losses of inappropriate implementation (e.g., if drying is done at critical stages of plant growth such as flowering). Avoiding the yield penalties associated with intermittent irrigation requires better awareness of good water scheduling approaches and also exploring new irrigation techniques with more easily controlled timing of drying. More importantly, promoting the adoption of intermittent irrigation requires developing effective policies to create an enabling environment and infrastructure for water delivery and control by farmers (Chirinda et al. 2017).

Yield-scaled pGWP (YpGWP) with intermittent irrigation systems was more than twice lower than those obtained with flooding in both northwestern Japan and southern Brazil (Table 3). Therefore, appropriate water management in intermittent irrigation systems can produce more than twice the amount of grain rice with current GHG emissions, or vice versa. Similar results were obtained by Zschornack et al. (2016) in southern Brazil. These authors observed a mean reduction of ca. 60% in YpGWP with intermittent irrigation relative to CF (0.33 vs. 0.81 kg CO₂ eq kg⁻¹ rice) over two rice growing seasons.

These results are very important for southern Brazil regions and other rice-growing LAC countries, where rice production can be increased without increasing GHG emissions by adopting intermittent irrigation – a technology Asian farms are enjoying already – on commercial farms.

5. Conclusions

Seasonal soil CH₄ emissions under CF in northwestern Japan are slightly higher than they are in southern Brazil as a consequence of the increased contents in organic carbon of Japanese soil and also, probably, of seedling transplantation to a previously flooded field expanding the period under flooding relative to direct seeding in dry soil in southern Brazil. Intermittent irrigation systems in paddy fields are similarly efficient in reducing soil CH₄ emissions as CF systems in southern Brazil and northwestern Japan. In both areas, irrigation systems failed to significantly affect seasonal soil N₂O emissions or rice grain yield. Intermittent irrigation reduced yield-scaled pGWP by one-half in relation to CF in both study areas. Intermittent irrigation is a key agricultural practice with a view to meeting a future increase in rice production not associated to an increase in GHG emissions in southern Brazil and other countries of LAC, where this technology is still to be widely adopted the farmers.

Acknowledgments

The authors are grateful to the staff of the Niigata Agricultural Research Institute and the National Institute for Agro-Environmental Sciences in Japan, and also in the Riograndense Rice Institute in Brazil, for their help with field and laboratory activities. They also thank the National Council for Scientific and Technological Development (CNPq) and Foundation for Research Support of Rio Grande do Sul State (Fapergs)  for funding this work and the Coordination for the Improvement of Personnel with Graduate Education (CAPES) for award of a sandwich doctorate scholarship.

Additional information

Funding

The authors also thank the National Council for Scientific and Technological Development (CNPq) of Brazil for funding this work and the Coordination for the Improvement of Personnel with Graduate Education (CAPES) for award of a sandwich doctorate scholarship.