Site-specific feasibility of alternate wetting and drying as a greenhouse gas mitigation option in irrigated rice fields in Southeast Asia: a synthesis

ABSTRACT

This study comprises a comprehensive assessment, integration, and synthesis of data gathered from a 3-year field experiment conducted at four sites in Southeast Asia, namely Hue, Vietnam; Jakenan, Indonesia; Prachin Buri, Thailand; and Muñoz, Philippines, to assess the site-specific feasibility of alternate wetting and drying (AWD) as a greenhouse gas (GHG) mitigation option in irrigated rice fields. AWD effectively reduced water use compared to continuous flooding (CF) but did not significantly reduce rice grain yield and soil carbon content in all sites. Methane (CH₄) emissions varied significantly among sites and seasons as affected by soil properties and water management. AWD reduced CH₄ emissions relative to CF by 151 (25%), 166 (37%), 9 (31%), and 22 (32%) kg CH₄ ha⁻¹ season⁻¹ in Hue, Jakenan, Prachin Buri, and Muñoz, respectively. In Prachin Buri and Muñoz, AWD reduced CH₄ emissions only during the dry season. Site-specific CH₄ emission factors (EFs) ranged 0.13–4.50 and 0.08–4.88 kg CH₄ ha⁻¹ d⁻¹ under CF and AWD, respectively. The mean AWD scaling factors (SFs) was 0.69 (95% confidence interval: 0.61–0.77), which is slightly higher than the Intergovernmental Panel on Climate Change (IPCC)’ SF for multiple aeration of 0.52 (error range: 0.41–0.66). Significant reductions in the global warming potential (GWP) of CH₄+nitrous oxide (N₂O) by AWD were observed in Hue and Jakenan (27.8 and 36.1%, respectively), where the contributions of N₂O to the total GWP were only 0.8 and 3.5%, respectively. In Muñoz, however, CH₄ emission reduction through AWD was offset by the increase in N₂O emissions. The results indicate that the IPCC’s SF for multiple aeration may only be applied to irrigated rice fields where surface water level is controllable for a substantial period. This study underscores the importance of practical feasibility and appropriate timing of water management in successful GHG reductions by AWD.

KEY WORDS:

• Water management
• methane
• nitrous oxide
• emission factor
• soil carbon
• water productivity

1. Introduction

Rice is the only crop grown in predominantly flooded fields, and the subsequent anaerobic soil conditions stimulate the production of methane (CH₄), a potent greenhouse gas (GHG) (Conen et al. 2010; Hou et al. 2000; Wang et al. 1993; Wassmann and Dobermann 2006). CH₄ emissions from rice cultivation contribute substantially to the national GHG budgets of Asian countries (Wassmann et al. 2000a, 2000b). The two major GHGs, carbon dioxide (CO₂) and nitrous oxide (N₂O), are also exchanged between rice fields and the atmosphere. Global and regional estimates of GHG emissions from rice fields vary greatly depending on the assumptions made on the relative importance of different factors affecting the emissions. Studies have shown spatial variations in CH₄ emissions from continuously flooded rice fields with varying soil properties including sand and clay contents, aeration, compaction, moisture, pH, organic matter, available N, C/N ratio, and also with changes in soil management, crop rotation, and climates (Adhya et al. 1994; Gogoi et al. 2008; Inubushi et al. 1997; Kimura et al. 1991; Kumar and Viyol 2009; Minami and Neue 1994; Wassmann et al. 2000a, 2000b; Yang and Chang 1998; Yagi and Minami 1990; Yao et al. 1999). The Intergovernmental Panel on Climate Change (IPCC 2006) has given a baseline emission factor for CH₄ (EF_CH4) of 1.3 kg CH₄ ha⁻¹ d⁻¹ (with error range of 0.8–2.2), estimated by statistical analysis of available data (Yan et al. 2005) from continuously flooded rice fields without organic amendments. This baseline EF_CH4 can be adjusted using scaling factors (SFs) to account for various conditions, e.g., water regime, organic amendments, and rice cropping practices.

Alternate wetting and drying (AWD) is a water management practice that was developed and being extended by the International Rice Research Institute (IRRI) and its partners to many rice-producing countries, to reduce the consumption of irrigation water (Lampayan et al. 2004). Water management practices such as AWD (Bouman et al. 2007), mid-season drainage, and intermittent irrigation (Minamikawa et al. 2014) mitigate CH₄ emissions by periodic aeration that inhibits the activity of CH₄-producing archaea. Several studies have demonstrated the effective mitigation of CH₄ emissions from rice fields through soil and water management (Ahmad et al. 2009; Harada et al. 2007; Ishibashi et al. 2009; Itoh et al. 2011; Jiao et al. 2007; Ko et al. 2002; Yagi et al. 1996). Sander et al. (2015) carried out a literature survey of 24 independent studies on GHG mitigation by water management in China, India, Japan, South Korea, Indonesia, and Philippines, and showed a mean CH₄ reduction of 1.26 kg CH₄ ha⁻¹ d⁻¹ resulting from multiple aeration or multiple drainage, and 1.15 kg CH₄ ha⁻¹ d⁻¹ from single aeration or single drainage, as compared to continuous flooding (CF). They also reported a SF, the ratio of CH₄ emission under target practice to that under baseline (CF) practice, of 0.57 (range: 0.19–0.86) and 0.63 (0.18–1.52) under multiple aeration and single aeration, respectively.

The N₂O emissions from rice fields mainly occur just after soil flooding for rice cultivation, during temporal drainage events, and during fallow periods, and thus are generally higher under water management practices as compared to CF. Akiyama et al. (2005) compiled available data in the literature on N₂O emissions from fertilized rice fields and reported that the fertilizer-induced N₂O emission factors for CF, mid-season drainage, and all water regimes were 0.22, 0.37, and 0.31%, respectively. The last value has been adopted as the IPCC’s default N₂O emission factor (EF_N2O) for flooded rice (0.3%) though with a large error range (0.0–0.6%) (IPCC 2006).

A field monitoring campaign was conducted under a joint international research project at four sites: 1) Hue, Vietnam, 2) Jakenan, Indonesia, 3) Prachin Buri, Thailand, and 4) Muñoz, Philippines (Fig. 1). They represent diversified paddy environments in Southeast Asia. Results from each site were reported in separate papers of this issue (Tran et al. 2018; Setyanto et al. 2018; Chidthaisong et al. 2018; Sibayan et al. 2018). They monitored the changes in soil carbon (C) content, which may be lost by intensive drainage events, in addition to CH₄ and N₂O emissions under AWD and CF. Here, we present a synthesis and integration of the results from the four sites to assess the generalities and site-specificities of AWD as a mitigation option for GHG emissions from irrigated rice fields in Southeast Asia. This study (1) analyzed the effects of environment and soil properties on the extent of GHG mitigation by AWD, (2) examined the potential trade-offs and co-benefits in terms of rice grain yield and water savings from AWD, and (3) compared the obtained EF_CH4, the ratio of N₂O-N emission to applied N as inorganic fertilizer, and SF for CH₄ under AWD (SF_CH4-AWD) with the IPCC’s default values (IPCC 2006).

Figure 1. Locations of the four experimental sites.

2. Materials and methods

2.1. Data source

The data set analyzed in this study was generated from field experiments conducted under the international research project known as MIRSA-2. The project covered six rice seasons including both dry and wet ones (DS and WS) in 2013–2016 at four sites, namely Hue, Vietnam; Jakenan, Indonesia; Prachin Buri, Thailand; and Muñoz, the Philippines (Fig. 1). The detailed information on site locations, basic soil properties, cropping seasons, weather conditions, and field management are summarized in Table 1. Photographs of soil profile at the four sites are shown in Fig. 2.

Table 1. Location, soil properties, season, weather, and field management of the four experimental sites.

Site name	Hue	Jakenan	Prachin Buri	Muñoz
Location
Country	Vietnam	Indonesia	Thailand	Philippines
Province	Thua Thien	Central Java	Prachin Buri	Nueva Ecija
City	Hue	Jakenan, Pati	Bansang	Muñoz
Latitude (°N)	16.43	6.78	14.01	15.67
Longitude (°E)	107.51	111.20	101.22	120.89
Soil properties^a
Texture	Loam	Loam	Heavy clay	Clay
Clay (%)	17.8	22.5	62.9	46.2
Sand (%)	49.3	48.6	10.4	21.7
Silt (%)	32.9	28.8	26.7	32.1
pH (H2O)	4.36 ± 0.02	5.04 ± 0.21	3.53 ± 0.03	5.89 ± 0.11
pH (H2O2)	2.94 ± 0.05	3.33 ± 0.21	2.10 ± 0.06	4.43 ± 0.02
Bulk density (g cm^−3)	1.01	1.54	1.43	1.21
Active Fe (g kg^−^1)	12.3 ± 0.0	3.6 ± 0.0	15.6 ± 0.1	12.6 ± 0.3
Active Mn (g kg^−^1)	0.03 ± 0.01	0.05 ± 0.00	0.07 ± 0.00	0.68 ± 0.01
Total C (g kg^−^1)^b	19.2 ± 1.1	5.3 ± 0.3	22.2 ± 1.2	14.6 ± 1.1
Total N (g kg^−^1)^b	1.88 ± 0.10	0.47 ± 0.02	1.80 ± 0.10	1.20 ± 0.10
C:N mole ratio^b	11.9 ± 0.2	13.3 ± 1.0	14.6 ± 0.4	14.0 ± 0.8
Season/weather
Crop calendar	Dry: January–March Wet: April–August WF: September–December	Wet: Nov–March Dry: March–July DF: July–November	Dry: January–April Wet: June–September WF: September–December	Dry: January–April DF: April–June Wet: June–September WF: September–December
Rainfall (mm)	Dry: 251 Wet: 283 WF: 1642	Wet: 966 Dry: 336 DF: 158	Dry: 99 Wet: 948 WF: 408	Dry: 25 DF: 253 Wet: 835 WF: 470
Air temperature (°C, min.-max.)	Dry: 11.3–37.3 Wet: 22.8–38.7 WF: 19.2–37.7	Wet: 24.0–39.0 Dry: 23.7–39.5 DF: 22.9–39.8	Dry: 24.0–35.5 Wet: 25.5–33.8 WF: 24.5–33.7	Dry: 21.8–31.1 DF: 24.5–35.9 Wet: 23.3–30.3 WF: 23.1–31.9
Field management
Variety	HT1	Cisadane	RD41	NSIC Rc238
Growth duration (day)	96–120	107–132	88–98	81–98
Crop establishment	Wet direct seeding	Wet: Direct seeding Dry: Transplanting	Pre-germinated seed broadcasting	Transplanting
Inorganic fertilizer
N (kg N ha^−^1)	92–120	120	70	90–120
P (kg P2O5 ha^−^1)	72	60	37.5	40
K (kg K2O ha^−^1)	62–78	90	37.5	40
Organic amendment	Microbial organic fertilizer	Farmyard manure	None	None
Residue management	Rice straw removal	Rice straw removal	Rice straw removal	Rice straw removal

^aRange of standard deviation.

^bThe average value of yearly samplings from all treatments during the experiment (see Fig. 4 for sampling time).

WF: Wet fallow; DF: Dry fallow.

Figure 2. Comparison of the soil profiles in the four experimental sites. The soil profiles are arranged according to the apparent soil redox condition with showing the USDA and FAO classifications.

Figure 3. Relationships between (A) predicted SF_CH4-AWD and observed SF_CH4-AWD and (B) observed SF_CH4-AWD and NNFD for the data in Hue and Jakenan (n = 24; 2 sites × 6 seasons × 2 practices (AWD and AWDS)). In A panel, dotted line indicates the linear regression between predicted and observed SF_CH4-AWD. In B panel, solid and dotted lines indicate the multiple regression line with MinWL of −15 cm and the linear regression between observed SF and NNFD, respectively.

Figure 4. Interannual variations in total carbon content at a soil depth of 0–20 cm among three water management practices in (A) Hue, (B) Jakenan, (C) Muñoz, and (D) Prachin Buri. Vertical bars indicate the standard deviation (n = 3–4).

The experimental plots in each site were laid out in a randomized complete block design with three treatments and three blocks except for Muñoz with four blocks. The treatments were (1) CF, (2) AWD, and (3) site-specific AWD (AWDS). In AWD, the water management followed what is called safe-AWD. The field was irrigated when the surface water level reached 15 cm below the soil surface after which the field was re-flooded to around 5-cm level above the soil surface throughout the cropping season except for the flowering stage (Sander et al. 2015). The method of implementing AWDS varied among sites, as briefly summarized below. In Hue, the threshold of re-flooding changed depending on rice growth stage. That is, the plots were irrigated when the water level dropped to 5 cm, 10 cm, and 15 cm below the soil surface at early tillering stage, late tillering stage, and grain filling and ripening stage, respectively (Tran et al. 2018). In Jakenan, multiple aeration 7 days before the 1st and 2nd N fertilizer topdressing was implemented as AWDS in the first two seasons, and in the succeeding seasons, the threshold of re-flooding changed to 25 cm below the soil surface (Setyanto et al. 2018). In Prachin Buri, the plots were re-flooded to 10-cm level under AWDS as compared to 5-cm level under AWD (Chidthaisong et al. 2018). In Muñoz, AWDS was implemented as a 7–10-day-long single aeration (i.e., mid-season drainage) in the first two seasons. In the succeeding seasons, it was implemented as the AWD with the threshold of 25 cm below the soil surface. The starting timing of AWD and AWDS was changed from 21 days after transplanting (DAT) to 10 DAT from the 5th season to achieve drier soil conditions (Sibayan et al. 2018).

Water use expressed as m³ ha⁻¹ (Table 3) is the amount of irrigation water used at each site that was estimated using a flow meter connected to the irrigation pump, plus the total amount of rainfall recorded for each season. Water productivity expressed as kg grain m⁻³ (Table 3) was estimated from the grain yield divided by the total amount of irrigation water including rainfall.

Table 2. Comparison of alternate wetting and drying (AWD) implementation among the four experimental sites^a.

Criteria	Water management	Hue	Jakenan	Prachin Buri DS	Muñoz DS	Muñoz WS
Minimum water level (cm)	AWD	−18.3 ± 6.2 (−25.4 to −7.7)	−23.1 ± 13.5 (−46.3 to −11.3)	−12.5 ± 5.6 (−15.4 to −4.0)	−13.9 ± 1.0 (−15.0 to −13.1)	−9.5 ± 4.2 (−12.5 to −6.5)
	AWDS	−13.0 ± 4.9 (−20.1 to −6.3)	−32.1 ± 18.0 (−63.8 to −19.2)	−13.0 ± 4.5 (−15.4 to −6.3)	−13.3 ± 6.3 (−19.1 to −6.6)	−8.6 ± 0.6 (−9.0 to −8.1)
Number of non-flooded days	AWD	30.2 ± 5.8 (22 to 39)	30.7 ± 7.9 (20 to 41)	24.3 ± 3.8 (20 to 27)	34.0 ± 10.5 (24 to 45)	10.5 ± 9.2 (4 to 17)
	AWDS	29.3 ± 8.5 (14 to 36)	27.2 ± 11.9 (14 to 44)	13.3 ± 4.7 (8 to 17)	31.7 ± 19.0 (12 to 50)	6.0 ± 2.8 (4 to 8)
Ratio of non-flooded days to total days^b	AWD	0.31 ± 0.08 (0.16 to 0.41)	0.28 ± 0.08 (0.16 to 0.39)	0.26 ± 0.03 (0.23 to 0.29)	0.39 ± 0.19 (0.21 to 0.60)	0.13 ± 0.11 (0.05 to 0.21)
	AWDS	0.30 ± 0.09 (0.16 to 0.41)	0.24 ± 0.11 (0.11 to 0.37)	0.14 ± 0.03 (0.09 to 0.17)	0.38 ± 0.27 (0.13 to 0.67)	0.07 ± 0.04 (0.05 to 0.10)
Average water level (cm)	AWD	−0.1 ± 2.2 (−3.0 to 3.6)	0.8 ± 0.6 (−0.1 to 1.7)	−0.5 ± 2.7 (−1.2 to 0.4)	−1.4 ± 1.1 (−2.7 to −0.9)	1.9 ± 0.7 (1.4 to 2.4)
	AWDS	−0.1 ± 1.9 (−3.4 to 2.4)	0.1 ± 1.3 (−1.5 to 1.7)	3.3 ± 1.0 (2.2 to 4.3)	−2.0 ± 4.8 (−7.4 to 1.9)	2.9 ± 1.1 (2.1 to 3.8)
Drainability^c (mm hr^−^1)		0.94 ± 0.28	1.85 ± 0.80	0.78 ± 0.29	0.67 ± 0.24	0.67 ± 0.24

^aMean value ± standard deviation across seasons with the minimum to maximum range in parentheses.

^bTotal days from seeding to final drainage.

^cObserved rate of decline of the surface water level, including evapotranspiration.

Table 3. Comparisons of the effects of AWD on greenhouse gas emissions, rice productivity, and water use among the four experimental sites.

	CH4 (kg CH4 ha^−1)		N2O (kg N2O ha^−1)		N2O contribution to GWP (%)		GWP (kg CO2 ha^−1)		Grain yield (Mg ha^−1)		Yield-scaled GWP (Mg CO2 Mg^−1 grain)		Water use (m^3 ha^−1)		Water productivity (kg grain m^−3)
Site	DS	WS	DS	WS	DS	WS	DS	WS	DS	WS	DS	WS	DS	WS	DS	WS
Value under continuous flooding (CF)
Hue	499.5	644.2	0.28	0.70	0.6	1.0	17,030	23,540	4.41	4.44	3.49	4.89	8127	8394	0.56	0.54
Jakenan	385.3	515.4	0.81	1.13	2.2	2.3	13,342	17,861	5.12	6.87	2.67	2.61	6635	12,521	0.83	0.57
Prachin Buri	13.4	23.0	0.97	0.22	39.3	5.6	746	1190	4.64	4.28	0.15	0.23	7119	11,332	0.71	0.45
Muñoz	69.9	328.9	1.60	0.51	20.6	1.4	2853	11,333	6.90	5.41	0.43	2.10	10,336	10,944	0.77	0.51
Change under AWD relative to CF (%)
Hue	−25.8	−23.8	−26.3	−26.3	−21.8	12.7	−30.8	−24.8	10.3	4.5	−33.6	−21.8	−14.4	−14.9	30.1	25.6
Jakenan	−36.7	−36.8	−13.5	12.1	26.4	85.9	−36.3	−35.9	−0.82	−1.4	−36.8	−35.4	−6.4	−5.7	4.6	3.3
Prachin Buri	−31.2	–	19.3	–	11.8	–	−11.5	–	−12.6	–	−3.2	–	−29.8	–	17.6	–
Muñoz	−32.1	14.1	91.3	4.7	95.8	25.1	−11.4	14.0	−0.13	1.9	−9.3	9.1	−47.1	−17.0	69.0	22.2
P-value
Site (S)	***		***		–		***		***		***		***		***
Dry or Wet (DW)	**		0.209		–		**		0.484		**		***		**
Water management (WM)	***		0.269		–		**		0.638		0.326		***		***
S × DW	***		***		–		***		***		***		***		***
WM × S	0.129		0.711		–		*		0.594		†		***		*
WM × DW	0.829		0.959		–		0.778		0.295		0.829		0.238		0.931
WM × DW × S	0.901		0.894		–		0.914		0.140		0.955		*		*

– data not available.

***, **, *, and † indicate significant difference at p = 0.001, 0.01, 0.05, and 0.1, respectively.

2.2. Gas measurement and soil analyses

The CH₄ and N₂O emissions were measured by a manual closed chamber method. Standardized protocols were commonly used to ensure precise and accurate gas flux measurements based on the guidelines given by Minamikawa et al. (2015). Gas sampling was done between 9 and 11 AM, once a week and 5 consecutive days after N fertilization. The total number of samplings conducted per season was 16–20 in Hue, 22–29 in Jakenan, 16–28 in Prachin Buri, and 15–21 in Muñoz. Gas samples were taken 5 times, at 1, 6, 12, 20, and 30 mins after chamber deployment and stored in evacuated glass vials until analysis using gas chromatography. Hourly gas flux was calculated using linear regression and the seasonal emission was computed as the sum of daily fluxes. Daily fluxes in between measurements were estimated from a linear interpolation of two consecutive measurements.

Soil (0–20 cm depth) samples were collected annually from the three treatments in each of the three replicate blocks at each site except in Muñoz with four replicate blocks. Air-dried samples were sent to IRRI in the Philippines after the 3-year experiment for analyses of pH in 1:1 soil:water suspension, pH after oxidation with 30% hydrogen peroxide (H₂O₂) (Ahern et al. 1998), total C and N, active iron (Fe), and active manganese (Mn), to eliminate the interlaboratory variability in accuracy and precision. Total C and N were analyzed by a dry combustion method. Active Fe and Mn were analyzed by sodium dithionite extraction and inductively coupled plasma optical emission spectrometry (Asami and Kumada 1959).

2.3. Calculation of emission-related parameters

The parameters included EF_CH4 under CF and AWD (EF_CH4-CF and EF_CH4-AWD), SF_CH4-AWD, the ratios of N₂O-N emission to applied N as inorganic fertilizer under CF and AWD, and GWPs under CF and AWD (GWP_CF and GWP_AWD). It is noted that EF_CH4-CF and EF_CH4-AWD calculated here are not exactly the same as the IPCC’s Tier-1 EFs (IPCC 2006). Rather, these are site-specific and management-specific EFs (e.g., in Jakenan, CF and AWD with the application of farmyard manure), and thus can be considered as the Tier-2 approach that is suitable for country-specific estimations.

Data from AWD and AWDS from each block were combined as AWD in calculating all the parameters as mentioned in the subsection 3.2. The EF_CH4-CF and EF_CH4-AWD (kg CH₄ ha⁻¹ d⁻¹) were calculated by dividing the seasonal (cumulative) emission by the growth duration. The ratios of N₂O-N emission to applied N as inorganic fertilizer under CF and AWD (%) were calculated by dividing the seasonal (cumulative) N₂O-N emission by the amount of inorganic fertilizer N applied. Site means across seasons were weighted against the variances obtained from three or four replicate (block) measurements in each season. For estimating weighted means and 95% bootstrapped confidence interval (CI) of SF_CH4-AWD, the LN (natural log or log base-e) response ratio (LN (EF_CH4-AWD/EF_CH4-CF) for each site and combined for all sites was calculated using MetaWin (ver.2.1; Rosenberg et al. 2000). Treatment effects were considered to be significantly different from one another if their 95% CIs did not overlap and were considered significant if the 95% CI did not overlap with zero.

The weighted means for Hue and Jakenan were based on six-season data, but for Muñoz, separate means were calculated for WS and DS due to the significant differences in emissions between them. In Prachin Buri, no data were generated in the 4th season as planting was canceled due to flooding. The data for AWD in the 6th season from Prachin Buri and Muñoz were excluded from analyses since soil drainage under AWD (i.e., surface water level below 0 cm) was not achieved during this season. In Hue, the N₂O data from the 5th and 6th seasons were excluded from the analyses due to the malfunction of gas chromatograph. For the same reason, N₂O data from the 2nd season in Prachin Buri were excluded from analyses.

The GWP_CF and GWP_AWD of CH₄ and N₂O emissions were calculated using the IPCC’s factors with inclusion of climate-carbon feedbacks (34 for CH₄ and 298 for N₂O; Myhre et al. 2013).

2.4. Quantitative assessment of soil drying under AWD and statistical analyses

We set the following five criteria to quantitatively assess the degree of soil drying under AWD: (1) minimum water level below the soil surface (MinWL), (2) number of non-flooded days (NNFD), (3) ratio of NNFD to total number of rice cropping days, (4) average surface water level during rice cropping period, and (5) soil drainability. The soil drainability was the apparent decline rate of surface water level that included plant evapotranspiration. Target period of the five criteria was from planting day until the last day before final drainage prior to harvest. Multiple linear regression analysis with stepwise selection was done using SAS software ver. 9.4 (SAS Institute Inc., Cary, NC, USA) to determine if any of the five measured criteria could be used to predict SF_CH4-AWD.

Analysis of variance (ANOVA) of data from the four sites was performed with a mixed model (‘proc mixed’) in the SAS software to assess the main effects of site (the four experimental sites), DS or WS (DW in Table 3), water management (WM in Table 3), and their interactions. A split-plot model was applied in which experimental site was treated as the whole-plot factor, DW as the split-plot factor, and WM as the split-split-factor. Variance components were estimated by the restricted maximum-likelihood method with the ‘nobound’ option, and the denominator degrees of freedom were estimated by the Kenward–Roger approximation (Kenward and Roger 1997; Littell et al. 2006). Because emissions of CH₄ and N₂O showed highly skewed distributions and violated normality and homoscedasticity assumptions, the Box-Cox transformation was conducted for CH₄, N₂O, GWP, and yield-scaled GWP using the powerTransform function in the car package of R (Box and Cox 1964; Fox and Weisberg 2011).

3. Results

3.1. Inter-site variations in soil properties, weather conditions, and field management practices

Soil texture differed from loam in Hue and Jakenan to heavy clay in Prachin Buri (Table 1). The total C and N contents ranged from 5.3–19.2 g kg⁻¹ and 0.47–1.9 g kg⁻¹, respectively, with the lowest values in Jakenan and the highest in Prachin Buri. The C:N mole ratio was comparable among Jakenan, Muñoz, and Prachin Buri, but that in Hue was relatively low. The active Fe content of Jakenan soil was the lowest among sites. Active Mn content was the highest in Muñoz.

The soil profiles at the four sites are shown in Fig. 2. Hue and Prachin Buri soils were wet and had reduced conditions as indicated by the bluish black color compared to Jakenan and Muñoz soils with relatively dry and more oxidized conditions revealed by the brown color. Prachin Buri soil differed from Hue soil in that the former had distinct dark yellowish brown mottles in the subsurface horizons indicating seasonal oxidation in the soil pores. Thus, among the sites, Hue soil was considered as having the most reduced condition followed by Prachin Buri. Prachin Buri soil was classified as acid, sulfic Endoaquepts originating from the deposition of seawater or brackish water sediment. Prachin Buri soil collected at the time of the experiment had an average pH(H₂O) of <4 (Table 1), which confirms that it is an actual acid sulfate soil (Ahern et al. 1998). Further, an average pH(H₂O₂) of <3 confirms the presence of sulfides that have not been oxidized. On the other hand, Hue soil with pH(H₂O) of >4 and pH(H₂O₂) of <3 indicates that it has the potential to become an actual acid sulfate soil. However, at the time of soil sampling, Hue soil may not be considered as acid sulfate. Muñoz soil exhibited vigorous reaction with H₂O₂ but the resulting pH(H₂O₂) was >4. The vigorous reaction may have been due to reactions of H₂O₂ with manganese, which had the highest concentration in Muñoz soil (Table 1).

Between Jakenan and Muñoz soils, the latter was thought to be under more oxidized conditions. Muñoz soil was classified as Ustic Epiaquert derived from Alluvium parent material and was poorly drained, while Jakenan soil, Aeric Endoaquepts, was a wet soil influenced by groundwater. But its relatively high chroma indicated either a shorter period of saturation of the whole soil profile with water or somewhat deeper groundwater than that in the soils of the Typic subgroup.

Cropping calendar differed among sites depending mainly on irrigation water availability and rainfall pattern (Table 1). In Hue and Prachin Buri, the fallow period coincided with months of continuous rainfall when inundation usually occurs. On the other hand, the fallow period in Jakenan coincided with dry months when irrigation water is scarce. In Muñoz, a 2- to 3-month dry or wet fallow period follows each crop of rice (Table 1).

Crop and fertilizer management practices also differed among sites (Table 1). Farmyard manure and an organic microbial fertilizer were applied in Jakenan and Hue, respectively, which are common in these areas. The amounts of inorganic N applied were the lowest in Prachin Buri. In Jakenan, a rice variety with relatively long growth duration was planted. In all sites, harvested rice straw was removed from the field while the rice stubbles were incorporated during soil puddling for the succeeding rice crop. In Muñoz, only in the 5th season, an earlier incorporation of rice stubbles was done under dry soil conditions (Sibayan et al. 2018).

3.2. Inter-site variation in soil drying under AWD

Table 2 shows the calculated averages, with ranges and standard deviations, of the five criteria of soil drying under AWD and AWDS in the four sites. The water level threshold of −15 cm for safe AWD was not always achieved (i.e., water level did not decline to −15 cm), but was often exceeded in Hue and Jakenan due to the drier weather conditions and soil properties. There were no substantial differences in the five criteria between DS and WS in Jakenan, and between winter-spring season and summer-autumn season in Hue, so the overall site means are shown for these sites. In Prachin Buri and Muñoz, AWD was not achieved during WS (data not shown for Prachin Buri). In Muñoz, remarkable differences between DS and WS were observed in the NNFD and the average surface water level. Thus, data for both DS and WS are shown. The NNFD covered 28–39% of the cropping period in Muñoz DS, Hue, and Jakenan whereas it covered only 13–26% in Muñoz WS and Prachin Buri. The highest soil drainability was observed in Jakenan.

The NNFD and MinWL were identified as the sole significant (p < 0.05) predictors of SF_CH4-AWD by stepwise selection and multiple regression analysis of data from the loamy soils of Hue and Jakenan (Tables S1 and S2). The two predictors explained 41% of the variability in SF_CH4-AWD in the loamy soils as indicated by the model R². Consequently, the following regression equation was derived from the multiple linear regression analysis for the loamy soils:

(1)

A significant linear relationship was obtained between the predicted and the observed SF_CH4-AWD (Fig. 3A). Solid line calculated using Eq. 1 shows the predicted SF at varying NNFDs when MinWL was fixed at −15 cm, which is the threshold set for safe AWD (Lampayan et al. 2004) (Fig. 3B). We did not obtain a significant multiple regression equation to predict SF_CH4-AWD for the clay soils and for the combined clay and loam soils.

Based on the calculated five criteria, there was no clear distinction between soil drying by AWD and AWDS at each site as indicated by the overlapping error ranges (Table 2). We therefore combined the data from AWD and AWDS in the following sections to represent AWD in general.

3.3. GHG emission, soil C and N contents, grain yield, and water use

The results of ANOVA showed significant (p < 0.05) effects of site, season (DW; DS vs. WS), and water management (WM) on the seasonal CH₄ emissions (Table 3). The highest CH₄ emission under CF was observed in Hue, while the emissions in Muñoz DS and Prachin Buri were much lower. The CH₄ emission reduction by AWD, which includes AWDS as mentioned above, relative to CF was observed in all the sites except for Muñoz WS, as indicated by the significant site × DW interaction. In absolute amounts, the emission reduction during DS in Hue, Jakenan, Prachin Buri, and Muñoz were 148, 190, 9, and 22 kg CH₄ ha⁻¹ season⁻¹, respectively, and those during WS in Hue, Jakenan, and Prachin Buri were 154, 141, and 8 kg CH₄ ha⁻¹ season⁻¹, respectively. In Muñoz WS, no CH₄ emission reduction was achieved through AWD due to the continuous heavy rainfall (Sibayan et al. 2018). The CH₄ emissions under CF were generally higher during WS than DS.

There was no significant effect of WM on the seasonal N₂O emission, but an emission increase under AWD was observed in Muñoz, Jakenan, and Prachin Buri (Table 3). The season effect (DS vs. WS) on N₂O emission was not significant but a significant site × season interaction was observed (Table 3). Significant differences in N₂O emissions during DS and WS were observed in Prachin Buri and Muñoz but not in Hue and Jakenan.

There were significant effects of WM and site on the GWP of CH₄ and N₂O with significant WM × site interaction (Table 3). The fraction of N₂O to the GWP under CF was much higher in Muñoz DS and Prachin Buri with clayey soils than in Hue and Jakenan with loamy soils. AWD increased the N₂O fraction in all the sites, except for Hue DS. Moreover, 32% CH₄ reduction through AWD in Muñoz DS was partially offset by 96% increase in N₂O emissions, resulting in lower GWP reduction by AWD (11.4%) as compared to the CH₄ reduction (32.1%) (Table 3).

The total C and N contents in the topsoil (0–20 cm) did not significantly differ among WM on each sampling date at each site (Fig. 4 for total C; data not shown for total N). Further, AWD showed no significant effect on rice grain yield and yield-scaled GWP (Table 3). Water use and water productivity, the ratio of grain yield to water use, were significantly affected by WM with the significant WM × site interaction (Table 3). The reduction in water use through AWD relative to CF was remarkably higher in Muñoz DS and Prachin Buri than in Hue and Jakenan. Water productivity was improved through AWD from 3.3% in Jakenan WS to 69.0% in Muñoz DS.

3.4. Site-specific EFs and SFs

The EF_CH4-CF and EF_CH4-AWD are shown in Fig. 5A. Data from DS and WS in Muñoz are separately presented to reflect the season-specific responses of CH₄ emission to AWD. The EF_CH4-CF and EF_CH4-AWD in Hue and Jakenan were significantly higher than those in Prachin Buri and Muñoz DS. The highest EF_CH4-CF was obtained in Hue followed by Jakenan and Muñoz WS. Lower EF_CH4-AWD than EF_CH4-CF was obtained in all the sites except Muñoz during WS. The weighted-mean EF_CH4-CF across sites and seasons including Muñoz WS (2.46 ± 0.56 kg ha⁻¹ d⁻¹) was slightly higher but not significantly different from the IPCC’s default EF for flooded rice (1.30 kg ha⁻¹ d⁻¹ with error range of 0.80–2.20; IPCC 2006). However, it should be noted that the EFs_CH4 reported here were estimated under site-specific management practices (e.g., the application of farmyard manure in Jakenan), while the IPCC’s default EF was derived from measurements without organic amendments. The baseline EF_CH4 in Jakenan that was measured under CF without organic amendments in a neighboring field (2.7 kg ha⁻¹ d⁻¹; Pramono and Setyanto 2015) was 34% lower than that reported with farmyard manure in this study.

Figure 5. CH₄ emission-related parameters in the four experimental sites. (A) Emission factors (EF) for CF and AWD. (B) AWD scaling factors (SF). Data in dry season are shown for Prachin Buri. Data in dry and wet seasons are shown separately for Muñoz. Mean data in dry and wet seasons are shown for Hue and Jakenan. Means overall sites and seasons with and without Muñoz wet season are separately shown. The horizontal bars for each site indicate the 95% confidence intervals. Bars for IPCC (2006) indicate the error range given in the IPCC (2006) guidelines. In A panel, the values for Prachin Buri CF and AWD are 0.024 ± 0.027 and 0.031 ± 0.041, respectively.

The SF_CH4-AWD ranged from 0.57 to 0.84 (excluding Muñoz WS) and did not significantly differ among sites (Fig. 5B). The weighted-mean SF_CH4-AWD of 0.69 (95% bootstrapped CI: 0.61–0.77) was estimated for across sites and seasons excluding Muñoz WS. This value was within the range of multiple aeration by Sander et al. (2015) (mean: 0.57; error range 0.19–0.86), but slightly higher (i.e., weaker mitigation effect) than the IPCC’s SF for multiple aeration of 0.52 (error range: 0.41–0.66; IPCC 2006).

The mean ratio of seasonal N₂O-N emission to inorganic fertilizer N applied across sites was 0.42% under CF and 0.49% under AWD including Muñoz WS (Fig. 6). These were comparable to those derived from the original data in Akiyama et al. (2005) (0.38 ± 0.41% under CF and 0.67 ± 0.22% under mid-season drainage; the error range expressed as 95% CIs recalculated from the original data). However, as indicated by the overlap of the 95% CIs, the ratio under AWD did not differ from that under CF in all sites. This agrees with the results of ANOVA that showed no significant effect of WM on N₂O emissions (Table 3).

Figure 6. The ratio of seasonal N₂O-N emissions to the total amount of chemical fertilizer N applied. Data for dry season and wet season are shown separately for Muñoz. Mean data in dry and wet seasons are shown for Hue, Jakenan, and Prachin Buri. Means overall sites and seasons with and without Muñoz wet season are separately shown. The horizontal bars indicate the 95% confidence intervals.

4. Discussion

4.1. Generalities and site-specificities of soil drying under AWD

This study used the five criteria to quantitatively describe site-specificities in AWD implementation (Table 2). AWD successfully reduced the seasonal CH₄ emission at varying levels, except for Muñoz WS and Prachin Buri WS (Table 3). In Prachin Buri and Muñoz, AWD was unsuitable during WS due to the frequent rainfall and the slow drainability of clayey soils. On the other hand, AWD was effective during WS as well as DS in Hue and Jakenan, both of which had a loamy soil.

Using the derived multiple regression equation for Hue and Jakenan data (Eq. 1), it was extrapolated that a 30% reduction in CH₄ emission (i.e., SF_CH4-AWD = 0.70) is more likely to be achieved with the safe AWD water level threshold of −15cm when NNFD is >32 (solid line in Fig. 3B). However, this relationship held only to the loamy soils in this study. Although large variabilities in CH₄ emission reduction through AWD or multiple aeration have been reported elsewhere (Sander et al. 2015), the corresponding data on surface water level were not often documented, making it difficult to explain the variability. Our results demonstrated, for the loamy soils of Hue and Jakenan, that the observation of surface water level and NNFD can improve the precision in SF_CH4-AWD estimation.

4.2. Environmental and soil properties affecting the extent of GHG mitigation by AWD

The loamy soils of Hue and Jakenan exhibited faster drainability of surface water than the clayey soils of Muñoz and Prachin Buri (Table 2), thereby enabling frequent short-term drainage events. For this reason, loamy soils may be regarded more suitable for CH₄ emission reduction by AWD implementation. Yagi et al. (1996) demonstrated that short-term (i.e., the period from disappearance of floodwater to re-flooding was short—only about a week) drainage practices strongly reduced CH₄ emissions from Japanese rice paddy fields and that CH₄ emission rates significantly decreased with an increase in the percolation rates (Yagi et al. 1998). On the other hand, the clayey soils of Prachin Buri and Muñoz DS achieved more water savings (17–47%) than the loamy soils of Jakenan and Hue (6–15%) through AWD (Table 3). Water use was the highest among the four sites in Muñoz DS under CF, but was the lowest under AWD. Thus, water saving efficiency under AWD was higher in the clayey soils than in the loamy soils.

Higher soil C content, organic fertilizer amendment, and lower C:N mole ratio that allows faster decomposition of soil organic matter are the major contributing factors to the highest CH₄ emission observed in Hue among sites. Moreover, Hue soil had the most reduced soil condition (Fig. 2). Although Prachin Buri soil had the highest soil C content, the observed CH₄ emission was the lowest (13.4 kg CH₄ ha⁻¹). The soil pH(H₂O) and pH(H₂O₂) measurements (Table 1) confirmed that it is an acid sulfate soil (Ahern et al. 1998). The soil active Fe content in Prachin Buri was also the highest among sites (Table 1). The presence of sulfate and active Fe slows down soil reduction and CH₄ production, by acting as oxidizing agents in paddy soils (Takai et al. 1963; Inubushi et al. 1997; Yagi et al. 1997). In Thai paddy fields, CH₄ emission during a rice growing period ranged from 53 to 787 kg CH₄ ha⁻¹ in freshwater alluvial soils whereas only 8–227 kg CH₄ ha⁻¹ in an acid sulfate soil (Jermsawatdipong et al. 1994). Jugsuhinda et al. (1996) suggested that CH₄ and CO₂ productions in flooded acid sulfate soils of Thailand were primarily governed by soil oxidation-reduction potential (Eh) and pH, and reported that the critical Eh and pH levels at which CH₄ emission occurred were −150 mV and 6.1, respectively. Therefore, low soil pH, which limits soil reduction and the subsequent CH₄ production, was the possible factor for the low CH₄ emission in Prachin Buri.

Jakenan soil had the lowest soil total C content among sites (Table 1) but the EF_CH4-CF was the second largest after Hue (Fig. 5A) due mainly to the application of farmyard manure. Without the manure application, EF_CH4-CF was 2.7 kg ha⁻¹ d⁻¹ in the same site (Pramono and Setyanto 2015). Muñoz soil had higher total C and active Fe and Mn contents than Jakenan soil (Table 1). Although higher soil organic C can increase CH₄ production and emission, higher active Fe and Mn contents, on the other hand, can reduce CH₄ emission (Ali et al. 2009; Smakgahn et al. 2009). Thus, in Muñoz DS, the high Fe and Mn contents would have primarily contributed to the low EF_CH4-CF. On the other hand, reasons for the high CH₄ emissions in Muñoz WS were provided by Sibayan et al. (2018) as follows: (1) higher mean air temperature in WS than in DS enhanced the release of rice root exudates (i.e., C substrates for CH₄ production) into the soil; (2) greater amount of rice residues (i.e., C substrates) coming from the previous DS crop was incorporated during WS; and (3) considerable amount of rainfall in WS resulted in higher surface water levels both under CF and AWD. In the same Muñoz site, Corton et al. (2000) reported 2–3 times higher CH₄ emissions during WS compared to DS.

Production of N₂O in the soil results primarily from microbial nitrification and denitrification processes. Major drivers of these processes are carbon and nitrogen substrate availability, temperature, pH, and moisture content (Mosier et al. 1998; Bouwman et al. 2002; Linquist et al. 2012; Rochette et al. 2008). A laboratory study showed that N₂O emission occurred at a redox value of 0 mV (Kralova et al. 1992) while Granli and Bøckman (1994) reported that the soil water content associated with maximum N₂O emission was close to the field capacity, suggesting that AWD implementation may stimulate N₂O emissions. Increased N₂O emissions under AWD were observed in Jakenan WS, Muñoz, and Prachin Buri but which were not enough to completely offset the reductions in CH₄ emission; while no increase in N₂O emission was observed in Hue even with AWD. Yagi et al. (1996) also reported very low N₂O emissions from a Japanese paddy field under intermittent irrigation similar to that in Jakenan. It has been reported that clayey soils tend to exhibit greater N₂O emissions than sandy soils (Brentrup et al. 2000). This agrees with the higher N₂O emissions observed in the clayey soils of Muñoz and Prachin Buri than in the loamy soils of Hue and Jakenan.

The contributions of N₂O to the total GWP of CH₄+N₂O under AWD were only 0.8 and 3.5% in the loamy soils of Hue and Jakenan, respectively, but were much higher in the clayey soils of Prachin Buri and Muñoz during DS (44 and 40%, respectively). The oxygen and moisture status in agricultural soils depends on soil texture and drainage. Fine textured soils have more capillary pores within aggregates holding water more tightly than do coarse soils. As a result, anaerobic conditions may be easily developed and maintained for longer periods within the aggregates, thereby creating more suitable conditions for denitrification than in coarse textured soils (Bouwman et al. 2002). In Muñoz, a large increase in N₂O emissions under AWD during DS occurred just after N fertilization, especially when the soil was not irrigated (Sibayan et al. 2018). In the loamy soils of Hue and Jakenan, the N₂O emissions were minimized by keeping soil flooded after N fertilizer application as shown in a study by Furukawa et al. (2007).

4.3. Trade-offs and co-benefits from AWD

The trade-off between CH₄ and N₂O emissions under AWD (i.e., increase in N₂O contribution to GWP) was observed in all the four sites, but the degree of which was not enough to offset the CH₄ emission reduction (Table 3). However, N₂O contribution to GWP was considerably high in Muñoz DS and Prachin Buri because the magnitude of CH₄ emissions was relatively low among sites and as compared to IPCC’s default EFs (Fig. 5A). This study also found that total C content in the topsoil was not significantly decreased by AWD implementation in all the sites after the 3-year field trials (Fig. 4). The dynamics in soil C is often overlooked in a field study, especially in tropical region; however, this field campaign demonstrated that multiple soil aerated conditions developed by AWD have no negative impact on the total C content in the topsoil at least for 3 years.

It should be noted that the AWD implementation did not significantly reduce rice grain yield (Table 3). Lampayan et al. (2004) argued that safe AWD with the threshold of −15 cm would not cause any yield reduction because rice roots would still be able to uptake water from the saturated soil. Our results therefore confirmed the findings of Lampayan et al. (2004). In addition, AWDS that modified to a greater or lesser extent from AWD also exhibited no significant yield penalty.

AWD successfully achieved the water saving and improved the water productivity in all the sites, especially in Muñoz DS (Table 3). This result supports the original purpose of implementing (safe-)AWD (Bouman et al. 2007). However, the degree of water saving differed between soil types (clayey vs. loamy). Furthermore, this study found that the degree of water saving was not always consistent with that of CH₄ emission reduction in the absolute value. That is, high water savings through AWD do not always translate to high CH₄ emission reduction due to a greater influence of soil properties. This raises a question that what benefits we expect for implementing AWD. Because AWD is known as a mitigation option for paddy GHG emissions so far, the primary role of AWD is now being questioned.

4.4. Implications to GHG inventories

The emission-related parameters estimated in this study represent practical ones since they were obtained in the 3-year field trial with varying weather conditions. These parameters can be applied to estimate CH₄ emissions in GHG inventories, when the emissions are calculated for a category of paddy fields in which AWD or similar intermittent flooding management were conducted, especially in tropical regions. Also, those parameters are expected to apply for evaluating the effects of alternative water management in rice cultivation as a GHG mitigation option in various GHG mitigation schemes, such as UNFCCC Clean Development Mechanisms, national appropriate mitigation actions, and other emission trading approaches.

The EFs_CH4 obtained in this study greatly varied against the IPCC’s default values (Fig. 5A) partly because they were obtained under site-specific management practices. This result also indicates that the effect of soil type (clayey or sandy, and acid sulfate soil) should be carefully reflected into the national GHG inventory in the relevant country.

The higher SF_CH4-AWD (i.e., lower CH₄ mitigation effect) than the IPCC’s default SF for multiple aeration partly resulted from varying weather conditions during the field experiment in four sites (Fig. 5B). Minamikawa et al. (2014) conducted a long-term (20 years) model simulation to estimate the effect of prolonged mid-season drainage on CH₄ emission reduction across nine sites in Japan and found that the simulated mean effect (20.1% reduction compared to normal mid-season drainage) was lower than the observed effect in the 2-year field experiment (30.5%; Itoh et al. 2011). The results of this study and Minamikawa et al. raise a question that the IPCC’s default SFs for water regime may be an overestimation of the actual CH₄ emission reduction. It is necessary to distinguish between the best-effort SF (as IPCC’s) and the practically feasible SF (as this study’s) for accurate GHG inventory.

Acknowledgments

This study was funded by the Ministry of Agriculture Forestry and Fisheries (MAFF) of Japan through the International Research Project ‘Technology development for circulatory food production systems responsive to climate change: Development of mitigation option for greenhouse gases emissions from agricultural lands in Asia (MIRSA-2).’ We thank the project members, Dr. Hoa Tran Dang from Hue University of Agriculture and Forestry, Vietnam; Dr. Prihasto Setyanto from Central Java Assessment Institute for Agricultural Technology, Indonesia; Dr. Amnat Chidthaisong from Joint Graduate School of Energy and Environment, King Mongkut’s University of Technology Thonburi, Thailand; and Engr. Evangeline Sibayan from Philippine Rice Research Institute for providing the data from the field experiments. We also thank Dr. Yusuke Takata (NARO, Japan) for evaluating the soil profiles from each study site. The SAS analyses were run on the supercomputer of Agriculture, Forestry and Fisheries Research Information Technology Center (AFFRIT), MAFF, Japan.

Supplemental data

Supplemental data can be accessed here.

Additional information

Funding

This study was funded by the Ministry of Agriculture Forestry and Fisheries (MAFF) of Japan through the International Research Project ‘Technology development for circulatory food production systems responsive to climate change: Development of mitigation option for greenhouse gases emissions from agricultural lands in Asia (MIRSA-2).’