Modeling methane emission from rice paddies with various agricultural practices



[1] Several models have been developed over the past decade to estimate CH4 emission from rice paddies. However, few models have been validated against field measurements with various parameters of soil, climate and agricultural practice. Thus reliability of the model's performance remains questionable particularly when extrapolating the model from site microscale to regional scale. In this paper, modification to the original model focuses on the effect of water regime on CH4 production/emission and the CH4 transport via bubbles. The modified model, named as CH4MOD, was then validated against a total of 94 field observations. These observations covered main rice cultivation regions from northern (Beijing, 40°30′N, 116°25′E) to southern China (Guangzhou, 23°08′N, 113°20′E), and from eastern (Hangzhou, 30°19′N, 120°12′E) to southwestern (Tuzu, 29°40′N, 103°50′E) China. Both single rice and double rice cultivations are distributed in these regions with different irrigation patterns and various types of organic matter incorporation. The observed seasonal amount of CH4 emission ranged from 3.1 to 761.7 kg C ha−1 with an average of 199.4 ± 187.3 kg C ha−1. In consonance with the observations, model simulations resulted in an average value of 224.6 ± 187.0 kg C ha−1, ranging from 13.9 to 824.3 kg C ha−1. Comparison between the computed and the observed seasonal CH4 emission yielded a correlation coefficient r2 of 0.84 with a slope of 0.92 and an intercept of 41.1 (n = 94, p < 0.001). It was concluded that the CH4MOD can reasonably simulate CH4 emissions from irrigated rice fields with a minimal number of inputs and parameters.

1. Introduction

[2] Methane is one of the principal greenhouse gases. Rodhe [1990] reported that CH4 has about 15–30 times greater infrared absorbing capability than CO2 on a mass basis and may account for 15–20% of the radiative forcing added to the atmosphere [Houghton et al., 1996]. Worldwide, rice fields are thought to contribute 9–30% to global CH4 emissions [Matthews et al., 1991; Sass, 1995; Houghton et al., 1996]. Projections based on population growth rates in countries where rice is the main food crop indicate that rice production must increase 65% by 2020 to meet the rice demand for the growing population [International Rice Research Institute, 1989]. This increase rice production will most likely be accompanied by an increase in methane emissions [Bouwman, 1991].

[3] Precise estimates of CH4 emissions from rice fields have been difficult to determine due to large regional differences in spatial and temporal variability in climate, soils and agricultural practices. In earlier efforts, estimates were made by extrapolating field measurements to a regional or global scale [Cicerone and Shetter, 1981; Holzapfel-Pschorn and Seiler, 1986; Schütz et al., 1989b], or by using a statistic relationship between methane emission and a certain variable such as the rice net primary production [Aselmann and Crutzen, 1989; Taylor et al., 1991; Bachelet et al., 1995], the rice grain production [Anastasi et al., 1992] and the organic matter input [Kern et al., 1995]. Nevertheless, large uncertainties might be introduced due to different regional factors in methane production, oxidation and emission processes. Thus models have become more and more important in estimating regional and global methane emissions [Intergovernmental Panel on Climate Change (IPCC), 2000].

[4] To obtain estimates of methane emissions from regional or global rice paddies via model method, it is essential to extrapolate the model from specific sites to a wider area by up scaling. Reliability of up scaling and thus the accuracy of the estimates mainly relies on two aspects, availability of a reliable input database and validity of CH4 models. It is most important for modelers to validate their models to insure that observed results can be realistically described under various soils, climates and agricultural practices. Several models have been developed over the last decade to estimate CH4 emission from rice paddies [Cao et al., 1995; Nouchi et al., 1997; Arah and Stephen, 1998; Huang et al., 1998; Li et al., 1994; Li, 2000; Matthews et al., 2000; Bodegom et al., 2001]. However, few models have been validated against field measurements with various parameters of soil, climate and agricultural practice because sufficient observations associated with these parameters were not available. Fortunately, since 1988 scientists have reported field observations of CH4 emission from Chinese rice paddies under a variety of soils, climates and agricultural practices [e.g., Shangguan et al., 1993; Wang et al., 1994; Wang, 1996; Khalil et al., 1998; Cai, 1999; Cai et al., 2000; Huang et al., 2001]. These studies offer a great opportunity to validate present models for the purpose of up scaling.

[5] With an understanding of the processes of methane production, oxidation and emission, Huang et al. [1998] developed a model to predict methane emission from rice paddy soils. The model associated these processes with rice growth, organic C depletion and environmental factors. Validated against independent field measurements of CH4 emission from rice paddy soils in Texas of USA, Tuzu of China and Vercelli of Italy, Huang's model provides a realistic estimate of the observed results. (Good Practice Guidance and Uncertainty Management in National Greenhouse Gas Inventories [IPCC, 2000]).

[6] It has been recognized that water management such as periodic drainage and intermittent irrigation during rice growing period significantly reduces CH4 emissions [Yagi et al., 1996; Mishra et al., 1997; Yang and Chang, 2001]. Decomposition of incorporated organic materials is the predominant source of methanogenic substrates in very early stages of rice growing [Watanabe and Roger, 1985] and the CH4 emission via ebullition during this period contributed significantly to the total emissions [Shangguan and Wang, 1993]. However, Huang's model was primarily developed for continuous flooding and not for intermittent irrigation rice paddies. Moreover, the CH4 emission via ebullition was not taken into account in their model. All of these factors may lead to significant errors if the model is employed to estimate CH4 emission from rice paddies with various agricultural practices including water regimes and organic matter amendments.

[7] In this paper, we pay specific attention to the effect of water regime on CH4 production/emission and the CH4 transport via bubbles for modifying the CH4 emission model developed by Huang et al. [1998]. The main objective of this paper is to realistically model field observations under various agricultural practices. A further objective is to focus on the estimates of CH4 emission from regional or global rice paddy soils by linking the modified model with available databases of soil, climate and agricultural practice.

2. Model Description and Modification

2.1. Description of the Original Model

[8] We accept the hypothesis from the original model [Huang et al., 1998] that methanogenic substrates are primarily derived from rice plants and added organic matter. Rates of methane production in flooded rice soils are determined by the availability of methanogenic substrates and the influence of environmental factors. Rice growth and development control the fraction of CH4 emitted through plant.

2.1.1. Substrates From Organic Matter Decomposition

[9] Decomposition of organic matter in soil was simulated with a first-order kinetics equation [Huang et al., 1998] as:

equation image

where COM is the daily amount of carbohydrate degraded from organic matter amendments (g m−2 d−1). The impact of soil texture and soil temperature on decomposition was quantified by the soil index (SI) and the temperature index (TI), respectively. OMN and OMS represent nonstructural and structural components of incorporated organic matter (g m−2, dry matter), respectively. Constants k1 and k2 represent the first-order potential decay rate for OMN and OMS with the values of 2.7 × 10−2 and 2 × 10−3 d−1, respectively [Huang et al., 1998]. The constant 0.65 is a reduction factor of field flooding on decomposition [Huang et al., 2002].

2.1.2. Substrates Associated With Rice Plants

[10] The amount of carbohydrates derived from rice plants was simulated by the following equation [Huang et al., 1998]:

equation image

where CR represents carbohydrate (g m−2 d−1) derived from rice plants and W is rice aboveground biomass (g m−2) on a given day. VI is a variety index identifying relative difference in methane production among rice varieties. Rice aboveground biomass was computed by a logistic growth equation [Huang et al., 1998] as:

equation image
equation image
equation image

where W0 and Wmax represent rice above ground biomass at transplanting and at harvesting, respectively. Time variable t (d) is scaled in days after transplanting. The GY is rice grain yield (g m−2). The constant r is an intrinsic growth rate for above ground biomass.

2.1.3. Influence of Environmental Factors

[11] The effect of soil texture on CH4 production was expressed by a dimensionless soil index (SI) that is linked with soil sand content (SAND) as equation (6) [Huang et al., 1998]. The TI was introduced to quantify the influence of soil temperature on CH4 production as equation (7) [Huang et al., 1998].

equation image
equation image

[12] Q10 is a temperature coefficient for a process involved in biochemical and microbial activities. Field measurements suggested that the Q10 for methane emission ranged from 2 [Khalil et al., 1991] to 4 [Schütz et al., 1989a]. A Q10 value of 3.0 was assumed [Huang et al., 1998].

[13] The effect of soil redox potential (Eh) on methane production was described by equation (8).

equation image

where FEh is a reduction factor of soil redox potential, 0 < FEh ≤ 1.0 [Huang et al., 1998].

2.1.4. Dependence of Methane Production on Substrates and Environment

[14] The net reaction of anaerobic carbohydrate fermentation with methanogenesis was assumed to be an overall reaction of C6H12O6 ⇒ 3CH4 + 3CO2. From this reaction, a conversion factor on a mole weight basis of C6H12O6 to CH4 is approximately 0.27 (3[CH4]/[C6H12O6] = 0.27). Rate of methane production, P (g m−2 d−1), is then determined by the availability of methanogenic substrates and the influence of environmental factors by the equation:

equation image

2.1.5. Methane Emission Via Rice Plants

[15] The fraction of produced methane emitted via rice plants, Fp, was simulated by [Huang et al., 1998]:

equation image

where W and Wmax has the same definition as in equation (3).

[16] The CH4 emission via plants (Ep) was then computed as:

equation image

2.2. Modifications to the Original Model

2.2.1. Methane Emission Via Bubbles

[17] Methane emitted via bubbles was observed in flooded rice paddies [Schütz et al., 1989a; Nouchi, 1994; Denier van der Gon and Neue, 1995; Wassmann et al., 1996]. When the flooded soil reaches the max solubility of CH4, produced methane in the soil will aggregate to form bubbles, travel straight upward to the water surface, and finally release into the atmosphere with very little oxidization. The methane ebullition occurs predominantly in the early phase of rice growth season, and trails off as the rice matures [Shangguan et al., 1993]. We simply adopted the equation by Li [1999] to simulate this process.

equation image

where Ebl represents the CH4 emission rate (g m−2 d−1) via bubbles. P is CH4 production rate (g m−2 d−1). P0 is a criterion when the bubbles occur and was quantified as 0.002 (g m−2 d−1) [Li, 1999]. Tsoil is the soil temperature (°C). Wroot is the rice root biomass (g m−2) that is described by the above ground biomass (W) as [Yoshida, 1981]:

equation image

For a given aboveground biomass of W, we set Wroot(0) = 0 to start the iteration and set Wroot(i+1)Wroot(i) < 0.1 as a limiting criterion.

2.2.2. Effect of Water Managements on Soil Eh

[18] Mishra et al. [1997] studied the effect of water regimes (continuously flooded, continuously nonflooded, alternately flooded) on CH4 flux from rice-planted soil. They found that the Eh was always high under nonflooded conditions, but dropped rapidly within several days after flooding. In intermittently flooded regimes, the Eh increased with a shift from flooded to nonflooded conditions [Mishra et al., 1997]. The original model simulated changes in soil Eh for the conditions from nonflooded to flooded but did not for the intermittently flooded regimes [Huang et al., 1998]. The change in soil Eh is also associated with soil type and the amount of fresh organic matter in the soil at the start of the season [Matthews et al., 2000]. Referring to the works by Mishra et al. [1997] and Matthews et al. [2000], we simulated the changes of soil Eh during flooding or drainage course by following differential equations.

equation image

where Eh(t) represents the soil Eh value at time t. The t represents the days after flooding or since drainage. The AEh and DEh are the differential coefficients. By applying a trial-and-error optimization method to the field measurements of Eh by Cao et al. [1997] in Jurong and by this group in Nanjing (unpublished data), the values of AEh and DEh were determined to be 0.23 and 0.16, respectively. The BEh is a limit criterion, which takes a low-limit value of −250 mv for the flooding course and an upper-limit value of 300 mv for the drainage course, respectively. It is a bit difficult to simulate the soil Eh change for the intermittent irrigation, since the irrigation frequency changes widely for different area. Thus we assumed the soil Eh fluctuates by 10–20 mv from a constant value of −20 mv.

3. Model Validation

[19] The modified model was run with a daily step and validated against independent CH4 emission measurements over the year from 1988 to 1999. These measurements were made in 9 sites covering five main rice cultivation regions in China, including different water regime, organic matter incorporation, and rice cropping system. A summary of the observations is given in Table 1.

Table 1. Summary of the Observations for Validating the CH4MOD
Observational SiteLocationSoil Texture (Sand Percentage)Observational SeasonsRice CultivationCasesReference
Guangzhou, Guangdong23°08′N, 113°20′Esandy loam (46.3)1994double4Institute of Atmospheric Physics, CAS
Changsha, Hunan28°09′N, 113°06′ESandy loam (62.0)1995–1997double20APN database
Taoyuan, Hunan28°55′N, 110°30′ESilty Loam (21.2)1992double8Shangguan et al. [1993]
Tuzu, Sichuan29°40′N, 103°50′ESand (78.5)1988–1994single7Khalil et al. [1998]
Chongqing29°48′N, 106°18′ESand (57.0)1995–1997single11APN database
Hangzhou, Zhejiang30°19′N, 120°12′Eloam (23.0)1995–1998single/double7/16APN database
Nanjing, Jiangsu31°51′N, 118°49′EClay (4.8)1999single4Program of TECO/NASA by this group
Fengqiu, Henan35°24′N, 114°24′EClay (2), loam (20), sand (80)1993–1994single6APN database
Beijing40°30′N, 116°25′ESand (55.0)1995–1997single11Wang et al. [1998]

3.1. Model Input

[20] Model parameter inputs include rice grain yield (GY), soil sand percentage (SAND), amount of organic amendment, initial fraction of the structural and non-structural carbohydrates of the incorporated organic matter, water management pattern, and daily air temperature (Tair).

[21] Rice grain yield was recorded for some but not for all of the measurements. We used the statistical average of the GY adjacent to the experimental site when it was not reported. The soil sand percentage was referred to the soil database developed by the Institute of Soil Sciences, Chinese Academy of Sciences when it was not reported. Types of organic matter amended into rice field at transplanting include animal manure, green manure, biogas residuals, and crop straw. Naturally addition of OM into the soil includes crop root and stubbles from previous season, and wild weeds when previous season was fallowed. If not explicitly specified, we assumed that root of the previous crop and stubbles (about 10% of the above ground biomass) were left in the soil. When the field remained fallow during the previous season, we assumed that wild weed residues would be present in the soil. According to the climate conditions, the amount of wild weeds varied from 0 (cold winter like in Beijing) to 2000 (warmer winter like in Hunan and Guangdong) kg ha−1 dry matter. In the double rice cropping system, 50% of the rice straw from the early-rice was assumed to be incorporated into soils of the late-rice season. The initial fractions of OMN and OMS for different types of organic matter are summarized in Table 2. Daily air temperature was obtained from local meteorological station. The soil temperature (Tsoil) was estimated by air temperature (Tair) as Tsoil = 4.4 + 0.76 × Tair [Huang et al., 1998].

Table 2. Initial Fractions of OMN and OMS in Incorporated Organic Matter
Organic MatterOMNOMS
  • a

    Calculated from Huang et al. [2003].

  • b

    Wild weeds included.

  • c

    Average value and strongly dependent on fermentation of the raw material.

Rice straw0.59a0.41
Rice root0.42a0.58
Wheat straw0.49a0.51
Wheat root0.31a0.69
Green manure0.80a,b0.20
Farm manure0.25c0.75
Bio-gas residues0.10c0.90

[22] Water management is one of the most important practices in rice cultivation. To reduce ineffective tillers, remove toxic substances and maintain healthy roots under reduced soil conditions, short periods of drainage for soil aeration during the vegetative growth period and intermittent irrigation during the reproductive growth period are commonly practiced in Japan [Yoshida, 1981] and China [Gao et al., 1992]. The extent of drainage and the interval between the cycles of irrigation-drainage-reintroduced water vary with soil characteristics and weather conditions. In high land (rain fed), rice paddies are kept flooded by rain without irrigation [Khalil et al., 1998]. Typical irrigation patterns in China according to Gao and Li [1992] and Su [2000] are summarized in Table 3. Seasonal changes in soil Eh for a given irrigation pattern were calculated by equation (14). The resulting calculated Eh was used in equation (8).

Table 3. Irrigation Patterns for Rice Cultivation in Chinaa
Pattern CodeIrrigation CoursesbDescription
1F—D—F—Msingle rice crop in the northern and eastern China
2F—D—Msingle and double rice crop in the southern and southwestern China
3F—Msimilar to pattern 2, without obvious drainage
4Fhigh land rice fields or salty soil field
5Mlowland, usually with high undergroundwater level

[23] Above ground biomass at transplanting (W0) was assigned a value of 15 g m−2 [Gao et al., 1992]. Huang et al. [1998] gave the value of 0.08 ± 0.02 d−1 for the intrinsic growing rate of rice plant (r). Values of r equal to 0.08 for single rice and 0.1 for early-/late-rice, respectively, were used to simulate rice growth. Huang et al. [1997] derived a rice variety index (VI) from field observations of CH4 emission, grain yield and the SI for ten cultivars under permanent flooding condition. They evaluated the value of VI to be 1.0 for the majority of cultivars (8 out of 10) and 1.5 for high emission cultivars (2 out of 10). However, measurements of CH4 emission for different cultivars under permanent flooding condition were not available in China. In a study by Huang et al. [1999], the VI was taken a constant of 1.0 for all cases to validate the model. Table 4 gives more detailed information about model symbols, definitions, and values/unites.

Table 4. Parameters and Constants and Their Description for CH4MOD Running
Rice Growth
WAboveground biomassModel result/g m−2
W0Initial above ground biomass at transplanting15/g m−2
WmaxAboveground biomass at harvestingModel result/g m−2
WrootRoot biomassModel result/g m−2
GYRice grain yieldModel input/g m−2
rIntrinsic growth rate of rice plant0.08/d−1 for single rice
  0.1/d−1 for early-/late-rice
Methanogenic Substrates
CRCarbohydrates derived from rice root exudationModel result/g m−2 d−1
COMCarbohydrates derived from OM decompositionModel result/g m−2 d−1
OMOrganic matterModel input/g m−2
OMNNon-structural component of OMModel result/g m−2
OMSStructural component of OMModel result/g m−2
k1First order decay rate of OMN2.7 × 10−2/d−1
k2First order decay rate of OMS3.0 × 10−3/d−1
VIRice variety index1.0/dimensionless
Methane Production and Emission
PCH4 productionModel result/g m−2 d−1
P0Criterion of CH4 production when CH4 bubbles occur0.002/g m−2 d−1
FpCH4 emitted fraction via rice plantModel result/dimensionless
EpCH4 emitted via rice plantModel result/g m−2 d−1
EblCH4 emitted via bubblesModel result/g m−2 d−1
Environments and Derivatives
SANDSoil sandModel input/%
TairAir temperatureModel input/°C
TsoilSoil temperatureModel result/°C
EhSoil redox potentialModel result/mv
Q10Temperature coefficient3.0/Dimensionless
SISoil texture indexModel result/Dimensionless
TISoil temperature indexModel result/Dimensionless
FEhSoil Eh indexModel result/Dimensionless
AEhDeferential coefficient of soil Eh0.23/mv
DEhDeferential coefficient of soil Eh0.16/Dimensionless
BEhLow and up limit for soil Eh−250/mv for flooding regime
300/mv for drainage regime

3.2. Validation for Single Rice Cultivation

[24] Single rice cultivation is prevalent in western, northern, and most of eastern China. Rotations of wheat-rice, green manure-rice, rapeseed-rice and fallow-rice are the main cropping systems in these regions. A total of 46 observations made in Beijing, Jiangsu, Zhejiang, Chongqing, Sichuan and Henan were simulated. Figure 1 shows the computed and the observed seasonal values of CH4 emission under different water regimes with or without organic matter amendments, indicating that the model can well capture the seasonal patterns of CH4 emissions.

Figure 1.

Comparison of simulated with observed seasonal patterns of methane emission from single rice paddies with diverse agricultural practices. (a) Nanjing1999, irrigation ptn-4, no OM amendment, wheat/rice; (b) Nanjing1999, irrigation ptn-4, wheat straw 4.5 t ha−1, wheat/rice; (c) Chongqing1996, irrigation ptn-3, farm manure 5.0t ha−1, waterlog/rice; (d) Chongqing1996, irrigation ptn-3, farm manure 5.0t ha−1, wheat/rice; (e) Beijing1995, irrigation ptn-2, pig manure 3.6t ha−1, wheat/rice; (f) Beijing1997, irrigation ptn-2, rice straw 2.6t ha−1, fallow/rice; (g) Hangzhou1996, irrigation ptn-3, no OM amendment, fallow/rice; (h) Hangzhou1995, irrigation ptn-1, green manure 1.1 t ha−1, fallow/rice.

3.3. Validation for Double Rice Cultivation

[25] Double rice cropping systems are mainly distributed in southern, southwestern and southeastern China where hydrological and thermal resources are more abundant. A total of 48 observations made in Guangdong, Hunan, Sichuan and Zhejiang were simulated. Figure 2 gives the simulated and observed seasonal changes in CH4 emissions from Changsha and Taoyuan of Hunan province, Guangzhou of Guangdong province, and Hangzhou of Zhijiang province, respectively. Results in Figure 2 suggest that the present model could well simulate CH4 emissions from early- and late-rice cultivation under different water regimes with or without organic matter amendments.

Figure 2.

Comparison of simulated with observed seasonal patterns of methane emission from double rice paddies. (a) Taoyuan1992, early-rice: green manure 3t ha−1, farm manure 1.9t ha−1, irrigation ptn-3; late-rice: rice straw 4.5t ha−1, farm manure 1.9t ha−1, irrigation ptn-3; (b) Changsha1996, early-rice: wild weeds 0.46t ha−1, irrigation ptn-3; late-rice: no OM amendment, irrigation ptn-3; (c) Guangzhou1994, early-rice: farm manure 3t ha−1, irrigation ptn-3; late-rice: no OM amendment, irrigation ptn-3; (d) Hangzhou1997, early-rice: Bio-gas residual 0.6t ha−1, irrigation ptn-3; late-rice: Bio-gas residual 0.6t ha−1, irrigation ptn-1.

3.4. Validation of Total Seasonal CH4 Emission

[26] Total simulated seasonal CH4 emission values were determined from the summation of simulated daily values. Observed and modeled total seasonal CH4 estimation for each case is given in Table 5. The observed seasonal amount ranged from 3.1 to 761.7 (kg C ha−1) with an average of 199.4 ± 187.3 (kg C ha−1). In consonance with the observations, simulations with the model result in an average value of 224.6 ± 187.0 (kg C ha−1), ranging from 13.9 to 824.3 (kg C ha−1). The regression of computed against observed emissions (Figure 3) yields an r2 of 0.84 with a slope of 0.92 and an intercept of 41.1 (n = 94, P < 0.001).

Figure 3.

Comparison of simulated with measured total seasonal methane emissions from rice paddies with diverse agricultural practices across China.

Table 5. Detailed Information on the Field Observations and Model Results
Case CodeTransplanting (yyyy-mm-dd)Harvesting (yyyy-mm-dd)Yield, g m−2I.P.aPrevious SeasonOM Amendment (t ha−1, dry weight)Observed CH4 (kg C ha−1)Modeled CH4 (kg C ha−1)
  • a

    Irrigation pattern.

  • b

    Estimated as local average.

BJ1995_T11995-06-041995-10-176492WheatPig manure 3.6273.74281.44
BJ1995_T21995-06-041995-10-176485WheatPig manure 3.6150.28181.07
BJ1995_T31995-06-041995-10-175614WheatPig manure 3.6361.19622.08
BJ1995_T41995-06-041995-10-175432Wheat 19.30110.88
BJ1996_T11996-05-251996-10-087702Fallow 16.8653.95
BJ1996_T21996-05-251996-10-086802Fallow 36.6753.95
BJ1996_T31996-05-251996-10-086902Fallow 32.8853.95
BJ1997_T11997-05-221997-10-067742FallowPig manure 2.6144.77199.72
BJ1997_T21997-05-221997-10-066672FallowCattle manure 2.632.32139.95
BJ1997_T31997-05-221997-10-066942FallowRice straw 2.6106.92200.12
BJ1997_T41997-05-221997-10-066942Fallow 4.3368.88
CS1995_HFe1995-05-081995-07-14600b3FallowWild weeds 1.0134.38182.54
CS1995_HFL1995-07-241995-10-20600b3Early-rice 206.59208.09
CS1995_HMe1995-05-081995-07-14600b3Green ManureGreen manure 0.75102.40166.65
CS1995_HML1995-07-241995-10-20600b3Early-rice 92.10196.09
CS1996_HFe1996-05-051996-07-14600b3FallowWild weeds 0.46252.28197.82
CS1996_HFL1996-07-311996-10-15600b3Early-rice 189.68191.81
CS1996_HMe1996-05-051996-07-14600b3Green ManureGreen manure 2.6219.50267.82
CS1996_HML1996-07-311996-10-15600b3Early-rice 52.63222.94
CS1996_HRe1996-05-051996-07-14600b3Rapeseed PlantRapeseed plant straw 5.4625.47395.24
CS1996_HRL1996-07-311996-10-15600b3Early-rice 304.53246.94
CS1996_HSe1996-05-051996-07-14600b3Waterlog 240.88151.45
CS1996_HSL1996-07-311996-10-15600b3Early-rice 159.98181.55
CS1997_HFe1997-050-61997-07-14600b3FallowWild weeds 0.9137.54173.68
CS1997_HMe1997-05-061997-07-14600b3Green ManureGreen manure 2.5232.55247.93
CS1997_HRe1997-05-061997-07-14600b3Rapeseed PlantRapeseed plant straw 2.2396.00237.01
CS1997_HSe1997-05-061997-07-14600b3Green Manure 191.14138.51
CS1997_HFL1997-07-311997-10-20600b3Early-rice 89.88215.06
CS1997_HML1997-07-311997-10-20600b3Early-rice 190.10221.74
CS1997_HRL1997-07-311997-10-20600b3Early-rice 461.86241.65
CS1997_HSL1997-07-311997-10-20600b3Early-rice 69.94196.33
FQ1993_Pn1993-06-281993-10-147655FallowFarm manure 1.13.1013.93
FQ1993_Pr1993-06-281993-10-148105FallowFarm manure 1.113.9831.10
FQ1993_Ps1993-06-281993-10-144715FallowFarm manure 1.18.8686.74
FQ1994_Pn1994-06-211994-10-077655FallowFarm manure 1.38.7917.33
FQ1994_Pr1994-06-211994-10-078105FallowFarm manure 1.312.2038.11
FQ1994_Ps1994-06-211994-10-074715FallowFarm manure 1.336.16105.07
GZ1994_T1e1994-04-031994-07-136402FallowFarm manure 3.045.3172.97
GZ1994_T1L1994-08-141994-11-186403Early-rice 29.2532.19
GZ1994_T2e1994-04-031994-07-136402Fallow 15.1033.20
GZ1994_T2L1994-08-141994-11-186403Early-rice 25.2232.42
HZ1995_T11995-05-301995-10-106632Fallow 162.18159.06
HZ1995_T21995-05-301995-10-106201FallowGreen manure 1.1152.04184.07
HZ1995_T31995-05-301995-10-106492FallowGreen manure 1.1242.67210.57
HZ1995_T41995-05-301995-10-106684FallowGreen manure 1.1334.36447.33
HZ1996_T11996-06-201996-10-305213Fallow 137.24130.48
HZ1996_T2e1996-05-081996-07-245153Fallow 68.3570.69
HZ1996_T2L1996-07-261996-11-075053Early-rice 75.03148.25
HZ1996_T31996-06-201996-09-265603Fallow 126.3199.71
HZ1996_T4e1996-05-081996-07-244933Fallow 63.5369.59
HZ1996_T4L1996-07-261996-10-305053Early-rice 84.16139.84
HZ1997_T1e1997-05-041997-07-206273Fallow 40.5559.29
HZ1997_T1L1997-07-221997-11-076331Early-rice 108.37136.59
HZ1997_T2e1997-05-041997-07-206273FallowFarm manure 0.944.8469.54
HZ1997_T2L1997-07-221997-11-076331Early-riceFarm manure 0.9136.96158.64
HZ1997_T31997-06-101997-09-206241Fallow 66.6178.37
HZ1997_T4e1997-05-041997-07-206273FallowBiogas residual 0.640.1261.73
HZ1997_T4L1997-07-221997-11-076331Early-riceBiogas residual 0.6115.36146.82
HZ1998_T1e1998-04-291998-07-176203Fallow 106.3967.63
HZ1998_T1L1998-07-211998-11-076313Early-rice 138.65166.48
HZ1998_T2e1998-04-291998-07-176163FallowRice straw 1.1168.3692.27
HZ1998_T2L1998-07-211998-11-076313Early-riceRice straw 1.1209.52239.65
HZ1998_T4e1998-04-291998-07-176133FallowRice straw 1.1150.1992.25
HZ1998_T4L1998-07-211998-11-076313Early-riceRice straw 1.1185.77239.73
NJ1999_D01999-06-211999-10-207502Wheat 102.5628.17
NJ1999_D21999-06-211999-10-207502WheatWheat straw 4.584.93197.81
NJ1999_F01999-06-211999-10-207504Wheat 90.68128.03
NJ1999_F21999-06-211999-10-207504WheatWheat straw 4.5148.34231.97
TY1992_T1e1992-05-011992-07-13584b3FallowGreen manure 3.099.5796.98
TY1992_T1L1992-07-151992-10-09584b3Early-riceFarm manure 0.9; Rice straw 1.5224.84194.21
TY1992_T2e1992-05-011992-07-13584b3FallowGreen manure 3.0; Farm manure 1.9118.35113.99
TY1992_T2L1992-07-151992-10-09584b3Early-riceFarm manure 1.9; Rice straw 4.5297.61349.45
TY1992_T3e1992-05-011992-07-13584b3Fallow 37.3046.63
TY1992_T3L1992-07-151992-10-09584b3Early-rice 50.2060.52
TY1992_T4e1992-05-011992-07-13584b3FallowBiogas residual 7.5123.1785.87
TY1992_T4L1992-07-151992-10-09584b3Early-riceBiogas residual 7.5153.18184.59
TZ19881988-04-201988-08-225003Oilseed plantFarm manure 6.1722.95781.41
TZ19891989-04-201989-08-264873Oilseed plantFarm manure 6.2490.54692.62
TZ19901990-05-011990-08-194703Oilseed plantFarm manure 6.4761.70727.65
TZ19911991-04-231991-08-255452Oilseed plantFarm manure 7.7460.63443.73
TZ19921992-05-011992-08-246723Oilseed plantFarm manure 10.0558.31602.73
TZ19931993-05-021993-08-295432Oilseed plantFarm manure 9.2690.71824.33
TZ19941994-04-261994-08-194713Oilseed plantFarm manure 5.3607.51628.46
CQ1995_T11995-05-151995-08-23613b3Waterlog 260.77327.79
CQ1995_T21995-05-151995-08-23613b3Waterlog 302.33327.79
CQ1995_T31995-05-151995-08-23613b3Wheat 89.76340.09
CQ1996_T11996-05-231996-09-02613b3WaterlogFarm manure 5.0662.92570.52
CQ1996_T21996-05-231996-09-02613b3Waterlog 608.02570.52
CQ1996_T31996-05-231996-09-02613b3WheatFarm manure 5.0444.18565.99
CQ1996_T41996-05-231996-09-02613b3Wheat 593.97565.99
CQ1997_T11997-05-101997-08-19613b3Waterlog 309.94362.32
CQ1997_T21997-05-101997-08-19613b3Waterlog 335.25362.32
CQ1997_T31997-05-101997-08-19613b3Wheat 345.53359.31
CQ1997_T41997-05-101997-08-19613b3Wheat 336.21359.31

4. Model Significance of the Bubble Flux for Overall Emissions and the Drainage Effect on CH4 Production/Emission

4.1. Model Significance of the Bubble Flux for Overall Emissions

[27] Methane emission via bubbles was observed from rice fields [Bartlett et al., 1988; Wilson et al., 1989; Shangguan et al., 1993]. Observations made in Hunan province of China by Shangguan et al. [1993] indicated that the contribution of CH4 emitted via bubbles to the overall emissions accounted for 24% for the early-rice and 55% for the late-rice cropping, respectively. Model simulation for 94 cases suggested that the bubble flux contribute 5%–45% to the overall emissions. Figure 4 shows the computed seasonal variations in plant-mediated CH4 emission and bubble flux. Calculations indicated that bubble flux from the single rice cropping system accounted for 11% in Hangzhou (Figure 4a) and 27% in Beijing (Figure 4b), respectively. With respect to the double rice cropping system, the bubble fluxes contributed 23.5% for the early-rice and 32.5% for the late-rice cropping in Taoyuan (Figure 4c), and 18.4% for the early-rice and 28.6% for the late-rice cropping in Changsa (Figure 4d), respectively. Clearly, the bubble fluxes simulated by the revised model fall into the range of field observations [Shangguan et al., 1993].

Figure 4.

Computed seasonal variations in plant-mediated CH4 emission and bubble flux. (a) Hangzhou1995, green manure 1.1t ha−1, irrigation ptn-4, fallow/rice; (b) Beijing1995, pig manure 3.6t ha−1, irrigation ptn-2, wheat/rice; (c). Taoyuan1992, early-rice: green manure 3.0t ha−1; farm manure 1.9t ha−1, irrigation ptn-3; late-rice, farm manure 1.9t ha−1 + rice straw 4.5t ha−1, irrigation ptn-3, (d). Changsha1996, early-rice: wild weeds 0.46t ha−1, irrigation ptn-3; late-rice: no OM amendment, irrigation ptn-3.

[28] It is well recognized that plant-mediated transport through micropores in rice leaf sheaths is the primary mechanism for CH4 emission [Nouchi, 1994]. In the early growing period, however, CH4 emission via rice plants might be restricted due to fewer micropores, and hence the bubble flux plays a key role, especially when a relatively high amount of organic matter is incorporated into the soil (Figures 4b and 4c). Higher temperature during the early growing period of late-rice cropping enhanced decomposition of organic matter, which is favorable for CH4 production and consequently emission via bubbles (Figures 4c and 4d). Obviously, the modified model captures the signals of bubble fluxes (Figure 4) and the simulated overall emissions agreed well with observations (Figures 1e, 2a, and 2b). Since the original model included no such function, overall emissions might have been underestimated by the original model, particularly when additional organic matter was incorporated and during periods of higher temperatures (Figures 4b–4d).

4.2. Model Significance of the Drainage Effect on CH4 Production/Emission

[29] It has been well documented that periodic drainage of irrigated rice paddies usually results in a significant decrease in methane emissions. Field observations indicated that intermittent irrigation system resulted in some 64% decrease in CH4 emissions [Yang and Chang, 2001]. Mishra et al. [1997] reported the reduction factors of 45% to 72% by setting different courses of periodic drainage in their pot experiment. According to the review of Cai [1997], mid-season aeration during the period of rice growth could mitigate CH4 emission by as much as 50%.

[30] We computed the CH4 emissions under conditions of irrigation pattern-1, pattern-2, pattern-3 and continuous flooding (see Table 3), respectively. These calculations indicate that CH4 emission from a single rice cropping system decreased 59% with irrigation pattern-1 in Hangzhou (Figure 5a) and 55% with irrigation pattern-2 in Beijing (Figure 5b), respectively. For the double rice cropping system in Taoyuan, calculations for irrigation pattern-3 resulted in 45% and 37% decrease in CH4 emissions for the early-rice and the late-rice cropping (Figure 5c), respectively. Apparently, the simulated reduction in CH4 emission induced by drainage events is comparable to observed values.

Figure 5.

Computed seasonal variations in CH4 emission with different irrigation courses.

5. Discussion

5.1. Advantages and Disadvantages of Mechanistic and Empirical Models

[31] Methane emission from rice cultivation is among the most uncertain estimates of the agricultural sector in rice-growing countries. Reduction in the uncertainties might be achieved by coupling field-scale model estimates to regional databases. A model developed for this purpose should be reliable, and allow extrapolation with input parameters that are commonly available. Several models have been developed to estimate CH4 emission from rice paddies over the last decade. Some of these models [Cao et al., 1995; Arah and Stephen, 1998; Li et al., 1994; Li, 2000; Matthews et al., 2000; Bodegom et al., 2001] are mechanistic, and others [Bachelet and Neue, 1993; Huang et al., 1998] are more empirical. Both mechanistic and empirical models have their advantages and disadvantages. Reliability of the models, either mechanistic or empirical, is more likely dependent on whether the observations in situ can be properly described or not.

[32] Mechanistic models combine available knowledge with the processes of CH4 production, oxidation and emission. The advantage of mechanistic models is that they have a theoretical foundation, and thus the model estimates should be reliable at least on a regional scale. However, like all models, mechanistic models have several limitations. Some mechanistic models need site-specific parameters, and some could properly simulate a given field measurement but the simulated components are not comparable with other field observations, which makes it hard to extrapolate these models to a wide scale. For example, a mechanistic model developed by Cao et al. [1995] needs model inputs of seasonal patterns of soil Eh and floodwater depth for simulating CH4 emission from rice paddies. Matthews et al. [2000] developed a mechanistic model in which CH4 production is associated with the soil pH, temperature, the presence of other ions (i.e., NO3, Fe3+, Mn4+, SO42−), and the concentration of O2 in the soil. The plant-mediated CH4 transport, emissions via diffusion and ebullition were simulated in their model. They validated their model against two cases of field measurement, and a generally good agreement between computed and observed overall CH4 flux was documented. However, the modeled proportion of CH4 produced that is then oxidized constitutes only some 7% of the seasonal total [Matthews et al., 2000], while experiments from Italian rice fields [Schütz et al., 1989b] and American rice fields [Sass et al., 1990, 1992] showed that the oxidized fraction varied from 0.45 to 0.95 over the growing season.

[33] Empirical models are mainly developed from statistical analysis and an integration of in situ signals. The advantage of empirical models is that they need only few input parameters. Results of empirical models are however difficult to extrapolate beyond the area for which they were developed or validated. In this study, validation against field observations covered main rice cultivation regions from northern (Beijing, 40°30′N, 116°25′E) to southern (Guangzhou, 23°08′N, 113°20′E) China, and from eastern (Hangzhou, 30°19′N, 120°12′E) to southwestern (Tuzu, 29°40′N, 103°50′E) China with various agricultural practices (Table 5). Resulting values suggest that the present model is of great potential for up scaling purposes when spatial databases on climate, soils, agronomic practices and crop grain yields are available.

5.2. New Features and Advantages of the Present Model

[34] New features of the present model are the incorporation of bubble flux process and the drainage events into the original model. Overall, the advantages of this model make it particularly applicable to the simulation of CH4 emissions from irrigated rice fields with few input parameters of irrigation patterns, type and amount of organic matter amendment, rice grain yield, air temperature, and soil sand percentage (Table 4).

5.3. Future Requirements to Obtain Greater Accuracy in Model Estimates

[35] It is well established that cultivar type can significantly affect CH4 emissions [Huang et al., 1997]. The model dependence of CH4 production and emission on rice cultivar (equation (2)) will be problematic in applying it on a large scale, since we are not able to quantify the VI due to unavailability of emissions data for different cultivars in China. However, the VI should be incorporated into the model to keep the model intact. When the VI is available, the model will be able to logically simulate the dependence of CH4 production and emission on rice cultivar, even though no variety effects are currently included. Recent work by Ding et al. [1999] indicated that plant height or certain aspects of the rice canopy geometry might be an indicator of the variety index (VI), which would allow the model to be more easily applied in cases where varietal data are lacking.

[36] Nitrogen fertilizers are commonly used in rice cultivation to increase crop yields. Urea and ammonium sulfate account for 80–90% of the total nitrogen fertilizer required in rice cultivation [FAO,]. The influences of these inorganic fertilizers on methane emission from rice fields, however, are not well understood and reported observations about it are not consistent. Lindau et al. [1991] showed an increased methane emission rate with increased urea application, but Wang et al. [1993] reported that no change over the control for urea application and a decrease in methane production with ammonium nitrate. Liou et al. [2003] reported that CH4 emission rates from the potassium nitrate plots were 1.5–3.7-fold higher than that from the ammonium sulfate plots throughout the growth period. Clearly, the effect of inorganic nitrogen on CH4 emission will need to be more carefully characterized before modeling of the process can be accomplished.

[37] As our knowledge of the processes involved in CH4 emission from rice paddies increase, models will become more mechanistic. The ultimate goal of this type of model should be to accurately calculate CH4 emissions on a regional or larger scale based on available geographic information system data sets and remotely sensed data.

6. Conclusion

[38] Compared with the original model developed by Huang et al. [1998], the modified model, CH4MOD, can reasonably simulate the effect of water regime on CH4 production/emission and the CH4 transport via bubbles. Model validation against independent observations demonstrated that the present model is capable of simulating CH4 emissions from irrigated rice fields with a minimal amount of inputs and parameters. A further conclusion is that the model is of great potential for up scaling as it has provided a realistic estimate of the observed results from various soils, climates and agricultural practices.


[39] This work was supported by grants from the Knowledge Innovation Program of the Chinese Academy of Sciences (approved KZCX1-SW-01-13), the National Key Basic Research Development Foundation (approved G1999011805) of China and the Hundred Talents Program of the Chinese Academy of Sciences. We thank the Asian-Pacific Network for Global Change Research (APN 2001-16) provided database for model validation. Thanks are also dedicated to Professors Mingxing Wang and Yuesi Wang in the Institute of Atmospheric Physics and Professor Dafang Zhuang of the IGSNRR for their contributions to this study, Professor Ronald L. Sass in Rice University, USA for his language check and three referees for their thoughtful comments.