Sources Of Atmospheric Methane: Measurements in Rice Paddies and a Discussion

We have made field measurements of methane fluxes from rice paddies, fresh water lakes, and saltwa- ter marshes to infer estimates of the size of these sources of atmospheric methane. The rice-paddy measurements, the first of their kind, show that the principal means of methane escape is through the plants themselves as opposed to transport across the water-air interface via bubbles or molecular diffusion. Ni-trogen-fertilized plants release much more methane than unfertilized plants but even these measured rates are only one fourth as large as those inferred earlier by Koyama and on which all global extrapola-tions have been based to date. We also compare our measured methane fluxes from lakes and marshes to similar earlier data and find that extant data and flux-measurement methods are insufficient for reliable global extrapolations.

We have made field measurements of methane fluxes from rice paddies, fresh water lakes, and saltwater marshes to infer estimates of the size of these sources of atmospheric methane. The rice-paddy measurements, the first of their kind, show that the principal means of methane escape is through the plants themselves as opposed to transport across the water-air interface via bubbles or molecular diffusion. Nitrogen-fertilized plants release much more methane than unfertilized plants but even these measured rates are only one fourth as large as those inferred earlier by Koyama and on which all global extrapolations have been based to date. We also compare our measured methane fluxes from lakes and marshes to similar earlier data and find that extant data and flux-measurement methods are insufficient for reliable global extrapolations.
In 1978 the tropospheric concentration of methane was about 1.72 ppm by volume in dry air in northern mid-latitudes and 1.62 ppm in southern mid-latitudes [Heidt et al., 1980]. While the first unambiguous detection of methane in the atmosphere appears to be that of Migeotte [1948], who observed its infrared absorption bands in the solar spectrum there has been very little sustained effort to observe its temporal and spatial variations. In the 1960's and 1970's a number of workers have conducted a sizable number of methane measurements mostly by flame-ionization gas chromatographic analysis of air sampled directly or stored in flasks. Absolute" calibrations for CH4 mole fraction in clean air differ slightly; Bush et al. [1978] •eported 1.63 ñ 0.12 for a 1978 worldwide average while Rasmussen and Khalil [1981] were closer to 1.60 in clean northern hemisphere air in 1978. At these concentrations methane is important in atmospheric chemistry and radiation: The oxidation of methane initiates key tropospheric reaction chains [Levy, 1971;McConnell et al., 1971;Wofsy, 1972;Crutzen, 1973;Chameides and Walker, 1973] that produce CO and H2 and strongly influence tropospheric chemistry. Also, the stratospheric oxidation of methane leads to approximately 50% of the water vapor and H2 found there and CH4 itself serves as the dominant terminator methane is biogenic is strongly supported by arguments such as those of Lovelock and Margulis [1974] who showed that a purely abiogenic thermodynamic equilibrium would predict 29 orders of magnitude less CH4 in the air than is observed. From Ehhalt's methane source surveys one can see that a large fraction of the biogenic sources are wholly or partly under man's control through his agricultural practices (mostly rice cultivation) and domestication of cattle, as are several other methane sources: natural gas handling, coal mining, etc. Coupled with possible anthropogenic influence on methane sinks [Chameides et al., 1977] these factors could lead to maninduced changes in atmospheric methane levels. Rasmussen and Khalil [1981] are measuring a methane increase in clean background air.
To provide a better data base on methane sources and to develop techniques for measuring fluxes of other largely biogenic gases we decided to devise field experiments to measure methane fluxes from several areas in California. These areas included several key environments of the type that influenced Ehhalt's surveys: rice paddies, freshwater lakes, and saltwater marshes. METHODOLOGY of chlorine-atom chains that destroy stratospheric ozone [Sto-Several problems arise in measuring fluxes of gases into the larski and Cicerone, 1974]. Further, the 7.7 /• band of CH• air from water bodies and soils. Though not completely docutraps a significant amount of outgoing planetary radiation so mented, many problems associated with gas collectors are nothat increases in atmospheric CH• levels would lead not only torious, including those of perturbing the turbulence fields in to chemical perturbations such as those mentioned by Cha-air or water, perturbing the thermal environment, or gaseous roeides et al. [1977] and Sze [1977] but also to surface temper-composition and introducing artificial gradients [see e.g., ature increases [Wang et al. 1976 to set the collector down, let it sit for 10 min, and take a held the collector stable with the rim of the carboy opening 5 cm below the water surface. The evacuated sample bottles were opened at prescribed intervals to draw samples from a central location beneath the collector. Numerous problems can be associated with this approach: (1) one establishes an artificial environment of CH4, CO2, H20 vapor and, possibly, elevated temperature; (2) one blocks any effect the wind might have on the flux; (3) one interferes with eddy transport in the water column (although our work was done in places where this was minimal), depending somewhat on how deeply the rim of the collector protrudes into the water and how the collector rim affects the liquid turbulence that renews the gas concentration near the water surface [see, e.g., Danckwerts, 1970]. Further, one must realize that methane transport by bubbles will be much less problematical in these regards than diffusive transport. To measure fluxes that involved plant material we used two kinds of collectors besides the glass carboy described above. To minimize damage to the plants, we employed a flexible saran bag attached to a 17.5-cm diameter aluminum collar illustrated in Figure lb. The volume was ~ 6 1 and the sampling In rice paddies we also used a third type of collector, a flowthrough bag shown schematically in Figure l c. This is a modified saran bag collector that had ~ 1 l/rain of ambient air drawn through it by a portable Bendix pump. The air passed from the collector directly to a portable gas chromatograph for immediate analysis. The procedure was to analyze a standard, analyze an ambient air sample, then a sample from the collector and repeat this several times. The flowing collector was used to maintain a near ambient environment around the rice plants beneath the collector. When the bag collectors were employed in rice paddies, no attempt was made to shade the bags from direct sun because in the collection interval (10-20 rain), no appreciable heating occurred. Indeed, no significant day-night differences were evident, as reported below. With the glass-carboy collector we shaded (usually but not assiduously) the sunward side of the collector with an aluminum foil wrap.
In the field, sample analysis was accomplished on an A.I.D. portable gas chromatograph equipped with a flame-ionization detector and gas-sampling valve. Three ml air samples were separated on a 5' x 1/8" Spherocarb column held at 100øC intervals were < 30 min. Evacuated 1 I spheres were con-with 25 ml/min N• carrier flow. The methane retention time nected to sampling tubes by feedthrough connectors on the was about 3 min. Signals were fed to a lmV recorder, and aluminum collar and served also to hold samples and keep the concentrations were calculated by using peak heights. The apparatus afloat. The inenness of these bags toward methane system min'.unum detectable methane concentration was 50 in air samples was tested by filling the bags with a higher than ppb and precision for repeated analysis was about 1%. Samples ambient secondary standard of methane in zero air and test-were compared to a standard prepared by Matheson of mething for methane loss. Less than 2% loss was observed after ane in zero air and calibrated in our laboratory by using a To enable us to run repeated analyses the atmospheric pressure samples were pressurized with 10 psig (68.9 kPa) of pre-purified N2 that had been previously checked for methane content. This dilution factor was applied to calculate the final methane concentrations. Once these concentrations were determined, the flux could be calculated by using the collector volume, the area covered by the collector, and the elapsed time. Dissolved oxygen was measured with an Orion Research model 399 A/F meter equipped with a model 97-08 oxygen electrode. The range was nominally 0-14 ppm (438/zM) with accuracy of :i: 0.05 ppm.

RESULTS
Rice paddy flux measurements were made in an experimental rice field located on the grounds of the University of California at Davis, (40.2øN, 122.1øW). The work was done in late summer with daytime temperatures ranging from 26 ø-32øC dropping to 17ø-20øC at night. Several parameters were measured with each collection. The water, which was 10-18 cm deep over the 350 m 2 area of the test paddy, was being replaced at a rate of ~3 l/min. It had a pH range of 6.9-8.1, while its temperature ranged from 18ø-26øC. The dissolved oxygen exhibited a cyclic diurnal behavior as was expected with daylight values supersaturated (off scale, over 440/zM) and nighttime values as low as 95/zM. Soil temperature was relatively stable at 28ø-29øC. By placing a collector over a test area and sampling the gas inside the collector after measured time intervals, we deduced the methane flux across the plane defined by the collector oririce. Figure     .36 x 10 -2 6.82 x 10 -2 4.13 x 10 -2 3.05 x 10 -2 3.83 x 10 -2 2.30 x 10 -2 3.49 x 10 -2 5.10 x 10 -2 1.32 x 10 -2 1.72 x 10 -2 1.91 • 10 -2 2.66 x 10 -2* 3.80 x 10 -2* 3.22 x 10 -2* The average over all nine of these data points is 3.28 x 10 -2 g CH4 m-2 d-•.
*Average. d, respectively, on average with no overlap at all in the range of observed fluxes (see Table 2). Proceeding down Table 2, rice fertilized with 140 kg N/ha/yr (as (NH4)2SO4) released muck base. It is covered by surface waters except in the end of the dry season, typically August and September. The Penasquitos location is similar but with much more swamp-grass vegetation. At Batiquitos the average flux was 2.16 x 10 -3 g Table 3.

CH4/m2/d for the five dates shown, while at Penasquitos the average flux was 3.65 x 10 -4 g CH4/m2/d for the 17 dates shown in
With similar methods we made 10 flux measurements at three southern California freshwater lakes, as summarized in  Table 5. In each case the collecslightly above the water line. Table 2 shows nearly equal tot without the screen showed a bigger flux, suggesting that fluxes, 0.053 and 0.044 g CH4/m2/d.

Finally, we observed no bubble ebullition is often the dominant means of escape for significant day-night differences in methane release rates from methane from the water column into the air. It can be sugfertilized rice plants. gested that the screen is impeding eddy transport in the water
We have also measured methane fluxes from several saltwa-column, but in the stagnant areas where this work was done ter marshes and lagoons in southern California. These mea-eddy transport would be weak anyhow. Further, in the course suremerits were made with the glass-carboy collectors in of similar studies we have done on N20 release from Michstanding water and are summarized in Table 3.  Koyama [1963Koyama [ , 1964 was performed by incubating rice paddy concentration in the collector over 24-hour periods rules out soils in the laboratory. On the basis of Koyama's estimate of back diffusion as a serious concern in these measurements. an annual global average methane flux of 206 g/m2/yr, Ehhalt Nonetheless, this technique is still susceptible to problems and Schmidt [1978] concluded that rice paddies account for such as those mentioned earlier. The Batiquitos location is a 2.8 x 1014 g CH4/yr globally or 33-49% of all biogenic atmoshallow, several km 2 saltwater lagoon with a thick organic spheric methane and 26-47% of all atmospheric methane from   At these locations, water pH ranged from 7.0 to 9.9 with highest pH values in daytime, lowest at night typically. Dissolved oxygen was near or supersaturated (>284/zM at 20øC) on bright days and went as low as 95/•M at night. all sources. Our measurements, if extrapolated in the same nations; even lower methane fluxes might be found elsewhere. Fertilizer type could also be an important variable. For example, because surfate-reducing bacteria compete with methanogenic bacteria for available H2 and acetate [Claypool and Kaplan, 1974;Martens and Berner, 1974] one might suspect that surfate-containing fertilizer could suppress methane generation initially. It is not clear why nitrogen-fertilized rice plants deliver more methane to the atmosphere. If it is because they are larger, more mature, and with more developed roots, then one can guess that plant maturity affects methane release: Variations during the growing season need investigation. The increasing usage of symbionts like .dzolla to provide fixed nitrogen in rice paddies in India and China should also be examined. Another likely factor is the organic content of rice paddy soils; Koyama's larger fluxes could have arisen if his research soils were more rich in metabolizable organic material than the University of California at Davis paddy soils. It is also likely that Koyama's sealed system, extremely low in oxygen, might have led to overestimates of methane release to the atmosphere. Even though his measured rates of methanogenesis could be correct, any oxidation of methane in the overlying water column of a natural rice paddy would have limited the methane release rate. For similar environments,

Ehhalt and Schmidt [1978] and Martens and Klump [1980]
have discussed the likelihood of methane oxidation in the water column above anoxic sediments; one might expect more oxidation in oxygen-supersaturated daytime waters. Our data ( Table 2) show (1) that direct transport through rice plants strongly dominates the net methane release to the atmosphere, (2) no day-night differences, and (3) fluxes 20% as large as Koyama's. Finally, Koyama's [1963, 1964  CHn/m2/yr and 5.9 x 10 •3 g CHn/yr globally. Our figure of 42 Our finding that methane transport through the rice plants g CHn/m2/yr arises from our measured methane flux of 0.18 g dominates other forms of release was surprising but not to-CHn/m2/d (see Table 2 and text) that applies to generously tally. Kozuchowski and Johnson [1978] observed that gaseous fertilized rice. Our global annual methane release (5.9 x 10 •3 g mercury compounds escape through the stomata of plants CHn) arose from assuming the same temperature dependence (common reeds) growing in mercury-contaminated lake sedifor methane generation, the same United Nations rice-cultiva-merits. Further, Dacey and Klug [1979] found that 50% or tion data, and 4-month growing season assumption that more of the methane escaping from a eutrophic lake in Mich-Koyama employed.
igan rose through water lily plants. The apparent lack of a Clearly, our measurements of methane release from rice diurnal variation in our data for methane fluxes through rice were more direct that Koyama's, and we found much less plants needs to be confirmed with more data because we have methane release, but our data are far from definitive. For ex-few data points and because Dacey and Klug [1979] and Kozuample, if the methane release depends as strongly on nitrogen chowski and Johnson [1978] found more gaseous release in fertilization as our data indicate (Table 2), then one must take daytime hours. Further, Dacey [1980] has reported tracer into account the lesser usage of N fertilizer in less developed studies that suggest a pressurized flowthrough system that  forces atmospheric oxygen to water lily roots and delivers CO•_ 0.1 by Ehhalt [1974], is very uncertain owing to lack of data to the plant leaves and methane to the atmosphere through and also owing to the great potential of plants to control the the leaves. Contrary to earlier ideas, Dacey [1980] found sig-release. As a last note on freshwater lakes, we saw no evidence nificant mass flows through plant lacunae. of diurnal variability in the methane release rate from open Our freshwater lake data (Table 4) Table 4 raises the average flux to 0.89 g CH•/m•-/d. Clearly, extrapolating from these data is perilous. Previously, four swamp-gas samples analyzed by Conger [1943], as interpreted by Ehhalt and Schmidt [1978], implied a flux of 0.32 g CH4/m•-/d from a location in Maryland. Baker-Blocker et al. [1977] trapped rising bubbles below the water surfaces of two farm ponds in Michigan and deduced fluxes between 0.09 and 1.1 g CH4/ m•-/d. As Baker-Blocker et al. and Ehhalt and Schmidt [1978] have discussed, it is desirable to deduce a temperature dependence for the methane evolution (and escape). It is questionable whether there are enough data at present to permit an accurate temperature dependence to be inferred. Further, because bubbles dissolve rapidly under some conditions [Martens and Klump, 1980], the methane escape rate to the atmosphere is likely to be less than the rate of methane release from the sediments. Additional complexities arise from the recent recognition [Dacey and Klug, 1979]  There are few data on fluxes from salt water marshes probably because it was established early on [Koyama, 1963] that the presence of salts inhibits methanogenesis. King and Wiebe [1978] show a wide range of 0.44-51 g CH4 m -•-yr -1 in Georgia salt marshes. Their measurements were made over soil with plant material and showed a seasonal dependence. The measurements we made in southern California yielded a 0.28 g CH4/m•-/yr average flux but were made in areas of standing water of 15 cm, which probably allowed a considerable amount of oxidation before the methane was released to the atmosphere. We saw no seasonal changes, but it should be noted that seasonal temperature changes are small in southern California (see Table 3). The possibility that methane and other gases escape from salt marshes directly through reedlike grasses remains to be investigated. Indeed, it seems clear that all future investigations of gas evolution from marshes, lakes, and rice paddies must recognize that areas covered with plants can be as important as open-water areas.