METHANE EMISSIONS FROM CALIFORNIA RICE PADDIES WITH VARIED TREATMENTS

. Two field experiments in California rice paddies are reported, one with a single treatment of a research plot and the other with varied treatments in a typical commercial rice field. Small total methane emissions, only 11 g CHJm 2, were measured for the entire growing season in the first experiment. In the second experiment, the addition of exogenous organic matter (rice straw), the presence or absence of vegetation, and the nitrogen fertilizer amounts were examined for their influence on methane emissions. The total methane emission over the growing season varied from 1.2 g CH4/m 2 (with no added organic matter) to 58.2 g CI-IJm (cid:127) (with largest organic matter treatments). Added organic matter was the major factor affecting methane emissions. Vegetation did not greatly affect total methane fluxes, but it did influence the mode and timing of release. Nitrogen fertilizer did not greatly affect the amount of methane emitted, but it influenced slightly the time course of the process. A diurnal effect in methane emission was observed during the early ontogeny of the crop. The variation of methane emission with time during the course of the growing season was very unusual in this experiment; only one peak was observed, and it was early in the season. During the period of largest emissions,/i(cid:127)3C values of the methane were measured to be -55.? + 1.8 %o in plots with added organic matter.


Methane has become one of the most interesting gases in the atmosphere partly because its concentration has increased worldwide at a rate of about 1% per year over the last 100-150 years [Rasmussen and Khalil
also interacts with chlorine chemistry through its reaction with CI atoms to produce HCI. Tropospheric reactions destroy perhaps 90% of the methane that enters the atmosphere; OH radicals initiate this attack. Atmospheric chemical species and processes involved in methane oxidation are also central to the control of the oxidation state of the atmosphere. These points are discussed more fully by Crutzen [1987] and Cicerone and Oremland [1988].
Thus it is important to discover the causes of the increasing trend of methane concentrations and to identify the sources and sinks of atmospheric methane and the processes that control them. Ehhalt [1974] presented the first quantitative analysis of the sources of atmospheric methane. On the basis of several indicators and Koyama's [1963] laboratory studies, Ehhalt postulated that the world's rice paddies would be a major methane source. Flooded soils are certainly good candidates to be methane producers: flooding results in a rapid lowering of oxidation-reduction potential with the consumption of all available electron acceptors including inorganic ions such as ferric and manganic ions, nitrate, and sulfate. This sequence terminates with the reduction of carbon dioxide and with the formation of methane.
Methane is also produced in the direct heterotrophic demethylation of lignin and other organics such as acetate. Soil chemists and microbiologists have shown clearly how methane is formed in such environments [e.g., Yamane and Sato, 1963;Takai, 1970;Ponnamperuma, 1972;Neue and Scharpenseel, 1984;Oreroland, 1988]. In soils that can remain flooded, and hence favor rice agriculture, rice is relatively shallow-rooted and diffusion of oxygen into the soil is slow, particularly when the soil is flooded. Thus any readily decomposed organic matter remaining in the soil at the time of flooding results in a rapid lowering of oxidation-reduction potential with the reduction of all available electron acceptors and soil transformations that produce methane, as outlined above.
Early field investigations of methane emissions from rice paddies were reported by Cicerone and Shetter [1981], Cicerone et al. [1983], and Seiler et al. [1984]; the latter two studies covered almost complete growing seasons. Cicerone et al. [1983] reported large changes in the rate of methane emission with peak emissions observed in September; 90% of the total emission for the growing season occurred in September. Seiler et al. [1984] found a strong but different seasonal variation and a distinct time-of-day (largest emissions in the late afternoon) variation. Time-of-day variations correlated positively with soil temperatures. In addition, these studies showed that methane transport through the rice plants greatly exceeded the rate of methane escape carried by diffusion across the water-air interface and by bubbles. Evidence that methane is transported primarily through the plants has become more convincing with time [ In this paper we report results of two further field experiments in California. The first was a relatively simple season-long observation of methane fluxes that we conducted in 1983.
The second was a more comprehensive experiment in 1985 in which, through the cooperation of a local rice grower, we were able to monitor and manipulate a portion of one of the more typical and more progressive large-scale rice producing activities in California.
In the next section, we describe the design of these experiments, along with the soils of the experimental sites and the analytical methods. Sections 3 and 4 of the paper re•rt and discuss the results of the field experiments. It is moderately well drained, somewhat atypical of rice fields where percolation rate usually is low. Its availability and proximity to research laboratories were the principal factors determining its application to rice. Methane emissions were measured through the use of static chambers placed over the rice plants. The chambers were rigid cylindrical sections of transparent polycarbonate material; the bases of the collectors were grooved, circular aluminum sections placed permanently into the paddy soil and beneath water level. During a sampling period, the polycarbonate collector (22 cm diameter) was placed into the grooved base, thus preventing gaseous exchange with the atmosphere. Sampling ports of stainless steel tubing extended from the top of the polycarbonate collectors downward to the midpoint of the collector. An open capillary tube in the collector top allowed internal and external pressures to equalize without appreciable air flow. Early in the season we used a shorter collector (45 cm high) and as the rice grew, longer sections (70 cm high). Samples were withdrawn into evacuated stainless steel flasks from the collectors l0 or 20 rain after emplacement as were air samples of ambient air Oust below rice-top height). Equipment and method of analysis were the same as those described by Cicerone et al. [1983]. Soil treatments and dates of treatments and of other significant events are listed in Table 1. A wooden pier was placed into the field before flooding. From this pier we were able to reach the first five rows of rice plants in the field without disturbing the water-covered soil.

_
The 1985 experiment was more complex and intensive using twelve test plots in a commercial October 1983 was unusually dry soil surface dry, cracks appear in soil.
Field was leveled in summer, 1982 and lay fallow until 1983. Rice was not cut or harvested during this experiment. Soil type was Capay silty clay; see text.
rice-growing field. During the previous year, rice was also grown commercially on this field; after harvest the remaining crop stubble was burned in place. This soil is less permeable to water than was the case for the sites of the 1983 experiment described in this paper or the 1982 experiment described by Cicerone el al. [1983]. It is classified as Sacramento Clay, a poorly drained montmorillonitic, noncalcareous thermic clay with a low water penetration rate. Although this heavy soil with limited drainage poses some management difficulties for wheat and maize, it is ideally suited to rice culture in many respects and is representative of many commercial rice-producing areas of the world where tillage characteristics and factors of water management tend to select for rice as the main agricultural cereal. One of the more notable differences between the management of this and similar California rice fields and many other rice growing regions lies in the use of mineral nitrogen fertilizers or urea as contrasted with higher carbon 'organic H sources. Table 2 lists the various events in the preparation and treatment of the 12 test plots and corresponding dates in 1985. Fertilizer types and amounts are listed there. Six plots (7-12) were given nitrogen fertilizer (urea) and the other six (1-6) were not. Four plots (1, 4, 7, 10) received no added organic matter. To four other plots (2, 5, 8, 11) were added 250 g/re' of dry organic matter (rice straw), and the remaining four plots (3, 6, 9, 12) received 500 g/m 2 of the same organic matter. The rice straw was reduced to particles of about 1 mm diameter (by grinding) before application. Rice was planted on the entire field, including all 12 test plots, by aerial broadcasting of seeds on May 10. Plants and roots were removed from plots 1, 2, and 3 on June 13 and from plots 7, 8 and 9 on June 19. Figure 1 illustrates the configuration of the 12 test plots.
Although the individual test plots were only 1 m 2, there was no evidence of interactions between plots. For example, the lines of demarkation between plots with and without added nitrogen fertilizer were quite sharp, as can be seen in Plate la.
Before the field was flooded we placed a wooden pier alongside the test plots (see Plate 1) to allow access to each of the 12 plots without causing excessive pressure or disturbance to the soil, water or plants. We measured the methane fluxes from the 12 plots using cylindrical chambers, or enclosures as shown in  During the previous summer (1984) the field was planted to rice and at the end of the season after harvest the residual straw and stubble were burned. Soil type was Sacramento Clay; see text.
base. Sampling ports were fed into these base sections; inside each section, copper sampling tubes (1/8 inch) rose to about 30 cm above water level. Gas samples were withdrawn by external teflon lines (1/8 inch joined to the internal tubes through a connector Sample-loop sizes were 3 em 3 for both GCs, and signals were displayed and processed by recording integrators (Hewlett-Packard Model 3390). Ultrapure nitrogen was used as a carrier gas; flame gases (hydrogen and oxygen) were prepurified by passing through molecular sieve traps.
Methane concentrations in whole-air samples (includes water vapor) were determined by peak areas ratioed to peak areas obtained from analysis of pressurized reference standards. Standards were prepared at four concentrations: 1.71, 3.9, 23 and ?2 ppm by volume. These reference standards were stored at high pressures in stainless steel flasks. Methane concentrations in each flask were compared against reference standards supplied by the National Bureau of Standards and GC respomes were essentially linear over the concentration range of the standards.
Methane emission fluxes F were calculated from the measured concentrations inside the collectors as follows:

F = M(T) [V/A] [•f/•t],
and T is temperature in degrees celsius. Mo is Lochschmidt's number, 2.69x10" molecules/cm'. As described above, initial methane concentrations were measured and were used to compute Af values. Supporting data were gathered during each sampling period on temperatures of air (inside and outside the enclosures), of water, and of soil. In addition, the heights of the rice plants were recorded along with surface wind direction and estimated speed and indicators of cloudiness or lack of. Visual scans were made for presence of bubbles rising from the soil surface upward through the water colunto. Water depth was also measured, and the lowest rim of each collector system was adjusted (if necessary) to remain below water.
On two dates, enclosure samples were collected for stable carbon isotope analysis of the methane. On June 27, 1985, samples were collected from each of the organic matter plots which had been treated with 500 g/m 2 of rice straw (plot numbers 3, 6, 9, and 12). The samples were obtained by withdrawing headspace air from the chambers into the evacuated cylinders after an enclosure time of 13 min. One additional sample was collected from plot 6 on September 6, 1985. This sample was collected from the enclosure using a 7.61-L stainless steel cylinder and metal bellows pump. Three 30-min enclosures of the rice plot were combined to obtain enough methane for sample analysis. Each time, about 7.61 L atm of headspace air were collected, the latter two collections being made using the pumping system. Between each enclosure time period, the rice was uncovered for a period of 40 min.
Samples for stable carbon isotope analysis were measured using gas chromatography techniques described in this paper. They were then prepared for mass spectrometer analysis. The procedure for methane combustion and subsequent measurement on a Nuclide 6-60 isotope ratio mass spectrometer are described in detail by Tyler [1986]. week during July and August, 1983 (although on each day two distinct measurements were made with paired chambers) and more frequently in September and early October. Altogether there were 86 observations in the l 11-day period from July 9 -October 27. Most measurements were made between 1100 and 1300 LT, but during September other times were sampled, including various times of night from dusk to dawn.
Between September 5 and 9, 1983, a period of relatively large methane emissions, we performed 13 daytime measurements and 10 measurements in darkness. The means (and standard deviations) of the 13 daytime and 10 nighttime fluxes were 0.152 (0.14) and 0.108 (0.03) g CH4/m2d, respectively. Although these results could indicate larger daytime fluxes than at night, the large standard deviation of the daytime fluxes renders the difference statistically insignificant. Of the 13 daytime fluxes, there were two that were much larger than average, about two standard deviations above the mean; these values were 0.41 and 0.49 g CH4/m2d at 1020 and 1650 local time on September 7, 1983. Such fluxes that are abnormally large compared to the daily averages of the period have been observed previously and they probably are due to sporadic bubbles rising through the water into the air of the collection chamber [Cicerone and Shetter, 1981; Holzapfel-Pschorn and Seiler, 1986]. When present, bubbles also cause larger variability in the data. Although we did not sight bubbles visually in the period September 5-9, 1983, the fact that the standard deviation of the daytime fluxes was about 4 times as large as the nighttime value suggests that bubbles were present at times during sunlight even if they were not easily seen. If we remove the two largest daytime fluxes from this data group, then the means (  day to night in either the mean fluxes or the variability of the fluxes during this period of about 4 days. During this period, rice-plant heights were between 58 and 72 cm so that the underlying soil was rather well shaded. Soil temperatures (at 5-to 10-era depth) varied between lows of 17o C (late night and early morning) and highs of 220 to 240 C in late afternoon. In any case, the point of these tests was not to investigate the mechanism underlying any diurnal variations but instead to estimate how the computed total flux over the growing season might be affected by the times of day of the individual observations. where the vegetation had been removed. The presence and the effects of bubbles were apparent from both visual observations and by occasional changes of slopes in (1) that were determined by sequential collections. This kind of behavior, especially for unvegetated plots, has been reported by Holzapfel-Pschorn and Seiler [1986]. Where rice plants were present there was some release of bubbles, particularly before tillering was extensive, but the major transport and release was probably by the vascular system of the plants.
Methane production was not appreciable until oxygen became limited, about 30 days after planting depending upon the amount of organic matter present.
Undoubtedly, a considerable fraction of added organic matter had been oxidized by this time. However, when methane production did become significant, it rose rapidly to a maximum value at about day 50 and then declined exponentially through the season, the timing of peak emission and the decay constant depending upon the treatment.
The influence of added organic matter, nitrogen fertilizer, and the presence or absence of vegetation can be discerned from Figure 3 and is discussed in greater detail below.

3.2.2.
The influence of added orcanit matter.
Added organic matter was the single most important factor influencing methane production in this system. Most of the readily decomposed organic material from the previous year had been oxidized prior to planting. Figure 4 compares the mean values for treatments 1, 4, 7 and 10 (no added OM) with those for treatments 3, 6, 9, and 12 (500 g/m 2 or 5 metric tons/ha ground rice straw). Those plots to which 500 g/m • of organic matter had been added had about a 20-fold greater methane emission (presumably due to proportionately larger methane production rates) than did the control plots. Figures 3 and 4. This leaves the impression that no methane was emitted from those plots to which no organic matter was added. When the data from plots with no added organic matter are shown on expanded ordinates (Figure 5), the same general pattern of methane emission as in the added organic matter treatments is revealed. Without vegetation to serve as a conduit for methane more variability is seen (Figure 5a, 5c), probably due to sporadic bubble release. The seasonal yield of the four plots without added organic matter was about 5% of the methane released from the four plots with 500 g/m 2 of added organic matter (see also Table 3). By the end of 125 days, methane emission had dropped off to generally low values for all plots and draining of the field preparatory to harvest was begun. When the soil surface had begun to show dry spots and some cracking in the clay, there was an increased release of methane; on day number 136, some plots showed high values comparable with the peak production earlier in the season. This probably represented a physical release of trapped methane and contributed significantly to the season total. Because of the random nature of cracking and timing of the peak release, it also contributed to variability in the data. When linear interpolation is made between data points, including those of this pulse at the end of the crop season, it is possible to make an estimate of the total seasonal methane release.  Table 2 and Figure 1). Only one peak period of methane emission was observed, near day 60, or 50 days after planting.

6b we used only data from vegetated plots because the added variability from unvegetated plots led to less good curve fits.) This may have been due to faster consumption of available organic maRer by methanogenic bacteria when N was less limiting.
Interpretation of these data is clouded by the probability that added nitrogen also accelerated organic decay in the first part of the season before plant growth and other oxygen demand had depleted oxygen supply. Under the highly reducing conditions of methane formation the fixation of nitrogen by autotrophic methanogens as well as other organisms can help provide a nitrogen source, but the energy requirement for spliRing the dinitrogen bond still must be met and places limitations on the system. Because the study was conducted with the cooperation and support of a commercial grower and because other logistic and work force limitations prevented large enough treatment areas to get statistically significant data on yield, we were not able to make reliable comparisons of methane production or other variables relative to yield. periodicity with peak emission in early afternoon and minimum in the early morning. No strong pattern was observed once the crop canopy was closed, and temperature fluctuations in the soil were less. Figure  7 gives mean values for all samples from vegetated areas for days 50 through 65 using 3 hour bins. It is apparent that daylight hours show a stronger activity; the daytime peak is as much as 70% larger than the nighttime minimum, at least for this time period (no other time period showed such clear variations in this experiment). It is assumed that this is largely a positive temperature effect in methanogenesis. Because release is presumably largely through the vegetation, stomatal opening may be a factor. Diurnal variations could also be influenced by diurnal changes in methane oxidation rates.
Because sample collection was at irregular times and intervals as dictated by logistic factors, it was necessary to normalize observed values for methane production according to collection time in order to estimate integrated production through the season. For this purpose the data for days 50 through 65 given in Figure 7 were used to calculate normalization factors. Table 3 gives calculated season totals for methane for the various treatments. The variability in recovery of added organic matter as methane is evident and an expression of the difficulty inherent in measurements of this sort. The low recovery of carbon as methane in treatments 2, 9, and 11 when compared with the remaining treatments is not easily explained. Variability in soil and vegetation as well as the difficulty experienced in getting uniform distribution of added organic matter and the random effect of bubbles all undoubtedly contribute to this inconsistency. Mean recovery of carbon as methane amounted to 15.6% of that applied for plots receiving 500 g of rice straw per meter square at the beginning of the season. For those plots receiving 250 g straw per meter square the recovery was 12.0%. Variability was sufficiently large that no significance is attached to this difference, although it is consistent with the expected influence of added organic matter. When all treatments are considered, the overall recovery of added carbon in methane was 14.4%.

3.2.7.
•5•3CH, in the 1985 exveriment. Data for _ the stable carbon isotope measurements of the rice paddy field methane emissions are shown in Table 4 Because relatively large quantities of sample are needed for isotopic measurements, the rapid drop off in methane flux soon after the end of June prevented us from obtaining systematic collections of methane for isotopic analysis throughout the growing season. However, the difference between the earlier June values and later September value is well outside the uncertainty of measurement. It might indicate a difference between the pathways of methane production and/or consumption during the growing season. This seems quite plausible, although with few exceptions the idea has not been tested in other published data sets.  In another seasonal study, Nakamura et al. [1990] measured carbon isotopes of both CO• and CH, from soil gases in paddy fields in northeast and central Japan (Konosu, Japan, 1976 to 1980). They investigated four types of fields each of which have undergone a long-term (over 28 years) fertilization experiment (varying treatments) for total inorganic carbon and related organic carbons. However, they report only a single •SI3CH, value of-55.1%oo for samples taken over the entire time period. This was for methane bubbles from submerged soil in the field receiving chemical fertilizer. They also report a value of-52.9 + 5.1%o for samples taken from incubation studies in the organic mature plot over a 6-week time period. This relatively large standard deviation could be due to changes in methane production and/or consumption, but the data as presented are not sufficient to draw conclusions about the seasonality of õ•CH, emissions from the paddy fields.

FURTHER DISCUSSION AND CONCLUSION
A major goal of this research and of many other investigators is to estimate the total annual emission of methane from the world's rice paddies. It has become apparent [e.g. Holzapfel-Pschom and Seiler, 1986] that it is no easy task to accurately measure the total arnourn emitted even by a single field over a full growing season; this is because of variations with time of day, time of season, soil treatments, and several other factors. Before accurate estimates of total annual emissions can be derived we must observe how methane emissions vary and determine the factors that control these variations. Now we discuss the patterns that were observed in this study and how they depended on experimental variables.
The seasonal variations that we observed in 1985, Figures 3, 4, and 5, were unanticipated. In almost all of the 12 plots methane emissiom displayed a single maximum on or before day 50 (measured from April 30), or 40 days after planting (see Table 2). The absence of middle and late season emissions (except for day 136 when the soils were dry with large cracks) was not a feature of our earlier seasonal studies in 1982 [Cicerone et al. 1983]  it was also seen by Seiler et al. [ 1984] in Spain and by Cicerone et al. [1983] in California. This maximum may be due to the action of methanogenic soil bacteria on organic material released by rice roots as root exudates. The third seasonal peak discerned by Schiitz et al. occurred in August in two of the three years in Italy but not in the third. An extremely large late-season peak was observed by Cicerone et al. [1983] in September in California in 1982 and major peaks were seen by Seiler et al. [1984] in late August in Spain and by Yagi and Minami in early September in Japan. Schlitz et al. suggested that this late peak may be due to microbial decomposition of rice roots.
It is important to learn whether there are three or more distinct peaks to be seen at each site and the causes for them. In our 1985 studies, only the first peak appeared. Why were the second and third peaks suppressed? The answers could be both low methane production rates at those times and high methane oxidation rates. These processes were not studied in our experiments. In any case, the strong relationship between methane emissions and above ground biomass that was observed by Sass et al. [1990] in one of two Texas fields through most of a growing season did not obtain in our 1985 experiments. Other factors such as water management practices and fertilizer applications can also cause variations in methane emissions during the growing season [Yagi and Minami, 1990], as can cultivar sensitivity to soil anaerobiosis (and hence to added organic matter) [Sass et al., 1991].  Table 3. The total methane emitted over the entire growing season from the four plots that received no added organic matter was 1.7 g/m 2 (average over four plots); the range was 1.2 to 2.9 g/m 2. For the four plots that received 250 g/m • of organic matter the average total methane emission was 15.7 g/m •, and the range was 6.3 to 27.8 g/m •. The four plots that received the largest application of organic matter, 500 g/m •, emitted 38.0 g CH4/m 2 on average; values ranged from 20.8 to 58.2 g/m •. These total methane emissions were calculated by integrating over the period of the observations, May 17 through October 3 (days 17 through 148) using the diurnal temporal dependence described in the previous section and Figure 7 to normalize observations from different times of day. The strong influence of organic matter additions over methane emissions has also been observed by Schlitz et al. [1989], Yagi and Minami [1990], and Sass et al. [1991]. Schlitz et al. found that emission rates of methane increased with increasing quantities of incorporated rice straw and reached maximum values at an application rate of 12 tonnes/hectare (t/ha). At this maximum, methane emissions were about 2.4 times higher than from an untreated control plot. In our 1985 experiments the enhancement was almost 10-fold (1.7 g/m: to 15.7 g/m :) between untreated control plots and those treated with 250 g/m • (2.5 t/ha) organic matter. At 5 t/ha application rate our emissions were enhanced by about another factor of 2 over the 2.5 t/ha case, 38 g CH4/m 2 compared to 15.7 g/m •. Yagi and Minami [1990] used application rates of between 6 and 9 t/ha, and they found enhancements of methane emissions of 1.8 to 3.5 times. Sass et al. [1991] found that methane emissions were enhanced by 2 to 3 times and rice yields decreased when 8 to 12 t/ha of rice straw were applied to test plots. The array of environmental, microbiological, and agricultural factors that control methane fluxes from rice paddies imo the atmosphere is quite large and complex. First, future efforts to minimize these emissions must be firmly based on mechanistic understanding of the controlling factors. Second, they must recognize the goals and needs of rice growers. In our 1985 study [and those of Schlitz et al. [1989], Yagi andMinami [1990], and Sass et al. [1990], added organic matter was the principal contributor to methane production and escape. The crop residue had been burned the previous year, and ample time had passed for the oxidation of remaining readily decomposable organic material to be oxidized in the disked field before the flooding and planting in 1985. The decline of methane fluxes that we observed through the growing season (except for the terminal burst when the field was drained) was very different from our observations of 1982 and 1983, and it has not appeared in other studies; generally, methane fluxes have been observed to increase through the season. Many factors could have combined to cause this seasonal behavior in 1985, including past history of the field, soil texture and water penetration rates. More complete studies of methane production, oxidation and escape rates will be needed along with manipulation of soil, plant, and water variables to understand how seasonal patterns unfold. Stable isotopes of carbon and hydrogen should be measured in the methane in soil, water and air as functions of time throughout the growing season; these isotopes could provide evidence of shifting pathways of methanogenesis and of methane oxidation. In addition, such data are needed before isotope-weighted methane flux budgets can be constructed.
Given the clear methane-emission consequences of adding organic matter to rice-paddy soils, a final word about the practice of burning crop stubble is indicated. In California (and elsewhere) it is common practice to dispose of postharvest field residues of straw and stubble by burning, but this practice is coming under increasing criticism because of airpollution problems in surrounding regions. Rice growers express other concerns such as the control of fungal diseases and other problems of sequential cropping, concerns that would be amplified if burning were to be prohibited. Further, if the organic residues are disked into the soil instead of being burned, their decomposition represents a potential source of methane, subject to the uncertainties of rainfall and temperature. Each mole of methane emitted into the atmosphere exerts a greenhouse radiative forcing that is about 5 times that due to a mole of carbon dioxide [Dickinson and Cicerone, 1986; Lashof and Ahuja, 1990; Rodhe, 1990], so burning the residues could reduce the greenhouse radiative forcing due to rice agriculture. Before deciding on such a course, however, each of these factors deserves some attention along with the effects of NOx and CO emitted by burning.