Effect of gas bubbles on the diffusive flux of methane in anoxic paddy soil

Limml. Oceonogr., 43(7), 1998, 15 11-1518 0 1998, by the American Society of Limnology and Oceanography. Inc. Effect of gas bubbles on the diffusive flux of methane in anoxic paddy soil Franz Rothfuss and Ralf Conrad Max-Planck Institut fiir terrestrische Mikrobiologie, Karl-von-Frisch-Str,, D-35043 Marburg, Germany Abstract The emission of CH, from paddy soil is driven by CH, concentration gradients in the submerged soil. CH, concentration gradients between anoxic methanogenic and oxic methanotrophic soil layers were measured in paddy soil microcosms by using gas diffusion probes with a spatial resolution of 1 mm. The CH, emission rate was measured by placing a microcosm into a chamber containing an atmosphere of either synthetic air (80% N,, 20% 0,) or N,. The CH,Aux was 1.6? 5.4 nmol cmmz d-l under synthetic air and 288 + 10 nmol cmmz d- under N,. The difference between the oxic and the anoxic CH, fluxes was due to CH, oxidation. The vertical CH, concentration gradients indicated CH, oxidation at 2-3 mm depth. Below this depth CH, concentrations increased steadily to about 10 mm depth, below which accumulation of gas bubbles was observed. The diffusive Aux calculated by Fick s first law from the linear part of the gradient was 166 + 14 nmol cm-.* d-l. Obviously, the flux calculated from molecular diffusion was smaller than the flux that was actually measured under N,. An important condition for the use of Fick s law is that the slope of the gradient used for the calculation is taken in the direction where the slope is steepest. This direction is not necessarily identical with the vertical direction if CH, concentrations also change in horizontal direction. Measurement of horizontal and vertical CH, profiles demonstrated that the gradients had a three-dimensional structure. The reason for this structure was that the isopleths of identical CH, concentrations followed the uneven surface of the gas bubble layer as the main direct source for the CH, diffusion gradients. We conclude that gas bubbles do not only directly cause a CH, flux by ebullition but also indirectly affect the diffusional flux of CH, in soil or sediment. When the diffusive CH, flux was calculated from the concentration gradient at 6-8 mm depth, it was larger (224? 70 nmol cmm2 d-l) than from that at 3-5 mm depth (125? 86 nmol cm-2 d-l). Thus, a transport process in addition to molecular diffusion seemed to be active in the upper soil layers, possibly bioirrigation. - Flooded rice fields and other wetlands are the most important sources in the budget of atmospheric CH,, a greenhouse gas, which has increased by about 1% per year over. the last decades (Cicerone and Oremland 1988; Khalil et al. 1993). Recently, the annual rate of increase has been slowing down (Dlugokencky et al. 1994). To understand CH, fluxes from wetlands, it is necessary to understand the involved processes and their interactions on a mechanistic level. The CH, fluxes from rice fields are largely controlled by microbial production and oxidation of CH, and by the processes involved in transport of CH, from the soil into the atmosphere, i.e., vascular transport, ebullition, and diffusive flux (Conrad 1989, 1993). Of particular importance is the diffusive flux through anoxic-oxic interfaces, i.e., the shallow oxic layers at the soil surface and around the rice roots where part of the produced CH, is oxidized (Holzapfel- Pschorn et al. 1985; Conrad and Rothfuss 1991; Frenzel et al. 1992; Epp and Chanton 1993; Banker et al. 1995; Gilbert and Frenzel 1995; Denier van der Gon and Neue 1996). In unplanted soil, CH, is only emitted by diffusion and ebullition. More than 80% of the diffusive CH, flux is attenuated by CH, oxidation in the oxic soil surface layers (Conrad and Rothfuss 1991). Similar conditions exist in other vegetated and unvegetated wetlands (reviewed by Chanton and Dacey 1991; Chanton and Whiting 1995; Hanson and Hanson 1996). Corresponding author. Acknowledgments This work was financially supported by the European Commission (EU5V-CT94-0499; BI04-CT96-0419) and the Fonds der Chemischcn Industrie. 1511 The diffusive flux of gases can be determined from their concentration gradients using Fick s first law. However, the actually measured CO, flux from a sediment was 2-28 times larger than that predicted from molecular diffusion, indicating the operation of additional transport processes such as bioirrigation (Middelburg et al. 1995). Three-dimensional transport processes larger than one-dimensional molecular diffusion were observed in sediments for 0, uptake rate, which was affected by the extension of the diffusive boundary layer above the sediment surface (Gundersen and Joergensen 1990). These examples show that the determination of diffusive fluxes at the sediment surface from concentration gradients in the sediment may be complicated by processes such as turbulence or bioirrigation. In contrast to diffusive transport, the ebullition of CH, bypasses the oxic soil layers and their methanotrophic activity (Chanton and Whiting 1995). The bubble flux of CH,can reach similar values as the diffusive flux (Chanton et al. 1989; Kelley et al. 1990). In addition, the emerging bubbles may produce channels in the sediment that allow higher rates of diffusive transport (Martens and Klump 1980; Martens et al. 1980). Furthermore, gas bubbles may partially dissolve during their migration toward the sediment surface. Martens and Klump (1980) observed that 15% of the CH, contained in rising gas bubbles dissolved in the water column. Thus, gas bubble formation and gas bubble transport may also af- fect the diffusive flux of CH,. However, information on direct effects of gas bubble formation on diffusive processes and the structure of CH, gradients is so far lacking. Recently, we have reported on the use of gas diffusion probes to measure concentration gradients of CH, in paddy soil at high resolution and to detect CH, bubbles within the

1512 Rothfuss and Conrad soil matrix (Rothfuss and Conrad 1994; Rothfuss et al. 1994, 1996). Here we compared actually measured CH4 fluxes to those calculated from the concentration gradients using microcosms of rice paddy soil, and studied the influence of gas bubbles on the CH4 concentration gradients. Materials and methods Paddy soil was sampled from Italian rice fields in February 1991 and stored under air-dry conditions as described by Conrad et al. (1987). The paddy soil microcosms were prepared in the following way. The soil was sieved (l-mm mesh size) and filled into glass vessels (36-n-m inner diameter [i.d.], 22-mm height; or 92-mm i.d., 52-mm height) to a height of 15 mm or 28 nun, respectively. The relatively shallow soil depth allowed CH4 production in deeper soil layers but prevented the excessive production of gas bubbles that could have caused frequent ebullition events. The third type of microcosm consisted of an aquarium (30 20 20 cm; length width height) filled to a height of 10 cm with soil. The soil in the microcosms was then flooded with degassed water and carefully mixed with a spatula to avoid entrapment of air bubbles. All microcosms were incubated under aerated water at 25 C. After 1 month of incubation, the CH4 gradients had reached the typical structure and a gas bubble layer had formed. This structure remained constant for up to 6 months during which time the microcosms were used for the experiments. In the two smaller microcosms, ebullition events were sufficiently infrequent to allow the measurement of diffusive fluxes from the linear increase of CH4 in an incubation chamber (see below) without disturbance by the sudden increase of CH4 due to the release of a gas bubble (Chanton and Whiting 1995). Vertical profiles of CH4 concentrations were measured by using gas diffusion probes made from stainless steel (Rothfuss et al. 1994). The sensitive part of the diffusion probe consisted of silicone membrane-covered windows (0.5-mm diameter) through which the CH4 could diffuse from the aqueous phase of the soil into the gas phase of the probe. The physical behavior of the probe and the way of handling was described earlier (Rothfuss and Conrad 1994; Rothfuss et al. 1996). The probe was routinely calibrated in an artificial sediment of glass beads (100 µm diameter), which was soaked with CH4 standard solutions. The detection limit was 10 µm CH4. The accuracy of the probe was determined by measuring linear CH4 gradients in sterilized paddy soil. The paddy soil microcosms were prepared as described above and then inactivated by exposure to 5 mm HgCl2. Linear gradients of CH4 were generated by the constant source technique (Duursma and Hoede 1967; Revsbech 1989) and were then measured with a precision of ±3.3% (Rothfuss and Conrad 1994). A similar technique was used to generate exponential CH4 gradients and to determine the diffusion coefficient in sterilized paddy soil (see below). The results of these experiments were described in detail by Rothfuss and Conrad (1994) and were consistent with the assumption that the paddy soil microcosms were homogenous with respect to physical conditions (porosity, diffusion coefficient) at least under sterile conditions in the absence of possible bioturbation. The probe was positioned with a micromanipulator that could be moved along three orthogonal axes (x, y, z; z = vertical axis). For measuring the two-dimensional distribution of dissolved CH4 at z = 6 mm depth, the diffusion probe was moved to a position near the glass wall of the soil microcosm. At this site the soil surface (z = 0) was defined by the position of where the windows of the diffusion probe were just below the soil surface. Because the soil was slightly curved up, the soil surface at the center of the microcosm was 1.5 mm higher than at the border. Isopleths of identical CH4 concentrations were drawn using the Microcal Origin 4.0 program. For measurement of randomly distributed concentration profiles, pairs of random numbers were used as horizontal coordinates. Diffusive fluxes (J) were calculated from the slopes (K/ az) of more or less linear parts of the CH4 concentration gradients, i.e., between 3 and 9 mm depth. The slopes were calculated by linear regression. For higher resolution, the profiles were divided into two depth zones (3-5 mm and 6-8 mm depth) resulting in an approximation of a nonlinear gradient by two linear sections analogous to the model of intersecting straight lines (Anderson and Nelson 1975). The diffusion coefficient of CH4 (Ds = 5.09 10-6 cm2 s-1) and the porosity (4 = 0.54) were determined in the same soil microcosms (Rothfuss and Conrad 1994). Because the microcosms were prepared with homogenized soil, Ds and 4 were a priori assumed to be constant. The CH4 flux was then calculated from Fick s first law (Cranck 1975): Image analysis on the distribution of gas bubbles at the glass wall of a microcosm (92 mm diameter) was done by using a dissection microscope with 400-fold magnification and equipped with a calibrated ocular grid. Four areas (27.6 27.6 mm each) were chosen on the glass wall. Within these areas the numbers of grid boxes covering visible gas bubbles were counted in horizontal layers of 1 mm each. The total number of bubbles was 126. Air-filled porosity was calculated from number of grid boxes divided by the total number of grid boxes in the layer. We assumed that the size distribution of the bubbles visible at the glass wall was representative for the bubbles in general. Gas from individual bubbles was sampled in the following way: A capillary (0.25 mm i.d., 0.35 mm outer diameter, 97 mm long) with hydrophobic surface (Microfil, Berlin, Germany) was mounted on a microliter syringe. The volume within the capillary and the syringe was filled completely up to the piston (without any residual gas bubble) with distilled water. The microliter syringe was fixed on the micromanipulator and positioned along the glass wall of the aquarium. A gas bubble of >2 mm in extension was selected and the capillary was moved down till the tip was inside the bubble (checked with a pocket lens). About 1 µ1 of the gas was sucked into the capillary. Then the capillary was moved up and 0.5 µ1 water was sucked into the capillary so that the gas sample was closed off. The gas sample was then transferred from the microliter syringe into a gas-sampling device (Rothfuss and Conrad 1994) containing 1 ml N2 (99.999%; Messer-Griesheim, Dusseldorf, Germany). After mixing, the (1)

Methane in anoxic soil 1513 CH4 and CO2 mixing ratios were analyzed by gas chromatography and corrected for dilution (Conrad et al. 1989). Emission rates of CH4 from soil microcosms were measured by placing a microcosm (36 mm diameter) into a Plexiglas chamber (85 ml volume). The flooding water had been removed from the microcosm so that the soil surface was just covered by a thin (<l mm) water film. The chamber was flushed with synthetic air (5 min) or N2 (1 h) for oxic or anoxic flux measurements, respectively. Then, the temporal increase of the CH4 mixing ratio was followed for 30 min by taking a gas sample every 5 min with a syringe. The gas headspace was mixed before each sampling by pumping with the syringe. Emission of CH4 from microcosms of 92 mm diameter were measured in the same way but using a larger flux chamber of 420 ml volume. The gas headspace in this chamber was mixed with a small ventilator. The CH4 emission rate was determined by linear regression of the temporal increase of CH4 inside the chamber using the gas volume of the chamber and the surface area of the soil. 1 Results Vertical CH4 concentration profiles were measured in a soil microcosm (36 mm diameter) that was incubated for 1 h under air or N2 (Fig. 1). In the oxic profile (air headspace), the CH4 concentrations in the upper 2-mm layer were below the detection limit of 28 µm due to CH4 consumption by methanotrophic bacteria. Between 2 and 3 mm depth the CH4 concentration gradually increased, reached its steepest gradient between 3 and 4 mm depth, and then became slightly less steep. In the anoxic profile (N2 headspace), the CH4 concentrations increased right from the surface because the methanotrophic bacteria were unable to oxidize CH4. In deeper layers, however, the CH4 concentration profile was similar to that under oxic conditions, reaching its steepest gradient between 3 and 4 mm depth. When the microcosm was incubated in a flux chamber, CH4 increased linearly under both oxic and anoxic conditions (Fig. 2). The emission rates (mean ± SD) were 4.5 ± 2.6 nmol cm-2 d-l and 390 ± 0.08 nmol cm-2 d-1, respectively, indicating that >98% of the anoxic CH4 flux was attenuated by CH4 oxidation. On the other hand, the diffusive CH4 flux that was calculated from the steepest gradient of the CH4 concentration in the soil using Fick s law was only 211-323 nmol cm-2 d-l, indicating that only 54-83% of the observed anoxic emission could be explained by molecular diffusion. Measurements that were repeated with larger soil microcosms (92 mm diameter) resulted in similar fluxes, i.e., 1.6 ± 5.4 nmol cm-2 d-l under oxic and 288 ± 10 nmol cm-2 d-l under anoxic conditions. The diffusive CH4 flux that was calculated from the steepest part of the CH4 profiles explained only 67% of the fluxes that were actually measured under anoxic conditions. The fact that the soil surface was slightly curved upward in the middle of the microcosm and thus increased the actual soil surface by about 1.5% was not an explanation for the discrepancy between calculated diffusive flux and actual emission. Also, we never observed an ebullition event when the microcosms with the relatively shallow soil depths (36 and 92 mm diameter) were used for Fig. 1. CH4 concentration profiles in microcosms (36 mm diameter) incubated for 29 d under aerated water at 25 C. The oxic profile was measured just after the overlaying water was removed and the CH4 emission rate under oxic conditions was measured (see Fig. 2). Then the microcosm was incubated under an N2 atmosphere For 1 h. Alterward the anoxic CH4 emission rate was measured followed by the measurement of the vertical CH4 profile. flux measurements. The linear increase of CH4 during the incubation in the flux chamber was never interrupted by a sudden increase in CH4 that would have been caused by the release of a gas bubble (Conrad and Schütz 1988). To check for statistical variations, n = 5 vertical profiles were randomly measured over the soil surface (Fig. 3). The linear regression at 3-9 mm depth of the pooled data resulted in a diffusive CH4 flux of 166 ± 14 nmol cm-2 d-l, i.e., only 58% of the actually measured anoxic CH4 emission rate. However, when the profiles were taken separately and analyzed according to the method of intersecting straight lines (Anderson and Nelson 1975), the diffusive CH4 flux that was calculated from the CH4 gradients at 6-8 mm depth was closer to the actually observed CH4 emission rates, i.e., 224 ± 70 nmol cm-2 d-l or 78% of the actually observed CH4 emission rates. For the soil layers between 3 and 5 mm depth the calculated diffusive CH4 flux was only 125 ± 86 nmol cm-2 d-l or 43% of the observed flux, indicating that molecular diffusion was not the sole transport process responsible for the observed CH4 emission. The different results obtained when calculating the flux from small or large parts of the vertical CH4 profile may

1514 Rothfuss and Conrad that this was the critical diameter at which the bubbles collapse due to their surface tension. The bubble layer was further investigated in a large mi-. crocosm that consisted of an aquarium that was filled with soil to a height of 10 cm. In this large soil microcosm we were able to find a sufficient number of gas bubbles with >2-3 mm in extension to be able to take gas samples from individual bubbles. The CH4 mixing ratio in individual bubbles was 44-82% for CH4 and 6.9-13% CO2 (Fig. 7). The bubbles at the upper border of the bubble layer contained less CH4 than those from deeper parts of the soil, indicating that the CH4 mixing ratio in the gas bubbles changed in parallel with the dissolved CH4 concentrations. To check whether the gas bubbles affected the shape of the CH4 concentration gradients in the upper part of the soil we searched for a large gas bubble at the glass wall of the aquarium and measured at this site a set of seven vertical CH4 profiles through this gas bubble (Fig. 8). Isopleths of identical CH4 concentrations were layered around the gas bubble thus forming CH4 gradients with a three-dimensional structure. The observed two-dimensional distribution of CH4 concentrations may in addition have been affected by gas bubbles that were located nearby but were not visible. Discussion Fig. 2. Time-dependent increase of the CH4 mixing ratio in the flux chamber containing a soil microcosm (36 mm diameter). The oxic flux was measured after removing the overlaying water. The anoxic flux was measured after 1 h incubation under an N2 atmosphere. One hour after the measurement of flux and vertical profile of CH, (see Fig. 1) were finished, the oxic flux was measured again. originate from a more complex three-dimensional structure of the CH4 gradient system. This hypothesis was confirmed by measuring the two-dimensional horizontal distribution of CH4 concentrations at a particular depth (Fig. 4). The CH4 concentrations that were measured at 6 mm depth did not vary stochastically but formed a distinct landscape-like structure with isopleths ranging between 240 and 510 µm CH4. We observed at the glass walls of our soil microcosms a layer of gas bubbles (Fig. 5). Assuming that the bubbles were not only present along the glass wall but that the bubble layer extended through the whole soil microcosm, the gasfilled porosity of the soil was determined from the relative percentage of the area of the glass wall that was covered by bubbles (Fig. 6). No bubbles were observed in the upper 10 mm of the soil. Below this depth the number and volume of the bubbles increased and reached a maximum at 14 mm depth. At this depth the air-filled porosity was 0.3 (Fig. 6). This was probably the critical value that just allowed the soil to hold the bubbles back against their buoyancy. Below 14 mm depth the number and volume of the bubbles decreased gradually to zero. The bubbles became increasingly spherical in shape when their diameter decreased. Bubbles smaller than 0.2 mm in diameter were never observed, indicating In anoxic paddy soil methane is produced in deeper anoxic layers and is transported upward by gas bubble movement and diffusion. Oxygen is only present in the upper 3-mm soil layers (Frenzel et al. 1992). Part of the CH4 that enters the oxic surface layers is oxidized by methanotrophic bacteria. Therefore, we observed a concave-shaped vertical concentration profile with low CH4 concentrations in the upper 3-mm soil layer. Similar profiles have been observed before (Rothfuss and Conrad 1994; Rothfuss et al. 1996). The shape of vertical CH4 profiles indicates a very narrow zone of CH4 oxidation at the lower boundary of O2 availability at 2-3 mm depth, because CH4 was often below the detection limit in the upper 2-mm layer. The oxidation of CH4 resulted under oxic conditions in a flux that was <10% of the potential CH4 flux that was observed under anoxic conditions. Similar rates were observed before (Conrad and Rothfuss 1991). If the methanotrophic activity is inhibited by lack of O2, the rate of CH4 emission should be equal to the rate of CH4 transport from the deeper soil layers toward the soil surface (under anoxic conditions) or toward the lower boundary of the oxic surface layer (under oxic conditions). The transport rate can theoretically be calculated from the CH4 diffusion gradient within the soil using Fick s first law. As expected, the CH4 gradients below the oxic surface layer were similar to those that were measured in the absence of an oxic surface layer. Nevertheless, we found that the emission rate of CH4 under anoxic conditions was always higher than the CH4 transport rate within the soil. In the small microcosms (36 mm diameter; Fig. 1) the CH4 gradient reached its maximum between 3 and 4 mm depth, indicating that the CH4 production zone started at 4 mm depth. In the larger microcosms (92 mm diameter; Fig. 3) the CH4 profiles exhibited the steepest slopes at depths between 5 and 10 mm, indicating

Methane in anoxic soil 1515-8- Fig. 3. CH4 concentration profiles at randomly selected sites chosen in a microcosm (92 mm diameter) after 58 d incubation under aerated water at 25 C; for clarity only three representative profiles are shown out of n = 5. that the CH4 production zone was localized below these depths. As an important border condition of Fick s first law (Eq. 1) the concentration gradient has to be taken in the direction where it is steepest (Berner 1980). This direction is the vertical direction, if the isopleths that connect identical concentrations are horizontally laminated (orthogonal to the vector). However, such an orientation was obviously not generally given, because the CH4 gradients in flooded paddy soil exhibited a three-dimensional uneven structure in which the isopleths seemed to follow the surface of the gas bubbles. This is the first report of such a complex CH4 gradient structure and was made possible by the use of gas diffusion probes with a relatively high spatial resolution (Rothfuss and Conrad 1994; Rothfuss et al. 1994, 1996). It is analogous to 0, gradients in microbial mats that sometimes also exhibit a landscape-like shape of horizontal concentration profiles that originate from O2-containing gas bubbles (Joergensen et al. 1983). Gas bubbles are formed when the production of CH4 raises the sum of the partial pressures of all dissolved gases above the hydrostatic pressure in the sediment (Chanton and Whiting 1995). Our observations indicate that a bubble is then not immediately released from the soil into the atmosphere but is gradually moving up within the soil. Otherwise, the X axis [mm] Fig. 4. Isopleths of the CH4 concentrations in a horizontal depth layer (z = 6 mm) of a microcosm (36 mm diameter). The centre of the round microcosm was at x = 8/y = 0 mm. The furthest point away from the center was at x = 18/y = 15 mm, i.e., close to the glass wall, where the soil surface was defined as z = 0 mm. The CH4 concentrations are given in µm. Sampling points are indicated by dots. bubbles would not accumulate in a distinct layer but would be more homogeneously distributed within the bulk soil where CH4 is formed. The gas bubbles in the upper part of the bubble layer had a lower CH4 mixing ratio and thus ap- -10-20 Or-- soil surface Fig. 5. Copy drawing of the extension of the gas bubble layer as observed at the glas wall of a soil microcosm (92 mm diameter) that was incubated for 81 d under aerated water at 25 C. No bubble was observed above the 10-mm depth.

1516 Rothfuss and Conrad 0 25 30 Fig. 6. Depth distribution of the air-filled porosity in a soil microcosm as obtained from the relative areas at the glass wall that were covered by bubbles (see Fig. 5). parently lost CH4 by diffusion into the porewater. This process obviously affected the gradient of dissolved CH4 above the bubble layer resulting in the complex three-dimensional gradient structure that we observed. Diffusive loss of CH4 from rising gas bubbles results in a decrease in the volume of the bubble and thus in its buoyancy. The loss in buoyancy results in the observed picture, i.e., that gas bubbles accumulated at the soil depth where the CH4 concentration gradient reached its plateau. Layers of gas bubbles are believed to sometimes act as a confining layer that traps CH4 bubbles rising from below (Romanowicz et al. 1995). In summary, our results demonstrate that gas bubbles affect the CH4 concentration gradients in the soil and the calculation of transport rates using Fick s first law. Similar problems were encountered by Archer and Devol (1992) when they attempted to calculate O2 uptake rates from the O2 concentration gradients within the diffusive boundary layer over the sediment surface. Oxygen uptake measured by placing a chamber over the sediment surface resulted in up to 3-4 times higher rates than those calculated from O2 gradients. One reason for the higher fluxes is the uneven structure of the diffusive boundary layer that is caused by the surface roughness of the sediment and results in a changing thickness of the diffusive boundary layer (Gundersen and Joergensen 1990). Indeed, if the actual surface structure was taken into account, a 2.5-times higher diffusive O2 flux was obtained (Gundersen and Joergensen 1990). However, in case of CH4 gradients within the submerged soil the problem is even more complicated. In contrast to the O2 gradients that are linear within the diffusive boundary layer, the CH4 gradients within the soil were not. Over short distances where the CH4 gradient system can be considered to be planar, the vector of diffusive flux can be divided into a horizontal and a vertical component. The average vertical component would represent the diffusive flux toward the surface. Fig. 7. Concentration of CO2 and CH4 in individual bubbles sampled by use of a microcapillary within a larger soil microcosm (aquarium). Notice that the depth scale starts at 15 mm. Above this depth the gas bubbles were too small for sampling or no gas bubbles were present. Fig. 8. Isopleths of CH4 concentrations in a vertical section of a soil microcosm (aquarium) 2 mm behind the glass wall at a site where a large gas bubble was observed within the soil. The CH4 concentrations are given in µm (steps of 150 µm). Sampling points are indicated by dots.

Methane in anoxic soil 1517 Hence, our set of randomly selected profiles should give representative data for calculation of the diffusive CH4 transport in soil. When calculating CH4 transport rates over a short depth interval (AZ = 2 mm) at a lower depth (6-8 mm depth) analogous to the method of intersecting straight lines (Anderson and Nelson 1975), the upper range of the thus calculated rates fell within the range of the actually measured CH4 fluxes. However, when the rates were calculated at a more shallow depth (3-5 mm depth), they were too low compared to the actually measured CH4 fluxes. There are several possible explanations: (1) The depth interval of 2 mm was still too long to be considered as being linear. (2) The effective molecular diffusion coefficient of CH4 changed with soil depth: (3) A transport process in addition to molecular diffusion enhanced the CH4 flux. Because CH4 oxidation was inhibited, the average diffusive CH4 transport rate at 6-8 mm depth must be the same at 3-5 mm depth and above. However, the CH4 gradients were such that the diffusive CH4 transport rate seemed to increase with soil depth. This observation can hardly be explained by mistaking curved for linear gradients. It is more likely that the effective molecular diffusion coefficient decreased with depth or that an additional transport process was operating of which the intensity decreased with depth. Because the microcosms were prepared with homogenized soil, such an additional transport process can only be due to bioturbation that established itself during the incubation. The effective molecular diffusion coefficient was determined in an earlier experiment with the same soil, but after inactivation of biological activity with HgCl2 (Rothfuss and Conrad 1994). The CH4 gradient that was observed in this earlier experiment had the ideal shape that was expected from molecular diffusion. The diffusion coefficient determined from these data was nearly exactly that which was expected from theoretical calculations. Compared to this earlier experiment the present experiments differed only in one way, i.e., the soil has not been inactivated. Therefore, bioturbation is a conceivable explanation for the observed discrepancy between calculated and measured CH4 fluxes. Bioturbation is usually decreasing in intensity with depth (Booij et al. 1994). In fact, we could observe some very small (1-2 mm length; <l mm diameter) worms in the oxic part of the soil. Although bioturbation should cause inhomogeneities in the concentration gradients within the upper part of the soil (Revsbech and Joergensen 1985; Frenzel 1990), such inhomogeneities were not observed for the CH4 gradients. Possibly, the bioturbation effects were on a smaller scale than the spatial resolution of our CH4 diffusion probe, i.e., <1 mm. Inhomogeneities in the concentration profiles should also occur, if the enhanced transport was caused by gas bubble tubes (Martens and Klump 1980) or by cracks in the texture at the soil surface. However, gas bubble transport was not observed. Instead bioturbation by small worms may have increased the effective diffusion coefficient and thus the flux of CH4 without causing detectable inhomogeneities in the CH4 concentration gradient. from deeper soil layers up to the surface. Vertical CH4 profiles showed that CH4 was consumed in the oxic surface soil layers at about 2-3 mm depth. Gas bubbles formed a distinct layer at the depth (10-20 mm depth) where the CH4 porewater concentrations reached its plateau. This gas bubble layer affected the CH4 concentration gradient. The curved surface of the bubbles resulted in curved CH4 concentration isopleths. The CH4 gradients had not only a vertical but also a horizontal component resulting in a nonplanar and uneven lamination of identical concentrations. Therefore, the effective diffusive area was larger than in a system of linear gradients with planar lamination of the isopleths. However, accounting for this complication was not sufficient to explain why measured CH4 fluxes were larger than those calculated using Fick s first law. Therefore, we reason that microbioturbation caused by small worms in the soil surface layers provided an additional transport process to the molecular diffusion of CH4. References Conclusions In the upper part of rice field soil microcosms a CH4 gradient was established that resulted in the diffusion of CH4