The apparent transfer velocities (kw) of CH4, N2O, and SF6 were determined for gas invasion and evasion in a closed laboratory exchange tank. Tank water (pure Milli-RO® water or artificial seawater prepared in Milli-RO®) and/or tank air gas compositions were adjusted, with monitoring of subsequent gas transfer by gas chromatography. Derived kw was converted to “apparent k600,” the value for CO2 in freshwater at 20°C. For CH4, analytical constraints precluded estimating apparent k600 based on tank air measurements. In some experiments we added strains of live methanotrophs. In others we added chemically deactivated methanotrophs, non-CH4 oxidizers (Vibrio), or bacterially associated surfactants, as controls. For all individual controls, apparent k600 estimated from CH4, N2O, or SF6 was indistinguishable. However, invasive estimates always exceeded evasive estimates, implying some control of gas invasion by bubbles. Estimates of apparent k600 differed significantly between methanotroph strains, possibly reflecting species-specific surfactant release. For individual strains during gas invasion, apparent k600 estimated from CH4, N2O, or SF6 was indistinguishable, whereas during gas evasion, k600-CH4 was significantly higher than either k600-N2O or k600-SF6, which were identical. Hence evasive k600-CH4/k600-SF6 was always significantly above unity, whereas invasive k600-CH4/k600-SF6 was not significantly different from unity. Similarly, k600-CH4/k600-SF6 for the controls and k600-N2O/k600-SF6 for all experiments did not differ significantly from unity. Our results are consistent with active metabolic control of CH4 exchange by added methanotrophs in the tank microlayer, giving enhancements of ∼12 ± 10% for k600-CH4. Hence reactive trace gas fluxes determined by conventional tracer methods at sea may be in error, prompting a need for detailed study of the role of the sea surface microlayer in gas exchange.
 Accurate quantification of a reactive trace gas flux between water and air requires an accurate estimate of its gas transfer velocity, kw, a kinetic parameter that cannot be directly measured. Most commonly, kw is estimated indirectly for inert volatile tracers by measuring their rates of evasion to air, with subsequent diffusivity based scaling of kw to reactive gases of interest [e.g., Watson et al., 1991; Wanninkhof et al., 1993, 1997; Nightingale et al., 2000]. For gases with predominantly invasive exchange (i.e., air to water) such as CO2, this approach necessarily assumes gas transfer to be fully reversible under identical conditions of turbulence; that is, kw for evasion equals kw for invasion. Importantly, this assumes control of gas transfer by exclusively physical means and precludes significant biogeochemical effects at the air-water interface.
 In this study we measured directly rates of invasive and evasive exchange of biogeochemically reactive CH4 and N2O, and inert SF6 at selected levels of interfacial turbulence in a custom-built closed laboratory gas exchange tank and calculated their apparent kw for both exchange directions. We chose to focus on the potential bacterioneuston control of CH4 exchange, for two reasons: first, because of the importance of CH4 in the global system and the fact that recent work suggests a larger marine source than previously thought. Second, we have ready access to a range of both marine and freshwater methanotroph strains, and a unique culturing expertise for these developed in our laboratories over several years. CH4 currently accounts for ∼12% of enhanced greenhouse forcing [Lashof and Ahuja, 1990], and it regulates tropospheric O3 and OH, and stratospheric H2O [Crutzen, 1991]. Tropospheric CH4 is increasing by ∼0.3% yr−1 [Steele et al., 1992; Dlugokencky et al., 1994a, 1994b], reflecting changes in poorly defined sources and sinks [Houghton et al., 1995]. The marine source of atmospheric CH4 is currently being revised upward, and recent estimates span a wide range, ∼0.4–50 Tg yr−1 [Bange et al., 1994; Houghton et al., 1995; Bates et al., 1996; Kvenvolden et al., 2001], equivalent to ∼1–10% of the global source total. Therefore, if the bacterioneuston plays a significant role in sea-air CH4 exchange, this may require accounting for in global budgets.
 We thus conducted gas exchange experiments in which we augmented the exchange tank water with selected strains of either live or deactivated methanotrophs, and bacterial cultures devoid of methane mono-oxygenase (MMO) activity. In other experiments, we examined the potential control of gas transfer by bacterially associated surfactants. N2O was used as a proxy for reactive trace gases unaffected by methanotrophy, and SF6 provided an index of gas transfer influenced solely by physical transport.
2.1. System Overview
 The experimental and analytical apparatus (Figure 1) comprises two gas chromatographs (Shimadzu GC 14B and Hewlett Packard, HP 5880A) and a gas equilibrator connected to a 100 cm (L) × 50 cm (H) × 50 cm (W) closed gas exchange tank, in a continuous gas-tight circuit. Valves V1-V3, V12, V13, V15, and V16 are two-position, software-driven chromatography valves (Valco, 316-grade stainless steel); V4, V6, V9, V10, and V14 are manually operated three-way switching valves; and V5, V7, V8, and V11 are manual shut-off valves (all Whitey, 316-grade stainless, TFE seating). Tubing between the exchange tank and valves V4–V6 is of 4 mm i.d. flexible nylon and all other system tubing is 2 mm i.d. stainless steel.
 The exchange tank is constructed from 10-mm acrylic and fitted with a 30-cm-diameter removable maintenance hatch, made gas tight by the compression of two Viton® O-rings (Figure 2). Possible contamination from the O-rings through gas adsorption and subsequent “memory effects” was rigorously evaluated and shown to be negligible. By adding water to the tank, a known gas, aqueous phase volume ratio (Va/Vw), can be created by reference to a graduated scale (phase volumes each ∼125 L, overall volume precisions ±40 cm3). The initial gas composition of each phase can be selected and modified by introducing gas mixtures of predetermined compositions. Gas and aqueous phase temperatures are monitored by means of calibrated thermometers. Homogeneity of gas phase composition is achieved with two small high-capacity fans (1.2 L min−1, RS components, UK). Aqueous phase turbulence is generated and controlled by means of a baffle (48 × 12 × 0.6 cm) driven by a variable voltage DC motor, for which the relationship between applied voltage and baffle cycling rate is well established [Frost, 1999]. A 30-L expansion vessel (Tygon® gas sampling bag, Cole Palmer Instruments) connected to an external water reservoir is used to maintain a constant internal tank pressure during the removal of water samples for dissolved gas analysis.
 Large gas leaks at the equilibration pump during routine operation induce changes in the transducer and flowmeter responses accompanied by slow inflation or deflation of the expansion vessel, and are therefore easily detected. Much smaller leaks not readily observable are nevertheless capable of compromising data integrity. Rigorous leak tests were therefore carried out routinely, immediately before and after each gas exchange experiment. These involved system pressurization to 2 atm with a mixture of 500 pptv SF6, 15 ppmv CH4, and 6 ppmv N2O in ultra-high purity (UHP) N2 (Air Products; <10 ppmv total gaseous impurities), with pressure monitoring over 48 hours (Budenberg Gauge Co., UK, model 315GP; measurement precision ±0.3%). Gas partial pressures monitored by gas chromatography at 3-hour intervals were stable within a typical measurement precision of ±1%.
2.3. Gas Exchange Experiments
 Invasive and evasive exchange of CH4, N2O, and SF6 was investigated in experiments using both Milli-RO® water (Millipore Corp; conductivity ∼10 μS, bacterial rejection >99%) and salinity 35 artificial seawater (ASW) [Kester et al., 1967] prepared in Milli-RO®. The use of Milli-RO® and ASW minimized potential modification of the experimental results by indigenous microbes. Experiments were as follows: ASW and Milli-RO® unamended; ASW and Milli-RO® amended with active methanotrophs; ASW and Milli-RO® amended with inactive methanotrophs; ASW and Milli-RO® amended with live bacterial cells (Vibrio) devoid of methane mono-oxygenase (MMO); ASW and Milli-RO® with added surfactant.
2.3.1. Bacterial Additions
 Experiments used one freshwater strain and four marine strains of methanotrophs (Table 1), and one strain of Vibrio. All methanotroph strains were grown in 50-mL batch cultures in sealed flasks (250 mL) containing a head space of 20% (v/v) CH4 in air. Growth was at 30°C on a gyratory shaker for 3–4 days to the late exponential phase of growth. Freshwater strain OB3B Methylosinus trichosporium was grown in the nitrate mineral salts (NMS) medium of Whittenbury et al. . Marine strains 9232 Methylococcus and O4 Methylobacter (donated by K. Nanba, Department of Aquaculture, University of Tokyo) were grown in NMS medium [Whittenbury et al., 1970] supplemented with 20 g L−1 NaCl. Marine strains YO Methylomicrobium and KO Methylomicrobium (donated by H. Fuse, Kure, Japan) were grown in basal seawater medium [Fuse e al., 1998]. Methanotroph enumeration was by standard viable count on agar plates supplied with 20% (v/v) CH4. The specific CH4 oxidation rates of the selected strains were all ∼10−6 mol CH4 d−1 mg−1 dry weight. (1 mg dry weight of methanotrophs is equivalent to about 2 × 1010 cells). Vibrio growth was in Vibrio enrichment broth (Difco); 500 mL of Vibrio enrichment broth was inoculated with 50 mL fresh seawater (salinity 35) and incubated for 24 hours at 30°C. Cell numbers in the resulting dense enrichment culture were estimated from optical density measured at 540 nm (OD540), where OD540 = 1 was equivalent to a cell concentration of 5 × 108 mL−1. All bacterial additions gave final tank water concentrations ∼5 × 108 cells L−1 (i.e., ∼6.25 × 1010 cells in a 125-L tank water volume).
Table 1. Methanotroph Species Used in the Gas Exchange Experiments
Isolated and donated by K. Nanba, Dept. of Aquatic Biosciences, University of Tokyo.
Manufactured from an equal part mixture of strains 9232, O4, YO, and KO.
 The potential for inactive methanotroph cells to modify gas transfer was evaluated in experiments using subsamples of methanotroph strain YO (Table 1), deactivated by the addition of 0.5 M aqueous HgCl2. Possible control of gas transfer by live bacterial cells devoid of methane mono-oxygenase (MMO) was investigated in experiments using freshly collected, 0.2-μm filtered coastal North Sea water (salinity 33) augmented with the Vibrio enrichment culture. These experiments are referred to as “environmental” in subsequent discussion.
 All bacterial additions were 12 hours prior to each experiment to allow for subsequent bacterial acclimatization in the tank water as appropriate. During this period, tank water surface tension was measured at intervals using a capillary tube-travelling microscope method [Nelkon and Parker, 1965]. Method accuracy was estimated at ±3% from comparisons of measured values for clean laboratory water (Milli-RO®) with values for pure water given by Lide . For all live cell additions, surface tension decreased rapidly to about 80–85% of the initial value during the first 2–3 hours of the acclimatization period, consistent with a migration of bacterial components to the tank microlayer.
2.3.3. Surfactant Additions
 The role of bacterially associated surfactants in gas transfer was examined through replicate additions of 2 mg L−1 N-acetyl-D-glucosamine (2-Acetamido-2-deoxy-D-glucose, Sigma-Aldrich, UK), a naturally occurring bacterial surfactant. The additions were designed to yield equivalent DOC concentrations for typical seawater, ∼1.4–3.2 g C m−3 in the microlayer and ∼0.9–1.1 g C m−3 below it [Hunter and Liss, 1981].
2.4. Experimental Protocols
 The exchange tank was rigorously cleaned prior to each experiment with 2.5M HCl followed by 70% ethanol and finally Milli-RO® water, in order to remove residual biological material. The tank water phase was subsequently equilibrated with ambient laboratory air for ∼12 hours using a high delivery rate pump (Millipore, >30 L min−1) and an aerator.
 To initiate an invasive experiment, valves V4, V6, and V12 were switched to connect the three GC sample loops in line with the tank air phase and simultaneously isolate the equilibration system from the circuit (Figure 1). Approximately 100 cm3 of a gas mixture containing 1% v/v CH4, 1% v/v N2O, and 2.14 ppmv SF6 in UHP N2 was injected via V5. The equilibration pump was then switched on for ∼30 s to drive the gas mixture into the tank headspace. Operation of the air phase stirrers for 5 min distributed the gas mixture evenly around the system and established initial air phase partial pressures of ∼8 μatm for CH4 and N2O, and ∼0.6 natm for SF6. For evasive experiments, a 1-L aliquot of tank water was removed, equilibrated with the gas mixture and then reintroduced to the tank using a low flow rate peristaltic pump (Watson-Marlow Ltd, UK), with gentle mixing of the tank water using the baffle set to its minimum rate of operation. With this procedure and allowing for gas solubilities, initial partial pressures in the water phase were ∼18 μatm CH4, ∼7 μatm N2O, and ∼10 natm SF6.
 Experiments were run at selected levels of water turbulence by modifying the applied baffle voltage. Periodic bubble formation was observed during some experiments. Because bubbles can lead to physical enhancements of gas invasion relative to evasion [e.g., Anderson and Johnson, 1992; Wallace and Wirrick, 1992], this possibility was minimized by limiting the maximum applied voltage to achieve a compromise between sufficient turbulence and minimal bubbling. Air and water phase samples were removed at intervals for analysis. During the experiments the system was arranged to run routinely with the tank air phase continually circulating through each of the three GC sample loops via the circuit pump. For air phase analysis, V3 and V13 were switched to equilibrate the sample loops to laboratory pressure for 3 s, prior to injection of their contents into the GC carrier gas flows via V1 and V15. For water sample analysis, a 1-L volumetric flask was gravity filled from the tank center (Figure 2) via a degassed, 4 mm i.d. silicon sampling line. Flask contents were stabilized against microbial reactions by the addition of 200 μL of 0.5-M aqueous HgCl2 and the flask closed with a secure stopper. The procedure maintains sample integrity with respect to CH4 and N2O for several days [Upstill-Goddard et al., 1996]. Filled sample flasks were equilibrated to constant temperature Teq (25 ± 0.1°C) in a water bath (not shown) prior to analysis at the end of each experiment. The maximum sample storage time was 6 hours.
 For water sample analysis, the tank was isolated from the remainder of the circuit via V4 and V6 and the equilibrator manifold was opened to the circuit via V12. All lines and sampling loops were flushed with an equilibrator gas (compressed air with predetermined CH4, N2O, and SF6 partial pressures close to ambient) prior to each analysis, in order to remove any traces of previous sample. The flask stopper was replaced with the equilibration manifold and displacing sample with more equilibrator gas generated a fixed volume headspace. This was equilibrated with the remaining water sample by recirculating it rapidly around a continuous closed circuit and through an aerator for 15 min, as in the procedure of Upstill-Goddard et al. .
2.5. Gas Chromatography
 Analysis of equilibrated flask headspace and tank air phase gases was by isothermal gas chromatography at 60°C, based on methods described previously [Upstill-Goddard et al., 1990, 1996]. Chromatographic separation of CH4 and N2O (Shimadzu GC 14B; V1 in Figure 1) was on 80–100 mesh Porapak-Q (CH4: 2 m × 2 mm i.d; N2O: 5 m × 2 mm i.d), whereas SF6 (HP 5880A; V15 and V16 in Figure 1) was separated on 80–100 mesh, 5A molecular sieve (5 m × 2 mm i.d.). Carrier gas was UHP N2, pre-dried on a molecular sieve/activated charcoal filter and passed through an O2 trap (Supelco). Flow rates were 25 cm3 min−1 for CH4 and N2O, and 50 cm3 min−1 for SF6. Water vapor produced during sample equilibration was removed using Mg(ClO4)2, and CO2 which can interfere with the ECD response was removed using NaOH [Upstill-Goddard et al., 1996]. Detection of CH4 was by flame ionization (FID) at 120°C. Detection of N2O and SF6 was by Electron Capture (ECD), both using a 3.7 × 108 Bq 63Ni source at 320°C, in constant current mode (1.0nA). SF6 analysis incorporated column backflushing (V15 and V16; Figure 1) to preclude O2 elution and increase sample throughput [Upstill-Goddard et al., 1990]. Typical retention times were ∼6 min for N2O, ∼0.6 min for CH4, and ∼1.5 min for SF6. The method was calibrated with mixed calibration standards prepared by pressure dilution (certified accuracies ±1.5% CH4, ±2% N2O, and ±1% SF6 [Upstill-Goddard et al., 1990, 1996]). The SF6 standards were independently calibrated against external standards prepared gravimetrically [Law et al., 1994]. Analytical precisions were typically ±1% CH4, ±0.8% N2O, and ±1% SF6. Detection limits were ∼2 ppbv N2O, ∼10 ppbv CH4, and ∼0.2 pptv SF6 [Upstill-Goddard et al., 1990, 1996].
2.6. Equilibration Theory
 For the water samples, volumetric gas mixing ratios measured in the equilibrated flask headspace, Ae, require correcting to account for solubility dependent phase partitioning. The initial gas partial pressure in the aqueous phase, pp, is given by
where Ai is the initial gas mixing ratio in the equilibrator gas, Vfa and Vfw are, respectively, the flask headspace and flask water sample volumes, βinsitu and βteq are, respectively, the Ostwald solubilities (v/v atm−1) at the tank water temperature and at the equilibration temperature, and P is the tank internal pressure (mbar) [Upstill-Goddard et al., 1996].
2.7. Mass Balance Theory
 The endpoint of an exchange experiment for a given gas occurs when its partial pressure reaches a common value in the tank air and water phases. For an inert gas such as SF6 this equilibrium value can also be calculated from a simple mass balance, and the observed and calculated equilibrium partial pressures should be identical within the associated errors. In contrast, for a biogeochemically reactive gas, active consumption or production during the experiments would cause a mismatch between the observed and calculated values. The underlying rationale for our experiments is that any such mismatch is reported as a corresponding mismatch between the estimated “apparent” kw for the inert and reactive gas, as determined through applying the following theory.
 The number of moles, a, of a gas in the tank air phase prior to an experiment is given by
Similarly, the corresponding number of moles, b, in the tank water phase is
where Va and Vw are, respectively, the tank air-phase and water-phase volumes and VmTP is the molar gas volume at in situ temperature and pressure. Hence the total number of moles in the system, yi, is
The molar gas ratio of the two phases at equilibrium, q, is fixed for all equilibrium values of pp (constant β(T) and Va/Vw), irrespective of the total gas content of the system,
 For an invasive experiment, a known number of moles, MI, of an individual gas is added to the tank headspace, given by
where MR is the individual gas mixing ratio in the total volume of added gas mixture, VI. The resulting number of moles in the system, yf, is then
At equilibrium yf partitions between the two phases according to q. Hence the final number of moles in the air phase, af, is
and the corresponding number of moles in the water phase, bf, is
The equilibrium gas partial pressure in both the air and water phases, Pge, is then
 For an evasive experiment, a known number of moles, MIW, of an individual gas is added to the tank water phase, given by
where Vo is the volume of gas-equilibrated water added to the tank.
 The value of Vw decreases stepwise during an experiment due to water sample removal, but Va is kept constant by action of the expansion vessel. Routine corrections were applied to the data to take account of corresponding stepwise changes in yf and q.
2.8. Gas Transfer Equation
 For gas transfer in an “open” system, such as a lake, the surface ocean, or an open wind/wave tank, water phase gas concentrations change continuously with time, whereas gas concentrations in the air phase are considered time invariant. For such situations the interfacial gas flux, F, is described by
where kw is the water side gas transfer velocity, and Cw and Ca are respectively, the gas concentrations in water and air [e.g., Liss and Merlivat, 1986; Wanninkhof, 1992]. Conversely, for gas exchange involving a small volume of enclosed air, such as in a free-floating gas exchange tank [Conrad and Seiler, 1988], Cw may be considered constant as compared to relatively large changes in Ca with time. For this situation,
where Cw/β is here the equilibrium volumetric mixing ratio at which the gas dissolved in the water is in equilibrium with the gas phase [Conrad and Seiler, 1988; Frost, 1999]. In the closed system used here, the quasi steady state equations (12) and (13) are invalidated by the simultaneous temporal changes in Ca and Cw, requiring an alternative expression for gas transfer. Considering the analogy with gas exchange in a constrained water body such as a lake, the gas flux from water to air is [Frost and Upstill-Goddard, 2002]
where MT is the total mass of gas in the water, AT is the area of the air-water interface, and t is time. Extending the analogy, during the experiments the tank water is always well mixed, so that [Frost and Upstill-Goddard, 2002]
where hT is the tank water depth. Following gas addition immediately prior to the start an exchange experiment,
where Cw0 and Cwt are, respectively, the gas concentrations in the tank water at the beginning and end of time interval Δt.
 During an exchange experiment at constant temperature and turbulence, kw derived from equation (17) has the same theoretical value for any selected Δt. In practice, however, the individual kw estimates vary due to errors in the measurement of Cwt (or Cw0). In order to overcome this problem and obtain a single, “best fit” value of kw for each experiment, Cwt can be substituted by CwtDER, the “derived” value of Cwt obtained by rearranging equation (17),
Applying equation (18) to each experimental Δt interval and manipulating the corresponding kw derived with equation (17) generates a time curve of CwtDER optimized to best fit the experimental values of Cwt. The single values of kw for CH4, N2O, and SF6 corresponding to their respective best fit lines were taken as representative values for each particular experiment.
 For evasive experiments, kw was also estimated from temporal changes in Ca, but for invasive experiments this was only possible for N2O because of solubility constraints. The low solubilities of CH4 (β = 0.032) and SF6 (β = 0.0053) as compared to N2O (β = 0.55) (all values at 25°C [Wanninkhof, 1992]) forced changes in Ca for CH4 and SF6 of similar order to the analytical precision, precluding their use.
where ka = kwβ [Frost and Upstill-Goddard, 1999] and Ca0 and Cat are, respectively, the tank air phase gas concentrations at the beginning and end of time interval Δt. Hence “derived” Cat is obtained from an expression analogous to equation (18),
Figure 3 shows typical partial pressure-versus-time curves for CH4, N2O, and SF6 in an UHP N2 headspace overlying Milli-RO® tank water previously equilibrated with ambient air and a calibration standard containing 60.6 pptv SF6. Use of the calibration standard was necessary due to the low solubility of SF6, which precludes its determination in the UHP N2 headspace at ambient water concentrations due to detection limit considerations.
3.1. Experimental Precision
 The inherent precision of the full experimental procedure was established from replicate invasive and evasive experiments using Milli-RO® water, with the tank baffle set to 0.9 Hz to induce water phase turbulence with minimum bubble formation. Running the replicates at intervals throughout the experimental program established system performance over time. Table 2 lists kw estimates for CH4, N2O, and SF6 determined with equations (17) (invasive data) and (20) (invasive and evasive data) and converted to k600, the corresponding value for CO2 in freshwater at 20°C, using
where Sc is the kinematic viscosity of water divided by the molecular diffusivity of the gas in water at a given temperature. Coherence of the k600 estimates (Table 2) is unaffected by the value of n used. For this purpose we therefore used n = −0.5 and Sc values derived using the Sc- temperature dependencies for experimental data given by Wanninkhof .
Table 2. Estimates of Invasive and Evasive k600 for Milli-RO® Water Derived From Measurements of CH4, N2O, and SF6 Exchange at a Baffle Frequency of 0.9 Hz
k600 - CH4
k600 - N2O
k600 - SF6
 There was no significant difference between k600 estimated from CH4, N2O or SF6, or between k600 determined evasively or invasively (analysis of variance, P > 0.5). Moreover, for the evasive experiments, k600 derived from either temporal changes in Cw (equation (17)) or Ca (equation (20)) were not discernibly different (analysis of variance, P > 0.5). The high internal consistency of these data demonstrates the robustness of our procedures and precludes the possibility of procedural artifacts affecting subsequent data interpretation.
where kw1 and kw2 are the transfer velocities of any two gases at their stated temperatures. Values of Sc were obtained from Wanninkhof . We restricted these estimates of n to experiments using MilliRo® water only, in order to preclude effects arising from possible chemical or biological modifications of gas exchange. Results are summarized in Table 3, where they have also been used to generate values of U10, the equivalent 10-m wind speed based on the empirical gas exchange relationships of Liss and Merlivat  and Wanninkhof . Comparison with previous n estimates for which equivalent wind speeds are available enables us to set our results into context by crudely scaling tank turbulence levels with gas exchange conditions in a field situation. Our estimate of n at the highest baffle frequency, ∼−0.2 (1.25 Hz), is close to the wave tank estimates of Wanninkhof and Bliven  for wind speeds >10 m s−1. Values around −0.5 to −0.6 are consistent with wind speeds >3 m s−1 observed in field studies [e.g., Watson et al., 1991]. Values lower than about −0.6 correspond to wind speeds <3 m s−1, consistent with the predictions of empirical models [Liss and Merlivat, 1986; Wanninkhof, 1992]. In contrast, our estimate of n at 0.91 Hz seems somewhat anomalous and is difficult to explain given the relative consistency of our other estimates. While it is conceivable that this anomaly is a consequence of additional turbulence due to tank wall or baffle effects, it is not possible to speculate further in the absence of more detailed measurements.
Table 3. Estimates of Cw-Derived k600 and the Schmidt Number Exponent, n, Based on Measurements of CH4, N2O, and SF6 Exchange in Milli-RO® Watera
All values are means of two invasive and two evasive experiments.
64 ± 6
66 ± 5
66 ± 4
−0.2 ± 0.1
18 ± 2
19 ± 1
19 ± 1
−1 ± 0.3
8.9 ± 0.7
9 ± 0.6
8.8 ± 0.7
−0.6 ± 0.1
2.4 ± 0.2
2.1 ± 0.1
2.4 ± 0.2
−0.4 ± 0.2
0.6 ± 0.08
0.6 ± 0.09
0.7 ± 0.08
−0.7 ± 0.1
3.3. Gas Exchange Experiments
Figure 4 shows estimates of “apparent k600” deriving from the experiments with ASW and added methanotrophs. As a result of baffle motor breakdown and its replacement prior to the two “community” experiments, these ran at a lower baffle frequency (0.3 Hz) than all of the others (0.45 Hz). Several important observations can be drawn from the data. First, for the individual strains in many cases, apparent k600 estimated invasively exceeds the value estimated evasively. Second, for both gas invasion and evasion in general, there are significant differences between the apparent k600 estimates derived with the various methanotroph strains. For example, for gas invasion the apparent k600 estimates for methanotroph strains 9232 and O4 exceed those for strains KO, YO, and the “community” by a factor ∼2 (Figure 4). Third, for all individual invasive experiments and the “community” evasive experiment, apparent k600 estimates deriving from CH4, N2O, and SF6 are identical within experimental error. Fourth, for each of the pure methanotroph strains during gas evasion, apparent k600 estimates derived from N2O and SF6 were statistically identical, whereas apparent k600 derived from CH4 was always significantly higher. The large inter-experiment contrasts in apparent k600 must be directly related to differences in an intrinsic bacterial property because no other experimental condition was varied (excepting the reduced baffle frequency for the “community” experiments noted earlier). In this context it is notable that among the pure strains, KO and YO, which are different strains of the same organism (Methylomicrobium), show a rather close agreement in apparent k600. For N2O, although modification of its apparent k600 through microbial production or consumption is entirely feasible, such effects are unlikely with methanotrophs. Moreover, the identical behavior that was observed for biologically inert SF6 argues strongly against such an explanation. An alternative possibility relates to microbial surfactant release. Physical suppression of kw by surfactants due to their hydrodynamic damping of turbulent eddies is well known, the degree of suppression being a function of surfactant type and concentration [e.g., Frew et al., 1990; Frew, 1997]. It is therefore entirely possible that the observed gross differences in apparent k600 between the various experiments (Figure 4) are due to differences in the types and amounts of surfactant released by the various methanotroph strains. However, further evaluating this possibility would require additional specialized measurements that are beyond the scope of this study. Notwithstanding the specific mechanisms that may underlay the gross inter-experiment differences in apparent k600, the consistently elevated values for apparent k600 derived from CH4 (hereinafter referred to as k600-CH4) for the pure methanotroph strains during gas evasion is the most important observation. This shows clearly the potential importance of the bacterioneuston in modifying reactive trace gas exchange. Important questions, however, are why such enhancements of k600-CH4 were not also seen for the “community” experiment during gas evasion or for any of the invasive experiments (Figure 4). It is conceivable that the lack of an enhanced k600-CH4 in the “community” experiment reflects microbial competition within the mixed population; however, it is inappropriate to speculate further here in the absence of relevant supporting data. The lack of an enhanced k600-CH4 during gas invasion is addressed in detail later (section 4) in the light of additional data presented below.
 The results of the bacteria specific control experiments are shown in Figure 5. As with the experiments with added methanotrophs (Figure 4), failure of a second baffle motor and its replacement prior to the “environmental” experiments (live cells devoid of MMO) at least in part accounts for the elevated values of apparent k600. For all six individual experiments (three invasive and three evasive), apparent k600 values deriving from CH4, N2O, and SF6 are essentially identical within the experimental precision. In other words the relative enhancements of k600-CH4 found for gas evasion with added methanotrophs (Figure 4) were not observed for any of the bacteria-specific controls. However, for the three bacterial controls individually, apparent k600 estimated invasively exceeded apparent k600 estimated evasively for all three gases (Figure 5), and similar enhancements were noted for some of the experiments with added methanotrophs (Figure 4). The fact that the observed invasive enhancements apply equally to inert SF6 as to CH4 and N2O demonstrates a wholly physical control of this phenomenon. Enhanced rates of gas invasion relative to gas evasion have been observed as a result of bubble formation in several gas exchange studies, and this effect is proportionately greater for low solubility gases [e.g., Anderson and Johnson, 1992; Wallace and Wirick, 1992]. Although in our experiments we tried to minimize this possibility by optimizing our selected baffle speeds, because some periodic bubbling was nevertheless observed, some manifestation of this phenomenon in our data seems likely. Even though the gases used in this work span 2 orders of magnitude in solubility [Wanninkhof, 1992], the close agreement between the estimates of apparent k600 deriving from the three gases individually strongly implies that any solubility related effect was negligible (Figure 5). Notwithstanding the details of any such modification of apparent k600 by physical means, what remains clear from these data (Figure 5) is that in contrast to the pure methanotroph strains (Figure 4), none of these experiments shows any significant enhancement of k600-CH4 relative to k600-N2O or k600-SF6 during gas evasion. Hence the two complimentary sets of experiments (Figures 4 and 5) are entirely consistent with the notion of an active metabolic control of k600-CH4 by added methanotrophs.
 In order to quantify the amount by which k600-CH4 for gas evasion was enhanced by methanotrophy, we divided these estimates by the corresponding estimates of k600-SF6 (Figure 6). Consistent with the above results, for a situation where CH4 behaves as an inert gas during water-air transfer, k600-CH4/k600-SF6 would equal unity. For the six bacteria-specific controls (Surfactant, Environmental, and Deactivated, Figure 6), each ratio derives from seven individual estimates for successive, equal Δt increments. Analysis of variance (ANOVA) showed no significant differences among the results from these six controls (P < 0.05). The mean of these data was then compared with the k600-CH4/k600-SF6 ratios derived from each of the 10 bacterial enrichment experiments using a modified t-test (Bonferroni method). This used the pooled estimate of variance from all groups (i.e., the residual variances from ANOVA analyses), with P value adjustment by multiplying by the number of paired comparisons (five evasive and five invasive). This is a conservative approach that is biased toward a nonsignificant outcome, but it is especially suited to the small number of comparisons involved [Altman, 1996]. Importantly, the analysis (P < 0.05) showed the results from all of the evasive bacterial enrichments to be significantly different from those of the bacterial controls (Figure 6), whereas for the invasive experiments, only the results from the KO experiment differed significantly from the controls (Figure 6). By contrast, values of k600-N2O/k600-SF6 for all 16 experiments (data not shown) did not differ significantly from unity.
 If active methanotrophy within the tank microlayer is postulated during the bacterial enrichment experiments, the true CH4 concentration gradient at the air-water interface would exceed that derived from routine sampling of water below the microlayer. Therefore k600-CH4 would be expected to exceed k600-SF6. Although the results of the bacterially enriched evasion experiments are consistent with this reasoning in so far as k600-CH4 > k600-SF6, the corresponding results for invasive exchange are not (Figure 6). However, the apparent paradox can be resolved through consideration of the experimental measurement constraints. For k600-CH4 estimated from Cw, the additional CH4 flux due to microbial consumption would not be observable in the bulk water due to its localized consumption in the microlayer. Hence an enhancement of k600-CH4 would not be detected invasively based on the measurement of Cw. The situation is analogous to the chemical enhancement of CO2, where its consumption at the sea-air interface leads to an enhanced sea-air transfer that would not be detected in CO2 measured in the underlying water [e.g., Wanninkhof and Knox, 1996]. Unfortunately, because of the solubility driven measurement constraints referred to earlier, invasive estimates of k600-CH4 based on Ca are precluded. Hence enhancements of k600-CH4 cannot be demonstrated for invasive CH4 transfer based on the measurement of either Cw or Ca.
 Our estimates of k600-CH4/k600-SF6 (Figure 6) are ∼1.12 ± 0.10 (range 1.04–1.27). It is instructive to examine the rates of microbial CH4 oxidation in the exchange tank that would be required to support these k600-CH4 values. Table 4 gives estimates of the total amount of CH4 that would be required to be removed in the tank microlayer in order to account for the observed k600-CH4/k600-SF6 values. These estimates were derived from the numerical difference between k600-CH4 and k600-SF6, which represents the fraction of k600-CH4 arising from microbial consumption, by substituting this value into equation (18) to derive the corresponding net change in CH4 concentration in the tank water phase. Assuming this amount of CH4 is removed exclusively within a 0.5-mm-deep microlayer (i.e., microlayer volume ∼0.25 L) yields the corresponding removal rates per unit mass of microlayer also listed. The required rates (∼10−8–10−6 mol CH4 kg−1 d−1) are on average about 2 orders of magnitude higher than typical rates of aqueous CH4 oxidation (Table 4). Nevertheless, assuming that the added methanotrophs partitioned between the tank microlayer and the underlying water in a ratio similar to that found in nature [e.g., Sieburth, 1971; Kjelleberg and Hakansson, 1977; Kjelleberg et al., 1979; Hardy, 1982; Hardy and Apts, 1984], the required rates seem to be realistic.
Table 4. Estimates of CH4 Removal in the Tank Microlayer Required to Account for the Observed Values of k600-CH4/k600-SF6, and Some Estimates of Marine CH4 Oxidation Rates
 In preliminary deployments of a free-floating flux box, Frost  determined k600-CH4 and k600-SF6 in the coastal North Sea. Such experiments are only feasible at relatively low wind speeds; however, it is precisely under such conditions that the sea surface microlayer is expected to be most well developed [GESAMP, 1995]. Frost  derived four estimates of k600-CH4/k600-SF6 from measurements of gas evasion, in the range 1.03–1.08 (mean 1.06). This preliminary work thus supports our laboratory findings and confirms the applicability of microlayer CH4 consumption to the natural environment. Using a similar seagoing flux box, Conrad and Seiler  observed an enhancement of invasive over evasive CH4 exchange, by following temporal changes in the CH4 partial pressure of the headspace. Although analytical constraints prevented us from identifying any enhancement of invasive k600-CH4 relative to invasive k600-SF6, our findings are nevertheless analogous to the field results of Conrad and Seiler  and Frost . All three studies strongly imply active CH4 consumption within the microlayer, the only plausible gas sink in the experiments. Such a consumption process would act to increase invasive CH4 exchange over that due solely to diffusion. In the case of evasive exchange, however, although the CH4 flux at the base of the microlayer would be similarly enhanced, Conrad and Seiler  could not have observed this from measurements in the air p hase because of its subsequent consumption in the microlayer.
 Current estimates of the open ocean CH4 flux to the atmosphere, 0.4–3.6 Tg yr−1 [Lambert and Schmidt, 1993; Bates et al., 1996] represent only a small fraction of the total global CH4 source [Houghton et al., 1995]. However, recent work that takes account of the potential emissions from estuaries and coastal waters, including shallow CH4 seeps on continental shelves, suggests that the marine CH4 source could even be as high as ∼30–50 Tg yr−1 [Bange et al., 1994; Upstill-Goddard et al., 2000; Kvenvolden et al., 2001]. On the basis of this upward revision and our laboratory estimates of k600-CH4/k600-SF6 (Figure 6), consumption by the bacterioneuston could conceivably account for ∼6 Tg CH4 yr−1 globally. Importantly, coastal regions are characterized by high productivity and hence high bacterial numbers; hence the associated bacterioneuston may be especially well developed. Combined with high ambient levels of dissolved trace gases, this could make the coastal zone an especially important region for trace gas processing by the bacterioneuston. Even so, scaling of the preliminary results of Frost  for the coastal North Sea using the CH4 source estimates above gives a comparatively modest estimate for CH4 uptake by the bacterioneuston, ∼3 Tg yr−1. To refine such estimates as these requires additional data for a range of marine provinces to be compiled.
 Future warming of the surface oceans could increase the metabolic activity of the bacterioneuston, leading to an increased turnover of trace gases. Conversely, future increases in incident UV-B radiation could result in bacterial cell damage and an overall decrease in microbial activity at the sea-surface [Zepp et al., 1998]. Whatever is the net effect of future global change, there are potentially important consequences for a range of biogeochemically reactive compounds. In the case of climatically active trace gases, reliable predictions of their future atmospheric growth require such possibilities as these to be effectively modeled, so that kw estimates deriving from inert tracer evasion can be effectively “scaled up.”
 To advance further our understanding of reactive trace gas cycling by the sea-surface microlayer, further refinements to the techniques described here are required. Such refinements should incorporate detailed studies of the distributions and activities of relevant microorganisms, both in the microlayer and in the underlying water. To our knowledge, there have been no published studies identifying the distribution and diversity of bacteria in the sea-surface microlayer. However, recent evidence from our own laboratories using gene probes specific for 16S rRNA and methane mono-oxygenase genes suggest that at least in the North Sea, marine methanotrophs such as those used in our laboratory tank experiments are indeed present not only in subsurface waters, but also in the microlayer. Moreover, it seems clear that these microlayer bacterial communities are quite distinct from those present even in waters less than a meter or so below the surface [Franklin, 2001]. Although it currently remains extremely difficult to accurately enumerate bacteria in both the bacterioneuston and the underlying seawater, advances in microbial molecular ecology techniques offer great potential in this direction and ongoing work in our laboratories is advancing toward this goal.
 We thank our many colleagues at both Newcastle and Warwick for their most valuable help and support. K. Nanba, Department of Aquatic Biosciences, University of Tokyo, kindly donated samples of marine methanotroph strains 9232 and O4, and H. Fuse, Kure, Japan, similarly donated samples of marine methanotroph strains YO and KO. This research was supported by the UK Natural Environment Research Council via grant award GR3/11144 to JCM, RCU-G and NJPO, and by the Newcastle University Research Committee, via the award of a Ph.D. studentship to T. F.