Higher Apparent Gas Transfer Velocities for CO2 Compared to CH4 in Small Lakes

Large greenhouse gas emissions occur via the release of carbon dioxide (CO2) and methane (CH4) from the surface layer of lakes. Such emissions are modeled from the air–water gas concentration gradient and the gas transfer velocity (k). The links between k and the physical properties of the gas and water have led to the development of methods to convert k between gases through Schmidt number normalization. However, recent observations have found that such normalization of apparent k estimates from field measurements can yield different results for CH4 and CO2. We estimated k for CO2 and CH4 from measurements of concentration gradients and fluxes in four contrasting lakes and found consistently higher (on an average 1.7 times) normalized apparent k values for CO2 than CH4. From these results, we infer that several gas-specific factors, including chemical and biological processes within the water surface microlayer, can influence apparent k estimates. We highlight the importance of accurately measuring relevant air–water gas concentration gradients and considering gas-specific processes when estimating k.


INTRODUCTION
Lakes cover less than 2% of the terrestrial surface area 1 and are estimated to emit carbon dioxide (CO 2 ) and methane (CH 4 ) in notable amounts with respect to the continental greenhouse gas exchange. 2,3 Transport of dissolved gases across the water surface to the atmosphere constitutes the main pathway for lake emission of CO 2 and is also a major flux pathway for CH 4 . The exchange of dissolved gases (Flux (F); mol m −2 d −1 ; units provided here and below are examples) between air and water depends on the concentration gradient expressed as the difference between the surface water gas concentration (C w ; mol m −3 ) and the gas concentration in equilibrium with the air (C air ; mol m −3 ), and the gas transfer velocity (k; m d −1 ), according to eq 1, Flux measurements by floating chambers (FCs) or eddy covariance have been used together with measurements of C w and C air to calculate k from eq 1, yielding local apparent k estimates. 4−7 There has also been several attempts to develop general k models, predicting k from external drivers. 8−11 A common assumption is that k can be converted between gases of interest if the key thermodynamic properties of the dissolved gases (i.e., the ratio of the kinematic viscosity and mass diffusivity, expressed through the "Schmidt number") are considered. 10 Usually, k-values are normalized to a Schmidt number of 600 (k 600 ), corresponding to k for CO 2 at 20°C. This Schmidt normalization procedure was derived for ideal well-defined k values considered predictable from fundamental physical principles. However, the apparent k used for estimating gas fluxes in nature is influenced by many factors, 12 such as turbulence within the water column, 9,11 physical mechanisms, 13 biological and chemical mechanisms, 14,15 or methodological uncertainties, 16,17 that all determine how welldefined and accurate the apparent k becomes.
Recently, mismatches between apparent k 600 for CO 2 and CH 4 in lakes and other aquatic systems have been observed. 4,5,7,13,18−20 In some cases, higher k 600 were found for CH 4 than for CO 2 , attributed to CH 4 microbubble formation and transport. 13,18 This explanation is linked to the assumption that gases with low solubility form or enter microbubbles that move faster across the water surface boundary layer than dissolved gases, but its importance for CH 4 depends on circumstantial indications. In one other case, higher k 600 for CO 2 than for CH 4 were observed, which could be attributed to hydrodynamic and processes in the surface boundary layer that favors production and transport of CO 2, 7 but the importance and general validity of this explanation for the k 600 difference between CO 2 and CH 4 is also unclear. Further, for both these types of observations, there may be other, not yet widely discussed, potential explanations (details in the Discussion).
The incompatibility between apparent k 600 for CO 2 and CH 4 generates questions regarding the assumption that apparent k determined for one gas can be easily converted to k for another gas via physical properties. If apparent k estimates are not directly comparable across gases, the general use of models to determine k beyond the physicochemical domains and the specific gas(es) represented by the measurements behind the model can be questioned. All of this could undermine the global estimates of gas emissions which are dependent on such models, and the interpretation of apparent k and comparability of apparent k 600 among gases are challenged by the above discrepancies. In this study, we investigate differences in apparent k 600 for CO 2 and CH 4 in four lakes, with different nutrient and dissolved organic carbon concentrations and discuss possible mechanisms for gas-specific variations in apparent k 600 . We hypothesized contrasting k 600 for CO 2 and CH 4 and that the k 600 ratio between CO 2 and CH 4 would differ depending on lake characteristics (trophic status).  23 The FCs consisted of plastic buckets with an opening of 0.062−0.075 m 2 facing the water surface and volumes of 5.4− 8.6 L, depending on the model used. FCs were covered with reflective aluminum tape to minimize internal heating and had floats attached to the outer edges such that the edges penetrated 2−3 cm below the water surface. In previous studies, this FC design yielded flux estimates similar to other methods that do not interfere with the water surface. 24,25 In OBJ, 5−6 FCs were deployed simultaneously for 15−30 min. FC deployments were made 1−3 times per day over 4 days, usually between 11:00 and 14:00, and one day between 20:45 and 21:00, yielding 37 individual FC deployments. Sampling of 30 mL gas from inside the chamber was made every 10−30 min during 30-minute-long deployments by syringe sampling via a 50 cm long polyurethane tube (inner and outer diameter of 3 and 5 mm, respectively) at the top of each chamber. This resulted in 2−4 samples for each FC deployment (13 deployments with two samples, 11 deployments with three samples, and 8 deployments with four samples). Samples were analyzed within 24 h on a greenhouse gas analyzer (LGR DLT 100, Los Gatos Research Inc. USA) equipped with a custom-made syringe injection system. Regressions between FC headspace concentrations and time generated the rate of gas accumulation in each FC (ppm d −1 ). Data were discarded when the regression R 2 for gas accumulation of either CO 2 or CH 4 was below 0.9 or if the data indicated leakage or sample handling errors (5 out of 37 measurements were discarded in total). After processing the data, 32 pairs of CO 2 and CH 4 accumulation rates were obtained.

MATERIALS AND
In SOR, BOL, and SOD, one FC was deployed for repeated continuous measurements of gas flux for 10−16 min. Measurements were made by connecting the top of the FC to two polyurethane tubes, which were inserted to the inlet and outlet of an ultra-portable greenhouse gas analyzer (UGGA; Los Gatos Research Inc. USA). The FC connected to the UGGA was left floating freely on the water surface, see Figure  S1. Between each measurement, the FC was lifted for ventilating it, allowing concentrations of CO 2 and CH 4 to reach background atmospheric concentrations. This provided data for a total of 16 individual FC deployments: seven in SOR, four in BOL, and five in SOD (no measurements were discarded). At least the first 60 s of gas accumulation was removed to allow time for mixing between the FC headspace  21 and MacIntyre et al. 11 Water chemistry for SOR, BOL, and SOD was sampled near the surface (0.1−0.5 m depth) during the ice-free season in 2020. DOC, TN, and TP denote dissolved organic carbon, total nitrogen, and total phosphorous, respectively. Trophic status is based on calculations from Carlson, 22 using TP. and the UGGA measurement cell. Ten-second (0.1 Hz) data were used in regressions for gas accumulation rates (ppm d −1 ) for a time interval set as the minimum time needed for a gas concentration increase rate with an R 2 > 0.7 to be established (between 40 and 100 s of data; R 2 threshold slightly lower than for OBJ above to account for the random noise in gas spectrometry). Gas fluxes were calculated from the gas accumulation rates in the chambers and the ideal gas law according to Rudberg et al. 26 (See Supporting Information, Text S1 for detailed information.) 2.2.2. Dissolved CO 2 Concentration. In SOR, BOL, and SOD, C wCO2 was sampled through a headspace extraction technique (e.g., Cole et al. 27 ). Near-surface water was collected with a Ruttner sampler deployed horizontally at ∼0.1 m. The outlet tube of the Ruttner sampler was inserted at the bottom of a 1.2 L plastic bottle and water was transferred, overflowing the bottle with at least two times the bottle volume. Then, the bottle was capped with a rubber stopper pierced by a long and a short polyurethane tube (inner and outer diameters of 3 and 5 mm), reaching the bottom of the bottle and the end of the stopper, respectively, each connected to closed 3-way luer-lock valves at the outer end. After connecting a 60 mL syringe (Becton−Dickinson) filled with atmospheric air to the short tube and connecting an empty 60 mL syringe to the long tube, air was pushed into the bottle via the short tube, while the added pressure pushed out a similar amount of water through the long tube, filling the empty syringe. The bottle was shaken for two minutes to equilibrate gases between the headspace and the water. The equilibrated headspace air was then extracted by the inverse procedure and transferred to evacuated vials. Separate air samples were collected to correct for initial CO 2 in the headspace.
For 21 of the flux measurements in OBJ, surface water gas concentrations of CO 2 (C wCO2 ) were derived from separate chambers (here termed as equilibration chambers) equipped with CO 2 sensors (Senseair K33 ELG, Sweden) according to Bastviken et al. 28 Equilibration chambers were deployed for 24 h prior to flux measurements, allowing CO 2 concentrations in the chamber headspace to equilibrate with the water underneath. Equilibration chambers were placed next to the FCs used for flux measurements to minimize differences in spatial variability between measurements of CO 2 flux and C wCO2 . To make sure that the equilibration chambers provided reliable measures for C wCO2 , snapshot samples for C wCO2 were collected using a headspace extraction technique described above. Parallel measurements with the equilibrated chambers and the headspace extraction technique of water samples differed by less than 5%. 11 For the additional 11 flux measurements in OBJ, CO 2 concentrations were measured with a common syringe headspace extraction technique by extracting 40 mL of water at ∼0.1 m depth and 20 mL of air with a 60 mL syringe. The sample was equilibrated by shaking, and the resulting headspace gas was transferred to a dry syringe and analyzed within 120 min. Separate air samples were collected to correct for initial CO 2 in the headspace.
2.2.3. Dissolved CH 4 Concentration. Dissolved CH 4 (C wCH4 ) was sampled using two approaches: for SOR, BOL, SOD, and for 21 of the flux measurements in OBJ, water was sampled next to the FCs at ∼0.1 m depth using a 10 mL plastic syringe (Becton−Dickinson). The sampled volume was adjusted to 5 mL and transferred to a 22 mL N 2 -filled glass vial (Agilent) containing 0.1 mL 85% H 3 PO 4 and sealed with a butyl rubber stopper and aluminum crimp. The initial overpressure of N 2 was removed prior to injection to adjust the vial pressure to ambient conditions. The acid preserved the sample and allowed the storage of CH 4 until analysis.
For the remaining 11 flux measurements in OBJ, water concentrations were sampled similarly to what was described in Section 2.2.2 for the additional 11 dissolved CO 2 concentration samples, with a common syringe headspace extraction technique.

Atmospheric Pressure and Wind
Speed. At OBJ, atmospheric pressure and wind speed were measured at 10 m height from a meteorological station located 300 m north-east of the lake using an Onset S-WCA-M003 and an Onset S-BPB-CM50, respectively. 18 For SOR, BOL, and SOD, atmospheric pressure and wind speed at 10 m were obtained from the Swedish Meteorological and Hydrological Institute (SMHI) MESAN model. The model interpolates measurements from nearby weather stations combined with a meteorological model to estimate hourly means on a 2.5 × 2.5 km grid. 29 Although MESAN estimates wind speed at 10 m height, previous results from two lakes showed reasonable agreement between hourly MESAN values and wind speed measured at lake level (R 2 = 0.65−0.74). 26 2.3. Analysis of Dissolved Gas. The samples in vials from all lakes were analyzed by gas chromatography (Agilent 7890) with a Poropak Q column and FID detection connected to a headspace autosampler (Agilent 7697 headspace sampler). Samples in syringes from OBJ were analyzed in the field on the Los Gatos DLT100 greenhouse gas analyzer as described above. Water concentrations were calculated using the partial pressures (the measured mixing ratio times the barometric pressure) combined with (i) the ideal gas law to calculate the amount of the gas in the headspace (n h ; mol) and (ii) the temperature adjusted Henry's law 30 to calculate the residual gas in the water within the vial/syringe (n aq ; mol). The sum of n h and n aq , after subtracting the background target gas amount in the gas forming the headspace (N 2 in vials and atmospheric air in the syringe extractions), divided with the volume of the water sample extracted yielded target gas concentrations. For CO 2 , we accounted for shifts in the carbonic acid equilibrium during equilibration according to Koschorreck et al. 17 and Rudberg et al. 26 2.4. Calculating k 600 . We used paired gas flux and concentration measurements of CO 2 and CH 4 to derive k from eq 1. These values were then converted to a standardized k with the same Schmidt number of 600 (k 600 ), i.e., the relationship between the kinematic viscosity of water divided by the diffusion coefficient of gas normalized to CO 2 at 20°C , 10 to allow comparison of gases 12 according to eq 2: where k is and Sc is are k and Schmidt numbers in situ, 600 is the reference Schmidt number, and n is a variable that is linked to the roughness of the water surface. Similar to other studies 4−7 and according to Jaḧne et al. 12 and Liss and Merlivat, 31 we have used a value for n of 2 / 3 and 1 / 2 in conditions when wind speeds, either measured at 10 m or corrected to a height of 10 m, were below and above 3.6 m s −1 , respectively.

Statistical Analysis.
Parameters and R 2 , adjusted to the number of observations (n) and predictor variables (p), according to eq 3, were derived from Ordinary Least Squares regression for linear relationships and from direct curve-fitting Adjusted R 2 values according to eq 3 are in the following sections, including text and figures, referred to as R 2 . Since k 600 data violated terms for parametric tests by not being normally distributed, the non-parametric Wilcoxon signed-rank (S-R) test and the Kruskal−Wallis analysis of variance rank test were performed in IBM SPSS statistics 28 to test for differences in k 600 values for CO 2 and CH 4 . We considered p-values below 0.05 as statistically significant to reject null-hypotheses in statistical tests and regressions.

RESULTS
All surface waters were supersaturated with CO 2 and CH 4 , with water to air concentration ratios (C w /C air ) ranging from 3.0 to 5.5 for CO 2 and from 95 to 335 for CH 4 . C wCO2 and C wCH4 ranged from 70 to 118 and from 0.39 to 1.38 μmol L −1 , respectively. This variability was mostly due to differences between lakes whereas within-lake variability in C wCH4 and C wCO2 was small (within-lake CV: 2−31% and 2−11% for C wCH4 and C wCO2 , respectively; Figure S2a). The measured fluxes ranged from 4 to 132 mmol m −2 d −1 and 0.02 to 0.98 mmol m −2 d −1 for CO 2 and CH 4 , respectively, with the highest fluxes observed in OBJ ( Figure S2b). Fluxes of CO 2 and CH 4 were highly related (p < 0.001, R 2 = 0.91; Figure S2b). The relationship was mainly driven by OBJ data where a greater flux range was observed, but it was also significant when OBJ data were excluded (p < 0.05, R 2 = 0.43).
Values of k 600 ranged between 0.12 and 2.24 m d −1 for CO 2 and between 0.08 and 1.84 m d −1 for CH 4 (Figure 1), with an overall mean and median k 600 ratio (k 600CO2 /k 600CH4 ) of 1.68 ± 0.77 (mean ± 1 standard deviation) and 1.37, respectively, with only small differences in mean values between the lakes (OBJ = 1.67 ± 0.75 m d −1 , SOR = 1.72 ± 0.76 m d − 1, BOL = 1.52 ± 0.24 m d −1 , and SOD = 1.76 ± 1.07 m d −1 ). This confirms the first part of our hypothesis that k 600 for CO 2 and CH 4 would differ (p < 0.001). However, the second part of our hypothesis, that lake trophic status would influence the CO 2 / CH 4 k 600 ratio, was not verified as this ratio was similar among the studied lakes (p > 0.05).
We observed that k 600CO2 and k 600CH4 were exponentially related to wind speed, which ranged from <0.4 to 3.9 m s −1 in OBJ and from 1.2 to 1.7 m s −1 elsewhere (R 2 = 0.69 to 0.75; Figure 2a; the relationship is primarily generated by OBJ data, where the wind speed range was greatest). There was no clear unidirectional relationship between wind speed and the k 600 ratio for CO 2 and CH 4 ( Figure 2b). The k 600 ratio was negatively and positively linked to CO 2 and CH 4 saturation (gas concentration in the water divided by the theoretical concentration in equilibrium with the atmospheric partial pressure), respectively ( Figure S3). Such trends were weak, yet significant considering all lakes combined (CO 2 ; p < 0.05, R 2 = 0.12, CH 4 ; p < 0.01, R 2 = 0.20), and driven mainly by the OBJ lake data (CO 2 ; p < 0.001, R 2 = 0.32, CH 4 ; p < 0.001, R 2 = 0.64).

Comparisons with Previous Studies.
Our mean and median apparent k 600 ratios for lakes, showing higher k 600CO2 by a factor of 1.7 and 1.4 compared to k 600CH4 , are consistent with one other study, suggesting that apparent k 600CO2 exceeded k 600CH4 . 7 More specifically, Rosentreter et al. 7 observed mean and median values for k 600 ∼ 2.8 and 1.6 times greater for CO 2 compared to CH 4 in a mangrove estuary. In contrast, observations from some other studies report higher apparent k 600 for CH 4 than for CO 2 (Table 2). Prairie and del Giorgio, 13 McGinnis et al., 18 and Rantakari et al. 19 found higher k 600 for CH 4 in 90−100% of their measurements, compared to 10 and 17% of the measurements in our study and Rosentreter et al., 7 respectively. These mixed results indicate variability in k 600 between gases, among systems and conditions for reasons not yet understood.
The results seem to indicate a hump-shaped relationship between the k 600 ratio and wind speed (Figure 2b). The results indicate that differences in k 600 between CO 2 and CH 4 can be substantial even at low wind speeds in both estuaries 7 and lakes and could be larger at intermediate wind speeds for unknown reasons. If correct, it can be speculated that some of the mechanisms behind differences in k 600 for CO 2 and CH 4 could have more effect at intermediate wind and intermediate k, while at higher wind speeds other factors become more important for gas flux and reduce the k 600 differences. It is worth noting that other processes and factors other than wind influence k 10, 11 and that the relationships shown in (Figure 2) should not be considered generally valid.

Possible
Explanations for Differences in k 600 among Gases. The mixed results from previous studies, where some observe higher k 600CO2 and others higher k 600CH4 , can be due to mechanisms specific to either CH 4 , CO 2 , or to both gases. To highlight how specific mechanisms may influence k for CH 4 and CO 2 , we discuss these mechanisms separately in the subsections below. Mechanisms related to the sampling methods are discussed in the section "Methodological reasons for gas k 600 differences". A summary of the mechanisms discussed can be found in Figure 3 and Table 3 Several studies reporting higher k 600CH4 than k 600CO2 considered the possibility of microbubble flux. 5,13,18 Due to the buoyancy of bubbles, allowing faster transport of dissolved gases, microbubbles could favor the transport of gases with low water solubility, resulting in higher apparent k for low-compared to high-soluble gases.
Microbubble flux, enhancing k for CH 4 , has also been suggested to be positively linked to the level of CH 4 supersaturation. 13 Although dissolved CH 4 concentrations in freshwater systems are usually supersaturated relative to atmospheric partial pressures (in the order of 2 μatm), the surface water concentrations are far from supersaturated relative to pure CH 4 (1 atm). Therefore, it is not realistic that the CH 4 itself should form microbubbles in surface water. However, microbubbles based on other gases can form due to entrainment of air in breaking waves in turbulent aquatic environments 32 and can remain entrained for several days. 33 Breaking waves and whitecap formation at wind speeds of 2−3 m s −1 has been suggested even at large-fetch systems. 34 The studies conducted by McGinnis et al. 18 and Rosentreter et al. 20 are, to our knowledge, the only freshwater studies that identified relations between potential freshwater microbubble flux and wind speed or current velocity. In contrast, studies conducted by Rantakari et al. 19 and Prairie and del Giorgio 13 that suggest higher k 600CH4 than k 600CO2 , did not observe such patterns. Studies by Beaulieu et al. 5 and Paranaíba et al., 6 also suggesting higher k 600CH4 than k 600CO2 , did not test for the above relationships. When it comes to our study, we did not experience any conditions with breaking waves or whitecap formation while sampling, thus limiting the potential of microbubble flux contribution.
Microbubbles, regardless of the formation mechanism, could transport all supersaturated gases, i.e., not only CH 4 but also CO 2 and other gases. The relative microbubble transport rates could be solubility dependent as suggested (favoring the transport of CH 4 over CO 2 ), but given the higher transfer rates of supersaturated CO 2 from water into a headspace ( Figure  S4), significant transport of CO 2 or other soluble gases via microbubbles cannot be excluded. Moreover, Beaulieu et al. 5 observed no significant differences between k for CH 4 and nitrous oxide (N 2 O) (t-test, p = 0.52), although solubility of N 2 O is similar to CO 2 and would be expected to have lower k than CH 4 if microbubble flux was occurring and was solubility dependent.
Our results may indicate that k 600CH4 relative to k 600CO2 decrease with CH 4 supersaturation (Section 3; Figure S2b), and our calculations on the potential microbubble flux (F mb ), according to Prairie and del Giorgio 13 (Text S2), show negligible effects on the enhancement of k for CH 4 . This does not support previous suggestions on microbubble formation and gas transport in boreal lakes 13 and shows that the relationship between microbubble flux and enhanced k for Eq and Hs denote measurements with flow through equilibrator and headspace extraction techniques, respectively, for surface water concentrations. The equilibrated chambers technique includes headspace extraction in floating chambers allowed to equilibrate. b Equilibrated chambers were used in 21 of the samples from OBJ, and headspace extraction was used in the remaining 11 samples from OBJ and in SOR, BOL, and SOD. Conceptual figure for potential effect of chemical reactivity and degradation processes in the surface microlayer on concentration gradients of CO 2 (blue dashed line) in relation to assumed concentration gradients for CO 2 or CH 4 without CO 2 reactivity (solid lines in blue and red, respectively) in waters supersaturated with both CO 2 and CH 4 . Please note that our intention with the figure is to present processes that contributes to the increase of apparent k 600 for CO 2 relative to CH 4 , hence relating to the findings from our empirical measurements. Therefore, we chose not to include processes in the figure that have the opposite effect, i.e., enhancing apparent k 600 for CH 4 , e.g., by the contribution of microbubbles or other processes that may be the result of sampling bias.  if not accounting for the chemical equilibration of the carbonate system inside the water sample when calculating C wCO2 in supersaturated waters, C wCO2 will be overestimated relative to C wCH4 . This in turn will underestimate apparent k 600CO2 .
a Here k 600 represents the gas transfer velocity normalized to the Schmidt number 600. The explanations are considering processes when k 600 is estimated from combined flux and surface water concentration measurements. Each process in the table is discussed further in the separate paragraphs in Section 4.2. The last column, study, is an attempt trying to link possible effects on apparent k 600 from each specific mechanism presented in Section 4.2 to the studies that are shown in Table 2.
Environmental Science & Technology pubs.acs.org/est Article CH 4 needs further consideration. Based on the abovementioned discussion, it is clear that the microbubble hypothesis, as an explanation for enhanced k for CH 4 relative to k for CO 2 , is not generally applicable and instead context dependent. The extent to which microbubbles influence apparent k 600 for different gases is therefore still an open question. Some of the alternative explanations outlined below may in many situations be more likely when explaining differences in k 600 among gases. 4 Production. CH 4 can be produced in waters under oxic conditions, 35,36 and there is support for such production associated with photosynthesis by cyanobacteria and other mechanisms. The extent of this surface water CH 4 production is debated, but if it occurs close to the water surface, it could contribute to CH 4 being formed in the surface boundary layer, making the real concentration gradient steeper than measured from deeper sampling, and leading to overestimating the apparent k 600CH4 when estimating k from concentration measurements below the boundary layer.

High Surface Primary Production.
Under some conditions, e.g., severe light limitation caused by algal blooms, it is possible (albeit unlikely�see "Surface microfilm processes" below) that the majority of the primary production may be restricted to the uppermost (few centimeters) of the water, resulting in limited productivity below this layer. 37 This in turn could lead to elevated CO 2 consumption near the water surface, and C wCO2 sampled some cm or deeper below the water surface may not represent the real C wCO2 driving the CO 2 gas exchange. In such cases, C wCO2 could be overestimated yielding underestimation of k 600CO2 , resulting in a higher apparent k 600CH4 than k 600CO2 . This explanation illustrates a case of method bias if gas concentration measurements do not represent the actual concentration gradient shaping the gas fluxes.

Potential Mechanisms
Leading to Higher k 600 for CO 2 than for CH 4 . 4.2.2.1. Chemical Reactivity. In contrast to dissolved CH 4 , CO 2 has two ways to pass the water-side boundary layer at the water-atmosphere interface. One way, which is shared with other gases, is via molecular diffusion. In addition, CO 2 can react with the water and form bicarbonate or carbonate ions. Accordingly, a part of the CO 2 can diffuse through the water boundary layer as bicarbonate (Figure 3). This generates a second way for CO 2 to cross the wateratmosphere interface that is not shared by dissolved CH 4 . At high pH, which is common in lakes with high primary productivity and low inorganic carbon concentrations, this is well acknowledged as chemical enhancement. 15,21 However, at low pH, when the carbonic acid equilibrium system favors a dominance of CO 2 as the net result of equilibrium reactions, there is still a continuous formation of bicarbonate that could potentially contribute to the transport of CO 2 across the surficial boundary layer. As a part of this mechanism and in the case with CO 2 -supersaturated waters, loss of CO 2 through emissions to the atmosphere will shift the inorganic carbon equilibrium balance in the surface water boundary layer to convert more bicarbonate to CO 2 , which leads to a higher resupply of CO 2 in the layer where the actual loss to the atmosphere occurs. This dual mechanism for passage across the diffusive boundary layer differs from gases not reacting with the water (such as CH 4 ). Thereby, water samples taken below the boundary layer may underestimate the concentration gradient of CO 2 and in turn overestimate k 600 for CO 2 in relation to gases such as CH 4 (Figure 3). A higher reactivity for CO 2 compared to CH 4 in the surficial boundary layer, that is not limited to conditions of high pH, might explain the higher apparent k-values of CO 2 that we observed during conditions of CO 2 supersaturation, and similar observations being consistent with this explanation have also been obtained in CO 2 -supersaturated parts of estuaries.

Surface Microfilm Processes
Stimulating CO 2 Production. The surface microfilm at the air−water interface of water bodies is often enriched in nutrients, particulate and dissolved carbon, phytoplankton, and microbes. 38−40 This is a zone where photochemical processes can degrade dissolved organic matter to CO 2 and labile compounds, providing additional substrate for microbes 41−44 (Figure 3). This is also a zone where photoinhibition may reduce primary production 37,45,46 and where buoyant particles aggregate and are respired, which enhances net production of CO 2 (Figure 3). Hence, C wCO2 measurements a few centimeters into the water may underestimate the C wCO2 and in such cases k 600CO2 will be overestimated and the apparent result would be that k 600CO2 is greater than k 600CH4 .
Biological and chemical processes in ocean surface microlayers have been suggested to influence k 600 , 14,47−49 but to our knowledge their effect on k 600 in inland waters is largely unknown. A study using FCs in the tropical Atlantic Ocean found mismatches in k 600 between CH 4 , carbon monoxide (CO), N 2 O, and hydrogen (H 2 ) and concluded that microbial gas consumption within the surface layer film was the only plausible explanation for such differences in k 600 . 14 Frost 47 came to the same conclusion when he identified a mismatch with up to 8% higher k 600 for CH 4 compared to sulfur hexafluoride (SF 6 ) in the North Sea, and results were corroborated by Upstill-Goddard et al., 50 who replicated similar patterns in a controlled experiment by adding methanotrophs to the surface microlayer film. In the subtropical Atlantic, Calleja et al. 51 found that k 600 for CO 2 was controlled by microbial metabolism in the uppermost (<2 cm) water layer, with sevenfold and tenfold differences in respiration and gross primary production, respectively, compared to the mixed photic layer below. If the above findings in ocean systems are valid for lakes, microbial communities within the uppermost water layer could potentially alter the concentration of CH 4 or CO 2 in lakes and bias our k determination method. It might explain the contrasting patterns found in different systems, where some studies indicate enhanced k 600CH4 and some show enhancement of k 600CO2 .

Methodological Reasons for Gas k 600 Differences.
Discrete sample headspace analysis is associated with a risk of introducing systematic error into calculations if chemical equilibration of the carbonate system inside the sample vial is not accounted for. Koschorreck et al. 17 found that such bias could lead to error, underestimating C wCO2 by a factor of 3 in highly undersaturated waters when using atmospheric air for the equilibration, but overestimating C wCO2 by less than 5% in supersaturated (>1000 μatm) waters. Bias is reduced if a high water to air volume ratio is used in the sampling syringe. We accounted for this error when calculating C wCO2 , but bias would otherwise only be ∼2% due to high supersaturation in our lakes and a water to air volume ratio of ∼19 when sampling. Nevertheless, it is possible that the findings from Koschorreck et al. 17 could influence the k 600 -ratios (CO 2 / CH 4 ) presented in Table 2 if not considered in all studies.
Environmental Science & Technology pubs.acs.org/est Article When using flow-through equilibrators, a mismatch between k for CO 2 and CH 4 could increase if differences in gas equilibration times are not accounted for. 16 The equilibration time for CH 4 is considerably longer than for CO 2 due to the greater supersaturation of CH 4 in water and its lower solubility, and thereby the greater proportional mass transport of CH 4 is needed from the water to the gas phase to reach equilibrium ( Figure S4). Accordingly, if measurements of C wCH4 are not fully equilibrated, this will underestimate C wCH4 and consequently result in overestimation of k 600 for CH 4 .
Regardless of choosing equilibrator or headspace extraction approaches, the depth of the water sampling is critical, as highlighted above. Potential explanations for the differences between apparent k 600 for CO 2 and CH 4 could be attributed to methodological inconsistency regarding the depth where water samples are taken to measure the gas concentration due to gradients in the very near-surface water. From this perspective, it could be argued that a more accurate way to estimate the surface water gas concentrations driving the flux may be to use long-term FC equilibrations, designed to make in situ chamber headspace gas concentrations reflect the gas concentrations in the uppermost water layer. Further exploration of such chambers designed for quickest possible equilibration rates (e.g., large surface area to volume ratio and constant water renewal under the chamber 28 ) and evaluating their pros and cons would be of interest to improve future k measurements.
Another methodological aspect regards the common temporal mismatch between flux measurements (integrating gas transport over 10−30 min) and instantaneous snap-shot concentration measurements or from the delay in the equilibration of chambers. This contributes uncertainty to kestimates as the flux and concentration data used to calculate k represents different time periods. The direction of this bias depends on how gas concentrations in the water change during the flux measurement. This time-mismatch may therefore not cause a systematic bias on k 600 differences among gases if considering many measurement periods and systems. However, reducing the related k uncertainty needs further consideration by, e.g., combining flux measurements with repeated or continuous concentration measurements during the whole flux measurement period or optimized FCs for rapid equilibration.

Future Considerations and Implications.
It is imperative to understand controls on k 600 to accurately estimate gas fluxes from surface waters. Our results from four boreal lakes add to the growing literature which indicates that current assumptions made when calculating k 600 from measurements designed for other gases, do not hold when tested in situ. Additional mechanisms than those outlined here may be possible. For example, variability in k values for CO 2 and CH 4 may also depend on conditions within the water column, and exploring potential patterns in k 600CH4 to k 600CO2 versus, e.g., local hydrodynamics influenced by heating, cooling, and wind speed and direction would be of interest. Clearly, several explanations to observed differences in k 600 among gases exist that can interact and differ in relative importance among studies. This is a matter of concern as kdependent gas exchange models are widely used in regional studies due to the perceived simplicity of collecting water samples to estimate gas concentration and modeling k to derive fluxes in comparison with making actual flux measurements. Many such studies, incorporating k-dependent gas exchange models, are used in global upscaling of fluxes, and the discrepancy-ranges observed in k 600 seem large enough to substantially affect aquatic greenhouse gas emissions estimates. Future studies should be designed to address potential biases in gas concentration measurements and account for the possible mechanisms that may affect k differently for CO 2 and CH 4 . Before these challenges are addressed, attempts to convert k from one gas to another based solely on physical properties may not be reliable beyond controlled laboratory conditions, and in situ empirical assessments of k for each gas of interest remain important for accurate flux assessments. ■ ASSOCIATED CONTENT
Gas calculations for CO 2 and CH 4 in lake systems SOR, BOL, and SOD (Text S1) and supplemental method description of microbubble flux calculations (Text S2), with figures showing the in situ sampling setup using UGGA and flux chamber for measuring fluxes of CH 4 and CO 2 ( Figure S1), the relationship between measured, lake dissolved CO 2 and CH 4 concentrations, and lake CO 2 and CH 4 fluxes ( Figure S2), the measured k 600 ratio (CO 2 /CH 4 ) and CO 2 saturation, and CH 4 saturation ( Figure S3), and gas specific equilibration time measurements for CO 2 and CH 4 in an aquatic environment ( Figure S4) (PDF) Source data contain data needed to reproduce results in the manuscript and include in situ measurements of the apparent k 600 for CO 2 and CH 4 , wind speeds approximated at 10 m height (U 10 ), apparent k 600 ratio (CO 2 /CH 4 ), dissolved CO 2 and CH 4 , gas fluxes of CO 2 and CH 4 , and apparent ratio of dissolved gases in the surface waters versus the above lying atmosphere (C w / C air ) for CO 2 and CH 4 , respectively (XLSX)