Coupled CH4 production and oxidation support CO2 supersaturation in a tropical flood pulse lake (Tonle Sap Lake, Cambodia).

Carbon dioxide (CO2) supersaturation in lakes and rivers worldwide is commonly attributed to terrestrial-aquatic transfers of organic and inorganic carbon (C) and subsequent, in situ aerobic respiration. Methane (CH4) production and oxidation also contribute CO2 to freshwaters, yet this remains largely unquantified. Flood pulse lakes and rivers in the tropics are hypothesized to receive large inputs of dissolved CO2 and CH4 from floodplains characterized by hypoxia and reducing conditions. We measured stable C isotopes of CO2 and CH4, aerobic respiration, and CH4 production and oxidation during two flood stages in Tonle Sap Lake (Cambodia) to determine whether dissolved CO2 in this tropical flood pulse ecosystem has a methanogenic origin. Mean CO2 supersaturation of 11,000 ± 9,000 μatm could not be explained by aerobic respiration alone. 13C depletion of dissolved CO2 relative to other sources of organic and inorganic C, together with corresponding 13C enrichment of CH4, suggested extensive CH4 oxidation. A stable isotope-mixing model shows that the oxidation of 13C depleted CH4 to CO2 contributes between 47 and 67% of dissolved CO2 in Tonle Sap Lake. 13C depletion of dissolved CO2 was correlated to independently measured rates of CH4 production and oxidation within the water column and underlying lake sediments. However, mass balance indicates that most of this CH4 production and oxidation occurs elsewhere, within inundated soils and other floodplain habitats. Seasonal inundation of floodplains is a common feature of tropical freshwaters, where high reported CO2 supersaturation and atmospheric emissions may be explained in part by coupled CH4 production and oxidation.

Globally, most lakes and rivers are supersaturated with dissolved carbon dioxide (CO2) relative to the atmosphere, highlighting their outsized role in transferring and transforming terrestrial carbon (C) (1–3). Terrestrial–aquatic transfers of C can include CO2 dissolved in terrestrial ground and surface waters (3–6), dissolved inorganic carbon (DIC) from carbonate weathering (7, 8), or organic C from various sources that is subsequently respired in lakes and rivers (9, 10). Initially, oceanic export was thought to be the only fate for terrestrial–aquatic transfers of C, but a growing body of research on sediment burial of organic C and CO2 emissions from freshwaters prompted the “active pipe” revision to this initial set of assumptions (11). Although freshwaters are now recognized as focal points for transferring and transforming C on the landscape, most of this research has been conducted within temperate freshwaters (2, 11, 12). Few studies focus on the mechanisms of CO2 supersaturation in tropical lakes and rivers, with most conducted in just one watershed, the Amazon (4, 13–15).

CO2 supersaturation within tropical freshwaters is likely influenced by their unique flood pulse hydrology. The canonical flood pulse concept hypothesizes that annual flooding of riparian land will lead to organic C mobilization and respiration (16). Partial pressures of CO2 (pCO2) have been measured in excess of 44,000 atm in the Amazon River (13), 16,000 atm in the Congo River (17), and 12,000 atm in the Lukulu River (17). Richey et al. (13), Borges et al. (18), and Zuidgeest et al. (17) have each shown that that riverine pCO2 scales with the amount of land flooded in these watersheds. Yet it was only recently that Abril and Borges (19) proposed the importance of flooded land to the “active pipe.” These authors differentiate uplands that unidirectionally drain water downhill (via ground and surface water) from floodplains that bidirectionally exchange water with lakes and rivers (19). They conceptualize how floodplains combine high hydrologic connectivity, high rates of primary production, and high rates of respiration to transfer relatively large amounts of C to tropical freshwaters (19).

Methanogenesis inevitably results on floodplains after dissolved oxygen (O2) and other electron acceptors for anaerobic respiration such as iron and sulfate are consumed (16, 19). Horizontal gradients in dissolved O2 and reducing conditions have been observed extending from the center of lakes and rivers through their floodplains in the Mekong (20, 21), Congo (22), Pantanal (23), and Amazon watersheds (4). CH4 production and oxidation occur along such redox gradients (4, 16, 19, 23). CH4 is produced by acetate fermentation (Eq. ) and carbonate reduction (Eq. ) within freshwaters (24, 25). CH4 production coupled with aerobic oxidation results in CO2 (Eq. and ref. 25), yet no studies have quantified the relative contribution of coupled CH4 production and oxidation to CO2 supersaturation within tropical freshwaters.

The relative contribution of coupled CH4 production and oxidation to CO2 supersaturation within tropical freshwaters can be traced with stable C isotopes of CO2 and CH4. Methanogenesis results in CH4 that is depleted in 13C (13C = −65 to −50‰ from acetate fermentation and −110 to −60‰ from carbonate reduction) compared to other potential sources of organic and inorganic C (13C = −37 to −7.7‰; see ) (24–26). The oxidation of this 13C-depleted CH4 results in 13C-depleted CO2 (24–26). At the same time, CH4 oxidation enriches the 13C/12C of residual CH4 as bacteria and archaea preferentially oxidize 12C-CH4 (25). This means that the 13C/12C of CO2 and CH4 can serve as powerful tools to determine the source of CO2 supersaturation within freshwaters.

Tonle Sap Lake (TSL) is Southeast Asia’s largest lake and an understudied flood pulse ecosystem that supports a regionally important fishery (21, 22, 27). Each May through October, monsoonal rains and Himalayan snowmelt increase discharge in the Mekong River and cause one of its tributaries, the Tonle Sap River, to reverse course from southeast to northwest (21). During this course reversal, the Tonle Sap River floods TSL. The TSL flood pulse increases lake volume from 1.6 to 60 km3 and inundates 12,000 km2 of floodplain for 3 to 6 mo per year (21, 27). Holtgrieve et al. (22) have shown that aerobic respiration is consistently greater than primary production in TSL (i.e., net heterotrophy), with the expectation of consistent CO2 supersaturation. But, the partial pressures, C isotopic compositions, and ultimately the source of dissolved CO2 in TSL remain unquantified.

To quantify CO2 supersaturation and its origins in TSL, we measured the partial pressures of CO2 and CH4 and compared their C isotopic composition to other potential sources of organic and inorganic C. We carried out these measurements in distinct lake environments during the high-water and falling-water stages of the flood pulse, hypothesizing that CH4 production and oxidation on the TSL floodplain would support CO2 supersaturation during the high-water stage. We found that coupled CH4 production and oxidation account for a nontrivial proportion of the total dissolved CO2 in all TSL environments and during both flood stages, showing that anaerobic degradation of organic C at aquatic–terrestrial transitions can support CO2 supersaturation within tropical freshwaters.

Results

pCO2 and pCH4 in TSL were consistently supersaturated relative to atmospheric equilibrium. pCO2 averaged 13,000 ± 6,000 atm (mean ± 1 SD) across sites during the high-water stage and 13,000 ± 12,000 atm during the falling-water stage (Table 1). pCH4 was significantly greater during the high-water stage (11,000 ± 2,000 atm) than during the falling-water stage (600 ± 300 atm) (P < 0.001, d = 1.8). By contrast, pCO2 and pCH4 at sea level are ∼400 and 1.8 atm, respectively.

Table 1.

Mean partial pressures (atm) ± 1 SD and

	pCO2 (μ atm)	pCH4 (μ atm)	δ13C-CO2 (‰)	δ13C-CH4 (‰) n	n
High
All environments	13,000 ± 6,000	11,000 ± 2,000	−40 ± 7	−36 ± 2	35
Open	12,000 ± 4,000	3,000 ± 2,000	−37 ± 4	−45 ± 9	6
Edge	14,000 ± 6,000	20,000 ± 10,000	−39 ± 6	−38 ± 6	6
Floodplain	13000 ± 6,000	11,000 ± 2,000	−41 ± 7	−34 ± 2	23
Falling
All environments	13,000 ± 12,000	600 ± 300	−38 ± 5	−62 ± 5	12
Open	14,000 ± 13,000	50 ± 20	−35 ± 3	−57 ± 6	6
Edge	14,000 ± 12,000	1,400 ± 800	−40 ± 6	−67 ± 7	6

Mean partial pressures ( atm) and 13C (‰) across all lake environments for the high-water and falling-water stages are also shown.

CO2 supersaturation exceeded dissolved O2 deficits, indicating sources of dissolved CO2 other than aerobic respiration (Fig. 1 ). CO2 supersaturation is expected to vary with dissolved O2 deficits in a −1/1 O2:CO2 ratio as one micromole of dissolved O2 is consumed for each micromole of dissolved CO2 produced. Instead, ratios of −0.1/1 were observed during both the high-water and falling-water stages. During the high-water stage, the greatest CO2 supersaturation occurred under the most hypoxic conditions (Fig. 1).

Fig. 1.

Dissolved O2 deficit and CO2 supersaturation, relative to atmospheric equilibrium in open water, edge, and floodplain environments of TSL (A) during the high-water and falling-water stages of the flood pulse (B). Dissolved O2 deficit and CO2 supersaturation are calculated as the difference between atmospheric equilibrium, according to Henry’s Law. Orange lines show atmospheric equilibrium at a dissolved O2 deficit and CO2 supersaturation of 0 mol ⋅ L−1. A slope (m) of −1.0 represents the equimolar consumption of dissolved O2 and production of dissolved CO2 expected during aerobic respiration (black dashed line). Instead, a slope of −0.1 was observed during both the high-water and falling-water stages. O2 deficits were strongly correlated to CO2 supersaturation during the high-water stage, but there was no such correlation during the falling-water stage.

The intercept of the relationship between 1/CO2 and 13C-CO2 can be used to determine the source of dissolved CO2 (Keeling Intercepts; ) (28, 29). In TSL, the inverse of pCO2 was strongly correlated with 13C depletion of CO2. The intercept of 13C-CO2 was as low as −51‰ during the high-water stage and −43‰ during the falling-water stage. This indicates a 13C-depleted source of dissolved CO2 relative to the other potential sources of organic and inorganic C measured, which ranged from −37 to −7.7‰ (Fig. 2). Observed 13C depletion of dissolved CO2 coincided with 13C enrichment of dissolved CH4 (Table 1 and Fig. 3 ). Acetate fermentation produces 13C-CH4 ranging from −65 to −50‰ and carbonate reduction produces 13C-CH4 ranging from −110 to −60‰ (24–26). By contrast, 13C-CH4 averaged −36 ± 2‰ during the high-water stage. During this flood stage, dissolved CO2 became more 13C depleted, and dissolved CH4 became more 13C enriched from open water environments (13C-CO2 = −37 ± 4‰, 13C-CH4 = −45 ± 9‰) to edge environments (13C-CO2 = −39 ± 6‰, 13C-CH4 = −38 ± 6‰) to floodplain environments (13C-CO2 = −41 ± 7‰, 13C-CH4 = −34 ± 2‰). Net fractionation between 13C-CO2 and 13C-CH4 (simply, 13C-CO2 – 13C-CH4) of typically <10‰ in TSL indicates substantial CH4 oxidation (25) (Fig. 3 ).

Fig. 2.

(A) Measured 13C-CO2 (blue) relative to other potential sources of organic and inorganic C during the high-water and falling-water stages of the flood pulse. “Other” potential sources of organic and inorganic C measured by this study in TSL include macrophytes, terrestrial C3 vegetation, periphyton, phytoplankton, and DIC. Emergent aquatic C4 grasses, measured by Hedges et al. (48) in the Amazon, and atmospheric CO2 in equilibrium with water (47) are also included. Isotopic values quantified by this study are in blue and white, and those quantified by other studies (24–26, 47, 48) are in gray. (B) Apparent fractionation between 13C-CO2 and 13C-CH4 () indicated substantial CH4 production through acetate fermentation with some carbonate reduction in TSL (24, 25). Therefore, a two-source, isotope-mixing model was created using 1) a continuous uniform distribution of 13C-CO2 known to result from the oxidation of CH4 produced by both pathways (gray boxes) and 2) a continuous, uniform distribution of 13C-CO2 from DIC and the aerobic respiration of potential organic C sources (white box).

Fig. 3.

13C-CO2 and 13C-CH4 in open-water, edge, and floodplain environments of TSL during the high-water (A) and falling-water stages of the flood pulse, modified from Whiticar (25) (B). Zones of CH4 production by acetate fermentation, CH4 production by carbonate reduction, and CH4 oxidation based on apparent fractionation between 13C-CO2 and 13C-CH4 (C) are shaded in gray. 13C depletion of CO2 was strongly correlated to independent measurements of net CH4 oxidation in the water column during the high-water stage (C), though not during the falling-water stage (D). 13C depletion of CO2 was also strongly correlated to gross CH4 production within sediments during the high-water stage (E), though not during the falling-water stage (F).

A two-source, stable isotope-mixing model for 13C-CO2 was used to estimate fractional contributions to dissolved CO2 by CH4 oxidation, compared with other potential sources of organic and inorganic C (Fig. 2). Assuming oxidation of CH4 produced by acetate fermentation only, the fractional contributions by CH4 oxidation to dissolved CO2 range from 63 to 85% across the distinct lake environments and flood stages of TSL (). Assuming oxidation of CH4 produced by both acetate fermentation and carbonate reduction, these contributions by CH4 oxidation fall to a more conservative 47 to 67%. Apparent fractionation between 13C-CO2 and 13C-CH4 (simply, 13C-CO2/13C-CH4) of typically <1.055 in TSL indicate substantial CH4 production by acetate fermentation with some carbonate reduction (24, 25) (Fig. 2).

13C-CO2 was strongly correlated to independent measurements of net CH4 oxidation in the water column during the high-water stage (Fig. 3 ). The same significant relationship was observed between 13C-CO2 and gross CH4 production within the sediments (Fig. 3 ). Despite these relationships, CO2 mass balance indicates that CH4 production and oxidation within the water column and underlying sediments contribute at most 9% to dissolved CO2 in TSL (). Of these two processes, CH4 production contributes one to two orders of magnitude more CO2 than CH4 oxidation. Other processing of C within the water column and underlying sediments, such as aerobic respiration, also contribute a relatively small share of total dissolved CO2 (13 ± 8%).

Discussion

Contributions of CH4 production and oxidation to CO2 supersaturation are understudied within tropical freshwaters, where extensive flooding, dissolved O2 deficits, and reducing conditions at aquatic–terrestrial transitions make such contributions likely. The subtropics and tropics are home to many high-order flood pulse rivers, such as the Amazon, Orinoco, Congo, Zambezi, and Mekong, which are collectively responsible for over 30% of global mean annual discharge (30). Along this tropical “active pipe” lays 52% of the world’s floodplains, transferring and transforming C at relatively high rates (20, 31). Using a combination of isotopic tracers and mass balance, we show that a substantial fraction this transfer and transformation of C occurs through coupled CH4 production and oxidation in TSL.

A majority of our measured 13C-CO2 fell between the 13C-depleted CO2 known to result from CH4 oxidation and the relatively more 13C-enriched phytoplankton, periphyton, macrophytes, and terrestrial C3 vegetation measured in TSL (Fig. 2). Because there is little fractionation during aerobic respiration of organic C, measured 13C-CO2 in lakes can be expected to fall inside the range of 13C observed for commonly considered sources of organic and inorganic C (31, 32). Instead, our observed 13C-CO2 fell outside of this range. Potential sources of organic and inorganic C in TSL ranged from 13C = −37‰ for macrophytes to 13C = −7.7‰ for atmospheric CO2 in equilibrium with water. De Kluiver and others (33, 34) have reported relatively 13C-depleted phytoplankton (13C = −41‰, ). However, the net heterotrophy and CO2 supersaturation consistently observed in TSL and other lakes (34) makes substantial contributions to dissolved CO2 from aquatic primary producers such as phytoplankton unlikely, because these ecosystems are inferred to receive greater inputs of terrestrial organic C than aquatic organic C. Accordingly, the 13C of dissolved CO2 measured in the same study by De Kluiver et al. (34) ranges from −21 to −9‰, suggesting that the aerobic respiration of relatively 13C-depleted phytoplankton in net heterotrophic lakes does not substantially impact 13C-CO2. Furthermore, our most 13C-depleted dissolved CO2 was sampled on the TSL floodplain, where the water column and underlying sediments are largely shaded by macrophytes and other emergent vegetation, limiting phytoplankton production (20). Ultimately, methanogenesis is the only possible source of the 13C-depleted, dissolved CO2 observed in TSL. We can therefore use a two-source, stable isotope-mixing model to estimate relative contributions to dissolved CO2 by 1) potential sources of organic and inorganic C and 2) CH4 oxidation. This mixing model shows that CH4 oxidation contributes between 47 and 67% of dissolved C-CO2 across the distinct lake environments and flood stages of TSL, which is unprecedented in the aquatic C-cycling literature.

High-CO2 supersaturation and an imbalance with dissolved O2 such as we observed in TSL (Fig. 1 ) have previously been attributed to autotrophic and heterotrophic respiration of macrophytes and other emergent aquatic vegetation on flooded land (15, 16, 17, 35). Macrophytes and other emergent aquatic vegetation fix primarily atmospheric CO2, acting more as terrestrial primary producers than aquatic primary producers. Melack and Engle (35) have shown that floating macrophytes dominate primary production and provide the bulk of organic C to an Amazon floodplain lake. Abril et al. (15) have further suggested that floodplain and riparian wetland vegetation in the Amazon could export fully half of its primary production on an annual basis. Data from TSL supports a more nuanced interpretation. The most 13C-depleted source of organic C in TSL was an individual macrophyte (13C = −37‰, mean 13C = −33 ± 4‰). Even so, 70% of our dissolved CO2 measurements were depleted in 13C below −37‰. As confirmed by our stable isotope-mixing model, this means that aerobic respiration of macrophytes can contribute to but not explain the C isotopic depletion of dissolved CO2 observed in TSL.

Corresponding 13C enrichment of dissolved CH4 indicated a fractionating loss process, further supporting the interpretation that CH4 oxidation supports CO2 supersaturation in TSL. Acetate fermentation within tropical lake sediments from the Amazon and Pantanal has been shown to produce 13C-CH4 values ranging from −86 to −61‰ (36, 37). The same studies showed concurrent carbonate reduction producing CH4 even more depleted in 13C (36, 37). By contrast, we measured an overall mean 13C-CH4 of −43 ± 9‰ in TSL, with some values as high as −11‰ (Table 1 and Fig. 3 ). Similar values were measured by Barbosa et al. (38) on Amazon River floodplains (13C-CH4 = −70.1 to −14.8‰). Independently measured rates of CH4 production and oxidation in TSL support this conclusion. Both net CH4 oxidation in the water column of TSL (Fig. 3 ) and gross CH4 production within the sediments (Fig. 3 ) were strongly correlated to 13C-CO2.

Despite these relationships, CH4 production and oxidation and aerobic respiration within the water column and underlying sediments typically contribute less than 15% of dissolved CO2 in TSL (). In our mass balance, we solve for CO2 advected from elsewhere, within inundated soils and other floodplain habitats, and infer that this is a far greater contributor to CO2 supersaturation. This was initially hypothesized by Junk et al. (18) and later combined with the “active pipe” by Abril and Borges (20). Yet it has been empirically tested using dissolved CO2 and CH4 in only two other locations (15, 17) and never with the C isotopic composition of these dissolved gases.

The 13C depletion of CO2, 13C enrichment of CH4, and their correlations to independently measured rates of CH4 production and oxidation suggest that these coupled processes support CO2 supersaturation in TSL. By extension, coupled CH4 production and oxidation are disproportionately responsible for CO2 emissions from TSL. Lauerwald et al. (13) estimate that >50% of global riverine CO2 emissions occur in the tropics, emphasizing the importance of tropical “active pipes.” Data on the stable C isotopes of both CO2 and CH4 are rarely reported for freshwaters, though 13C-enriched dissolved CH4 (>−50‰) reported in tropical and temperate lakes, wetlands, peatlands, and the Amazon River implies widespread oxidation of CH4 to CO2 (). Coupled CH4 production and oxidation have thus been understudied but may support CO2 supersaturation and CO2 emissions from other tropical freshwaters with large amounts of seasonally or perennially flooded land. The extent of this flooding will most likely change under the twin stressors of hydropower development and climate change in the tropics (21), impacting the future role of floodplains in the transfer and transformation of C from terrestrial to aquatic ecosystems.

Materials and Methods

Field Sampling.

Field sampling was conducted during the high-water and falling-water stages of the annual flood pulse in October 2015 and March 2016, respectively, representing the typical hydrological range in TSL. Flood stages were assessed using historical data from a gauging station at Kampong Luong () (21). Sampling focused on three locations in the southwest (Kampong Preah), central (Anlang Reang), and northwest (Prek Konteil) basins of TSL. Transects designed to capture horizontal gradients in dissolved O2 and reducing conditions were established at each location. These transects consisted of six points extending through the distinct open water (Transect Point 1), edge (Transect Point 2), and floodplain environments of TSL (Transect Points 4 to 6). The edge environments were characterized by a transition from open water environments to emergent, permanently rooted floodplain vegetation.

Partial Pressures of CO2 and CH4.

Partial pressures of CO2 and CH4 at each transect point and flood stage in TSL were quantified as the average of three duplicates collected at 0.1 m below the water surface and at 0.5 m above the lake bottom where water depth exceeded 0.5 m (n = 143 duplicates, n = 47 replicates). Water was collected into 74-mL gas-tight serum bottles using a van Dorn sampler, preserved in the field with 74 L of 50% mass/volume zinc chloride solution, and placed on ice for transport to the Royal University of Phnom Penh, where they were stored at 4 °C until analysis. For analysis, samples were displaced with helium to roughly equal parts headspace and water, left to equilibrate for 12 h, and analyzed for headspace pCO2 and pCH4 using gas chromatography (SRI 8610c GC) by referencing to certified standards of known concentrations.

Stable C Isotopes of CO2 and CH4.

Following analysis for partial pressures, samples were resealed with Apiezon grease, inverted, placed on ice, and transported to the University of Washington for C isotopic analysis (n = 47). A 20-mL headspace sample was analyzed for the 13C/12C of CO2 and CH4 simultaneously using a cavity ring-down spectrometer (Picarro G2201i) with a small sample introduction module (Picarro A0314 SSIM). Following Malowany et al. (39), a column of reduced copper shavings was installed on the small sample introduction module to eliminate interference by hydrogen sulfide with isotopic measurements. Samples exceeding 300 atm CH4 were diluted with ultra-high purity nitrogen to further eliminate interference by high concentrations of this gas with isotopic measurements. Stable C isotopes of CO2 and CH4 are each expressed in delta () notation relative to Vienna Pee Dee Belemnite by referencing to certified CO2 and CH4 standards of known concentrations and 13C/12C.

Stable C Isotopes of Organic and Inorganic C.

Grab samples of floating macrophytes (Eichhornia species), terrestrial C3 vegetation, periphyton, and phytoplankton were collected across the distinct lake environments and flood stages of TSL, combined, and considered a single, lake-wide sample with a minimum of four replicates. Phytoplankton were collected using a Wisconsin net sampler (Wildco 40-A50), and periphyton was scraped from the benthos and the surfaces of floating macrophytes and emergent, aquatic vegetation. Macrophytes (13C = −33 ± 4‰, n = 4), terrestrial C3 vegetation (13C = −29 ± 2‰, n = 7), periphyton (13C = −28 ± 4‰, n = 18), and phytoplankton (13C = −24 ± 4‰, n = 6) in TSL were freeze dried, ground, and analyzed for bulk 13C/12C using an elemental analyzer (CE Instruments 2500 NA) interfaced with an isotope ratio mass spectrometer (DeltaV IRMS). Laboratory working standards were glutamic acid 1 (13C = −28.3‰ versus VPDB), glutamic acid 2 (13C = −13.7‰), and sockeye salmon (13C = −21.3‰). DIC (13C = −13.8 ± 0.4‰, n = 98) samples from another sampling effort across the same lake environments and flood stages were acidified, displaced with a helium headspace, analyzed on a DeltaV IRMS, and considered a lake-wide sample, as described previously.

Depth-Integrated Gross Primary Production and Aerobic Respiration.

Gross primary production (GPP) and aerobic respiration were modeled across the distinct lake environments and flood stages of TSL using diel-dissolved O2 data in the “LakeMetabolizer” R package (n = 16) (40, 41). Model inputs include hourly dissolved oxygen (millimoles per liter), hourly water temperature (degrees Celsius), and hourly photon flux for photosynthetically active radiation (PAR; microeinsteins per second ⋅ per square meter). Continuously logging dissolved O2 and water temperature sensors were deployed for a minimum of 20 h (Precision Measurement Engineering miniDO2T Logger, accuracy ±0.16 mg, O2 L−1, and ±0.1 °C). Accuracy of dissolved O2 sensors was verified prior to field deployment using the Winkler titration method. PAR was not measured directly but calculated from full-spectrum irradiance based on latitude, longitude, aspect, slope, transmissivity data, and the “astrocalc4r” function in the “fishmethods” R package (42). GPP and aerobic respiration were converted to millimoles of CO2 per cubic meter per day using an assimilation efficiency of 1.2 for photosynthesis (43, 44) and a conversion efficiency of 1.0 for respiration.

Volumetric rates were multiplied by mixing depths to obtain areal rates in terms of millimoles of CO2 per square meter per day. Mixing depths were evaluated with dissolved O2 profiles at each site using a multiparameter sonde calibrated just prior to deployment with water-saturated air (YSI 6920). Dissolved O2 data were plotted over depth (m), smoothed using a loess-spanning function of 0.2, and interrogated for inflection points in R (41). The depth of these inflection points at each transect was considered the mixing depth.

Depth-Integrated CH4 Production and Oxidation.

Gross CH4 production within lake sediments was quantified as the average of three duplicate sediment incubations. At each transect point and flood stage in TSL, sediment cores were taken with a stainless-steel corer. The upper 1 cm3 of each core was sealed inside a 74-mL gas-tight serum bottle (n = 72 duplicates, n = 24 replicates). The remaining volume of the bottle was filled with bottom water collected 0.5 m above the sediments. Three additional bottles were filled with bottom water only and three with water collected 0.1 m below the water surface. All bottles were incubated at ambient air temperatures (25 to 33 4 °C), which were typically <4 °C different from water temperatures in TSL and sampled daily from a helium headspace for 7 d. pCH4 was analyzed as described previously and corrected for progressively decreasing headspace:water ratios. Net CH4 oxidation in surface waters was multiplied by mixing depths to obtain areal rates as in Depth Integrated Gross Primary Production and Aerobic Respiration. Net oxidation in bottom waters was added to net CH4 production measured in the bottles containing a combination of sediment cores and bottom water and considered gross CH4 production. Following incubation, each sediment core was dried at 100 °C for 3 h and weighed. Gross CH4 production rates were then corrected for sediment core weight and scaled to nanomoles of CH4 per cubic meter per day. Previously published studies of CH4 production in lake sediment cores show that rates measured at the sediment–water interface are consistent to a sediment depth of 0.1 m (25, 45). Volumetric rates of CH4 production were thus multiplied by 0.1 m to obtain areal rates in terms of nanomoles of CH4 per square meter per day. Because one mole of CO2 is produced for each mole of CH4 produced during acetate fermentation, presumed to be dominant in TSL (Fig. 2) and within freshwaters more broadly (24, 25), rates were also considered in terms of nanomoles of CO2 per square meter per day. Each transect sampled included negative control incubations amended with a 74 L of 50% mass/volume zinc chloride solution.

Mass Balance.

A mass balance for dissolved CO2 in TSL was created from processes resulting in a gain or loss of CO2:where is the CO2 gained from modeled aerobic respiration, is the CO2 lost from modeled GPP, is the CO2 gained from measured gross CH4 production within sediments, and is the CO2 gained from measured net CH4 oxidation in the water column, each in millimoles per square meter per day. is the pCO2 measured within the water column of TSL and multiplied by a temperature dependent Henry’s constant and mixing depth to yield millimoles of dissolved CO2 per square meter on the day of sampling. Diffusion of CO2 from TSL to the atmosphere was modeled using following Cole and Caraco (46). CO2 diffusion reflects an atmospheric loss subsequent to and was ultimately excluded from the mass balance. is the remaining CO2 in the mass balance assumed to result from aerobic respiration and anaerobic degradation of organic C elsewhere, within inundated soils and other floodplain habitats under steady-state conditions (millimoles per square meter per day). Mean daily ± 1 SE was quantified using normal distributions—based on sample size, mean, and SD—of other terms in the mass balance over 10,000 Monte Carlo simulations in R (41).

Stable, Isotope-Mixing Model.

The C isotopic composition of CO2 measured in TSL fell between the 13C/12C produced by 1) the oxidation of 13C-depleted CH4 to CO2 (−110 to −50‰) and 2) that of other potential organic and inorganic sources of CO2 (−37 to −7.7‰). Here, the sole concern is the fraction of CO2 derived from CH4 oxidation. Thus, a two-source (“Methane” versus “Other”), stable, isotope-mixing model was deemed appropriate. The model also accounted for CO2 losses from primary production and atmospheric diffusion and took the form:where . was modeled as a continuous, uniform distribution of 13C-CH4 values produced by methanogenesis, ranging from −110 to −50‰ (Fig. 2) (24–26). is the fraction of CO2 resulting from CH4 oxidation. Because encompasses the range of 13C values produced by both acetate fermentation (−65 to −50‰)—presumed to be dominant in TSL and within freshwaters more broadly (24, 25)—and carbonate reduction (−110 to −60‰), the model results in a conservative estimate of for this freshwater lake (Fig. 2).

was also modeled as a continuous, uniform distribution of 13C values encompassing other potential sources of organic and inorganic C (). This distribution ranges from the most 13C-depleted source of organic C measured in TSL, macrophytes (13C = −37‰), to the most 13C-enriched source of inorganic C, atmospheric CO2 in equilibrium with water (13C = −7.7‰) (47). therefore encompasses the 13C of terrestrial C3 vegetation, periphyton, phytoplankton, and DIC measured in TSL and the 13C of emergent, aquatic C4 grasses (13C = −12.2 ± 0.3‰) measured by Hedges et al. (48) in the Amazon. With multiple sources of organic and inorganic C that overlap in 13C and no prior information on the relative importance of each, the most parsimonious option was to treat them as a group with equal probability across the full range of 13C values. However, multiple alternative models were also tested ().

and mass-dependent fractionations for photosynthesis () and diffusion to the atmosphere () in the model were quantified by this study and its mass balance (). The kinetic fractionation factors for photosynthesis and diffusion, and , are −19 and −1.1‰, respectively (47). Following the IsoSource-mixing model by Phillips et al. (49), and were assigned possible values between 0.00 and 1.00 by 0.05, and was solved for iteratively in R (41). If the resulting and (allowing a minimum of 20%), then was saved. Sensitivities of the continuous, uniform distributions generated by the model were quantified over 10,000 Monte Carlo simulations in R (41). The mean of all saved values was then reported as the fraction of CO2 resulting from CH4 production and oxidation. Variance around these saved values is based on different, continuous uniform distributions generated at random by the mixing model and was ultimately not reported.

Statistical Analyses.

Normality in the data were assessed using quantile–quantile plots and Shapiro–Wilk tests. Homogeneity of variance in the data were assessed using Levene’s tests. pCH4 and 13C-CO2 followed nonnormal distributions and were log transformed for parametric comparisons along with pCO2 and 13C-CH4 across the distinct lake environments and flood stages of TSL using ANOVA. Multiple pairwise comparisons between means in the open water, edge, and floodplain environments during the high-water and falling-water stages were carried out subsequently using Tukey Honest Significant Differences. Our Bonferroni-corrected, critical alpha value for multiple pairwise comparisons was 0.025 (for linear regression, our critical alpha value remained 0.050). To assess whether differences between means were independent of sample size, we also calculated effect sizes using Cohen’s d, where d = 0.2 to 0.4 corresponds to a small effect and low support for differences between means, d = 0.5 to 0.7 corresponds to a medium effect, and d > 0.9 corresponds to a large effect and high support for differences (50). All statistical analyses were conducted using R (41).