Ebullition of oxygen from seagrasses under supersaturated conditions

Abstract Gas ebullition from aquatic systems to the atmosphere represents a potentially important fraction of primary production that goes unquantified by measurements of dissolved gas concentrations. Although gas ebullition from photosynthetic surfaces has often been observed, it is rarely quantified. The resulting underestimation of photosynthetic activity may significantly bias the determination of ecosystem trophic status and estimated rates of biogeochemical cycling from in situ measures of dissolved oxygen. Here, we quantified gas ebullition rates in Zostera marina meadows in Virginia, U.S.A. using simple funnel traps and analyzed the oxygen concentration and isotopic composition of the captured gas. Maximum hourly rates of oxygen ebullition (3.0 mmol oxygen m−2 h−1) were observed during the coincidence of high irradiance and low tides, particularly in the afternoon when oxygen and temperature maxima occurred. The daily ebullition fluxes (up to 11 mmol oxygen m−2 d−1) were roughly equivalent to net primary production rates determined from dissolved oxygen measurements indicating that bubble ebullition can represent a major component of primary production that is not commonly included in ecosystem‐scale estimates. Oxygen content comprised 20–40% of the captured bubble gas volume and correlated negatively with its δ18O values, consistent with a predominance of mixing between the higher δ18O of atmospheric oxygen in equilibrium with seawater and the lower δ18O of oxygen derived from photosynthesis. Thus, future studies interested in the metabolism of highly productive, shallow water ecosystems, and particularly those measuring in situ oxygen flux, should not ignore the bubble formation and ebullition processes described here.

O 2 concentrations are independent of seawater pH as it does not undergo acid-base speciation. These traits make O 2 an advantageous tracer of aquatic metabolism as its production or consumption is easily measured by dissolved O 2 sensors. However, low O 2 solubility often leads to wide variations in saturation state, particularly in low volume systems that have high rates of primary production and respiration. Such large shifts in saturation state often result in the formation of bubbles. The direct flux of O 2 transported in bubbles to the atmosphere therefore results in an underestimation of photosynthetic rates as determined from measurements of changes in dissolved O 2 over time (Jørgensen et al. 1979;Revsbech and Jorgensen 1983;Borum et al. 2007).
Although substantial qualitative evidence exists for the formation and ebullition of bubbles in photosynthetic systems, little attention has been applied to quantifying seagrass oxygen ebullition rates and the in situ conditions that promote bubble formation. Acoustic seagrass mapping and detection of free bubbles has been explored as a tool for estimating photosynthesis (Wilson and Dunton 2009;Wilson et al. 2010Wilson et al. , 2012Felisberto et al. 2015), optical detectors have been used to quantify bubble spatiotemporal distribution and size (Delwiche and Hemond 2017a,b), and models using noble gas concentrations have been developed to estimate ebullition rates (Howard et al. 2018). However, these methods have not conducted direct measurements of bubble O 2 content, which is needed to accurately determine the fraction of total photosynthetic O 2 released as bubbles. Therefore, a quantification of photosynthetic O 2 production by bubble ebullition methods requires independent quantification of bubble O 2 concentration to determine the fate of photosynthetic O 2 in shallow surface waters.
Inverted funnel traps have been used to evaluate bubble ebullition and composition, primarily to investigate microbial production of methane, nitrous oxide, nitrogen, and carbon dioxide in freshwater systems (Keller and Stallard 1994;Casper et al. 2000;Huttunen et al. 2001;Varadharajan et al. 2010;Gao et al. 2013). More recently, these techniques have been used to investigate O 2 release from sediments and microphytobenthic communities (Cheng et al. 2014;Koschorreck et al. 2017). The simplest design employs an inverted funnel to capture gas, followed by manual collection and measurement of the gas volume (Odum 1957;Martens and Klump 1980;Keller and Stallard 1994;Cheng et al. 2014;Koschorreck et al. 2017). These bubble traps can be easily deployed to quantify seagrass ebullition as conditions favorable for bubble formation have been identified (e.g., shallow water, high irradiance, low flow, high oxygen saturation) (Zieman 1974;Hargraves 1982).
In this study, we used bubble traps to quantify rates of seagrass ebullition and the subsequent direct flux of gas to the atmosphere. Ebullition was measured over shallow Zostera marina seagrass meadows in Virginia, U.S.A. at two sites where intense sampling was conducted during conditions likely to favor ebullition. Our goals were to (1) explore the relationship between rates of bubble formation, irradiance, and tidal stage that controls the overlying water depth, (2) evaluate the isotopic composition (as δ 18 O) and concentration of O 2 in the bubble gas to quantify the photosynthetic source of O 2 and to estimate rates of bubble equilibration with the surrounding seawater, (3) estimate the transport of photosynthetically derived O 2 directly to the atmosphere via bubbles relative to rates of O 2 flux derived from dissolved O 2 measurements, and (4) characterize the temporal relationship between the timing of photosynthetic O 2 production and vertical transport of bubbles to the sea surface.

Study sites
Photosynthetic bubble formation was investigated in two eelgrass (Z. marina L.) meadows in shallow coastal bays (Spider Crab Bay: 37.342617 N, −75.802853 W and South Bay: 37.272783 N, −75.806097 W) on the Virginia portion of the Atlantic side of the DelMarVa Peninsula, U.S.A. The meadows extended for hundreds of meters from both measurement locations in all lateral directions. The sites were within 100 m of the Virginia Institute of Marine Sciences water quality stations (https://stormcentral.waterlog.com/public/vims) where water column O 2 , temperature, salinity, and depth were monitored every 15 min using a factory-calibrated EXO2 sonde with accuracies of AE 1% O 2 , AE 0.01 C, AE 0.1 salinity, AE 0.004 m, respectively. Photosynthetically active radiation (PAR) was measured at each site using an Odyssey PAR logger (Odyssey, NZ) which was calibrated, in situ by the methods of Long et al. (2012) to a factory-calibrated 4-channel HR4 spectroradiometer system (HOBI Labs HydroRAD-4).

Seagrass characteristics
Seagrass density at both locations was quantified by counting all shoots in twenty to thirty 0.05 m 2 quadrats randomly located within a 20 m radius of the bubble traps. One shoot was harvested from each quadrat for determination of leaf size-frequency distribution, and leaf area index (LAI). Leaves were cleaned of periphyton, as our visual observations indicated that most bubbles were forming on relatively clean younger leaves, and not the heavily fouled older leaves or the associated periphyton ( Fig. 1), by gentle scraping with a razor blade (Fig. 1). The clean shoots were dried at 60 C for 14 d and weighed on a top-loading balance (precision 0.001 g).

Bubble traps
Four to eight bubble traps where deployed, each 3 m apart, in July 2017 at the two sites over 4 d in Spider Crab Bay and over 5 d in South Bay. The bubble traps consisted of an inverted funnel (maximum diameter = 30 cm, overall height = 25 cm) constructed from a metal ring, a clear vinyl skirt and 13 cm diameter (wide end) plastic laboratory funnel. The clear vinyl was cut into a cone shape and sealed inside the smaller funnel with silicon glue and rivets (Supporting Information Fig. S1). A plastic stopcock with a Luer-Lok™ tip was mounted to the narrow end of the plastic funnel. The buoyant funnel was anchored to the bottom by four adjustable lines, one from each quadrant of the metal ring such that the stopcock was level with the water surface.
Each trap was submerged, cleaned of any debris, and cleared of all bubbles each day. The traps were sampled approximately hourly when on-site during the daytime, with longer sampling intervals occurring overnight (see Supporting Information). The captured gas was collected through the stopcock at the top of the funnel-mounted syringe using a glass, gas-tight syringe with the volume determined by the syringe draw required to remove all headspace gas from the trap. The gas volume was recorded and, for a subset of samples (see Supporting Information Table S1), a 2 mL aliquot of the sampled gas was injected into an evacuated 2 mL glass crimp vial. Oxygen content of the 2 mL aliquots of bubble trap gas was measured using a fixed-needle O 2 optode and meter (Pyroscience, GE) (Koschorreck et al. 2017). The optode was calibrated using high-purity nitrogen gas (Airgas, 99.99%) and aviation grade O 2 gas (Airgas, 99.95%) by injecting aliquots of each gas into 2 mL crimp vials to produce O 2 standards (2 mL each) of 0%, 25%, 50%, and 100% O 2 . Standards, blanks (no gas added), and gas samples were analyzed by piercing the vial septa and allowing the optode to equilibrate.

Oxygen isotopes
Oxygen isotope ratios and O 2 :Ar ratios were measured with a multicollector IsoPrime100 isotope ratio mass spectrometer coupled to a gas chromatograph with a manual injection port (Sutherland et al. 2018). The same 2 mL crimp vials samples, which were previously evaluated by the O 2 optode, were used for this analysis as the optical measurements did not require modification of the gas sample. Prior to analysis, samples were slightly pressurized with an aliquot of high purity helium to prevent the sample from mixing with air while sampling. The sample was introduced to the injection port using a helium flushed, gas-tight syringe. Downstream of the injection port the sample was passed through a 2 m molecular sieve (5 Å) gas chromatography column (Restek; OD 1/16 00 ) for separation of O 2 and Ar from N 2 . Moisture was removed from each sample by a 2 m Nafion dryer with a dry, helium-purged jacket (Permapure). Signal intensities for mass/charge (m/z) ratios of 32, 34, and 40 were monitored simultaneously to determine the O 2 isotope ratio and O 2 :Ar, and m/z of 28 and 29 were used to monitor N 2 and validate sample integrity. Oxygen isotopic compositions were expressed using delta notation with δ 18 O values in units of per mil (‰) with respect to Vienna Standard Mean Ocean Water (VSMOW). Oxygen isotope measurements were standardized to lab air, taken as +23.88‰ (Barkan and Luz 2005). All O 2 :Ar measurements were standardized to dissolved gas taken from water in equilibrium with lab air at room temperature, which was collected after introducing a helium headspace above the water in a sealed serum vial and shaking for a minimum of 30 min. Oxygen concentration measurements were derived from sample O 2 :Ar ratio. Generally speaking, the oxygen fraction of each sample is as follows: where n represents the number of moles of each gas including O 2 , Ar, and all other gases. If we make the simplifying assumption that all dissolved gases diffused into the bubble at approximately the same rate and the sources of dissolved gas are either atmospheric or photosynthetic, we can solve for X O 2 as a function of the observed O 2 :Ar ratios: Reproducibility of δ 18 O and O 2 :Ar for lab air standards in this study were 0.06‰ and 0.2%, respectively (1 standard deviation, n = 18). In the theoretical treatment of this equation, we assume the end-members of 24.5‰ and 0‰ for dissolved oxygen in equilibrium with seawater and photosynthetic O 2 , respectively (Benson and Krause 1984;Luz and Barkan 2000;Barkan and Luz 2005). It is important to note that water may undergo some oxygen isotope fractionation during photosynthetic O 2 production (Luz and Barkan 2011a). Photosynthetic O 2 was observed to range from~0‰ for cyanobacteria to as high as~6‰ in some eukaryotic algae. This effect is small relative to the magnitude of the difference between photosynthetic oxygen and atmospheric O 2 , and thus δ 18 O of dissolved oxygen is useful for fingerprinting the addition of isotopically light photosynthetic O 2 into the system. We also note that this mixing construction does not consider the influence of respiration on δ 18 O, and therefore cannot be strictly used for interpretation of a two end-member mixture. Qualitatively, respiration will decrease the O 2 concentration, and enrich the residual O 2 in 18 O (Luz and Barkan 2011b), thereby dampening the dynamic range of a mixture between atmosphere and photosynthetic O 2 . As such, we note that δ 18 O alone is insufficient to quantitatively disentangle these three processes.

Results
Extensive eelgrass meadows were present at both sites, with mean biomass density two times higher in South Bay (315 AE 56 g dry weight m −2 [gDW m −2 ]) than in Spider Crab Bay (163 AE 40 gDW m −2 ) ( Table 1). The leaf biomass was 197 AE 46 g DW m −2 and 106 AE 26 g DW m −2 and the LAI was 3.25 AE 0.37 and 1.82 AE 0.19 for South Bay and Spider Crab Bay, respectively. Both sites are heavily influenced by tidal exchange with the open waters of the Mid-Atlantic Bight, as indicated by large variations in water depth, temperature, and O 2 saturation (Table 1). Gas bubble formation on eelgrass leaves and ebullition through the water column were visibly present at both sites at low tide (e.g., Fig. 1).
The flux of gas bubbles (< 0.1 mmol gas m −2 h −1 ) and O 2 concentrations (< 14% O 2 , Figs. 2-3) were low during overnight periods. Bubble gas fluxes collected during the daytime were much larger (up to 7.3 mmol gas m −2 h −1 ) and the captured gas contained higher O 2 concentrations (up to 41% O 2 ) with maxima   occurring at the coincidence of low tide and high light conditions. Low rates of gas ebullition were observed (< 1.0 mmol gas m −2 h −1 , Figs. 2-3) when high tides occurred around the noon period of high irradiance (e.g., 14-18 July). When low tides coincided with high noontime irradiances, maximum gas fluxes were 7.3 mmol gas m −2 h −1 and 4.5 mmol gas m −2 h −1 with O 2 concentrations of 41% and 29% at South Bay and Spider Crab Bay, respectively (Table 1). The product of the maximum rates of gas ebullition (7.3 mmol gas m −2 h −1 ,  (Table 1). The daily gas fluxes, estimated from the summation of hourly fluxes and their mean O 2 content over each 24 h period, were 0-9.6 mmol O 2 m −2 d −1 and 0-10.7 mmol O 2 m −2 d −1 for Spider Crab and South Bay, respectively. The gas flux at both sites increased with irradiance, O 2 saturation, and temperature (Fig. 4). Gas fluxes decreased with increasing depth, likely due to decreased light availability and increased gas solubility and decreased buoyancy with increasing pressure. Gas sample concentrations determined from the O 2 optode and the isotope ratio mass spectrometer were significantly correlated (see Supporting Information Fig. S2) and both methods produced a positive relationship between O 2 gas concentration and ebullition rate. Furthermore, δ 18 O values decreased with increasing O 2 content of the gas (Supporting Information Fig. S2, Fig. 5). δ 18 O values plot along a theoretical mixing line between atmospheric O 2 in equilibrium with seawater (δ 18 O ffi 24.5‰) and that of photosynthetically derived O 2 (δ 18 O ffi 0-6‰) (Benson and Krause 1984;Barkan and Luz 2005; Luz and Barkan 2011a). The δ 18 O values suggest that, during periods of low ebullition and bubble O 2 concentrations near saturation, the gas in the slow-forming bubbles is in near isotopic equilibrium with the surrounding water (ffi 24.5‰) (Fig. 5). While a slight mismatch between the theoretical mixing line and our δ 18 O data may reflect some influence by respiration and/or isotope effects associated with mass transfer of gas between dissolved and gas phases (Knox et al. 1992), the negative relationship between δ 18 O and concentrations of O 2 in the collected gas further suggests that O 2 ebullition rates were directly correlated with photosynthetic production. The influence of respiratory consumption of O 2 notwithstanding, this simple two-end-point mixing model indicates that photosynthetically derived O 2 transported via bubbles ranged from 2.7% of the  evolved gas at the lowest measured O 2 concentrations to 39.4% of the evolved gas at the highest measured O 2 concentrations (Fig. 5).
At both sites, there was an apparent time lag between the optimal conditions for bubble production and when gas bubbles were captured within the traps (Fig. 6). Plotting measured gas fluxes (during periods of intense sampling on 20 June and 21 June) vs. water depth, irradiance, O 2 saturation, or temperature with a range of time lags for these latter quantities showed that an expected and significant linear relationship between gas flux and these quantities was observed with a time lag of at least~1.5-2 h between observed environmental conditions (water depth, irradiance, O 2 saturation, or temperature) and measured gas fluxes.
The data were well fit by a simple logistic regression (Supporting Information Eq. S1) indicating that low bubble O 2 concentrations coincided with low ebullition rates (R 2 = 0.82, Fig. 7). When gas concentrations within the bubble approached equilibrium with the air (i.e.,~21%), the gas flux rapidly increases through the K s (28.6% O 2 ) and approached ϕ max . Although the data fit well to this simple relationship, a lack of data during high flux periods reduces the confidence of this relationship to parameterize maximum gas fluxes and maximum bubble O 2 concentration (i.e., 95% confidence intervals, Fig. 7) and this relationship is expected to be site-and condition-specific.

Discussion
The results of this study reveal that ebullition represents a significant flux of photosynthetic O 2 from eelgrass meadows-a flux that is not captured by measurement techniques relying only on concentration measurements of dissolved O 2 . These results are consistent with other shallow photosynthetic systems where ebullition of O 2 can represent a significant fraction of ecosystem O 2 exchange (Koschorreck et al. 2017;Howard et al. 2018). The presented gas ebullition rates, O 2 concentrations, δ 18 O compositions in the evolved gas, and the conditions of their production revealed a photosynthetic origin of bubble production and highly variable O 2 concentrations due to constantly varying biological and physical conditions. Under high irradiance at slack low tide, ebullition can represent nearly half of the photosynthetic O 2 flux from seagrass meadows. Importantly, in ecosystems where autotrophic biomass approaches  Table S2 for detailed statistics). Fig. 7. Measured gas flux (ϕ) and measured O 2 percent from the optode from both sites. The optode O 2 and gas flux measurements were fit to a logistic regression to illustrate the relationship between O 2 concentration and the initiation of bubble production and the gas fluxes. Supporting Information Eq. S1 (red lines) and the 95% confidence intervals (red shading) and the fit parameters were estimated using an iteration algorithm. the carrying capacity of the system, the high daily rates of net production that lead to bubble ebullition are balanced by nighttime respiration. Therefore, the unquantified O 2 flux via bubble ebullition can be a critical component for achieving measurement closure with model estimates of daily metabolic balance needed to determine the potential for seagrasses to persist and serve as blue carbon sinks (Koweek et al. 2018).
The highest daily oxygen ebullition observed in this study (up to 11 mmol O 2 m −2 d −1 ) is consistent with ebullition rates estimated from temperate marsh ponds (up to 8 mmol O 2 m −2 d −1 , Howard et al. 2018) and directly measured in eutrophic lakes (5 mmol O 2 m −2 d −1 , Koschorreck et al. 2017) and shallow permeable sands (7 mmol O 2 m −2 d −1 , Cheng et al. 2014). Previous measurements of seagrass O 2 flux conducted at this site, determined from dissolved O 2 eddy covariance, report summer gross primary production rates (~2.7-16.6 mmol O 2 m −2 h −1 ; Hume et al. 2011, Rheuban et al. 2014 which, on the low end, are similar to the maximal hourly O 2 ebullition flux of 3.0 mmol O 2 m −2 h −1 at the South Bay site. Eddy covariance measurements conducted at the same time at South Bay measured gross primary production rates of 3.9 AE 0.7 mmol O 2 m −2 h −1 (Long et al. 2019), which are very similar to the presented maximal hourly rates of bubble ebullition. This suggests that during periods conducive to bubble formation, O 2 ebullition has the potential to represent a large and generally unquantified fraction of seagrass photosynthesis. Indeed, the large range of fluxes determined in earlier studies could be partially due to the inability to capture O 2 ebullition rates with dissolved O 2 measurements. Most importantly, methods relying on dissolved oxygen measurements determined the sites to be either net heterotrophic ( The fraction of the photosynthetically derived O 2 stored or transported in bubbles is not included when metabolic rates are derived solely from changes in dissolved O 2 concentration. Therefore, the resulting low fluxes that occur frequently during mid-day at low tide may be misinterpreted as photoinhibition or photorespiration effects (Ramus and Rosenberg 1980;Kosinski 1984;Hanelt 1996;Silva and Santos 2003). Neglecting bubble ebullition in field measurements may also help explain why short-term laboratory measures of seagrass leaf photosynthesis vs. irradiance (performed in well-mixed chambers at O 2 concentrations below saturation to prevent bubble formation) often show no evidence of photoinhibition at irradiances reported to induce photoinhibition in the field (Mazzella and Alberte 1986;Zimmerman et al. 1991;Touchette and Burkholder 2000). Furthermore, the short-term storage of O 2 within bubbles trapped on leaf surfaces during brief periods of ideal bubble formation may be followed by redissolution as the dissolved O 2 concentration decreases. This may produce a lag in measured dissolved O 2 flux that is misinterpreted as a late afternoon recovery of photosynthetic capacity following the mid-day depression. Similarly, changes in other physical parameters such as temperature, salinity, or tidal fronts with different dissolved O 2 concentration could lead to dissolution of gas bubbles, resulting in dissolved O 2 changes that are incorrectly attributed to biological processes such as photoinhibition or photorespiration.
Photosynthetic O 2 production is responsible for creating the condition of gas supersaturation that leads to the formation of bubbles, but molecular diffusion across the bubble interface also acts to equilibrate the gas composition inside the newly formed bubble with that of the surrounding seawater. Consequently, O 2 never represented more than~40% of the entire gas volume because the bubble composition depends on the partial pressures of all of the dissolved gases in the surrounding seawater, which largely include dissolved N 2 , O 2 , and CO 2 that diffuse into the bubble. Consistent with these multiple sources of gas to the bubbles, the fraction of the bubble gas with a photosynthetic origin, determined from measurements of δ 18 O (which clearly reflects these two sources of gas into the bubbles), suggests that photosynthetically derived O 2 only accounts for up to 39% of the O 2 in the captured bubbles during periods when ebullition was highest. Bubbles produced at night may represent the delayed dislodgement of bubbles produced on seagrass leaves during the daytime, in addition to bubbles of methane, nitrous oxide, carbon dioxide, or other trace gases released from the underlying anoxic sediments where they have been produced by microbial processes (Martens and Klump 1980;Huttunen et al. 2001;Gao et al. 2013). During nighttime bubble ebullition, the lower O 2 concentrations in the bubbles are likely the result of increased time for equilibration with the lower dissolved O 2 concentrations observed at night (~50-75% saturation), or possibly that sediment-derived gas bubbles strip O 2 out of the water column during their transport through the water column (Koschorreck et al. 2017).
The occurrence of ebullition depends on a number of biological and physical parameters whose interactions are not well characterized. At both sites, bubble ebullition over several days varied from zero to rates comparable to the gross primary production of seagrasses (Hume et al. 2011;Rheuban et al. 2014;Long et al. 2019). Bubble formation was observed to occur at low tide during the daytime when the difference between the internal seagrass aerenchyma gas pressures and the water pressure would increase due to the decreasing water depth and active photosynthesis (Larkum et al. 1989;Borum et al. 2007). While rates were maximal when low tide was coincident with high irradiance, ebullition rates were also dependent on the time of day since the diel maxima of O 2 saturation and temperature were during the afternoon. Further complicating the attribution is the~1.5-2 h time lag between these ideal conditions for bubble formation and their subsequent release from the leaf surface, which is triggered by bubble size, adherence to the leaf, and current and wave driven turbulence that can act to dislodge bubbles. Using tidally driven depth change (dz/dt) as an analog for flow velocity may explain why the largest fluxes were observed as flow increased following low tide, likely enhanced by current movement of the seagrass leaves that overcame the bubble adhesion to the leaves. However, increasing depth also leads to higher hydrostatic pressure and therefore reduced bubble size and buoyancy. These observations indicate that there is likely to be a dynamic relationship between increasing hydrostatic pressure, flow velocity, and bubble transport.
The logistic regression illustrates the relationship between gas production rates and gas concentrations that include a number of site-specific factors that influence the production of bubbles including water O 2 saturation, gas concentration gradients between the plant and the bubble, gas diffusion rates from the water into the bubble (and vice versa), hydrostatic pressure, and the volume at which the bubble becomes buoyant and detaches from the leaf. During periods of highest ebullition, the bubble gas still contains a majority of other gases (i.e., O 2 < 41% [approx.]) and indicates that the bubble growth rate on the leaf surface over time (~1.5-2 h) facilitates gas diffusion into the bubbles, prior to ebullition to the atmosphere. The majority of the ebullition, which occurred during periods of high O 2 saturation and irradiance, were consistent with previous field observations (Zieman 1974;Drifmeyer 1980;Hargraves 1982;Roberts and Caperon 1986) and indicate the importance of ebullition to ecosystem O 2 budgets (Koschorreck et al. 2017;Howard et al. 2018). These results clearly show that bubble ebullition represents an important flux of photosynthetic O 2 and presents key insights into the physical and biological dynamics of metabolism in highly productive seagrass ecosystems.