Lake Grevelingen is a coastal marine lake located in the south-west of the Netherlands ( Supplementary Figure 1 ; Hagens et al., 2015; Sulu-Gambari et al., 2017). The lake is a former estuary that was dammed on the landward side in 1964 and on the seaward side in 1971. The lake has a surface area of 115 km² and average water depth of 5.1 m but is intersected by several former, deep tidal channels with a maximum depth of 45 m. Since 1978, a sluice in the dam on the seaward side of the lake allows for exchange of water with the North Sea, explaining the relatively constant salinity of the lake water of ~29-33. The main former tidal channel of the lake experiences stratification in summer with bottom waters becoming either hypoxic or anoxic since at least 1978 (Wetsteijn, 2011; Hagens et al., 2015; Sulu-Gambari et al., 2017).

Here, we focus on the deepest part of the lake, the Scharendijke basin ( Supplementary Figure 1 ), located in the main channel (45 m water depth; 51.742°N, 3.849°E). Because of its depth the basin is only representative for the deeper part of the lake. In Scharendijke basin, an oxic-anoxic interface (oxycline) typically is present in the water column in summer following an increase in surface water temperatures in spring ( Supplementary Figure 2 ). Stratification and anoxia persist in the main channel throughout the summer, typically ending in early autumn when atmospheric temperatures decrease and the water column mixes (Hagens et al., 2015). Such seasonal stratification and anoxia is common in many coastal systems worldwide (e.g. Kemp et al., 2009). The sediments in Scharendijke basin are characterized by an unusually high sedimentation rate of ~13 cm yr^-1 (Egger et al., 2016). Rates of organic carbon input, as estimated from a reactive transport model applied to a geochemical dataset, are also exceptionally high at ~250 mmol C m^-2 d^-1. The elevated input of organic matter supports high rates of S O 4 2 - reduction and methanogenesis near the sediment-water interface, with CH₄ by-passing the sulfate-methane-transition-zone (SMTZ) as described previously (Egger et al., 2016). The sediment in the Scharendijke basin is therefore a source of methane to the water column. In this study we assess if the removal processes in the water column are efficient enough to prevent methane emissions to the atmosphere.

In the current study, we sampled both the sediment and the water column in Scharendijke basin during two cruises with RV Navicula in March and September 2020. Depth profiles of dissolved oxygen, temperature and salinity in the water column during 19 monitoring cruises between January and December in 2020 were obtained from the Dutch Ministry of Infrastructure and the Environment (https://waterberichtgeving.rws.nl/monitoring/tso-metingen/grevelingenmeer). Depth profiles of dissolved oxygen, temperature and salinity in the water column were additionally obtained with a CTD (Seabird SBE 911 plus) equipped with a Seabird oxygen sensor (SBE43) during the cruise with RV Navicula in September 2020. Most oxygen sensors, including the SBE43, suffer from a background signal in the absence of oxygen. Since Winkler titrations also suffer from artefacts at low oxygen concentrations, these cannot be used for calibration (Grégoire et al., 2021). Here, we assume that oxygen was absent from the water when oxygen concentrations were low and the profile became a vertical line.

A total of four sediment cores was collected in March and September 2020 with a UWITEC gravity corer using transparent PVC core liners of 120 cm length (6 cm inner diameter). Samples for porewater CH₄ and its isotopic composition were collected from the first core directly after collection using a liner with holes predrilled at 5 cm intervals that were covered with tape prior to coring. From each depth, 10 mL of sediment was sampled with a plastic syringe, previously cut to remove its top part, and directly transferred into 65 mL glass prefilled with saturated NaCl solution. Subsequently, the bottles were topped up with NaCl solution, to ensure no air bubbles remained inside, and closed with rubber stoppers and aluminum caps and stored upside down until analysis. Note that during sampling, CH₄ can degas from cores while collecting sediment, leading to underestimations of methane concentrations. This degassing effect intensifies with increasing CH₄ concentrations (Egger et al., 2017; Jørgensen et al., 2019). While the degassing impacts CH₄ concentrations it does not significantly affect its isotopic composition (e.g. Wallace et al., 2000; Egger et al., 2017). Samples for porewater and solid-phase analysis were collected from a second sediment core, which was sectioned in a glovebag under a nitrogen atmosphere in a temperature-controlled laboratory on board ship at in-situ temperature. The core was sliced at a depth resolution of 1 cm. The slices were placed in 50 mL centrifuge tubes. The tubes for selected samples (each sample for the 0-10 cm interval, every second sample for the 10-50 cm interval and every fifth sample for the 50-90 cm interval) were then removed from the glove bag, centrifuged at 4500 rpm for 20 minutes to extract porewater and then returned to the glovebag. The supernatant was subsampled for analysis of H₂S, N H 4 + , N O 2 - , N O 3 - , dissolved Fe and Mn (d Fe and d Mn, respectively), S O 4 2 - and alkalinity. Samples for H₂S analysis (0.5 mL) were stored in a glass vial with a 2% zinc acetate solution (2 mL) at 4°C. All other samples were stored in polypropylene vials. Samples for N H 4 + , N O 2 - and N O 3 - , were stored at -20°C. Samples for d Fe and d Mn were acidified with 35% suprapur HCl (10 µL of acid per mL of sample) and stored at 4°C. Samples for S O 4 2 - and alkalinity were stored at 4°C. All sediment samples were stored at -20°C in an aluminum gas tight bag under a nitrogen atmosphere. A third sediment core was sectioned at a resolution of one centimeter to determine the porosity. A fourth core was used to determine high-resolution depth profiles of oxygen in the porewater using microelectrodes (50 µm tip diameter, detection limit<1 µmol L^-1) operated with a motorized micromanipulator (Unisense, A.S. Denmark) directly on board. Calibrations were performed as described in Sulu-Gambari et al. (2016).

Water samples from discrete depths (1-5 m depth resolution) were collected directly from a 10 L Niskin bottle during the RV Navicula cruises in both March 2020 and September 2020. Samples were obtained for total dissolvable metals and dissolved metals (March and September), CH₄ and its isotopic composition and N H 4 + , N O 2 - and N O 3 - (September). No samples for H₂S were collected but there was a strong smell of H₂S during sampling of the water from 37 m downwards in September. Depth trends in dissolved Mo are used as an additional indicator for the presence of H₂S here, since Mo drawdown in an anoxic water column points towards scavenging by H₂S (e.g. Helz and Vorlicek, 2019). CH₄ samples were collected in 125 mL serum bottles, closed with butyl rubber stoppers, crimped with aluminum caps, poisoned with 1 mL of saturated ZnCl₂ solution and stored upside down in the dark. Samples for N H 4 + , N O 2 - and N O 3 - analysis were filtered through 0.2 µm nylon filters and stored at -20°C. Samples for total dissolvable metals (Fe, Mn and Mo) and dissolved metals (d Fe, d Mn and d Mo) were collected in 250 mL acid-washed LDPE Nalgene bottles using Tygon 5000 acid-washed tubing. The samples for dissolved metals were filtered through a 0.2 µm nylon filter (in March) and through a 0.2 µm Sartobran 300 cartridge (in September). Tests in our laboratory show no difference between metal concentrations in water filtered with these two types of filters. All metal samples were acidified with concentrated ultrapure HCl (200 µL of acid per 100 mL of sample). The difference between total dissolvable and dissolved metal was termed particulate metal.

Prior to analysis of CH₄, a headspace of 5 mL (water samples) or 10 mL (sediment samples) of nitrogen was added to each bottle while allowing the same volume of liquid to escape. After equilibration of the gas and water phase (>2h for water column samples, 7 days for sediment samples), CH₄ concentrations were measured with a Thermo Finnigan Trace™ gas chromatograph (Flame Ionization Detector; limit of detection 0.02 µmol L^-1). The isotopic composition of CH₄ (δ¹³C–CH₄ and δD–CH₄) was analyzed by Continuous Flow Isotope Ratio Mass Spectrometry (CF-IRMS) as described in detail in Brass and Röckmann (2010) and Sapart et al. (2011). Insufficient sample was available to determine the δD–CH₄ in the bottom water in March.

Porewater H₂S was measured spectrophotometrically using the phenylenediamine and ferric chloride method (Cline, 1969). Concentrations of N H 4 + were determined using the indophenol blue method (Solorzano, 1969). N O 3 - and N O 2 - were determined with a Gallery™ Automated Chemistry Analyzer type 861 (Thermo Fisher Scientific, USA). SO₄ ^2- was determined with ion chromatography (detection limit<50 µmol L^-1). Alkalinity was measured through titration with 0.01 M HCl within 24 h of sampling. Dissolved Fe and Mn in the porewater were determined by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES, Perkin-Elmer Avio 500; detection limit 0.1 ± 0.04 µmol L^-1 and 0.03 ± 0.01 µmol L^-1 and recovery rate of 101% and 99% for Fe and Mn, respectively). Water column dissolved and particulate Fe, Mn and Mo were measured by ICP mass spectrometry (ICP-MS, Nexion (Perkin Elmer)). The water column samples were concentrated and the seawater matrix was removed using a SC-DX SeaFAST S2 (Elemental Scientific) prior to analysis (Lenstra et al., 2022). Samples acidified to pH 2 were buffered with 4.8 M glacial acetic acid 99.7% (Baseline, Seastar) and 4.1 M ammonium hydroxide 29% (Baseline, Seastar) and subsequently loaded onto the pre-concentration column for 20 seconds (Lagerström et al., 2013). The analyte was injected directly into the device and measured online. The average limit of detection, calculated as the standard deviation of the analysis of oligotrophic seawater with known trace metal concentrations, multiplied by 3, was 0.2 ± 0.06, 0.03 ± 0.01 and 8.47 ± 2.82 nmol L^-1 and the recovery was 101%, 101% and 99% for Fe, Mn and Mo, respectively.

Freeze-dried sediment samples were ground using an agate mortar under a nitrogen atmosphere. To determine the organic C content around 300 mg of the powdered material was decalcified with 1 M HCl (2-step wash) (Van Santvoort et al., 2002), dried, weighed and powdered. A subsample was analyzed with an elemental analyzer (Fisons Instruments NA 1500 NCS) and the organic C was calculated after correction for the weight loss due to decalcification. The accuracy and precision of the analyses was determined based on measurements of the internationally certified soil standard IVA2. The certified value for IVA2 is 0.732 wt.% C. The mean value that was obtained in this study for IVA2 (n=24) was 0.722 wt.% C, with a standard deviation of 0.009 wt.% C. The analytical uncertainty based on duplicates (n=15) was 0.11 wt.% C for organic C. A second subsample (between 50 and 100 mg) of the powdered material was subjected to a 5-step sequential extraction procedure to determine the Mn (Lenstra et al., 2021) and Fe speciation (combined methods of Raiswell et al. (2010) and Claff et al. (2010)). The extraction steps were as follows: (1) 0.057 M ascorbic acid, 0.17 M sodium citrate and 0.6 M sodium bicarbonate (pH 7.5) which extracts easily reducible Fe and Mn oxides, (2) 1 M HCl which extracts reducible crystalline Fe oxides, Fe carbonate, FeS and Mn carbonate, (3) 0.35 M acetic acid, 0.2 M Na citrate and 50 g L⁻¹ Na dithionite, pH 4.8, which extracts crystalline Fe and Mn oxides, (4) 0.2 M ammonium oxalate and 0.17 M oxalic acid which extracts recalcitrant Fe oxides and non-reactive Mn phases (5) 65% HNO₃ which extracts pyrite and Mn associated with pyrite. Fe and Mn in the solutions of step 1 and 5 was measured by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES, Perkin-Elmer Avio 500; with a detection limit of 0.1 µmol L^-1 and 0.02 µmol L^-1 for Fe and Mn, respectively). The average recovery for Fe and Mn was 106% and 100%, respectively and the average analytical uncertainty based on duplicates (n=16) was 3.2% and 2.6%, respectively. Fe in the solutions of step 2, 3 and 4 was determined colorimetrically using the phenanthroline method (APHA, 2005). Average analytical uncertainty based on duplicates (n=16) was< 13.4% for all fractions of the three sequential extraction procedures. The sum of the Fe extracted in steps 1, 3 and 4 is assumed to represent Fe oxides. We exclude step (2) as a contributor to the Fe oxide pool since analysis of selected samples showed that nearly all Fe in the HCl solutions at this site was present as Fe(II). The Mn extracted in step 1 is used as a measure of reactive Mn oxides.

Water samples were filtered through 0.22 µm Supor PES filters using a vacuum pump within one day after sampling and stored at -80°C. DNA was extracted from the filters following the protocol of the DNeasy Power water DNA Isolation kit (Qiagen, Venlo, the Netherlands). For a further description of the data acquisition and analysis we refer to Venetz et al. (2022).

A 1-dimensional multicomponent reactive transport model, modified from that of Egger et al. (2016), was used to investigate the production and removal of CH₄ in the sediment at our Scharendijke site. The model describes the transport and reaction of 12 solutes and 9 solids ( Supplementary Table 1 ). Here, we focus specifically on modeling the depth trends in porewater O₂, d Mn, d Fe, SO₄, H₂S, N H 4 + , CH₄ and alkalinity and sediment organic C and Mn and Fe oxides. All profiles were used to fit the model to the data. Depth trends in N H 4 + , SO₄ ^2- and alkalinity in particular provide a strong constraint on CH₄ production and removal in the sediment. Since bubble formation is not included in the model, our CH₄ concentrations at depth in the sediment may be overestimates. To estimate alkalinity, we use modeled dissolved inorganic carbon (DIC). We assume the modeled DIC are similar to those of alkalinity which at high bicarbonate concentrations is a reasonable assumption. Solutes and solids are transported by advection (burial). Solutes are additionally transported by molecular diffusion. Biogeochemical reactions implemented in the model and their respective rate laws are given in Supplementary Tables 2, 3 . Parameter values were taken from the literature or constrained using the model ( Supplementary Tables 4, 6 ). The model was written in R (version 3.6.2.) and makes use of the marelac (Soetaert and Meysman, 2012) and reactran (Soetaert et al., 2010) packages for chemical and physical constants and the calculation of transport coefficients, respectively. The model describes the upper 100 cm of the sediment, which is divided into 1000 model layers. The boundary conditions at the sediment-water interface for solutes and solids are defined as concentrations and fluxes, respectively ( Supplementary Table 5 ). The model was run with fixed boundary conditions for 60 years until steady state was reached, as in Egger et al. (2016) for the same site. The model was then run for 20 additional years in a scenario in which the bottom water O₂ concentration and input of organic matter, Mn oxides and Fe oxides as well as the sedimentation rate were varied following a seasonal cycle ( Supplementary Figure 3 ). The key purpose of the model application was to assess the production and (potential) removal pathways of CH₄ in the sediment and estimate the diffusive release to the overlying water. Given the complexity of the system and strong temporal dynamics, a perfect fit of the model to the data is not expected.

Vertical fluxes of O₂, N O 3 - , N O 2 - , CH₄ and N H 4 + at the oxycline in the water column were calculated using the following equation:

Where F_z is the vertical flux of each solute in mmol m^-2 d^-1, K_z represents the vertical eddy diffusivity in m² d^-1 and dC/dz is the concentration gradient in mmol L^-1 m^-1. Depth profiles of K_z were estimated as follows (Gargett, 1984; Lewis and Landing, 1991):

where a₀ is a system-specific constant related to the internal wave energy in m² d^-1, g is the gravitational constant in m² d^-1, ρ_z is the density at depth z in kg m^-3 and dρ/dz is the density gradient in kg m^-4. Here, we estimate a₀ based on Equation 2 using an averaged K_z value for Lake Grevelingen (van Haren, 2019) assuming that a₀ remains constant with time. The vertical fluxes of electron acceptors (O₂, N O 2 - , N O 3 - , Fe and Mn oxides) were calculated based on the downward concentration gradient (negative), towards the oxycline while the vertical fluxes of electron donors (CH₄, Mn²⁺, N H 4 + ) were calculated based on the upward concentration gradient (positive), towards the oxycline. The downward fluxes of particles (Fe and Mn oxides) were calculated as follows (Neretin et al., 2003):

where F_part is the downward vertical flux of particles in mmol m^-2 d^-1, ω is the sinking speed of particles in m d^-1 and C is the measured concentration of particulate Fe or Mn at 30 m depth in mmol L^-1. The upward and downward vertical fluxes were summed in order to determine the electron balance and to evaluate the roles of different electron donors and acceptors at the oxycline.

The flux of methane at the water-atmosphere interface was calculated using the following equation (Barber et al., 1988):

where F_atm is the flux of methane at the water-atmosphere interface in mmol m^-2 d^-1, K_L is the gas exchange coefficient in m d^-1, C_W is the concentration of methane dissolved in the surface water (1 m depth) in µmol L^-1 and C_O is the calculated methane concentration in equilibrium with the atmosphere in µmol L^-1. The K_L value was estimated based on the wind speed according to Barber et al. (1988). The values of the minimum and maximum wind speed for the day of sampling (6th of September 2020) were obtained from The Royal Netherlands Meteorological Institute data base (https://www.knmi.nl/nederland-nu/klimatologie/daggegevens) and were used to calculate the range of F_atm.

The fractionation factor for the water column samples was derived from the equation as described in Grant and Whiticar (2002):

Where δ¹³C-CH₄ is the isotopic signature of the methane, α is the fractionation factor, f is the fraction of methane remaining and (δ¹³C-CH₄)₀ is the isotopic composition of methane before oxidation. The parameter f was calculated based on the concentrations of methane at depth 35 m (before oxidation) and 25 m (after oxidation).

Salinity showed relatively little change with water depth in Lake Grevelingen in 2020. Overall, salinity varied between 30 and 33 with a trend towards highest values (>32) from August onwards ( Figure 1A ). This temporal trend is likely indicative of increased inflow of North Sea water into the lake in the late summer and fall (Hagens et al., 2015). While in January surface water temperatures were lower than bottom water temperatures (6.1 vs. 7.3°C, Figure 1B ), in March, a constant lake temperature of 7.2°C was observed, indicative of a fully mixed water column. Warming of the surface water in spring led to stratification that persisted throughout the summer, with a thermocline that gradually deepened from 20 m to 35 m. The water column fully mixed again after our sampling in September. In October, water column temperatures were constant at around 15°C. Temperatures then gradually decreased to 8°C in December. The temperature-induced stratification strongly impacted water column oxygen concentrations. While the water column was fully oxygenated from January to March, bottom waters were depleted in oxygen in May ( Figure 1C ). The oxycline stabilized at a depth of around 20 m until July, followed by a deepening of the oxycline to a depth of 35 m, linked to the downward movement of the thermocline. Upon mixing in the fall, the water column became fully oxygenated again.

In March, oxygen penetrated into the sediment to a depth of 0.5 cm ( Figure 2A ). In September, in contrast, the bottom water and hence, also the sediment was completely anoxic. A distinct seasonality was observed in dissolved Mn and Fe in the upper 15 cm of the sediment, with high concentrations near the sediment water interface in March and low concentrations in September ( Figures 2A, B ). Sulfate concentrations rapidly decreased with depth in the sediment in both seasons, with the strongest decrease in September ( Figure 2C ). While in March porewater H₂S concentrations reached a maximum at a depth of 15 cm (~2 mmol L^-1), much higher concentrations of H₂S (~5 mmol L^-1) were observed near the sediment-water interface in September ( Figure 2D ). Porewater N H 4 + and alkalinity both increased with depth, and profiles showed higher values (up to 2 and 17 mmol L^-1, respectively) in the upper 15 cm of the sediment in September when compared to March ( Figures 2E, F ).

High porewater CH₄ concentrations were observed in the sediment in both March and September ( Figure 2G ). However, both profiles are strongly affected by degassing of CH₄ during sampling and concentrations are therefore underestimated. Methane near the sediment water interface in March was enriched in ¹³C ( Figure 2H ). In September, in contrast, a distinct minimum in δ¹³C-CH₄ was seen in the surface sediment and a small enrichment in ¹³C of CH₄ was only observed in the bottom water. A decrease in δD-CH₄ values was found near the sediment water interface in both months ( Figure 2I ).

The upper 10 cm of the sediment was enriched in organic C and depleted in Fe oxides and Mn oxides in September when compared to March ( Figures 3A–C ). A seasonal pattern of higher and lower organic C content is preserved at depth. While Fe and Mn oxide contents also show such variations at depth, they are no longer out of phase with those of organic C because of diagenesis as shown previously for this site (Egger et al., 2016).

In March the water column was fully oxygenated, while in September dissolved oxygen became depleted at 35 m ( Figures 1C , 4A ). In September, N O 3 - and N O 2 - profiles only showed minor changes with depth ( Figures 4B, C ; no data for March). Dissolved and particulate Mn concentrations showed little change with depth in March ( Figures 4D, E ). In September, dissolved Mn accumulated in the anoxic bottom waters, while a maximum in particulate Mn was observed above the oxycline. In March, dissolved Fe was not detectable in the water column whereas in September concentrations of 0.7 µmol L^-1 were observed below 30 m ( Figure 4F ). Particulate Fe concentrations gradually increased with depth in both months, with an especially sharp increase below the oxycline in September ( Figure 4G ). In March, dissolved Mo remained constant throughout the water column ( Figure 4H ), whereas in September a depletion in dissolved Mo was observed from 32 m downwards, which indicates the presence of H₂S in the deeper water (Helz and Vorlicek, 2019). In September, N H 4 + and CH₄ accumulated in the anoxic bottom waters with the concentrations near the sediment-water interface reaching 73 and 227 µmol L^-1, respectively ( Figures 4I, J ; no data for March). Both solutes strongly decreased in concentration towards the oxycline. Above the oxycline, CH₄ concentrations were ~0.2 µmol L^-1 while N H 4 + remained present at concentrations between 25 and 50 µmol L^-1 throughout the water column. A strong increase in δ¹³C-CH₄ from -63.2 to -32.4 ‰ and in δD-CH₄ from -276.8 to 15.7 ‰ was observed between 35 and 25 m ( Figures 4K, L ).

Calculated vertical fluxes of key solutes and solids ( Table 1 ) indicate that oxygen was the main electron acceptor at the oxycline. Sinking Mn and Fe oxides also contribute substantially to the electron budget. Given that the profiles of N O 2 - and N O 3 - do not show a gradient and that these solutes may be directly removed through denitrification, we cannot estimate their contribution. The main electron donors at the oxycline were N H 4 + and CH₄, with a lesser role for dissolved Mn²⁺. Our results indicate a deficiency of electron acceptors relative to electron donors, which likely is even larger than calculated here since we cannot include H₂S in the budget due to lack of data. The flux of methane from the surface water to the atmosphere, as calculated with Equation 4, ranges from 0.09 to 0.37 mmol m^-2 d^-1 considering the minimum and maximum wind speed above the water surface.

Amplicon sequencing of bacterial and archaeal 16S rRNA genes indicated that aerobic methanotrophs of the Methylomonadaceae family were present throughout the water column, with the highest relative abundance at the oxycline (35 m; Figure 5A ). Importantly, Methylomonadaceae were also present in the anoxic and sulfidic waters below the oxycline ( Figure 5B ). Anaerobic archaeal methanotrophs of the Methanoperedenaceae family were mostly present at and below the oxycline.

The modeled porewater and sediment profiles generally correspond well with the measured data ( Figures 2 , 3 ). Key exceptions are the CH₄ profile which is affected by degassing during sampling and the Fe oxide profiles at depth where the amplitude of change is overestimated. We note that the N H 4 + and alkalinity profiles are also not modeled perfectly. This could be related to temporal variations in parameters that are not included in the model (e.g. C/N ratios). However, for our purposes here, i.e. obtaining first-order estimates of the production and removal rates of methane in the sediment, the current approach is appropriate. The model results illustrate that actual CH₄ concentrations in both months were much higher than those measured. Model-calculated rates of organic matter degradation coupled to S O 4 2 - reduction integrated over the upper 100 cm of the sediments are higher in September than in March ( Table 2 ). Depth-integrated rates of CH₄ production show little change. However, the model results do suggest an upward shift in the zone of methanogenesis from March to September and a decreased removal of the CH₄ in the sediment through anaerobic oxidation coupled to S O 4 2 - reduction ( Figure 2 , Table 2 ; Supplementary Data ). In the model, the contribution of other electron acceptors than S O 4 2 - (i.e. O₂, N O 4 - , N O 3 - , Fe and Mn oxides) to CH₄ removal in the sediment is negligible. (e.g. 0.005 mmol m^-2 d^-1 for aerobic oxidation of CH₄ in March). Model calculated rates of CH₄ diffusion across the sediment-water interface are 1.1 mmol m^-2 d^-1 and 2.7 mmol m^-2 d^-1 in March and September, respectively.

Eutrophication in coastal systems can lead to a shoaling of the zone of methanogenesis and SMTZ, thereby increasing the chance of release of CH₄ to the water column (Egger et al., 2015). This CH₄ release is exacerbated in systems that suffer from seasonal hypoxia (Gelesh et al., 2016). At our study site in Lake Grevelingen, a coastal marine system subject to both eutrophication and seasonal hypoxia, we find a very shallow SMTZ at a sediment depth of ~15 cm. While the depth of the SMTZ does not differ between March and September, changes in the curvature of the S O 4 2 - and N H 4 + profiles in the porewater point towards higher rates of organic matter degradation coupled to S O 4 2 - reduction in September when compared to March ( Figures 2C–E ; Table 2 ). We attribute this to an increased input of organic matter in summer linked to the increase in primary production in the lake and increased inputs of algae, likely mostly Phaeocystis, from the North Sea during the same period (Hagens et al., 2015; Sulu-Gambari et al., 2018).

Rates of methanogenesis are apparently less sensitive to the changes in organic matter input and remain high throughout the year. The CH₄ produced by methanogenesis in porewater is usually depleted in ¹³C and D (Whiticar, 1999), with the various pathways of microbial CH₄ formation producing characteristic isotopic signatures in δ¹³C and δD. This can be visualized in a so-called co-isotope plot ( Figure 6 ). In our case, the isotopic signature of the CH₄ suggests that methanogenesis occurs both via carbonate reduction and utilization of methyl-type substrates such as acetate (Whiticar, 1999). Methanotrophy typically results in a gradual increase in δ¹³C-CH₄ and δD-CH₄ due to preferential oxidation of CH₄ that is isotopically light (Whiticar, 1999; Reeburgh, 2007). We see little evidence for substantial CH₄ oxidation in the sediment at Scharendijke from the isotopic signatures ( Figures 2H, I , 6 ). The minor increase in δ¹³C-CH₄ in the upper 20 cm of the sediment in March indicates that there is some removal of CH₄, but a clear isotopic signal of removal is only observed in the bottom water. We attribute this to aerobic oxidation of CH₄ near the sediment-water interface in March. The decrease in δ¹³C-CH₄ in the upper 15 cm of the sediment in September is likely the result of an increase in methanogenesis relative to methanotrophy linked to slow growth of ANME as proposed earlier for this location (Egger et al., 2016) and other sites (Knab et al., 2009; Egger et al., 2015).

The combination of high rates of CH₄ production and inefficient CH₄ oxidation leads to release of CH₄ into the water column in both March and September. Our calculated benthic release rates of CH₄ (1.1 and 2.7 mmol m^-2 d^-1) for these two months are in line with the values of 0.6 - 2.2 mmol m^-2 d^-1 previously calculated for Scharendijke (Egger et al., 2016) and are higher than the average diffusive flux of methane from inner shelf sediments determined in a recent data compilation (0.57 mmol m^-2 d^-1; Egger et al., 2018). Methane also will have escaped from the sediment via ebullition. We base this on the fact that the porewaters were strongly oversaturated with CH₄ ( Figure 2G ) and on visual observations of the emergence of bubbles in the surface water during sampling, especially in September.

The fate of the CH₄ in the water column depends strongly on the extent of stratification of the water column and the resulting oxygen dynamics. In March, when the water column was fully mixed and oxygenated, CH₄ was removed aerobically near the sediment-water interface, likely both in the surface sediment and bottom water. In September, when the water column was stratified, CH₄ diffused out of the sediment and accumulated in the anoxic bottom waters to concentrations of ~73 µmol L^-1 ( Figure 4J ). These values are at the high end of the range of concentrations observed in waters of other eutrophic coastal environments. For example, maximum CH₄ concentrations of only ~1.5 µmol L^-1 and 0.3 µmol L^-1 were recently reported for bottom waters in Mariager Fjord and in Boknis Eck, respectively (Steinle et al., 2017; Steinsdóttir et al., 2022). In Chesapeake Bay, however, bottom water CH₄ concentrations are more comparable at values of ~ 40 µmol L^-1 (Gelesh et al., 2016). Importantly, the bottom water salinity at Scharendijke is much higher than that at the sites mentioned above (~31 versus 20-24). Thus, even in coastal areas with high sulphate concentrations, intense eutrophication can lead to a strong accumulation of CH₄ in stratified water columns.

Water column CH₄ can be removed through aerobic and anaerobic oxidation. At our study site, we found a distinct increase in δ¹³C-CH₄ and δD-CH₄ between 35 and 25 m in September ( Figures 4K, L ) pointing towards intense CH₄ oxidation near the oxycline ( Figure 6 ). The strong decrease in oxygen concentrations in this zone suggests that substantial CH₄ removal occurred via aerobic pathways. The isotopic fractionation factor (α) calculated for this zone assuming Rayleigh distillation was 1.008, which falls within the range reported for aerobic CH₄ oxidation (1.002 – 1.035; Grant and Whiticar, 2002). These results imply that 98% of methane is oxidized in the water column. However, we note that the isotopic signature of methane may be affected by ebullition and bubble dissolution. Hence the calculated fractionation factor and removal efficiency should be seen as approximations only. Importantly, N H 4 + removal also occurred near the oxycline, presumably through nitrification. However, no accumulation of N O 3 - was observed near the oxycline, which is likely due to its immediate reduction via denitrification (Venetz et al., 2022). Nitrate removal could also potentially be linked to anaerobic CH₄ oxidation as the zone of decreasing CH₄ concentrations overlaps with the zone where N H 4 + is removed and, hence, where N O 3 - is produced and reduced. The profiles of particulate Fe and Mn suggest that both Fe and Mn oxides are present in the zone of CH₄ removal and hence could contribute to CH₄ oxidation as well. Where metal oxides meet H₂S, Fe and Mn reduction coupled to CH₄ oxidation will be limited because of rapid metal oxide reduction with H₂S. Particulate Fe below the oxycline presumably consists mostly of FeS₂.

The mass balance calculations at the oxycline show that the downward flux of electron acceptors is lower than the upward flux of electron donors ( Table 1 ). This imbalance must be even larger as the calculations do not include two other electron donors: hydrogen sulfide and sinking organic matter. Such an imbalance could indicate an upward movement of the oxycline in the water column. However, the annual trends in water column oxygen ( Figure 1C ) show an opposite trend with a deepening of the oxycline in September followed by its disappearance in October when the water column becomes fully mixed. Hence, the most likely explanation is an underestimation of the downward flux of key electron acceptors, in this case, oxygen and metal oxides. For oxygen, occasional lateral influxes of more saline oxygenated waters from the North Sea (e.g. Hagens et al., 2015) or periods of more intense vertical mixing could play a role. Furthermore, we may have underestimated the sinking speed of the Fe and Mn oxides, since this is strongly dependent on the diameter and density of the particles and hence can strongly vary between systems (Neretin et al., 2003). In this case, a 4-fold higher sinking speed could explain the current imbalance in the budget. This also implies that there is potential for metal oxide reduction in CH₄ oxidation in the water column.

The strong oxidation signal in the CH₄ isotopes and decrease in CH₄ concentrations around the oxycline in September point to CH₄ removal. However, not all the CH₄ is removed since we find a diffusive flux of CH₄ to the atmosphere in a range of 0.09 to 0.37 mmol m^-2 d^-1. This atmospheric CH₄ flux range is higher than the range reported for the Boknis Eck (0.0025 – 0.0062 mmol m^-2 d^-1 in September 2014; Steinle et al., 2017) but is comparable to that observed in the Belgian coastal zone in summer and fall (0.43 mmol m^-2 d^-1 and 0.07 mmol m^-2 d^-1, respectively; Borges et al., 2016). In addition to the diffusive flux, CH₄ is transported to the atmosphere via ebullition. Bubbles reaching the surface were frequently observed during the September sampling trip. Hence, the elevated concentrations of methane and slight depletion in δ¹³C and δD in the upper 6 m of the water column could be a result of bubble dissolution, although we cannot completely rule out an additional contribution from aerobic production of methane by phytoplankton (e.g. Karl et al., 2008). Regardless of the mechanism, we conclude that seasonally stratified, eutrophic coastal marine waters may be more important as potential sources of CH₄ to the atmosphere than generally thought (Naqvi et al., 2010; Steinsdóttir et al., 2022). Since eutrophication is increasingly leading to anoxia in coastal systems worldwide (Breitburg et al., 2018), further research on their role as a source of methane to the atmosphere is warranted.

Methylomonadaceae ( Figure 5A ) were most abundant near the oxycline indicating that these microorganisms could remove CH₄ at the expense of oxygen or other electron acceptors. In the upper and lower parts of the oxycline one would expect Methylomonadaceae with a high affinity to CH₄ and oxygen, respectively. In the middle of the oxycline, more versatile and flexible Methylomonadaceae would likely be dominant with intermediate affinity to both O₂ and CH₄ (Steinsdóttir et al., 2022). In all cases, the methanotrophs would have to compete with heterotrophs for available electron acceptors. Below the oxycline, the methanotrophs could switch to other modes of metabolism using N O 3 - or metal-oxides as terminal electron acceptors, reserving the limiting amounts of oxygen to activate CH₄ to methanol (Kits et al., 2015; Zheng et al., 2020). Methanol or other fermentative products could serve as an electron donor for N O 3 - respiration either in full or partial denitrification. Hence, aerobic methanotrophs could be involved both in CH₄ and N O 3 - removal near the oxycline.

Despite the presence of N O 2 - at some depths, no 16S rRNA gene sequences of the intra-aerobic Methylomirabilis were detected although their presence in the water column of marine oxygen minimum zones has been reported (Padilla et al., 2016; Thamdrup et al., 2019). At and below the oxycline, however, 16S rRNA gene sequences of archaeal methanotrophs (Methanoperedenaceae) were identified ( Figure 5B ), indicating that in the anoxic waters, CH₄ removal might be coupled to reduction of N O 3 - or Fe and/or Mn oxides, although the contribution to total CH₄ removal is likely small. Some of the produced N O 2 - and/or N H 4 + might be used by anammox bacteria, although only a few 16S rRNA sequences of anammox were recovered in the amplicon set (Venetz et al., 2022). To our knowledge this is the first report of Methanoperedenaceae in marine waters in a zone of CH₄ oxidation. The source of CH₄ in the water column, the pathways of CH₄ removal and key microorganisms involved, namely Methylomonadaceae and Methanoperedenaceae, are summarized in a schematic ( Figure 7 ).