The survey was performed on board of the R/V Mare Nigrum, operated by GeoEcoMar Romania, in April 2019 at two shallow sites in the Black Sea Romania territorial water. Over a period of 5 days (3–7 April), we surveyed an area of ∼ 12.5 km² at the deeper site (44.233°N, 30.737°E, 100-km long survey), with water depths between 110 and 128 m, and ∼ 3.5 km² at the shallower site (44.057°N, 29.491°E, 19-km long survey), with water depths between 53 and 58 m (Figure 1).

At the deeper site, water temperature and electrical conductivity were on average 9.5°C and 18.5 mS cm⁻¹ at the sea surface, and 9.0°C and 20.5 mS cm⁻¹ near the seafloor. At the shallower site, temperature and electrical conductivity were 9.5°C and 18 mS cm⁻¹ at the sea surface and 8.0°C and 18.5 mS cm⁻¹ at the seafloor. The deeper site showed an oxycline between 60 and 80 m water depth, while the concentration of dissolved oxygen near the seafloor at the shallower site was 17% lower than at the surface (360 mmol L⁻¹) (CTD, Conductivity, Temperature and Depth, data are reported in the Supplementary Datasheet S1).

Two near-surface horizontal profile surveys (HP01, mean depth 5.2 m, min 0.4 m, max 18 m, and HP03, mean depth 4.2 m, min 1.5 m, max 9 m) on dissolved CH₄ were conducted. HP01 was performed for 11 h at the deeper site on April 4th and is composed of ten parallel ∼5.5-km long lines spaced by ∼260 m, for a total surface area of ∼ 12.5 km². HP03 was a 2-h long survey at the shallower site performed on the April 6th, covering an area of ∼3.5 km² (see Figure 1). During these two surveys, the MILS and METS sensors were deployed simultaneously. The MILS probe was configured to improve the sensitivity of the measurements, by minimizing the flow of carrier gas (see the method description below for further details). This decreased the dynamic range of the measurement, while increasing the precision at low concentrations, to the detriment of a slightly longer response time (t₉₀ of 30 s, corresponding to 75 m resolution at the highest speed of 2.5 m s⁻¹ reached during the surveys).

Vertical profiles (VPs) with the MILS and METS sensors were performed at different locations at the two sites. Because of the lacking in dynamical positioning, the vessel was located in the vicinity of a flare (hydroacoustic anomalies in echosounder records attributed to the presence of gas bubbles) or a cluster of flares, and down- and up-casts were performed. It should be noticed that because of the strong dependency of the METS sensor to dissolved oxygen content and change in hydrostatic pressure, the recorded vertical profiles are not reported in this work.

Hydro-casts (HYs) for discrete water sampling were conducted at different time with respect to in situ measurements. The locations should have been the same as the in situ measurements, but this has proven to be challenging due to the lack of dynamical positioning of the vessel.

During the cruise, a 70 kHz split beam echosounder (Simrad EK80 with ES70 transducer) was used to hydro-acoustically detect and locate bubble releasing methane seeps. With an opening angle of 18°, it has a footprint at the seabed of ∼22 m and ∼48 m diameter at 55 m and 120 m water depths, respectively. The pulse length was set to 0.256 ms over the entire cruise. This turned out to be suitable for the shallower site, while for the deeper site the noise level remained visibly higher. However, since this noise level was acceptable, for a better inter-comparability between the two study sites, the pulse length was unchanged for both surveys. At an average vessel speed of ∼2 m s⁻¹, the distance between two pings was around 32 cm (14 cm) for 120 m (55 m) water depth at a ping rate of 0.16 s (0.07 s). To obtain precise backscatter values for the bubble quantification method, the echosounder was calibrated prior to the cruise with a 38.5 mm Tungsten sphere for the applied pulse length (MacLennan and Svellingen, 1989).

The idea to detect gas seepage locations using echo-sounding techniques was adopted from a series of former studies and has been accepted as an efficient approach to identify submarine gas flares. The method has been described in detail in e.g. Greinert et al. (2010) and Veloso et al. (2015).

A membrane inlet laser spectrometer (MILS) prototype allowing the combination of fast response and in situ dissolved methane measurements was used. The instrument relies on a patent-based membrane extraction system (Triest et al., 2017). It is described in detail in a previously published paper (Grilli et al., 2018), where a laboratory comparison with measurements of discrete water samples at different water temperatures and salinities was conducted. It was deployed successfully during two campaigns, over a methane seepage area in western Svalbard in 2015 (Jansson et al., 2019b), and at Lake Kivu in Rwanda in March 2018 (Grilli et al., 2020). The instrument allows a dynamic range from a few nmol L⁻¹ up to a few μmol L⁻¹. The MILS was powered by a battery pack (SeaCell, STR Subsea Technology and Rentals) and deployed together with a CTD SBE 911plus (Seabird) for measuring temperature, conductivity and water depth.

Because of the dynamic and fast profiling capability of the instrument, the spatial and temporal synchronization of measurements needs particular attention. For this, a first dynamic correction of the time-lag due to the flushing time (the time it takes the gas sample to go from the membrane extraction system into the measurement cell) was applied. By knowing the total gas flow (sum of the carrier gas plus the dry and wet gas permeating the membrane) and the volume of the gas line between the extraction system and the measurement cell, this time lag was calculated and varied between 15 and 60 s during the campaign (depending on the total gas flow). The instrument response time is related to the time necessary to replace the gas inside the measurement cell; it depends on the measurement cell volume (∼20 cm³ at standard temperature and pressure), the working pressure (20 mbar) and the total gas flow. This parameter can as well be calculated and it varied between 8 and 30 s during the campaign. Both time-lag and response time are affected by the total gas flow (Grilli et al., 2018). With the instrument towed behind the vessel, the distance between the instrument and the ship also varies as function of the ship speed and water depth of the sensor. Instrument location was therefore dynamically corrected simulating the mooring of our sensor using the “Mooring Design and Dynamics” matlab routine (Dewey, 1999). This dynamic correction allows to apply a time (and therefore a position) correction of the sensor which ranged from a few seconds and a few meters at the sea surface and at minimum speed, up to 100 s when the system was towed at 100 m water depth. This corresponds to maximum horizontal correction of ∼80 m, since typical ship speed during vertical profiles was ∼0.8 m/s. Water currents were neglected for this position correction. The vessel position was provided by a Garmin GPS 18x, with an accuracy of 15 m (1σ).

Measured concentrations are reported in mixing ratio with respect to the total dissolved gas pressure, which is assumed to be 1 atm for this setting. Therefore, a value of partial pressure, pCH₄, in the gas mixture can be retrieved, which is then converted into dissolved methane concentrations, C_CH4, expressed in mol per liter of water. This conversion is performed by considering the solubility of the gas in the water under given physical conditions as well as its fugacity. C_CH4 is related to the pCH₄ through the following equation: C CH 4 = K ( T , S , P ) p C H 4 φ CH 4 ( T , P ) , (1) where φ _CH4 is the fugacity coefficient (assumed to be one in this case) and K is the solubility coefficient, i.e. the ratio between the dissolved methane concentration and its fugacity. The solubility coefficient K (mol L⁻¹ atm⁻¹) of CH₄ as a function of temperature T (K) and salinity S (g/kg) is calculated using the following equation: ln ( K ) = A 1 + A 2 ( 100 T ) + A 3 ⁡ ln ( T 100 ) + S [ B 1 + B 2 ( T 100 ) + B 3 ( T 100 ) 2 ] , (2) Where A_i and B_i are empirical parameters from Wiesenburg and Guinasso. (1979).

The solubility coefficients need to be corrected for local pressure P (Pa) at the sampling water depth (sum of hydrostatic and atmospheric pressure), using the following equation (Weiss, 1974): K ( P ) = K e [ ( 1 − P ) v CH 4 R T ] , (3) where R = 8.31446 J mol⁻¹ K⁻¹ is the ideal gas constant, and ν_CH4 is the partial molar volume (cm³ mol⁻¹) of CH₄ calculated from Rettich et al. (1981).

Calibration of the instrument was performed in the laboratory using the calibration system described in Grilli et al. (2018). Experiments were performed at atmospheric pressure for different water conditions (temperature from 5 to 25°C, and salinity from 0 to 30.5 g/kg) and at different concentrations of CH₄ (0–1,000 ppm). The gas mixtures were obtained using two mass flow controllers (Bronkhorst, EL-FLOW) and mixing zero air (ALPHAGAZ 2, Air Liquide) with synthetic air containing CH₄ (8920 Labline, 1,000 ppm of CH₄ in air, Messer).

A Franatech METS sensor was used to measure anomalies in a concentration of dissolved methane in water. The measurement principle is based on a SnO₂ semiconductor detector (Seiyama et al., 1962) working at ∼500°C (Ippommatsu et al., 1990). Its principles can be summarized as follows: first, oxygen (O₂) is absorbed on the SnO₂, then, the dissolved CH₄ diffuses through a membrane to the measurement cell and interacts with O₂ molecules, causing their desorption and increasing the conductivity of the SnO₂ material. This technology is however known for its lack in gas selectivity and its dependency on the amount of O₂ present in the measured environment (Boulart et al., 2010; Chua et al., 2016). The METS sensor can be operated at water depths down to 4,000 m and temperature ranging between 2°C and 20°C. Prior to its deployment, the sensor was calibrated by the manufacturer (in February 2019) at atmospheric pressure and methane concentrations ranging between 100 nmol L⁻¹ and 40 µmol L⁻¹. Although the sensor can be operated over a larger methane concentration range, the manufacturer calibrated the sensor in a narrower range in order to preserve its linear response (Franatech Pers. Comm.).

Discrete water sampling was conducted using a CTD-rosette with 16 Niskin sampling bottles (8 L), a SBE 911plus CTD (Seabird), an altimeter (Teledyne Benthos PSA 916), and an oxygen optode (Aanderaa Optode 4831F). The sensors were connected through telemetry for real-time monitoring of the water depth of the assembly. This allowed to adjust the sampling strategy during the profile according to the anomalies recorded by the echosounder. The Niskin bottles were sampled during the up-cast, and water subsampling was performed onboard for laboratory gas analysis (both PT and HS). For all the samples, a few mg of sodium azide (NaN₃) were added to the vials and glass bulbs before adding the water sample. Filled vials for HS analysis were then stored upside down.

The samples were analyzed by two different techniques: the purge and trap (PT) and the headspace (HS) techniques. These methods have already been compared and validated in the laboratory (Donval and Guyader, 2017). The PT method used here is based on Swinnerton et al. (1968) and modified by Luc Charlou et al. (1988). Briefly, 125 ml glass bulbs devoted to the analysis of methane by PT method were used. The bulbs were overflowed with twice the volume of seawater. Particular care was taken to exclude air bubbles during sampling and avoid contamination. Once in the laboratory, CH₄ was stripped from seawater for 8 min using helium as carrier gas (quality 99.9995%), trapped on activated charcoal at −80°C, and detected and quantified with a flame ionization detector after separation on a packed column (GC Agilent 7890A/column Porapak Q 2 m). The calibration was performed by injection of gas standards containing 108 ppm of CH₄ in air ±5% (Restek). The limit of detection was 0.03 nmol L⁻¹, the precision based on five replicates from the same rosette bottle was within ±2% (±1σ), while the accuracy corresponds to that of the gas standards (±5%). The HS method (Donval et al., 2008; Donval and Guyader, 2017) was performed on 10 ml vials by analyzing the composition of the headspace in equilibrium with the water. At the beginning of the cruise, the vials were flushed with zero air (Alphagaz 2, Air Liquide) to avoid introducing methane into the initial gas phase. With a gas tight syringe, 5 ml of seawater were transferred into the vial while a second needle was introduced to keep the pressure close to atmospheric pressure inside the vial. The analysis was performed by means of a headspace sampler connected to the same GC unit used for the PT method. The limit of detection was 5 nmol L⁻¹ and the precision was ≤10% for concentrations below 100 nmol L⁻¹, and ≤5% for higher concentration (±1σ). Further details in the comparison between the results of two methods can be found in the Supplementary Datasheet S1.

In this section we compare results from different techniques with the aim of testing the reliability of the measurements, identifying possible artefacts or weaknesses of each technique used, and validating the finds on the dissolved methane distribution in the water column at the studied sites. This data validation is important for comparing the distribution of dissolved methane at the two locations as reported in the discussion section.

On the echogram, large areas of high backscatter (>−30 dB) of a flare-like shape have been identified as methane gas seeping areas (Figure 2, color-coded in orange to red). Broader high backscatter areas are related to vessel movement. In our case, the ship remained over the methane seep location for some time and as the echogram displays backscatter over time, the seepage area widens. During the cruise, 36 and 13 gas emissions in the deeper and shallower site, respectively, were identified from the 70 kHz echosounder data. The signal produced by the scattering of the acoustic wave by the gas bubbles provides only qualitative information about the distribution of dissolved CH₄ in the water column. This is due to the fact that gas bubbles shrink and change their chemical composition during their rise through the water column, and that free bubble-forming gas is mobile whereas dissolved methane is more stationary. Therefore, a quantitative analysis would require different assumptions on the initial bubble size distribution, the bubble rising speed and the gas exchange ratio between the two phases. Results from the quantitative analysis of the acoustic signal is beyond the scope of this paper, and will be the subject of another study. Here, an example of the qualitative comparison between the acoustic signal and the dissolved CH₄ concentrations from MILS is reported in Figure 2 for the VP03 profile performed on April 5th at the deeper location (44.224° N, 30.730° E). The concentrations of dissolved methane determined by MILS were dynamically corrected for the position of the instrument with respect to the vessel. This correction allows to synchronize/match the two time-series, accounting for the fact that the echosounder passed over a target area prior to the towed MILS. A more detailed figure reporting original and synchronised data can be found in the Supplementary Datasheet S1.

In this inter-comparison, one should consider that the MILS probe is measuring a specific location behind the vessel, while acoustic data has a larger footprint on the seafloor as well as in the water column. This may induce discrepancy between the two signals, for instance in the case of a bubble plume located a few tens of meters on a side of the MILS instrument that would be spotted acoustically but not observed by the MILS (this could be the case for the signature at 12:22 local time in Figure 2, more visible in Supplementary Figure S1-3). The results may inverse if water enriched in CH₄ (by a bubble plume that is outside the acoustic lobe) is laterally transported by currents. This may be the case for the increase in dissolved CH₄ at 12:28 local time (visible in Supplementary Figure S1-3) that was observed by the MILS sensor without the corresponding acoustic signal.

Despite its evident limitations, this comparison allowed us to verify the qualitative agreements between these two datasets and validate the dynamic correction of the position of the MILS probe with respect to the vessel during the profiles.

For comparison, the MILS and the METS sensor were deployed simultaneously during the vertical and horizontal profiles. However, due to dependency of the METS sensor to the hydrostatic pressure, salinity and oxygen (Newman et al., 2008), this comparison only focus on the 12-h long horizontal profile (HP01). The profile was conducted at ∼5 m depth, and the water inlets of the two sensors were ∼10 cm distance from each other. The resulting profiles used for comparison are reported in Figure 3. One can clearly see that most of the peaks in methane concentration for both instruments agree with each other. However, the methane concentrations measured by the METS are more than one order magnitude higher than the concentrations obtained with the MILS. Previous field studies have shown significant differences between measurements of discrete water samples of seawater methane concentration from well-proven laboratory methods and the METS outputs (Heeschen et al., 2005; Newman et al., 2008). This has led Heeschen et al. (2005) to interpret their METS in situ measurements in a qualitative way. Here, the differences observed between the MILS and the METS measurements may be explained by: 1) the fact that the METS sensor was used below the calibration range (100 nmol L⁻¹–40 μmol L⁻¹ at atmospheric pressure) certified by the manufacturer and therefore it cannot provide reliable quantitative measurements at the sea surface; 2) the METS suffers from dependency to hydrostatic pressure, salinity and oxygen content (Newman et al., 2008), which makes near-surface horizontal profile also challenging. Moreover, the wide drifts observed during this 12 h continuous near-surface profile (Figure 3) can be explained by the fact that small height changes in near-surface depth lead to large relative changes in hydrostatic pressure that considerably affect the sensor response. Despite the large discrepancy on the methane concentration from the METS sensor, both in situ instruments agreed on the presence of highly localized peaks of CH₄ at the sea surface. The METS sensor, even when it is used outside the calibration range and under severe conditions (changing oxygen concentration that affect its detection system) can provide valuable qualitative information on the location of dissolved methane concentration spots. Further laboratory tests, together with deployment at deeper water depth within the anoxic layer, would be required to provide a thorough assessment of the sensor for quantitative and reliable dissolved methane measurements.

Eleven well localized dissolved CH₄ peaks (numerated and marked with stars in Figure 3) were identified during the horizontal profile HP01 and are discussed in the next section. Furthermore, the sharp concentration decrease recorded at 01:35 local time (highlighted in Figure 3 by the red arrow and also visible in the 2D colored map graph of the dissolved methane concentrations measured by the MILS instrument and reported in Figure 4) that could have been associated with a possible instrumental (MILS) drift, was confirmed as a real signal since observed by both in situ sensors.

The MILS vertical profiles recorded at the two sites during the campaign are reported by grey and orange dots in Figures 7A,B. For the deeper site, a total of four down- and up-casts (VP01, 02, 03, and 04) were used, whereas for the shallower location, the vertical profile VP05, composed of a series of seven down- and up-casts over the seepage area was used. The 2-m average curves are in black and blue for the deeper and shallower sites, respectively. Because the gas bubbles originate from the seafloor, the vertical profiles were compiled in distance from seabed (rather than in water depth), allowing a better comparison of the vertical distribution of the dissolved CH₄ within the two sites. For each datapoint, the distance from the seafloor was calculated using the depth of the seafloor measured by the echosounder and the water depth of the instrument provided by MILS-implemented CTD and the MILS instrument itself. The systematic decrease in dissolved CH₄ concentration near the seafloor may be due to the fact that the position of the vessel was not dynamically maintained. In most of the vertical profiles recorded, the instrument passed over the bubble plume either during its descent or ascent (this is visible in the time series reported in Figure 2). The MILS instrument was therefore within the uppermost part of the gas bubble plume most of the time a few tens of meters above the seafloor. For a finer visualization of this effect, the stronger profile recorded by the MILS at the deeper location (corresponding to the VP04) is reported in orange dots (Figure 6A). The two black arrows indicate the time direction during the descent. The maximum dissolved CH₄ concentration of 924.5 nmol L⁻¹ was reached at 15 m from the seafloor (or 93 m water depth). Then, despite the probe continuing its descent, the concentration decreased, probably due to the fact that the probe was moving out of the bubble plume. Another possible reason for this trend may be directly related to bubble dissolution. The seafloor topography can influence bottom current (Weber et al., 2000; Stow et al., 2009), which in turn affect bubble trajectory in the water column and shape the plume morphology. Hence, the rim of the bubble plume is widened few meters above the seafloor, enhancing the spreading of dissolved CH₄ at this height, and placing the maximum level of dissolved methane concentration further above the seafloor. Lastly, note that the distance between the instrument and the ship was calculated by mooring simulation (as mentioned in the materials and methods section) which may add uncertainty on the position of the MILS sensor with respect to the one of the bubble plumes. From the maximum concentrations of the averaged curves, a difference in emission intensity of ∼80 fold between the shallower and the deeper site was estimated.

In Figure 6C, all the vertical profiles were averaged producing one data point every 2 m water depth, and the concentration of dissolved CH₄ was normalized with respect to the maximum averaged dissolved CH₄ concentration. At both sites, an exponential trend was observed which follows the expected dissolution function of the bubbles into the water column (Jansson et al., 2019a). By fitting the exponential curves on the whole vertical profile, exponential time constants (corresponding to the distance from the seafloor required for decreasing the intensity by 1/e – e-folding – of the value at the sea bottom) of 6.8 and 6.3 m were obtained for the deeper and shallower sites, respectively. They are close, but the difference remains visible in Figure 6C, with a faster decreasing in concentration of the profile at the deeper site while moving away from the seafloor. This emphasizes the larger storage capability of dissolved CH₄ at the bottom water level of the deeper site (i.e. a better tendency of CH₄ to escape towards to the sea surface at the shallower site). The reason for this difference is likely a combination of factors, including: 1) the difference in hydrostatic pressure and bubble-size distribution, leading to a different bubble/water exchange (more precisely related to the difference between the buoyant rise time of the bubble and its diffusive equilibrium time (Leifer and Patro, 2002)); 2) decoupling between bottom and surface water at the deeper location marked by the presence of the oxycline, which prevents the rise of CH₄ towards the sea surface; 3) a possible local production of CH₄ in the anoxic water of the deeper site (Artemov et al., 2019), which would increase the concentration of dissolved CH₄ below the thermocline. Discriminating between the different scenarios would involve a more intensive investigation of methane distribution in all three water layers, combining a larger number of horizontal and vertical profiles, and both molecular and isotopic concentration measurements of CH₄; this was not performed during this campaign. From this analysis, we can conclude that at the deeper site, going from the seafloor towards the surface, the dissolved CH₄ rapidly decreases within the first 45 m (∼7 times the exponential time constant), and remains uniform in the upper part. On the other hand, at the shallower site, dissolved CH₄ concentrations corresponding to ∼20% of the maximum average dissolved CH₄ concentration on the water column are still present at the sea surface (at ∼52 m from the seafloor).

Two near-surface horizontal profiles were conducted during the cruise: HP01 at the deeper site and HP03 at the shallower site. The 2D distributions of dissolved CH₄ are reported in Figure 4. The average concentration at the deeper site was 2.23 ± 0.78 nmol L⁻¹ (1σ, min 0.78, max 5.16 nmol L⁻¹). For the measured temperature and electrical conductivity of the surface water (9.5°C and 18 mS cm⁻¹), the atmospheric CH₄ concentration of 2 ppm (part per million) would correspond to an equilibrium dissolved CH₄ concentration of 3.5 nmol L⁻¹ (calculated from Equations 1, 2 and 3). This is slightly higher than the average concentration measured by the MILS, but it still lays within the range of the measurements made by the MILS sensor near the surface. On the other hand, from the comparison with measurements of discrete water samples (Figure 6), concentrations measured by the MILS seems to be systematically ∼2 nmol L⁻¹ lower. We can therefore conclude that at the deeper site, dissolved CH₄ in the water is close to equilibrium with the atmosphere. At the shallower site, the average dissolved CH₄ concentration was 5.6 times higher than observed at the deeper site (average concentration 12.5 ± 2.76 nmol L⁻¹ 1σ; min 5.16, max 19.7 nmol L⁻¹) and almost four times higher than the expected concentration in equilibrium with the atmosphere. These concentrations are close to previously reported measurements by Amouroux et al. (2002) and Reeburgh et al. (1991) in north-western and central part of the Black sea, respectively. This, together with what we observed in the vertical profiles (i.e. the fact that, at the shallower site, 20% of the amount of dissolved CH₄ present at the bottom was found near the surface), confirms the occurrence of methane fluxes from the oxic layer to the atmosphere at shallow water depths, which is also in agreement with previous findings (Reeburgh et al., 1991; Schmale et al., 2005; McGinnis et al., 2006). The difference in dissolved CH₄ concentration at the sea surface of the two sites may support the hypothesis of an efficient vertical transport of dissolved methane from the seafloor up to the surface at shallow depths. Alternatively, other unidentified sources may be the cause of this oversaturation of the surface water, either through lateral diffusion or CH₄ generation in the upper water layer.

Average concentrations from measurements of discrete water samples (reported in Table 1) were obtained from only three data points at the deeper site and two at the shallower, with values of 6.6 and 15.1 nmol L⁻¹, respectively. This confirms the results of the in situ technique showing a generally higher concentration of methane at the shallower location.

From the near-surface dissolved CH₄ data of the deeper site, eleven isolated peaks were identified (light-blue dots in Figure 4 corresponding to the peaks marked with stars in Figure 3, also numerated on both figures) with peak intensities ranging from 0.6 to 2.3 nmol L⁻¹ and average width of ∼500 m (full width at half maximum, FWHM). All the selected peaks were observed by both in situ instruments except for the second peak in the time series, which was recorded only by the MILS instrument but with a peak intensity of 1.26 nmol L⁻¹ and therefore a relatively good signal to noise ratio. Peaks observed by the METS but not by the MILS sensor were disregarded, since they may be due to a measurement artefact because of the low selectivity of the sensor. The intensities were calculated after subtraction of the background concentration, by means of a Gaussian fit. The two stronger and larger peaks (FWHM of ∼1,500 and ∼640 m) are the N° 3 and 7, both located at the southern edge of the grid survey. These events at the sea surface are however difficult to correlate with the locations of the identified flares (red dots in Figure 4), and the mismatch suggests that: 1) the vertical transport of dissolved CH₄ from the seafloor up to the surface may strongly be affected by lateral transport of methane through currents; 2) other factors or other unidentified sources (i.e. microbial activities) may play a role in the occurrence spots of high methane concentration near the sea surface. Despite our achievements, we are still far from computing all the processes for conclusively assessing the fate of methane in the Black Sea water column. The fact that the shallower site has a higher dissolved CH₄ concentration at the sea surface with respect to the deeper site agrees with previous findings (Reeburgh et al., 1991; Amouroux et al., 2002). This reinforces the hypothesis of methane transport from the seafloor to the atmosphere at shallow sites, although contributions of other sources cannot be firmly discarded. This is further supported by the fact that the methane oxidation rate is lower in the oxic layer (Reeburgh et al., 1991), promoting its persistence in the water during the ascent. Our results agree with Schmale et al. who, in 2005, concluded that only shallow seeps (depths <100 m) seem to affect the methane concentration of surface water and direct local emissions into the atmosphere, while high-intensity seep sites below this boundary do not show regional influence on the surface layer. Greinert et al. (2010) highlighted how methane fluxes rapidly increase with increasing bubble size and decreasing water depth, underlining that shallow sites may represent a significant source of methane to the atmosphere. Other factors have been proposed to explain marine methane transfer to the atmosphere. Indeed, Shakhova et al., 2017 and Pohlman et al., 2017 also discussed the large release of CH₄ from the sediments at shallow sites in the Arctic regions and its potential impact on the climate. They claimed that upwelling and erosion are, for instance, possible mechanisms promoting gas transfer to the atmosphere.