Physical characterization of the water column

For the investigations of the physical properties of the water column during M87/4, we performed studies in the upper 100 m with an SBE 911plus conductivity, temperature, depth (CTD) system (Sea-Bird Electronics, U.S.A.) consisting of a SBE 9plus CTD unit with redundant sensor packs for CTD and oxygen and a SBE 11plus deck unit. Fluorescence was measured with a FLNTURTD sensor from Wet Labs (U.S.A.). Real-time data acquisition and post-measurement data processing were performed using the SBE Seasoft V2 software package. The sensors were mounted on an SBE 32 carousel water sampler with a set of 24 × 10 L free-flow bottles from Hydro-Bios (Germany) for water column sampling. A similar system was used on RV Elisabeth Mann Borgese in August 2012.

In addition, high-resolution microstructure measurements were undertaken with a free-falling MSS90-L microstructure profiler (SN 055) to obtain vertical eddy viscosity profiles. The instrument was equipped with precision CTD sensors for pressure, temperature, and conductivity, two airfoil shear probes (PNS06), and a SeaPoint turbidity sensor. Sampling frequency was 1024 Hz for all sensors. The instrument was operated in free-falling mode with a sinking velocity of 0.5–0.6 m s⁻¹. The two shear probes provided estimates of the vertical current shear on a millimeter-scale, from which the dissipation rate of turbulent kinetic energy (TKE) was derived. The vertical eddy diffusivity or eddy viscosity (A_z) was estimated according to Osborn (1980) as follows:

(1)

where A_z is the vertical eddy diffusivity, Γ is the mixing efficiency, ɛ is the TKE dissipation rate, and N is the stratification of the water column expressed as Brunt-Väisälä frequency. For the mixing efficiency Γ, the commonly accepted value of 0.2 was used. The technical details of the micro structure probe (MSS) profiler and for the data processing are described in Prandke and Stips (1996) and Prandke et al. (2000).

The vertical eddy viscosity profiles were combined with a high resolution methane profile taken on 18^th July to exemplarily calculate the vertical turbulent methane transport in the EGB (see “Discussion” section). The corresponding four microstructure profiles were gathered within 2 h after the water sampling for methane analysis. To minimize the impact of short-term vertical displacement by internal waves, averaging of the eddy viscosity profiles and fitting to the CTD profile was performed along the density vector.

Water column methane analysis

Methane concentrations and their stable carbon isotope ratios were analyzed to identify methane enrichments in the upper water column and to discuss their origin. Water samples were collected in high resolution in the upper water column every 3 m with the carousel water sampler. For methane concentration measurements, water samples (600 mL) were taken directly from the free-flow sample bottles and transferred into pre-evacuated 1100 mL glass bottles. Within 3 h after sampling, the dissolved methane was extracted from the water samples, using a partial vacuum extraction line, and subsequently analyzed with a gas chromatograph equipped with a flame ionization detector (average precision of the method ± 3%, Schmale et al. 2010a). An aliquot of the gas sample was transferred into 10 mL vials, and the carbon isotope ratio (δ¹³C CH₄) was determined with online isotope-ratio mass spectrometry (average precision of the method ± 1‰, Schmale et al. 2012). Methane analysis of the water column during the sampling time of the microbiological samples was performed using a purge and trap method, see Jakobs et al. (2014).

Surface-water methane concentration and sea–air fluxes

During the entire cruise, we employed a system for the continuous measurement of the partial pressures of methane and carbon dioxide based on coupling of an off axis integrated cavity output spectrometer to an air–water equilibration unit. Only the methane measurements are used in this study. The principle of operation, the sensor/equilibrator setup, and further data processing have been described in Gülzow et al. (2011). In this study, data acquisition of the sensor/equilibrator system data together with the ship's UTC time stamp and GPS coordinates were undertaken at 0.2 Hz. For T-correction, the data were merged with the ship's thermosalinograph data, and all data were bin averaged over 1 min. For calibration, we used a standard gas of 1.975 ppm methane in natural air, fine calibrated at the Max Planck Institute for Biogeochemistry in Jena (now integrated carbon observation system [ICOS] Central Analytical Lab, Flask and Calibration Lab), which was analyzed several times during the cruise. Between 30^th June and the end of the cruise, the instrument was not recalibrated, as control measurements with the standard gas were always within 50 ppb of the target value. The accuracy of the method was better than 1%, while the precision was better than 0.02 nM.

Methane sea–air fluxes were quantified by (1) deriving the equilibrium concentration with the ambient atmosphere of 1.895 ppm (NOAA station Barrow for July 2012; https://www.esrl.noaa.gov/gmd/dv/data/index.php?site = brw), (2) calculating the concentration difference between sea surface and equilibrium concentration, and (3) applying the air–sea exchange parameterization of Wanninkhof et al. (2009). The wind speed that is used in this calculation was derived from the ship's meteorological station. Details of the calculation are given in Gülzow et al. (2013). Methane air–sea fluxes were calculated for each individual (1 min) data point, followed by averaging all data within the 1 h time intervals.

Plankton studies

Phyto- and zooplankton samples were used to identify differences in the plankton community structure between the two study sites that could help to explain variations in the development of the shallow methane anomaly (de Angelis and Lee 1994).

Phytoplankton was sampled in the upper water column in both basins (WGB down to 15 m and EGB down to 45 m water depth) with the carousel water sampler from discrete depths. The water was filled into glass bottles (200 mL), preserved with 1 mL acetic Lugol solution and analyzed for phytoplankton composition and cell-carbon biomass according to the Utermöhl method as described in the manual of HELCOM (2012).

Zooplankton was sampled from both basins above the thermocline (20–0 m) using a WP2 net (mouth opening 0.25 m², mesh size 100 μm) during nighttime according to HELCOM (2012).

The zooplankton accumulation below the thermocline and the dial vertical migration of the zooplankton organisms were monitored with hydroacoustic studies using a broadband R2SONIC 2024 with 200–400 kHz frequency and variable pulse length (Cox et al. 2013). These studies were targeted on the seston fraction > 100 μm, typical for the Baltic Sea zooplankton (HELCOM 2012). To cover the entire cruise period, we recorded the back scatter signal by means of continuous video monitoring. From these video data, we extracted 5 min mean profiles of the water column down to 105 m. The upper 6 m were discarded to remove the ship disturbances. Due to variable weather/ocean conditions, we had continuously to recalibrate the back scatter signal to keep the signal-to-noise ratio constant. Therefore, the resulting data allow for qualitative analysis only.

Microbiological investigation

Microbiological studies were conducted on samples collected on expedition 06EZ1213 on RV Elisabeth Mann Borgese in the WGB and EGB to identify methanogenic Archaea in the shallow oxygenated water column.

Water samples of 1 L were filtered onto 0.22 μm pore-size polycarbonate filters (Millipore GVWP, 47 mm diameter, Darmstadt, Germany), immediately frozen in liquid nitrogen and subsequently stored at −80°C until further analysis. For the nucleic acid extraction, samples from the upper water column that showed highest methane concentrations were used. For analyses in the EGB, a filter sample from 50 m depth (8.9 nM of ambient methane) and for the WGB, a filter sample from 40 m and 50 m depth (10 nM of ambient methane) were used.

DNA was extracted from the filters with an acidic phenol extraction (Weinbauer et al. 2002), using 30 μL of diethylpyrocarbonat (DEPC) H₂O in the final elution step. The DNA concentration was measured with a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, Delaware, U.S.A.). For the generation of Archaea specific 16S ribosomal ribonucleic acid (rRNA) polymerase chain reaction (PCR) products, 1 μL of DNA (∼ 40 ng) was combined with 1 μL of 10 pmol μL⁻¹ Met630F and Met803R primers (Hook et al. 2009), 0.125 μL Taq Polymerase (Thermo Fisher Scientific, St. Leon-Rot, Germany), 2.5 μL 10× PCR buffer (Thermo Fisher Scientific, St. Leon-Rot, Germany), 0.5 μL 40 mM dNTP mix (Thermo Fisher Scientific, St. Leon-Rot, Germany), 2.5 μL 25 mM MgCl₂ (Thermo Fisher Scientific, St. Leon-Rot, Germany), and 15.875 μL sterile nuclease-free water with a total reaction volume of 25 μL. The following PCR conditions were used: initial denaturation at 95°C for 15 min, followed by 40 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 30 s and extension at 72°C for 1 min, finishing with a final extension of 72°C for 10 min.

The PCR products were analyzed on a 1.2% agarose gel, stained with ethidium bromide (Boehringer Mannheim, Mannheim, Germany), to confirm the presence of a single, correctly-sized band. To clean the PCR products the NucleoSpinGel and PCR Clean-Up kit (Macherey-Nagel, Düren, Germany) was used following the manufacturer's instructions. 16S rRNA gene fragments were cloned using StrataClone PCR Cloning Kit (Agilent Technologies, Waldbronn, Germany) as described in the manufacture's procedures. From each clone library, random colonies were picked and for a total of 96 colonies the 16S rDNA gene insert was sequenced with the vector specific primer M13-24F-BLUE, using the MTP Sanger sequencing service (LGC, Berlin, Germany). The resulting sequences where checked afterward for chimeras with the DECIPHER's Find Chimeras web tool (Wright et al. 2012).

All sequences were submitted to the European Nucleotide Archive under accession numbers LT671514 through LT671576 (see Supporting Information Table 1). Raw sequence data was analyzed using BioEdit Sequence Alignment Editor (Hall 1999) and vector fragments were removed from each end of the sequence. All good quality sequences were aligned using the SINA Online Aligner (SILVA Incremental Aligner, v1.2.11) in the web interface of www.arb-silva.de/aligner/ according to Pruesse et al. (2012). The aligned partial clone sequences (∼ 190 bp) were added to the existing All-Species Living Tree (LTPs123, September 2015, Yarza et al. 2008) using parsimony in arb (version 6.0.2.; Ludwig et al. 2004).

Incubation experiments

Incubation experiments were carried out at the EGB station, to test our hypothesis that methane enrichments in the shallow oxygenated waters (in about 30 m water depth) are connected to a copepod-dominated seston fraction (> 100 μm). Animals for the experiments were sampled in the layer between 25 m and 45 m, where a maximum concentration of methane and seston particles was observed during daytime by the hydroacoustic measurements (see “Results” section). Specifically, we used a WP2 net (mouth opening 0.25 m², mesh size 100 μm; HELCOM 2012) equipped with a nonfiltering cod end that was gently towed (0.1 m s⁻¹) from 45 m to 25 m, closed and brought onboard. We pooled seston samples from a total of four tows for the incubation experiments to avoid dropping below the detection limit for the methane concentration measurements with our gas chromatographic method (0.1 nmol). After each tow, the animals were gently poured, without rinsing, into a 10-liter bucket filled with sea water from 35 m depth that was obtained from the prior CTD cast. After 1 h, a subsample of 3450 mL was decanted from the bucket into a 4000 mL beaker to reduce the amount of dead animals and particles. Fifty milliliters of this highly concentrated seston suspension was conserved in 4% buffered formaldehyde to analyze the zooplankton community structure in the 45–25 m layer, and the rest of the sample was used for the methane production experiments (see process chart in Fig. 2).

For the incubations, we prepared four different seston dilutions with increasing adult T. longicornis densities from the seston suspension sampled and concentrated from the 45–25 m depth layer. We quickly and gently added 2 × 3780 adult T. longicornis (equivalent to 2 × 300 mL of seston suspension, Bottle 5 and 6), 3 × 6300 adult T. longicornis (equivalent to 3 × 500 mL of seston suspendion, Bottle 7, 8, and 9), 1 × 7560 adult T. longicornis (equivalent to 1 × 600 ml of seston suspension, Bottle 10), and 1 × 8820 adult T. longicornis (equivalent to 1 × 700 mL of seston suspension, Bottle 11) to 2.3 L polycarbonate bottles (Nalgene®) using a 1 L graduated cylinder. The bottles had been half-filled with water from 35 m depth before the animals were added, were filled completely after the respective animal suspension was added and were subsequently closed with Teflon coated butyl-rubber septa caps (Montoya et al. 2004; Voss et al. 2006). The vitality of the animals was controlled daily by visual inspection of the animals' movement in the incubation bottles and the absence of detritus accumulations on the bottom of the bottles. A sterile filtered control sample (filter size 0.2 μm, Bottle 1 and 2) as well as a natural water sample obtained from 35 m water depth (including undefined microplankton and particles of a size < 100 μm, Bottle 3 and 4) were included in this incubation experiment. A defined headspace volume of 50 mL (atmospheric air) was added to the gas-tight sample bottles and the bottles were stored in the cold store at in situ temperature (about 6°C). The bottles were gently shaken several times over the day to limit the development of an oxygen gradient in the bottle. Methane concentrations were measured daily by taking a gas sample (1 mL) from the headspace and determining the methane mole fraction in the headspace by gas chromatography (for method details, see above and Schmale et al. 2010a). A pressure balance was achieved by the injection of an equivalent injection of filtered sea water. The total methane content in the incubation bottles was calculated using the solubility coefficients established by Wiesenburg and Guinasso (1979). At the beginning of our experiment, the water samples were equilibrated with the head space of the incubation bottles and the methane concentration was then determined. This value was then subtracted from the concentrations measurements that followed in order to obtain the methane accumulation over the course of the experiment for each incubation bottle. Oxygen concentrations in the incubations were controlled during the experiment using an oxygen microprobe (Microx TX3, PreSens, Germany) to verify an oxic environment.

Mass balance approach

We developed a mass balance to assess if methane production rates from T. longicornis can maintain the methane enrichment below the thermocline and the flux at the air–sea interface. These calculations were exemplarily conducted on the dataset obtained on the 18^th July, comprising of a high resolution methane profile, a high resolution eddy diffusivity profile, and methane production rates derived from our seston incubation experiments. The vertical methane transport (T_CH4) was estimated based on a nonadvective Fickian turbulent diffusion model on the gradient of measured methane concentration (∂C_CH4/∂z) and vertical eddy viscosity (A_z).

(2)

In this model, the upward (downward) directed transports are defined as positive (negative) values. Since our approach focuses on processes that sustain the prominent subthermocline methane peak, calculations were based on a smoothed methane profile that neglects fine scale methane fluctuations. Assuming quasi steady conditions and no lateral fluxes, the divergence/convergence of the methane transport profile has to be balanced by local sources/sinks of methane. The net sources and sinks of methane (P_CH4) in the water column were calculated as divergence of the estimated vertical methane transport as follows:

(3)

Physical properties of the water column

The hydrographic conditions at the two main stations were characterized by the typical summer season stratification of the Baltic. The upper mixed layer had a temperature of about 16°C, and was separated from the underlaying cold intermediate water (about 4°C) by a strong seasonal thermocline between 15 m (WGB) and 25 m (EGB) depth. The halocline at 50–70 m depth represents the transition to the oxygen depleted deep-water layer of the central Baltic. Despite this general structure the particular stations in the eastern and WGB depicted some important differences (Fig. 3). In the EGB, a double thermocline was found, with steep gradients at 15 m and 25 m. This was caused by strong drop in wind speed 2 d before the sampling. The reduced mixing depth together with high surface heat flux established the secondary thermocline at 15 m depth. In contrast, in the WGB, the temperature gradient was weaker. Thus, the transition between the mixed layer and the winter water was smoother than in the EGB.

The strong stratification in the EGB was also reflected in the eddy viscosity profile (Fig. 4). Due to the low wind speed, the eddy viscosity was weak in the surface layer (10⁻⁴ m² s⁻¹). Below the surface mixed layer the dissipation of TKE was very low and close to the noise level of the MSS. Thus, the estimated eddy viscosity (Fig. 4) is an upper limit of the true value. The strong stratification at the thermocline reduced the eddy viscosity below 15 m by two orders of magnitudes to about 10⁻⁶ m² s⁻¹. At the halocline, a further drop in eddy viscosity to 10⁻⁷ m² s⁻¹ was observed.

Water column methane distribution

At both stations, the surface-water methane concentrations and the stable carbon isotope ratios of CH₄ (δ¹³C-CH₄) were close to atmospheric equilibrium (∼ 3 nM, −47‰; Fig. 3). However, strong differences between these two stations were observed in the methane and δ¹³C-CH₄ depth profiles. In the WGB, the vertical methane concentration profile was quite stable with maximum concentrations of 7.7 nM at 63 m water depth. The methane profile obtained on 18^th July in the EGB showed a strong increase from the sea surface to 17.1 nM below the thermocline at 27 m water depth. This distinct concentration anomaly came along with a shift in the stable isotope values toward light δ¹³C-CH₄ values of −67.6‰. In contrast, δ¹³C-CH₄ values in the WGB were rather uniform with maximum ¹²C-CH₄ enrichment at 44 m water depth (δ¹³C-CH₄ of −52.2‰). The described methane anomaly in the EGB clustered between 20 m and 40 m water depth and constantly persisted during the field days (Fig. 5, detected on 10 working days between 5^th July and 18^th July), although strong variations in the methane concentrations occurred (range of 15–77 nM).

Surface-water methane concentration and sea–air fluxes

We compiled all data gathered within one nautical mile of the station in the EGB during the entire cruise, summing up to a total of 8189 data points (minute means) recorded between July 7^th and 20^th. Surface-water methane concentrations were very constant over the entire campaign, ranging between 3.1 nM and 3.7 nM, with an average value of 3.45 nM. Data was bin averaged for 1 h intervals to scrutinize the data set for any diurnal trend. No time interval was represented by less than 200 values. The mean hourly values showed a general slightly decreasing trend between 3 a.m. and 5 p.m., a stagnant phase between 5 p.m. and 8 p.m., and an increase from 9 p.m. to 3 a.m., though some deviations from this trend occurred on an hour to hour basis (Fig. 6a). The difference between the maximum mean concentration at 3 a.m. and the minimum at 5 p.m. was ∼ 0.105 nM or 105 pM.

The mean equilibrium concentration derived from the same data set was 2.93 nM. The mean oversaturation of 18% is in line with data from the summer situation in the Gotland Basin in 2010 reported in Gülzow et al. (2013), displaying an oversaturation of ∼ 20%. The averaged sea–air flux was 1.90 μmol m⁻² d⁻¹. The variation in the mean hourly fluxes is larger than that of the surface concentration, reflecting the stronger variation of wind speeds (Fig. 6b).

Plankton community

Our hydroacoustic investigations in the EGB revealed dynamic changes in the position of the layer with highest seston > 100 μm accumulation (Fig. 7). During daytime, we identified the maximum density layer of seston > 100 μm below the thermocline, at a water depth of 25–45 m. This coincided with the location of the methane concentration peak (Figs. 3, 5) and a mesoplankton community > 100 μm that was dominated by different stages of T. longicornis (Fig. 8c). The copepod T. longicornis has been found to perform dial vertical migration between 40 m (daytime) and 10 m (nighttime) in the central Baltic Sea in August (Hansen et al. 2006). Assuming that a dial vertical migration of T. longicornis was largely responsible for the change in the position of the seston > 100 μm layer (Fig. 7), the hydroacoustic data indicates that the copepods started to enter into the mixed layer depth at about 9 p.m. and returned below the thermocline at about 2 a.m. (Fig. 7).

A comparison of the nighttime mixed layer zooplankton communities from the EGB and the WGB revealed the dominance by T. longicornis at the EGB station, while Acartia spp. dominated at the WGB station (Figs. 8a,b). At the EGB station, the nighttime surface zooplankton community from 9^th July at the EGB station (Fig. 8b) was almost congruent to the community found in the 45–25 m layer (sampled 18^th of July, Fig. 8c) and subsequently used for the incubation experiments (Fig. 8c).

The highest chlorophyll a concentration was within the upper 20 m of the water column (Fig. 3). This was confirmed by the biomass data of the phytoplankton as shown in Fig. 9. Even though in the WGB only the upper 15 m were sampled, we assume that the phytoplankton concentration followed the shape of the fluorescence profile and decreased below 12 m water depth as observed in the EGB (Figs. 3, 9). It became evident that the dinoflagellate genus Dinophysis was almost exclusively represented by Dinophysis norvegica, and that the abundance of this plankton species was higher in the EGB compared to the WGB. D. norvegica was concentrated at a depth of 10–20 m, and represented up to 79% of the dinoflagellates in the WGB (at 15 m water depth), and 97% in the EGB (at 20 m water depth). Other groups such as Prymnesiales and Cyanobacteria (Nostocophyceae) seemed to prefer the upper mixed layer.

Identity of methanogenic Archaea

16S rRNA gene analyses based on DNA filters from the WGB and EGB revealed that 63 clonal sequences (no chimeras) matched within the domain Archaea. Sixty-two of these sequences were spread throughout the Archaea, outside of the known methanogenic clusters (Fig. 10): 15 of these sequences clustered with the clades Thermoplasmatales and Methanomassiliicoccaceae, with 11 retrieved from the WGB and four from the EGB, respectively. The other 47 clonal sequences clustered with the clade Nitrososphaeraceae, with 45 sequences retrieved from the WGB and two from the EGB. One 16S rRNA sequence (LT671516), generated from the EGB, was affiliated with the clade Methanomicrobiaceae and thus truly identified as a methanogen.

Methane production in incubation experiments

Methane production was observed in all flasks during the incubation experiments (Fig. 11). In all set-ups, the trend of methane accumulation was very similar and reflected highest production rates on the first day and decreasing methane production within the following 2 d. Visual inspection of the animals revealed vivid movement in most flasks, except for bottles 10 and 11 on days 2 and 3 when dead animals started to accumulate at the bottom of the flasks and oxygen levels decreased into a suboxic state on day 2 (concentration of O₂ about 50 μM) and anoxic state on day 3 (concentration of O₂ below 1 μM). The experiments clearly showed that the seston concentration correlated positively with the methane production rates. Maximum production rates after 1 d were observed in the incubation that contained the highest seston concentration (Bottle 11 with 2.2 nmol CH₄ d⁻¹). Also, in the set-up with a natural water sample that included undefined microplankton and particles, methane concentrations increased. Also here, the highest rates were observed during the first day (0.9 nmol CH₄ d⁻¹). Similar to the other incubations, methane levels remained nearly constant after the second day. In contrast to these incubations, no methane production was observed in the control flask that contained sterile filtered sea water (filter size 0.2 μm). Because potential artifacts like decreasing oxygen values, food limitation and effects of the unnatural high zooplankton abundance got more pronounced during the course of the experiments, we rely on the observed methane production rates of the first day for further interpretation.

Oxygen concentrations on day 2 of the experiment were on average 295 μM in the sterile filtered (Bottle 1 and 2) and in the natural water sample (Bottle 3 and 4), and 304 μM at the end of the incubation experiment (day 3). Oxygen concentrations in Bottle 5 and 6 declined from 204 μM on day 2 to 173 μM on day 3. A similar trend was observed in Bottle 7, 8, and 9, with a decrease from 136 μM to 77 μM. The strongest oxygen depletion was observed in Bottle 10 and 11 from 50 μM on day 2 to values below 1 μM at the end of the incubation experiment.

Potential sources of methane in subthermocline oxic waters

Seasonal methane studies in the central Baltic Sea conducted over a time frame of 2 yr showed that methane concentrations in the upper water column were always highest during August (Jakobs et al. 2014). The strongest methane enrichment in the shallow oxygenated water column was observed in the eastern Gotland Sea with a concentration up to 27 nM at a 40 m water depth in the year 2011 and 25 nM at 50 m in 2012. In contrast, methane concentrations in the western Gotland Sea were relatively low with concentrations of about 10 nM at a 50 m water depth in both years. These observations are in accordance with this study, which was conducted in July 2012 (Figs. 3, 5). The high sampling resolution applied in our study showed that the methane enrichment was only developed in a narrow band. Thus, the concentration peak can be easily missed. Lower concentrations above and below this layer hint at a source at this depth level. The stable carbon isotope ratios of methane in the EGB clearly indicate it's in situ biogenic production (−67.6‰; Fig. 3, Reeburgh 2007). Jakobs et al. (2014) showed that methane oxidation within the oxic/anoxic transition zone at about 100 m water depth leads to a ¹³C enrichment resulting in δ¹³C values of about −40‰ in this specific zone in both basins. Since our studies show δ¹³C values of about −67.6‰ within the methane concentration peak, we can exclude that methane from the ¹³C-enriched deep pool contributes to the shallow methane enrichment.

Different methane sources in the oxygenated water column are currently under debate (de Angelis and Lee 1994; Karl et al. 2008; Keppler et al. 2009; Damm et al. 2010; Grossart et al. 2011). We assume that subthermocline methane production by the microbial break-down of MPn, as proposed for phosphate-stressed environments (Karl et al. 2008), played a rather subordinate role in the free water column, as phosphate (PO₄) was available below the thermocline during our time of sampling (PO₄ conc. ∼ 0.5 μM; Nausch et al. 2013, 2016). Laboratory experiments showed that the uptake of phosphate is preferred during the microbial growth when phosphate and MPn are concurrently available (Karl et al. 2008; Beversdorf et al. 2010), and methane production was inhibited when inorganic phosphate was added to seawater amended with MPn containing semi-labile dissolved organic matter (DOM) (Repeta et al. 2016). Furthermore, the δ¹³C-CH₄ value of −67.6‰ detected in the methane concentration maximum (Fig. 3) in our study goes against a relevant contribution of methane through MPn-decomposition. For methane released from MPn, Repeta et al. (2016) report δ¹³C-CH₄ values of −39‰. These values are considerably more enriched in ¹³C than methane that is typically associated with anaerobic methanogenesis (−50‰ to −110‰, Whiticar 1999). The strong discrimination against ¹³C displayed in our vertical methane profile suggests the reduction of CO₂ by Archaea (Whiticar 1999; Valentine et al. 2004). Other studies indicate that methane production in the oxygenated surface water is associated with primary production (Bogard et al. 2014; Tang et al. 2014), but the low concentration of phytoplankton biomass below 20 m water depth (Fig. 9) rules out a major contribution of algal linked methanogenesis to the subthermocline methane enrichment. For instance, Grossart et al. (2011) suggest that the attachment of photoautotrophs to hydrogenotrophic methanogens allows anaerobic growth and a transfer of diazotroph-derived hydrogen between these microorganisms. Our dataset displays a spatial separation of the Chl a maxima (Fig. 3) and phytoplankton biomass maxima (Fig. 9) above the thermocline from the methane maximum below the thermocline (Fig. 3) that clearly points to a methanogenic source that, in contrast to the Grossart et al. (2011) study, was not congruent with phytoplankton density. Bogard et al. (2014) found indications for an algal-driven acetoclastic methane production in surface waters, a process that appears to be of minor relevance during our studies since the light δ¹³C-CH₄ values detected within the methane peak can hardly be explained by acetoclastic methanogenesis that only shows relatively small isotope fractionation effects (between −5‰ and −35‰, Goevert and Conrad 2009). Another potential source for methane in the water column is the reduction of DMSP (Damm et al. 2010), an algal osmolyte that is abundant in marine phytoplankton and enriched in the mixed layer (Matrai and Keller 1993). Anaerobic methane production by the microbial metabolism of DMSP and its degradation products (e.g., methanethiol) is known from anoxic sediments (Kiene et al. 1986; Finster et al. 1992; van der Maarel et al. 1995). The associated methyl reductase Mcr gene complex was so far only detected in methanogenic Archaea. In the present analysis, we could not detect any known Archaea that metabolize MPn, though further analysis is needed. Nevertheless, methane production could be detected in microcosm experiments with DMSP spiked oxygenated water, in which the abundance of Archaea was negligible (Damm et al. 2010). To explain these contradictory results, Damm et al. (2015) developed a theoretical approach to demonstrate that bacteria are able to maintain an anoxic cytoplasm in which the DMSP metabolism might take place. Recent works showed that microalgae might also contribute to surface-water methane supersaturation by demethylation of organosulfur compounds (Lenhart et al. 2015). However, methane formation from methionine mediated by coccolithophores as described from laboratory experiments (Lenhart et al. 2016) can be excluded, since calcifiers are absent in the central Baltic Sea (Tyrrell et al. 2008).

According to this line of reasoning, we assume that the main contribution for subthermocline methane came from its production in anoxic microniches within particulate biogenic material (e.g., fecal pellets; Karl and Tilbrook 1994) or gastrointestinal tracts of zooplankton (de Angelis and Lee 1994). This assumption is supported by 16S rRNA sequences generated from the EGB, clustering with the clade Methanomicrobiaceae (Fig. 10). The family of Methanomicrobiaceae uses H₂/CO₂ or formate as a major energy source and some species also use primary or secondary alcohols as an electron donor for the methanogenesis (Oren 2014). Several members of the Methanomicrobiaceae have been described in marine sediments such as Methanogenium marinum, Methanolacinia paynteri, Methanogenium marisnigri, or Methanogenium cariaci (Romesser et al. 1979; Oren 2014). For the WGB, we could not detect any methanogens based on our 16S rRNA approach. However, the present analysis was conducted on a relatively small scale and therefore, it is possible that other methanogenic Archaea were undetected due to the methodological approach rather than their actual absence. Based on our hydroacoustic findings, which showed that during the daytime the density of the mesozooplankton fraction was highest in the depth interval where methane was enriched, we propose that microniches in the guts of copepods or their faeces may host the methanogenic Archaea. This hypothesis obtains support from the fact that putative hydrogenotrophic (Methanogenium, Methanobacterium) and methylotrophic (Methanolobus) methanogens have already been identified in fecal pellets of copepods (Cynar and Yayanos 1991; Marty 1993; Ditchfield et al. 2012). In addition, Bianchi et al. (1992) observed methane production associated with both copepods and egested fecal pellets, which would be consistent with the presence of methanogens in the gut of copepods. The authors speculated that methanogens, active within fecal pellets originated from the digestive tract of the zooplankton and are incorporated into the pellet at the time it is produced.

The anoxic potential of different marine aggregates was investigated in detail by Ploug et al. (1997, 2008), Ploug and Jørgensen (1999), and Ploug (2001). According to their oxygen microsensor studies, only “carbon and nutrient rich” aggregates > 600 μm in size fulfill the prerequisite to become an anoxic microenvironment. In a study by Ploug et al. (2008), the oxygen environment in small copepod fecal pellets (88–168 μm long, 40 μm wide) was suggested to be too unstable for the relatively slow growth of obligatory anaerobic microorganisms such as methanogenic Archaea. Furthermore, Ploug et al. (1997) used Archaea specific primers but could not identify methanogens in copepod fecal pellets from experimental Acartia tonsa grazing on the crypthophyte Rhodomonas baltica. Their results from the Acartia study are in contrast to the study by Ditchfield et al. (2012) that identified 16S rRNA sequences in T. longicornis and Acartia clausa fecal pellets closely related to methanogens.

De Angelis and Lee (1994) conducted laboratory experiments to study the potential of different copepod species to produce methane under oxygenated conditions. This study identified T. longicornis as a key player in zooplankton associated methane production. Their finding strongly supports our hypothesis, since T. longicornis was the dominant copepod species in the EGB (Fig. 8), the area where methane was enriched in the upper oxygenated water column (Fig. 3). In contrast to the EGB, the abundance of T. longicornis in the WGB was relatively low and the copepod community was dominated by Acartia spp., an animal that did not produce any methane in the experiments carried out by de Angelis and Lee (1994). De Angelis and Lee (1994) further studied the influence of the food source on methane production with the outcome that T. longicornis had their highest production rates when the diet consisted of dinoflagellates (Prorocentrum minimum). The influence of the diet on the redox conditions in copepod guts was also supported by oxygen microprofiles (Tang et al. 2011). Interestingly, during the time of our studies, the abundance of the mixotrophic dinoflagellate D. norvegica was relatively high within the thermocline in the EGB, whereas the abundance of this phytoplankton species was distinctly lower in the WGB (Fig. 9). Carpenter et al. (1995) showed that in summer D. norvegica is the predominant dinoflagellate within the thermocline of the EGB and no cells of this species could be detected in the mixed layer depth. Based on these reports, it appears that the observed methane maxima in the EGB and the absence of such subthermocline methane enrichment in the WGB are likely to be related to the relatively high abundances of T. longicornis and D. norvegica.

Subthermocline methane production rates

Our measurements show that the methane enrichment coincided with the density maxima of mesozooplankton (Figs. 3, 5, 7). To assess methane production rates from this plankton fraction, we applied field-incubations with different seston concentrates obtained from the relevant depth interval. According to our assumption that T. longicornis was mainly responsible for the methane increase within the first day in our incubations (Fig. 11), we calculated an average methane production rate of 0.3 pmol CH₄ d⁻¹ per adult T. longicornis. This production is relatively low compared with the rate measurements conducted by de Angelis and Lee (1994), who obtained rates between 3.9 pmol CH₄ T. longicornis⁻¹ d⁻¹ and 20.4 pmol CH₄ T. longicornis⁻¹ d⁻¹. A factor that might explain the difference in the production rates is the size deviation between T. longicornis in the Atlantic Ocean (1300 μm) that was used in the de Angelis and Lee (1994) study and T. longicornis in the Baltic Sea (700 μm). The copepod size and the extension of the digestive tract influence the oxygen gradient within the gut (Tang et al. 2011), and one could assume that the activity of oxygen sensitive methanogenic Archaea might be suppressed in smaller copepods. However, Tang et al. (2011) could demonstrate that not only the size but also the diet can cause oxygen depleted conditions in the intestinal tract that would allow methane production even in smaller copepods. Other studies indicate that the integrity of the peritrophic membrane, that encloses the food within the copepod gut and surrounds the fecal pellets, also influences oxygen diffusion and that an intact membrane allows oxygen depleted conditions already within 100 μm (Alldredge and Cohen, 1987). Investigations on methanogens showed that these microorganisms can be metabolically active under oxygenated conditions by overexpressing oxygen detoxifying genes, which was demonstrated for biological soil crust samples (Angel et al. 2011). We think that food limitation was the most probable cause for the low methane production rates in our experiment. This assumption was supported by the reduced phytoplankton biomass concentrations below 20 m depth and by the very low biomass below 30 m depth compared to the phytoplankton biomass in the mixed layer (Fig. 9b). While we probably underestimated the rate of methane production in our incubation experiment, a positive association between methane production and the abundance of T. longicornis was evident (Fig. 11).

We also observed relevant methane production rates in the incubation of the natural water sample obtained from 35 m water depth, for which sources, other than copepods, must be responsible, since mesozooplankton was not observed in these approaches. It might be speculated that during these incubations, methane was also produced by methanogenesis within anoxic microniches in particles (such as fecal pellets, Karl and Tilbrook 1994). For instance, studies on fecal pellets showed that these aggregates can be highly enriched in DMSP (Kwint et al. 1996) and that the descent of pellets can play a critical role in the vertical flux of DMSP (Daly and DiTullio 1996). Tang (2001) showed that within 24 h up to 95% of the DMSP in freshly produced fecal pellets, is lost by microbial turnover and diffusion from the pellet into the surrounding water. Thus, within anoxic microniches of sinking fecal pellets, methane production from DMSP by Archaea may also have taken place over the course of our experiments and might contribute to the methane enrichment below the thermocline.

A mass balance to describe sources and sinks of methane in oxic waters

Our mass balance calculations estimated a maximum vertical methane transport of about −4 × 10⁻⁶ μmol m⁻² s⁻¹ closely below the methane peak at 28 m depth (Fig. 4), which was driven by the strong methane concentration gradient. The upward transport at the upper edge of the peak was considerably lower (1 × 10⁻⁶ μmol m⁻² s⁻¹), since the strong thermocline reduced the eddy diffusivity. Further, the results show the expected surplus of production for the methane peak at 28 m. The integrated methane net production rate (sources minus sinks) in the depth interval between 26 m and 29 m, where the methane enrichment was detected, is 0.33 μmol d⁻¹ m⁻² (i.e., 110 pmol L⁻¹ d⁻¹). Since the concurrent methane consumption (oxidation) rates are unknown, a calculation of absolute production is not possible and the transport-derived production rate displays a somewhat lower limit.

During our field studies, the abundance of T. longicornis was 3.5 individuals L⁻¹ on average in the depth interval between 20 m and 60 m. Based on the rate of 0.3 pmol CH₄ d⁻¹ per adult, T. longicornis determined in our experiments, this results in a zooplankton-associated methane production of about 1 pmol L⁻¹ d⁻¹. Therefore, methane production rates from our incubation experiments are two orders of magnitude lower than predicted from our mass balance approach (110 pmol L⁻¹ d⁻¹), and it must be assumed that also other methane sources are necessary to explain the subthermocline methane enrichment or that there is a bias in our budget calculation. As discussed above, a significant discrepancy exists between the rates measured in our experiments and the studies conducted by the de Angelis and Lee (1994) that might be explained by the chosen design of our incubation. Using the production rates determined by de Angelis and Lee (1994, 3.9 pmol CH₄ copepod⁻¹ d⁻¹ and 20.4 pmol CH₄ copepod⁻¹ d⁻¹), who carried out their experiments under more natural conditions (e.g., natural abundance of copepods), we calculate a zooplankton-associated methane contribution of 14–70 pmol L⁻¹ d⁻¹, indicating that zooplankton has the potential to contribute significantly to the subthermocline methane enrichment. Furthermore, our multibeam survey displays temporal variations in the position and density of the seston > 100 μm layer below the thermocline (Fig. 7) that implies a patchy structure of the distribution of zooplankton, a behavior that is well-known from T. longicornis (Seuront and Lagadeuc 2001; Dutz et al. 2010). From these observations, it can be expected that the in situ abundance of T. longicornis can differ substantially from its average value determined for the 20–60 m depth interval. Such an inhomogeneous distribution will affect the source strength of zooplankton-associated methane production and might explain the variations in the subthermocline methane concentrations observed during the time of our field studies (methane conc. maximum 77 nM and minimum 15 nM).

Apart from these insights into the distribution and strength of the sources of methane, our mass balance approach also delivers critical information about methane sinks in the oxic water column. Our calculations indicate that total production and consumption of methane are balanced in the water column between 20 m and 60 m depth (Fig. 4). Therefore, it can be assumed that methane produced below the thermocline is consumed before it is transported into the mixed layer. Microbial methane consumption within the upper water column has not yet been investigated, but methane oxidation and methanotrophic bacteria were identified in August 2012 within the depth interval of 80–120 m water depth at the oxic/anoxic transition zone of the EGB (Jakobs et al. 2014). Our mass balance suggests that methane oxidation at this depth might not only be maintained through the upward flux of methane from the deep anoxic waterbody but also from the downward flux of methane from the subthermocline methane accumulation. Such a subthermocline balance between sources and sinks of methane most likely exists during the summer months when the upper water column is strongly stratified by a pronounced temperature gradient, a situation that changes in autumn when a decreasing surface-water temperature and an increasing wind stress lead to a destabilization and an upward mixing of water from intermediate depths. Such a ventilation of subthermocline methane is indicated in surface water equilibrator measurements that showed elevated methane saturation values during the August–September period (Gülzow et al. 2013).

Since the transport of methane from below the thermocline into the mixed layer is unlikely to happen during the period of our investigations, the observed methane supersaturation in the surface water (Figs. 3, 6) can only be explained by methane sources that are located in the upper 20 m. Our air–sea flux measurements indicate that an additional source in the mixed layer of 1.9 μmol CH₄ m⁻² d⁻¹ (synonymous with an average net production rate of 95 pmol CH₄ L⁻¹ d⁻¹ in the 0–20 m depth interval) is required to maintain the methane loss into the atmosphere. From the vertical migration behavior of T. longicornis (Dutz et al. 2010), it can be expected that these copepods also contribute to the methane production while they are grazing in the mixed layer at night (Fig. 7). Such zooplankton related methane production in the mixed layer is supported by the weak diurnal cycle of the methane surface concentrations which shows an increase in methane for the time-interval between 9 p.m. and 2 a.m., during which the zooplankton was present in the mixed layer (Figs. 6, 7). Since the daily residence time of T. longicornis in the mixed layer depth is only about 5 h, the methane production rate in the mixed layer of 95 pmol L⁻¹ d⁻¹ appears to be too high to be sustained by zooplankton associated methane production alone.

In contrast to the subthermcoline waterbody where the phytoplankton biomass is relatively low (Fig. 9), it is likely that methane sources associated with primary production play a central role in the phytoplankton-rich surface water (Bogard et al. 2014; Tang et al. 2014; Lenhart et al. 2015). Mid-summer blooms of nitrogen fixing cyanobacteria are a common feature in the central Baltic Sea during the time from July to August (Schneider et al. 2014). Different to the subthermocline water, where phosphate is accumulated, nitrogen-fixation can promote upper-ocean phosphate limitation in the central Baltic Sea (Nausch et al. 2008) and it might be speculated that such an environment supports the microbial utilization of MPn resulting in an accompanied release of methane into the surface water (Karl et al. 2008; Repeta et al. 2016).