Sampling campaigns in the EGB were conducted on the RV Alkor cruise AL422 in August/September 2013, RV Poseidon cruise POS487 in July/August 2015 and RV Alkor cruise AL473 in March 2016 (Table 1; Figure 1). A previous cruise on Alkor AL355 taking place under euxinic conditions of the deep basin in May/June 2010 has been described by Noffke et al. (2016); results from that study are not re-tabulated here. In subsequent sections this cruise will be referred to as pre-inflow (early summer) cruise. Data from AL422 are representative of late summer stagnant, euxinic conditions in the deep basin at a time of cyanobacterial bloom development referred to here as pre-inflow (late summer) cruise. Data from POS487 is referred to as post-inflow (summer) cruise and were taken at the same time of year under ventilated deep basin conditions. Hence, comparison of data from these cruises should allow the effect of the MBI on sediment fluxes to be addressed directly without being too confounded by seasonality, although some degree of inter-annual variation in primary production and benthic respiration is to be expected. Data from winter 2016 [cruise AL473, referred to a post-inflow (winter) cruise] are presented here to show the evolution of benthic fluxes under prolonged ventilation, but are not the focus of the present study. This cruise took place in March 2016 and captured the moderate MBI that was triggered in November 2015.

For this study we will adopt the definitions suggested by Noffke et al. (2016) based on water column O₂ profiles in the Baltic Proper under stagnant conditions. The depth range 60–80 m was identified as the oxycline where O₂ in the bottom water is mostly present although highly variable. The range > 80–120 m was characterized as the HTZ with O₂ levels typically < 30 μM but which can be variable. The zone below 120 m and a well-expressed redoxcline defines the deep anoxic and sulfidic basin. During MBIs, this zone becomes oxygenated (Table 1).

Conductivity, temperature, depth (CTD) measurements were performed during casts of a Seabird CTD system equipped with a water-sampling rosette. These casts were made at water depths between 50 and 223 m along the redox depth transect shown in Table 1 and Figure 1B. Immediately after retrieval, water samples from Niskin bottles were analyzed for the nitrogen species nitrate (NO3-), nitrite (NO2-), and NH4+. Phosphate and hydrogen sulfide measurements include all dissolved species. Their ionic forms PO43- and HS⁻ are referred to in this study.

In situ fluxes were determined at eight sites during the pre-inflow (late summer) cruise AL422 (Table 1, Figure 1) covering the entire oxic to anoxic/sulfidic gradient. Flux measurements were repeated at 110 and 123/124 m, giving a total of 11 deployments. The same sites were investigated on the post-inflow (summer) cruise POS487, with a further site at 161 m to achieve a better resolution during this ventilation period. During the post-inflow (winter) cruise AL473, in situ measurements were conducted at six sites along the transect.

Sea floor images were obtained using the towed camera system OFOS (Ocean Floor Observation System) equipped with a video and still camera (Nikon D70s), two Xenon lights (Oktopus) and a flashlight (Benthos). The system was towed ~1.5 m above the sea floor at ~0.3 knots. Seven OFOS deployments were conducted during the pre-inflow late summer cruise along the depth transect where in situ fluxes were determined (Figure 1B). Additionally, 11 OFOS deployments were conducted to the north and south of the depth transect to confirm presence of microbial mats in the HTZ as described by Noffke et al. (2016). During the post-inflow (summer) cruise, OFOS deployments completely surveyed the depth gradient where in situ fluxes were performed covering a distance of about 27 km of continuous video surveillance, spanning water depths from 70 to 240 m. A second depth transect was investigated to the north of the main working area covering a distance of ~18 km and a depth range of ~63–195 m (not shown). During the post-inflow (winter) cruise OFOS deployments were conducted at the depths where flux measurements were made.

In situ fluxes were determined using benthic chambers mounted in two Biogeochemical Observatories (BIGO) (Sommer et al., 2009; Pfannkuche and Linke, 2003). Each BIGO contains two chambers. The water volume enclosed by the chambers ranged from ca. 9 to 20 L and incubation times ranged from 29 to 56 h. Solute fluxes in each chamber were determined from concentrations measured in discrete water samples removed periodically with eight glass syringes (each ~46 mL). Immediately after retrieval of the observatories, the water samples were stored (max. 4 h) in the onboard cold room (about 6°C) before geochemical analyses. Furthermore, chamber water was sampled into 5 slender glass tubes (each ~15 mL) for onboard determination of DIC. TOU was determined from in situ O₂ concentration measurements in the chambers using optical sensors (Aanderaa Instruments, Norway; Model 3830, Tengberg et al., 2006). Optode performance was cross-checked with O₂ concentrations determined in the water samples by Winkler titration. Solute fluxes were calculated from the linear increase or decrease of concentration versus time, corrected for the surface area to volume ratio of each chamber.

The landers are also equipped to recover the upper incubated sediment layers (~10–15 cm), which serves as a check for sediment disruption during seafloor operations and chamber insertion. The sediment surface for all lander deployments during the pre-inflow (late summer) cruise was intact and undisturbed. On post-inflow cruises the sediments close to the chamber wall of BIGO-II-3 at the 178 m site and of BIGO-I-3 at 151 m were slightly disturbed. It is not known whether this was caused during insertion of the chambers into the sediment or during lander retrieval. The concentration data did not indicate any artifacts during the flux measurements, and the initial concentrations inside the chamber at the start of the incubation were close to the bottom-water concentrations. Consequently, we have no reason to disregard the fluxes determined from these deployments.

Concentration measurements of dissolved inorganic nitrogen (NO3-, NO2-, NH4+), PO43- and HS⁻ were performed on board. Nutrients (NO3-, NO2-, PO43-) were determined on a QuAAtro autoanalyzer (Seal Analytical) using standard photometrical methods (Grasshoff et al., 1999). NH4+ and HS⁻ were measured using photometry with a Hitachi U2800 photometer. Analytical details are described by Sommer et al. (2016). NO2- fluxes were always < 10% of NO3- fluxes, and here we report NO3- fluxes as ΣNO3- + NO2-. Dissolved O₂ in Niskin bottles was determined using automated Winkler titration with a detection limit of 3 μmol L⁻¹. Concentrations below the detection limit were considered to be anoxic for purposes of defining the prevailing bottom water O₂ regime.

DIC measurements were performed using a quadrupole membrane inlet mass spectrometer (MIMS, GAM200, In Process Instruments). The instrument was equipped with inline sample acidification to shift the carbonate system entirely to the volatile CO₂ species, which then was measured on the MIMS at a mass to charge ratio of 44 (Bell et al., 2011). The general setup is described in detail in Noffke et al. (2016). In short, the water samples taken during the BIGO deployments were pumped using a peristaltic pump (Ismatec, REGLO Digital MS-4/6) at 1.0 mL min⁻¹ through a membrane inlet (Sommer et al., 2015), where extraction of dissolved gases takes place. Gas flow from the inlet to the mass spectrometer was conducted in a steel capillary supported with helium that was supplied through a fused silica capillary. The distance between the inlet and the ion source of the quadrupole was about 80 cm. An in-line cryo-trap (−35°C, ethanol) between the inlet and the mass spectrometer was used to reduce water vapor. The water samples were first analyzed without acidification and recovered in Labco Exetainers capped without a headspace. Subsequently, an acidification inlet was integrated into the sample flow in front of the membrane inlet and 4 M hydrochloric acid added at a rate of 0.15 mL h⁻¹ using a precision syringe pump (KDS-100-CE, kd Scientific) resulting in a sample-acid ratio of 402:1 transforming all DIC into CO₂. The DIC measurements were calibrated with a Na₂CO₃ standard (Merck) that was diluted with a sodium chloride solution (10 g L⁻¹) to standards in the range of 1.50–2.25 mM. The standards were prepared at the respective in-situ temperature and were measured before and after each measurement session.

Seafloor imaging during all cruises showed that the sediment surface in the HTZ was densely covered with white filamentous microbial mats, tentatively identified as belonging to the family Beggiatoaceae. Also the stations to the south and north of the main working area showed occurrences of microbial mats confirming earlier observations made during cruises R/V Poseidon (POS369 July–Aug. 2008), R/V Alkor (AL346 Sept.–Oct. 2009) as well as R/V Alkor cruise AL355 (Noffke et al., 2016). Notably, under euxinic conditions only sporadic occurrence of microbial mats was observed below 120 m water depth and they became absent with increasing depth. In contrast, during ventilated conditions the microbial mats extended down to water depths of 240 m and extensively covered the sediment surface. Mats were dominated by different size-classes of filaments, with smaller filaments of 4–5 μm in diameter being more abundant at oxic stations and larger filaments with diameters of 22–32 μm dominating at stations where bottom water O₂ concentrations approached zero.

As known from previous MBIs in 1993 and 2003 (e.g., Nausch and Nehring, 1994), the ventilation in 2015/2016 profoundly changed the redox landscape of the water column below ca. 125 m. Deep-water renewal resulted in similar low NH4+ and PO43- concentrations to those observed after the inflow event in 1993 (Nausch and Nehring, 1994) as well as elevated NO3-. The water column HS⁻ inventory was completely oxidized and not detectable, in stark contrast to euxinic conditions (240 μM, this study). This entails very efficient HS⁻ oxidation mechanisms in the water column as well as at the sediment surface. Oxidation of reduced compounds likely becomes enhanced in the water column following ventilation due to bottom boundary mixing. Furthermore, during propagation of dense bottom currents, entrainment of less saline water also leads to interleaving of incoming oxygenated water masses with the ambient water mass at the level of neutral buoyancy and concomitant ventilation at different depths (Reissmann et al., 2009, cf. their Figure 2). In the sediments, sulfate reduction below the oxic sediment surface still produces high amounts of HS⁻ whose oxidation represents a strong sink for O₂. Mohrholz et al. (2015) estimate that the total amount of O₂ transported into the Baltic Sea during 2015 amounts to ~2.04 × 10⁶ t. Given a benthic HS⁻ flux of 6.7 mmol m⁻² d⁻¹ during pre-inflow (Table 2), and a deep basin area of ~19,000 km² (Table 3), one can estimate that the sediments alone could consume this flux within 8 months. This is a maximum lifetime of O₂ in the basin with respect to benthic respiration in the deep basin since the O₂ inventory in the Gotland Basin will be lower than the total inflow into the Baltic (Schmidt, 2014) and a fraction of O₂ will be consumed by respiration of reduced dissolved and particulate compounds in the water column. Indeed, a fast decrease of O₂ levels was recorded in September and December 2015 during time-series observations conducted at 2 and 20 m above the seafloor, respectively (Mohrholz et al., 2016). This calculation, whilst necessarily first-order, nonetheless emphasizes that HS⁻ oxidation at the sediment water interface strongly contributes to the rapid depletion in O₂ following deep basin ventilation.

In contrast to elevated P release under euxinic bottom water conditions (this study; Jilbert et al., 2011; Viktorsson et al., 2013a; Noffke et al., 2016; and references therein), PO43- was taken up by the sediment at the sites at 142–161 m water depth under ventilated conditions during July/August 2015. A significant removal pathway for phosphorus occurs by way of its adsorption onto iron oxides while pH < 9 (Berner, 1973; Van Cappellen and Ingall, 1996), and reactive iron might become increasingly available in surface sediments under oxic conditions. Remarkably, though, despite oxic conditions during both inflow cruises POS487 (summer) and AL473 (winter) at the deepest sites, average PO43- release (0.17 and 0.23 mmol m⁻² d⁻¹ respectively) was similar to respective fluxes during euxinic conditions in July/August (0.21 mmol m⁻² d⁻¹, this study) or even higher compared to May/June (0.15 mmol m⁻² d⁻¹, Noffke et al., 2016). This might be related to the time course of the inflow spreading into the Gotland Basin. O₂-rich intrusions of North Sea waters propagate as dense bottom currents, such that the deepest sites were likely ventilated first followed by ventilation of the shallower sites sometime later as the deep water is progressively replaced. Hence, reactive iron for P sequestration might already have become limited at the deeper sites since organic P degradation rates are relatively elevated here resulting in a low P retention capacity of the sediments (Noffke et al., 2016). During both post-inflow cruises, PO43- release from the deepest station resulted in the gradual accumulation of PO43- in the water column as observed in the deepest part of the EGB (compare Figures 4, 5).

At the upper boundary of the deep basin at 124 m water depth, the PO43- release during both post inflow cruises in summer and winter was as high or almost as high as under pre-inflow euxinia. An O₂ time series recorded at this site over the entire cruise POS487 revealed strong bottom water O₂ fluctuations between 0 and 15 μM and extended periods of O₂ levels < 5 μM (not shown). The O₂ profiles and NO3- distributions (Figures 2, 4) suggest that the bottom water at this site has the lowest availability of electron acceptors following ventilation. Hence, this site can be considered as strongly hypoxic, thereby providing a partial explanation for the ongoing high PO43- release rates. These O₂ fluctuations are likely caused by internal waves and seiches (Reissmann et al., 2009), which drive water mass movement across the 124 m site resulting in the periodic entrainment of colder anoxic water from above (HTZ) as well as warmer North Sea water rich in O₂ and NO3- from below.

Seafloor imaging revealed that sulfur bacteria belonging to the family Beggiatoacea extensively colonized the sediment surface even down to the deepest part of the EGB. So far these organisms were only observed in water depths of about 70–120 m where electron acceptors are at least temporarily available (Noffke et al., 2016; Sommer et al., unpublished results). These chemotrophic organisms gain their energy from HS⁻ oxidation using O₂ as terminal electron acceptor (e.g., Teske and Nelson, 2006). However, some members of the Beggiatoaceae have the capacity to switch to a NO3- based HS⁻ oxidation, releasing NH4+ to the environment as a metabolic waste product (reviewed by Jørgensen and Nelson, 2004). In contrast to denitrification and anammox, this process known as dissimilatory nitrate reduction to ammonium (DNRA) retains DIN in the ecosystem and has been argued to be an important benthic process under low O₂ conditions in the HTZ (Noffke et al., 2016). It has also been reported for sediments from Eckernförde Bay (Dale et al., 2011, 2013), the northern Baltic Proper, and the coastal Gulf of Finland (Kuparinen and Tuominen, 2001; Hietanen and Lukkari, 2007; Jäntti et al., 2011; Jäntti and Hietanen, 2012). When all the data is considered collectively, NH4+ release rates under ventilated conditions were similar to those measured during euxinic conditions and thus apparently not greatly enhanced by DNRA. The NH4+ release rates measured during post inflow conditions are consistently below the expected based on Redfield organic matter degradation (Figure 7A). However, it is possible that a fraction of NH4+ produced by DNRA may be consumed by close coupling with nitrifying bacteria in adjacent sediment layers, thereby explaining the low N:C ratio of the fluxes. Only a few sites, including those from our previous study in 2010 (Noffke et al., 2016), hint toward an excess of NH4+ release from the sediment, and thus an excess of DNRA over nitrification. It is important to note that during the post-inflow cruise in March (AL473), larger filaments of Beggiatoaceae with internal NO3- concentrations of 2.5 ± 1.8 mM dominated within the HTZ at 100 and 110 m water depth as well as at the upper boundary of the deep basin (124 m) (Schulz-Vogt, unpublished). This strongly indicates the potential capacity for DNRA in this region. Due to higher concentrations of O₂ in the deep basin, aerobic respiration by HS⁻ oxidizing bacteria may have been energetically more preferable than anaerobic, NO3- respiration.

Filamentous HS⁻ oxidizing microbes may also play an important role in benthic P cycling. They are able to perform luxury uptake of PO43- from the porewater and bottom water under oxic conditions that is subsequently stored as polyphosphates in their vacuoles (Brock and Schulz-Vogt, 2011). During anoxic conditions, these organisms degrade the polyphosphates and release the stored PO43- back into the porewater. Evidence for this metabolic mechanism has been provided for the genera Thiomargarita spp. and Beggiatoa spp. in sediments from the Namibian upwelling system (Schulz and Schulz, 2005; Goldhammer et al., 2010; Winkel et al., 2016). Breakdown of polyphosphates may account for the fact that sediments in the deep basin again acted as a PO43- source in March 2016 where O₂ in the bottom water had decreased to < 40 μM. Interestingly, though, during March 2016 filaments of Beggiatoaceae were void of polyphosphates, whereas smaller granules of polyphosphates were abundantly observed in the surface sediment (Schulz-Vogt, unpublished). The source of these particles remains an open question, and could conceivably be attributed to authigenic formation at the former pelagic oxycline, dead phytoplankton, or polyphosphate accumulated by smaller benthic bacteria (Diaz et al., 2008, 2012). The absence of polyphosphate in the filamentous Beggiatoacea is an additional indication that polyphosphate accumulation is not a trait shared by all Beggiatoaceae under all environmental conditions. Members of the genus Marithioploca, (former Thioploca) prevailing in sediments off the Chilean coast were also reported not to contain visible polyphosphate granules (Høgslund et al., 2009).

A plot of the PO43- vs. DIC flux from all cruises demonstrates that the sediments in the EGB behave in a highly non-Redfield manner (Figure 7B). Most sites with a positive PO43- flux lie far above the expected Redfield ratio (solid black curve). Even accounting for preferential remineralization of organic P relative to C by a factor of two (Dale et al., 2016), additional P sources are needed to explain the data distribution (dashed black curve). Off Peru, Lomnitz et al. (2016) observed a clear correlation between PO43- flux and cell number of the filamentous bacterium Marithioploca. In principal, as described above, transient P storage and release from bacterial mats of Beggiatoaceae spp. that almost completely cover the sediment surface in the HTZ could conceivably contribute to the measured P fluxes if they had accumulated polyphosphate during conditions prior to March 2016. As mentioned, smaller less conspicuous sediment bacteria may instead be responsible for polyphosphate mediated P cycling. In addition, iron-bound P dissolution could be significant and deserves greater attention to constrain potential PO43- fluxes from this source (e.g., Jilbert et al., 2011). Clearly, future work must now focus on process-based studies to fully address the factors leading to the highly dynamic nature of N and P fluxes in the EGB.

The O₂ depth profiles obtained during post inflow summer conditions indicate that the MBI only affected water masses deeper than ~120 m. Above this, the O₂ levels were close to zero and reached ~30 μM at 80 m at the base of the oxycline. During winter inflow the O₂ profile shows a similar trend, yet with slightly higher O₂ levels in the HTZ. An identical O₂ distribution has been previously described during euxinia and, along with the distribution of microbial mats, was used as a major criterion to define the HTZ (Noffke et al., 2016). Further similar observations (e.g., Schmale et al., 2016), coupled with the observation that the sediments were also densely covered with mats of sulfur bacteria, imply that the HTZ can be considered as a typical and persistent feature of this water depth range. Pre- and post-inflow solute fluxes in the HTZ were variable but generally indistinguishable. These findings have important implications because the HTZ has been recognized to be a major region strongly contributing to the internal loading of PO43- and NH4+ (Noffke et al., 2016).

The above discussion has demonstrated that the deep basin and HTZ behaved differently during the most recent MBI events. In order to assess the importance of the expected basin wide reduction of nutrient release during ventilation, we approximated the entire benthic nutrient load for the Baltic Proper before and after ventilation (Table 3) using the approach put forward by Noffke et al. (2016). Here, the Arkona Basin has been excluded from the analysis since this is a seasonally hypoxic setting and the presence of more extensive mats of filamentous sulfur oxidizing bacteria there is in doubt. The Baltic Proper was divided into three depth zones according to the vertical distribution of O₂ in the water column during pre-inflow conditions (Figure 1A, Table 1). To arrive at regional nutrient fluxes, the areas of each depth zone were multiplied with the respective averages of the benthic PO43-, NH4+, and NO3- (NO3- + NO2-) fluxes measured in each depth zone. Our flux measurements were only conducted in water depths below 60 m, which approximately separates deeper muddy from sandy sediments in the Baltic Proper (www.helcom.fi/baltic-sea-trends/data-maps).

For ease of comparison with other published rates (e.g., Viktorsson et al., 2013a; Noffke et al., 2016) the fluxes are reported on a yearly basis. However, the ventilation events are transient and the strong MBI triggered in December 2014 only lasted from March to September 2015 when O₂ was already consumed 2 m above the seafloor (Mohrholz et al., 2016). The second moderate ventilation event captured during AL473 only lasted for about 3 months (Mohrholz et al., 2016). Hence, we made the assumption that the fluxes we measured in the deep basin in response to the deep-water renewal prevailed only for 6 and 3 months respectively. For the remaining time of the year we assumed fluxes as measured during pre-inflow conditions. This neglects the dynamics of the inflowing water that progressively replaces the anoxic and sulfidic deep water before the shallower sites of the eastern flank of the basin. It also neglects the time needed for the microbiological community to become established in response to oxygenation, although observations of rapid establishment of Beggiatoaceae in response to anoxia in Eckernförde Bay implies colonization times of < 1 month (Dale et al., 2013).

Ventilation during the MBI from December 2014 (post-inflow summer cruise POS487) reduces the yearly PO43- release from the deep basin sediments by about 33 and about 10% for NH4+ (Table 3, Figure 8). In parallel to the flux study during cruise POS487, benthic P fluxes were measured in the ventilated deep basin at 170 and 210 m water depth in July 2015 on the Swedish site of the EGB (Hall et al., 2017). Basin-wide P release extrapolated for the deep basin was 17.4 ktons year⁻¹ compared to 50.3 ktons year⁻¹, which was measured by these authors during euxinic conditions corresponding to a reduction of 65%. During the moderate inflow event from November 2015 (post-inflow winter cruise AL473) there was only a 3% reduction of the PO43- release compared to flux measurements made during euxinic conditions (pre-inflow summer cruise AL422). In contrast, NH4+ release was even higher compared to all previous cruises.

Importantly, since the fluxes in the HTZ were apparently not greatly affected by the inflows, the HTZ remained a major nutrient release site. With the exception of data from the pre-inflow early summer cruise AL355 (Noffke et al., 2016), differences between pre- and post-inflow conditions are indistinguishable from the range of variability measured in individual chambers. A major source of variability of the fluxes within the HTZ is contributed by the 80 m site located just beneath the oxycline. As already shown (Noffke et al., 2016, cf. their Figure 3), O₂ levels are highly variable there and range between close to zero and 160 μM, which affects mobilization of PO43- as well as N cycling. Despite the reduction of the seabed nutrient release from deep basin sediments, the overall, basin-wide, reduction in nutrient release for the strong MBI (post-inflow summer cruise) only amounts to 5 and 12% for PO43- and NH4+ respectively. For the moderate MBI (post-inflow winter cruise), there was no reduction with regard to NH4+, yet there was an overall reduction of the PO43- release by 42%, which was caused by a drop of PO43- release within the HTZ and thus not greatly affected by the inflow. This implies that changes in benthic fluxes induced by the inflow may simply lie within the natural spatial and temporal variability exhibited by the basin sediments. We excluded the sites from oxycline and shallower waters from our analysis since nutrient release there is highly variable based only on one site.

The analysis shows that seabed nutrient release in the basin is controlled by processes in the HTZ rather than the deep basin as previously assumed. Even the largest MBI recorded from December 1951 (225 km³), which compares to 198 km³ of the MBI in 2014, would be very likely insufficient to ventilate the HTZ and suppress seafloor nutrient release. Despite the uncertainties involved, this simple extrapolation highlights that the effect of deep basin ventilation on the reduction of benthic nutrient release can be considered as minor in the context of the entire nutrient budget. It should be noted that the above extrapolation was conducted under the assumption that the HTZ is present throughout the Baltic Proper (Figure 1). During the MBI triggered in December 2014, an O₂ deficient zone was also retained at the Swedish side of EGB with a still-elevated average P release rate of 0.34 mmol m⁻² d⁻¹ (Hall et al., 2017). Up-scaling of this P flux yields a P-load of 94 ktons year⁻¹, which is very similar to that estimated for the HTZ during euxinia as well as during post inflow conditions presented in this study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.