In this paper we divide the North Sea into three different hydrogeographical zones: the southern North Sea (SNS), the northern North Sea (NNS), and the combination of the Skagerrak and the Norwegian Trench (SKNT). The SNS (surface area: 279.2 km2) encompasses the Southern Bight (between Belgium–Netherlands and the UK), the shallow Wadden Sea (running along the Dutch and German coast up to Esbjerg in Denmark), the German Bight, and the central part of the North Sea (Fig. ). The major difference between the SNS and the NNS is the stratification regime. Whereas the shallow SNS has a fully mixed water column throughout the year, the NNS is thermally stratified in summer but fully mixed during winter, due to strong wind forcing and surface cooling . As the operational border between the SNS and NNS, we use the absence/presence of stratification as observed in September 2011. Thus, all non-stratified stations are classified as SNS, while all stratified stations belong to either the NNS or the SKNT. The third hydrogeographical zone is the Skagerrak and the Norwegian Trench (SKNT), which forms one of the major sediment depositional areas of the North Sea . The Skagerrak is part of the transition area that connects the North Sea with the Baltic Sea. It is a rather small strait (200 by 100 km) between the southeastern coast of Norway, the southwestern coast of Sweden, and the Jutland Peninsula of Denmark. The average depth of the Skagerrak is about 210 m. The Skagerrak is strongly stratified in summer, but also features a weak stratification in winter driven by Baltic freshwater inputs . The Skagerrak is connected to the Norwegian Sea through the Norwegian Trench, with a sill depth of 270 m. The Norwegian Trench itself is a deep sedimentary basin (250–700 m) that reaches from the Oslofjord in the southeast to the Stad Peninsula in the upper northwest . Like the Skagerrak, the Norwegian Trench is characterized by haline stratified water masses .

The water transport in the NNS is dominated by the large open boundary with the North Atlantic. Water entering in the west through the Shetland Channel and the Faire Island Channel turns eastwards and leaves the North Sea via the Norwegian Trench in the east. The residence time of this water is about 1 year. Generally, water entering the NNS does not influence the SNS, as just 5 % of the North Atlantic water entering the northern boundary reaches the SNS . The water transport in the SNS is mainly determined by inflow from the Atlantic Ocean through the channel, which mixes with low-salinity water coming from rivers and moves along the eastern coastlines towards the northeast. Baltic Sea water entering through the Kattegat and coastal run-off are important in maintaining the Norwegian Coastal Current that initiates in the Skagerrak. This northwards-directed current is the only net outflow of the North Sea and thus balances all incoming water inputs as described above. Additionally, the shallow SNS is influenced by strong tidally induced currents and mixing . The tides in the SNS are diurnal, whereas maximum surface currents at spring tide occur in the western and southern parts of the SNS. More to the north and into the German Bight, the tidal current velocities decrease .

The seafloor of the North Sea predominantly consists of permeable sediments. Medium and fine sand is the main sediment type and occupies the largest part of the North Sea basin. Coarse sand is found at confined locations throughout the entire North Sea basin, with larger areas of coarse-grained sediments present along the English coast and in front of the German and Danish coasts. Mud and sandy mud are mainly found in the deep trenches along the Norwegian coast, off the coast of Scotland, in particular locations along the Belgian coast, in smaller areas north and east of the German Bight, and near Helgoland in the German Bight .

During a cruise onboard R/V Pelagia in September 2011, we sampled a total of 19 stations across the whole North Sea basin (Table ). On a second cruise in June 2012, we sampled seven stations along a transect perpendicular to the Dutch coast from the Wadden island of Terschelling up to the Oyster Ground in the central North Sea (1b–7b in Fig. ). Sediment cores were collected at each site using a Reineck box corer. Polymethyl methacrylate (PMMA) core barrels (19 cm inner diameter) were subsequently inserted into the sediment of the box core to a depth of 10 to 15 cm enclosing 15 to 20 cm of overlying water. The cores were excavated from the box core, closed off with a lid at the bottom, and immediately transferred to a water-filled reservoir in a thermo-controlled container that was kept at in situ bottom water temperature. These sediment cores were subsequently used for closed core flux incubations as described below.

During the September 2011 cruise, small sediment cores were retrieved for solid-phase analysis and microsensor profiling, and these were taken from the same box core as the flux cores. For solid-phase analysis, acrylic core barrels (5 cm i.d.) were inserted into the sediment of the box core. The upper 10 cm of each core were sliced into 1 cm intervals and sediment samples were analyzed for porosity and grain size distribution. Porosity was determined by weight loss after freeze-drying, accounting for salt precipitation in the saline pore water. Grain size distribution was determined using a Malvern Mastersizer 2000 particle analyzer.

For O2 and pH microsensor profiling, acrylic core liners (3 cm i.d.) were inserted into the sediment of the box core, and afterwards, the sediment was brought level to the rim of the core liner. Cores were subsequently placed in an aquarium containing bottom water at in situ temperature, which was constantly bubbled with ambient air, to ensure constant water flow over the sediment. Bottom water was retrieved by casts with 25 L Niskin bottles (Ocean Test Equipment, Fort Lauderdale, USA) retrieving water at approximately 1 m above the sediment surface. A conductivity, temperature, depth (CTD) was mounted on a standard rosette frame together with 24 Niskin bottles. The CTD was equipped with a SBE3+ thermometer, a SBE4 conductivity meter and a SBE43 dissolved oxygen sensor (Seabird, USA).

Flux chamber incubations are potentially susceptible to various methodological artifacts resulting from sediment enclosure, which relate to the sensitivity of benthic fluxes to changes in ambient hydrodynamics and altered benthic faunal activity . In permeable sediments, fluxes are particularly susceptible to the imposed stirring regime and the associated local pressure gradients that are generated in the flux chamber, which drive the advective pore-water exchange between sediment and overlying water . The flux chamber type employed here was based on the design in Huettel and Gust (1992), which was specifically developed for flux studies in permeable sediments. The sediment cores retrieved were closed off with a PMMA top lid that was equipped with a large central stirring disc (diameter: 14 cm). The rotation of the disc mixes the overlying water and establishes a specific radial pressure gradient, which drives pore-water exchange in the chamber. A major challenge for flux chamber studies in permeable sediments is the selection of the appropriate stirring regime. Without knowledge of the local in situ hydrodynamics, it is impossible to predict a priori which stirring regime is appropriate for a given site. Here we used two different stirring rates at each station (40 and 80 rpm) to mimic a range of interfacial pressure gradients and solute exchange conditions . By taking this approach, we are able to discern how sensitive the fluxes at each station are to advective exchange, and thus we get an idea of the uncertainty in our flux estimates for the permeable sites that we visited.

Prior to the start of the flux measurements, the overlying water in each core was replaced with ambient bottom water to ensure that the chemical composition of the overlying water closely resembled in situ conditions. Each set of flux measurements began by securing gas-tight lids equipped with O-ring seals on each core. Core lids contained two sampling ports on opposite sides for subsampling during the incubation, and these ports were carefully purged with bottom water prior to the start of each incubation to remove any air bubbles trapped. Fiber-optical oxygen sensors (FireSting OXF1100) were inserted into a third opening in the top lid. The chambers were closely inspected to ensure that no gas bubbles remained inside the chamber. Two or three replicate chamber incubations were made per station.

The temporal evolution of the oxygen concentration in the overlying water of the flux chambers was continuously monitored using oxygen optodes at a sample interval of 1 min. Optodes were pre-calibrated on the same day using a two-point calibration with ambient seawater at 0 % (saturated with sodium sulfite) and 100 % O2 saturation (bubbled with air). Optode results were verified with measurements of the O2 concentration by Winkler titration in discrete water samples before and after incubation .

A water subsample (∼50 mL) was withdrawn from the chambers at 4–6 h intervals for solute analysis as described below. When a water sample was extracted via one sampling port, an equal amount of ambient bottom water entered through the replacement tube connected to the other sampling port. Samples were collected in plastic syringes for alkalinity (AT) and in glass syringes for dissolved inorganic carbon (DIC) analysis. Water samples for AT (∼10 mL) were filtered (0.45 µm Millex-HA syringe filter) and stored in the dark at 4 ∘C. DIC water samples (∼12 mL) were not filtered, but were poisoned with 10 µL HgCl2 and stored submerged at 4 ∘C in a fridge.

The total oxygen uptake (TOU) rate and the flux of DIC and AT were determined from a linear regression of overlying water concentrations versus incubation time (Eq. ). J=VowAdCowdt

In this, Vow is the volume of overlying water, A is the surface area of sediment, and Cow is the concentration in overlying water. At the end of the incubation, the top lid was removed, the height of the overlying water (H=Vow/A) was measured at four points along the side of the core using a ruler, and the mean height was calculated. suggested that an oxygen decrease by more than 10–15 % from the initial conditions can already stimulate processes that cause a nonlinear decrease in oxygen concentrations, which also might affect other solute fluxes across the sediment–water interface. Here we found that the linear regressions used to calculate the oxygen uptake were insensitive to an oxygen decrease of 30 % or more (for examples, see Fig. ).

Total AT was determined via an open-cell titration procedure, using a Metrohm Titrando 888 system with a combined Metrohm glass electrode (Unitrode) following the SOP3a procedure as described in . Samples (10 mL) were placed in a temperature-regulated open cell (25 ∘C) and titrated with a solution of hydrochloric acid (0.1 N) in a two-stage process. First the sample (10 mL) was acidified to a pH close to 3.5 and then titrated in small steps down to a pH of 3.0. Subsequently, AT was calculated using a nonlinear regression approach based on SOP3a from . Two replicate measurements were carried out for each sample analyzed. Titrations (n=10) of certified reference materials (CRM batch 116 provided by A.G. Dickson) were on average within 4 µmol kg-1 of the nominal value, with a precision of 5 µmol kg-1.

DIC was determined using an AS-C3 DIC analyzer (Apollo SciTEch, USA), in which the sample (0.8 mL) was acidified and the released CO2 was detected using a solid state infra-red CO2 detector (LI-7000, LI-COR Biosciences, USA). Two replicate measurements were carried out for each sample analyzed. Quality assurance of the DIC analysis was also based on CRM (batch 116; accuracy and precision: 3 µmol kg-1).

Microsensor profiling was performed using commercial Clark-type O2 and potentiometric pH microsensors operated with a motorized micromanipulator (Unisense A.S., Denmark). To minimize the error by ship movement, the microprofiling setup was placed in the center, the most stable location of the ship. Under strong wind conditions, microprofiling was not performed, in order to prevent sensor damage. Vertical depth profiles of O2 were recorded using an electrode with a tip size of 100 at 250 µm steps, beginning at 2 mm above the sediment–water interface until either anoxia or 20 mm depth. The O2 microsensors were calibrated with a two-point calibration made in air-saturated seawater (100 % saturation) and at depth in anoxic sediment (0 % saturation). Depth profiles of pH were measured using microsensors with a tip size of 200 µm in 1 mm steps, beginning 4 mm above the sediment surface until 35 mm depth. Measurements were always started within 1 h after sampling. The pH microsensors were calibrated using NBS buffers (pH 4 and 7) and TRIS buffers , and the pH is reported on the total scale.

The diffusive oxygen uptake (DOU) of the sediment was calculated from the O2 depth profiles as DOU=-DO2d[O2]dz, where z is the depth and [O2] denotes O2 concentration. The slope d[O2] / dz was determined from the gradient in the diffusive boundary layer , and while profiling, the overlying water was bubbled with air to ensure a water flow over the sediment surface. The molecular diffusion coefficient of O2 in seawater (DO2) was calculated as a function of the bottom water salinity and temperature using the CRAN:marelac extension package in the R open-source programming language .

Results are reported as the mean ± 1 standard deviation (SD) of n replicate measurements. Nonparametric statistics were used in the interpretation of results, including the Mann–Whitney U test (p) for comparison of the mean of two independent groups of measurements. The Spearman rank correlation coefficient (ρ) was used as a measure of the statistical dependence between two variables X and Y. The more frequently used Pearson correlation provides a measure of the linear relationship between two continuous random variables, while Spearman's correlation allows nonlinear correlation. More precisely, a perfect Spearman correlation results when X and Y are related by any monotonic function, while for the Pearson correlation, a perfect correlation only results when X and Y are related by a linear function. As the environmental variables tested were nonlinearly related (see Fig. ), we used the Spearman correlation. Statistical analyses were conducted in R using the CRAN:stats package.

Bottom water salinity, temperature and oxygen concentration as obtained by CTD profiling are listed in Table , and characteristic temperature depth profiles are displayed in Fig. . Bottom water salinity and temperature of stations from the basin-wide North Sea cruise ranged from 35.12 PSU (NNS) to 33.01 PSU (SNS) and from 16.36 ∘C (SNS) to 6.13 ∘C (NNS). The bottom water in the NNS and SKNT was generally colder and more saline than bottom water in the SNS (Mann–Whitney; temperature: p=0.001; salinity: p<0.001). A thermocline was formed between 20 and 50 m water depth in all stations of the NNS and SKNT (Fig.  and Table ). As noted above, the absence of stratification was used to classify stations within the SNS. The bottom water oxygen concentrations did not however exhibit significant differences between the SNS and other parts of the North Sea (Mann–Whitney; p>0.5).

The median grain size for all stations visited during the basin-wide North Sea campaign in 2011 ranged from 28 to 499 µm (mean median grain size: 215 µm). In the SNS the median grain size ranged from 138 to 499 µm (mean median grain size: 263 µm), and according to the Wentworth scale, the SNS sediments can be classified as fine to medium sand (Table ). Sediments in the NNS (median grain size range: 119–314 µm; mean: 225 µm) were not significantly different from the SNS (Mann–Whitney; p=0.8), hence also being classified as fine to medium sand. In contrast, sediments from the Skagerrak and Norwegian Trench (SKNT) were considerably finer (Mann–Whitney; p<0.01), with the median grain size ranging from 21 to 62 µm. Thus, sediments of SKNT are classified as fine to coarse silt on the Wentworth scale (Table ).

Porosity for all stations in the North Sea basin varied from 0.32 to 0.74. Between sediments of the SNS (mean: 0.35) and NNS (mean: 0.35), no significant differences in porosity were found (Mann–Whitney; p=0.31). However, sediments of the SKNT displayed a significantly higher porosity (mean: 0.68) than sediments from the SNS or NNS (Mann–Whitney; p<0.05).

The concentration changes over time of O2, DIC, and AT in the overlying water from flux chamber incubations are presented in Fig.  for representative stations from each of the three zones (SNS, NNS, SKNT). Stations were chosen to represent average benthic fluxes for the respective region of the North Sea, thus not considering stations that are likely to underestimate or overestimate benthic fluxes (e.g., Station 20). The oxygen data display a linear decrease, while the DIC and AT data display a increasing trend, albeit with greater variability due to the limited number of subsamples (four or five). Fluxes were only included in further analysis if their calculation is based on a linear regression of at least four points combined with an r2 value greater than 0.8. Furthermore, fluxes were set to zero in case the slope of the linear regression was not significantly different from zero.

During the basin-wide campaign in September 2011, TOU rates ranged from 3.1 to 28.7 mmol m-2 d-1 for the SNS, 0.7 to 6.2 mmol m-2 d-1 for the NNS, and 2.9 to 5.7 mmol m-2 d-1 for the SKNT. During the Dutch transect cruise in June 2012, TOU rates ranged from 6.5 to 25.1 mmol m-2 d-1 (Table ). Thereby, the highest oxygen uptake rates were measured in the SNS, followed by the SKNT, while the lowest TOU rates were measured in the NNS (mean SNS values: 10.3 mmol m-2 d-1 in 2011 and 13.5 mmol m-2 d-1 in 2012; SKNT: 3.9 mmol m-2 d-1; NNS: 3.2 mmol m-2 d-1). In general, TOU rates in the SNS are significantly higher than TOU rates of the NNS and the SKNT (Mann–Whitney; NNS: p<10-5; SKNT: p< 0.001). Between TOU rates from the NNS and SKNT, no statistical difference could be identified (Mann–Whitney; p=0.6). Furthermore, TOU rates measured in the SNS in June 2012 are not significantly different from TOU rates measured in September 2011 (Mann–Whitney; p>0.05).

Within the permeable sediments of the SNS and NNS, flux chamber incubations were performed at two different stirring speeds (40 and 80 rpm). The TOU values at these two stirring speeds were linearly correlated (Fig. 4a). In general, the TOU rates were about 80 % higher at the higher stirring speed, indicating that the pore-water transport was dominated by physical advection, as expected in permeable sediments. On the other hand, a correlation between median grain size of the sediments and TOU rates could not be confirmed.

In general, AT and DIC concentrations increased linearly with time in the overlying water of the flux incubations for most stations. AT fluxes for both campaigns ranged from 0 to 21.4, 0 to 3.0 and 1.4 to 9.9 mmol m-2 d-1 for the SNS, NNS and SKNT, respectively. The highest AT fluxes were observed in the SNS (mean: 6.6 mmol m-2 d-1 in 2011 and 5.7 mmol m-2 d-1 in 2012) and in the SKNT (4.3 mmol m-2 d-1). The lowest mean AT flux was calculated for the NNS (1.7 mmol m-2 d-1). Only the AT fluxes of the SNS are significantly higher than AT fluxes of the NNS (Mann–Whitney; SNS: p<0.001; SKNT: p=0.3). AT fluxes measured during the basin-wide cruise in 2011 are similar to those measured along the Dutch transect in 2012 (Mann–Whitney; p=0.3).

DIC fluxes varied between 1.5 and 29.1, 0 and 6.2, and 4.2 and 7.2 mmol m-2 d-1 for the SNS, NNS, and SKNT, respectively. The trends in DIC flux were similar to those obtained for the TOU rates and AT fluxes. The highest DIC fluxes were measured in the SNS, followed by the SKNT and the NNS (mean: SNS: 11.7 mmol m-2 d-1 in 2011 and 12.3 mmol m-2 d-1 in 2012; SKNT: 6.1 mmol m-2 d-1; NNS: 0.8 mmol m-2 d-1). DIC fluxes were significantly higher in the SNS than in the NNS, but showed no differences to the DIC fluxes measured in the SKNT (Mann–Whitney; NNS: p<0.01; SKNT: p=0.3), while no significant difference was found either between DIC fluxes of the SKNT and the NNS (Mann–Whitney; p=0.08). As found for TOU rates and AT fluxes, the DIC fluxes in the SNS were similar between both campaigns (Mann–Whitney; p=0.7). This allows us to examine all solute fluxes recorded in the SNS as one single group in the discussion section.

Representative examples of O2 and pH depth profiles in stations from all three zones of the North Sea are presented in Fig. . In some stations pore-water profiles could not be measured due to the coarse grain size and the presence of carbonate shell fragments, which induce a high risk of damaging the microsensors. In general, pore-water O2 declines as a result of O2 consumption associated with organic matter degradation. From the O2 depth profiles the DOU rates of the sediments were calculated as detailed in Sect. . In the SNS, DOU rates ranged from 0.20 to 6.95 mmol m-2 d-1, while in the NNS and SKNT, ranges of 1.47 to 4.33 and 0.83 to 2.49 mmol m-2 d-1 were found (Table ). The lowest mean DOU rate was measured in the SKNT, followed by the DOU rates of the NNS and SNS (SKNT: 1.77 mmol m-2 d-1; NNS: 1.92 mmol m-2 d-1; SNS: 3.08 mmol m-2 d-1).

The oxygen penetration depth (OPD) is defined as the thickness of the oxic zone in marine sediments , and was operationally defined as the depth below which the O2 concentration drops below 1 µmol kg-1. In some permeable sediments, the oxygen did not fully deplete over the measured depth profiles (first 2 cm), and so the OPD could not be determined. In the remaining cores, the OPD was as shallow as 4.7 ± 2.8 mm for the SNS, 18.1 ± 1.4 mm for the SKNT, and 9.8 ± 3.9 mm for the NNS.

Table  compares the TOU with the corresponding DOU for the basin-wide campaign in September 2011. The TOU / DOU ratio ranged from 50.1 (SNS) to 1.1 (NNS), and DOU rates were significantly lower than TOU rates (Spearman; p<0.01), indicating that physically or biologically driven advective transport strongly enhanced the sedimentary oxygen uptake.

Depth pH profiles of representative cores are presented in Fig. . Depth profiles generally show a decline of pH in all profiled cores, which can be attributed to the release of CO2 resulting from organic matter degradation. Furthermore, the cores from the NNS and SKNT were typically characterized by a sub-surface minimum in pH, while the depths of these pH minima matched the corresponding OPD. Hence, we interpret these sub-surface pH minima as resulting from the aerobic oxidation of reduced substances transported upwards from deeper sediment layers. Generally, the pH increased again at depth in the cores, most likely due to AT generation associated with the anoxic degradation of organic matter (see discussion below).

Several studies have previously addressed benthic oxygen consumption in the North Sea, although most studies are restricted to the southern North Sea (Table ). All TOU rates presented in the following section are not subdivided between both stirring speeds, but are discussed together. The TOU rates found in the present study for the North Sea (range 3.1–28.7 mmol O2 m-2 d-1) fall within the range of previously published TOU rates (range 0–57.1 mmol O2 m-2 d-1), which is however large. Due to temporal variability (seasonality), spatial variability (e.g., substrates ranging from cohesive mud to highly permeable sand), but also differences in methodology (see the discussion below on the impact of the stirring rate), it remains difficult to intercompare the oxygen consumption rates that have been obtained in different studies. So despite this growing database of TOU rates, a more accurate assessment of the spatial and temporal variation of the oxygen consumption in the North Sea remains an important challenge.

In our basin-wide campaign in 2011, we obtained the highest mean TOU rate in the SNS (10.3 mmol m-2 d-1), followed by the SKNT (3.9 mmol m-2 d-1), while the lowest TOU rates were measured in the NNS (3.2 mmol m-2 d-1). One important environmental factor in controlling the TOU rates is the amount of primary production in these different regions. Part of this organic matter produced during photosynthesis is respired in the water column, while the remaining part sinks down to the sediments and undergoes respiration there. As oxygen serves as the ultimate electron acceptor for almost all respired organic carbon , high primary production rates tend to relate to high TOU rates in the sediments. Note that besides primary production rates, the water depth and thereby the amount of organic carbon that reaches the sediments are also of great importance. Because of the shallow water depth, the fraction of organic carbon that reaches the sediments in the southern North Sea is considerably higher than in the northern North Sea, hence further increasing the TOU rates in the SNS . Based on measured surface chlorophyll concentration in the southern North Sea, estimated primary production rates for a period from August 1988 to October 1989, and found clear regional differences in primary production in the SNS ranging from 18 mmol C m-2 d-1 along the British coast to 59 mmol m-2 d-1 in the German Bight. A model study by covering the whole North Sea confirmed the large regional variation in depth-integrated annual primary production rates ranging from 21 mmol C m-2 d-1 in the NNS to 79 mmol C m-2 d-1 in the German Bight of the SNS. TOU rates observed are congruent with these previous observational and modeling studies, with high values in the southern North Sea, and in particular in the German Bight, and lower values in the northern North Sea.

A second important environmental factor controlling the north–south gradient in TOU rates is the bottom temperature, which exhibits a similar gradient in summer, with higher temperatures in the south. In a recent seasonal flux study of shallow coastal sediments from the North Sea, a positive correlation was found between benthic solute fluxes and water temperature . These authors proposed that higher bottom water temperatures in summer are a main driver for higher benthic fluxes, as increased temperatures enhance the metabolic activity in the sediment. In September 2011, the thermocline was formed between 20 and 50 m (Fig.  and Sect. ), and the whole NNS and SKNT were stratified, while in contrast, the SNS was fully mixed (which is the basis of our station classification). As a result, the bottom water in the NNS and SKNT was substantially cooler than bottom water in the SNS (≈11 ∘C difference between SNS and NNS/SKNT – Table ). Adopting a temperature Q10 coefficient of ≈2 for benthic respiration , one would expect a nearly doubling of the TOU between the NNS and SNS due to bottom water temperature, all other environmental factors being the same. Accordingly, lower bottom water temperatures in the NNS and SKNT could indeed partially explain the lower TOU values recorded compared to the SNS (Table ).

To further identify the drivers of the observed spatial variability across the North Sea, we examined the correlation of TOU rates with water depth, porosity and grain size (Fig. ). We found significant correlations between TOU values and water depth (Spearman; p<0.01), suggesting that increased water depth reduced the benthic pelagic coupling in the NNS and SKNT. As more organic matter is remineralized upon the longer transit through the water column, less detritus reaches the seafloor, and this hence decreases the contribution of benthic mineralization in the overall respiration. Finally, we also found both a negative correlation between TOU and porosity (Spearman; p<0.02) and a significant negative correlation between TOU and median grain size (Spearman; p<0.01). This contradicts the classical picture of sedimentary diagenesis, where high oxygen consumption rates and intense biogeochemical cycling are typically linked to fine-grained organic-rich sediments. Instead, we found the highest TOU values in the permeable sediments of mostly the SNS (low porosity, high median grain size), which supports the more recent ideas of permeable sediments as active bio-catalytic filters , which actively trap suspended detritus by means of advective currents through the upper sediment layers and rapidly mineralize this trapped organic matter. As a result, these permeable sediments display low standing stocks of organic matter but high TOU values.

The importance of wind- and tidal-induced advective transport for the benthic oxygen dynamics is emphasized by the deep oxygenation of the surface sediment at permeable sites (Fig. a), the strong impact of benthic chamber stirring speed on the TOU (Fig. ), and by the difference between TOU values as measured by benthic incubations and the corresponding DOU values as obtained by microprofiling (Table ). We found that DOU values were always smaller than TOU values for all stations throughout the North Sea. In general, the total oxygen uptake can be decomposed as TOU = DOU + BMU + AMU , where BMU represents benthos mediated O2 uptake resulting from burrow irrigation and the respiration of infauna, and AMU represents the advection mediated O2 uptake by purely physical transport processes, such as pore-water advection induced by currents over bottom topography and oscillatory pore-water mixing induced by waves .

A second signature of physical and or biological pore-water irrigation is the deep oxygenation of the surface sediment. As already noted above, in some highly permeable sites, the surface sediment remained completely oxygenated over the whole surface layer that was examined by O2 microsensor profiling (first 20 mm), and so no oxygen penetration depth (OPD) could be determined. For those stations that did allow one to determine the OPD (i.e., the pore-water O2 signal decreased to zero), we employed the analytical model of , which provides an inverse relation between OPD and DOU (Fig. a) (Eq. ). L=2ϕDs[O2]bwFO20

In this expression, L is the theoretical OPD of the sediment, ϕ is the porosity, and Ds is the effective diffusivity of O2 in the pore water (i.e., corrected for tortuosity); [O2]bw is the bottom water concentration of O2 and FO20 is the benthic oxygen flux (i.e., the DOU). This relationship is based on the balance between diffusive oxygen fluxes and oxygen consumption rates under steady-state conditions and negligible advection. The above relation was used to estimate the theoretical OPD from the measured DOU as shown in Fig. b. The actually measured OPD is systematically larger than the predicted OPD for all stations (with SKNT stations 52 and 65 being the exception to this rule). As Eq. () holds for sediments that only experience diffusive transport, the discrepancy suggests that non-diffusive transport (i.e., bio-irrigation or physical advection) increases the oxygen availability and penetration in the sediment .

The benthic DIC release in the North Sea follows the same spatial pattern as the benthic oxygen uptake: the highest DIC effluxes were recorded in the SNS (mean: 11.7 mmol m-2 d-1 in 2011 and 12.3 mmol m-2 d-1 in 2012), followed by the SKNT (mean: 6.1 mmol m-2 d-1) and the NNS (mean: 0.8 mmol m-2 d-1; Table ). However, while TOU values were comparable in the NNS and SKNT, the DIC flux in the SKNT was significantly higher than in the NNS. As for the TOU, we found a significant positive correlation between the DIC efflux and water depth (Spearman; p=0.02), and a weak negative correlation between the DIC efflux and the porosity (Spearman; p=0.07) and a significant negative correlation between the DIC efflux and median grain size (Spearman; p<0.01; Fig. ). Overall, this suggests that the same environmental factors (pelagic primary production, temperature, water depth, sediment permeability) that are driving the TOU are also controlling the spatial pattern of the sedimentary DIC release in the North Sea.

Overall, we found DIC fluxes to be positively correlated with TOU rates (Fig. a; Spearman; p<0.0001). The ratio between the DIC efflux and the TOU at a given site represents the respiratory quotient (RQ), and we found that the mean RQ value for the SNS (0.95), NNS (0.94), and SKNT (1.00) to be all similar. No significant differences were detected between the three zones (Mann–Whitney; p>0.05). In general, the RQ can be used as an indicator of the degree of reoxidation of reduced compounds associated with anaerobic remineralization . Assuming that (1) no carbonate dissolution occurs in the sediment, (2) that the stoichiometry of organic matter follows the Redfield ratio (C : N : P = 106 : 16 : 1; ), and (3) that organic matter is oxidized using oxygen as the sole electron acceptor, one would obtain a RQ of 0.77 (solid line in Fig. a; ). Note that a RQ close to the Redfield ratio serves as a baseline for comparison rather than an expected result in this study, due to the anticipated impact of anoxic respiration and carbonate dissolution. In principle, the RQ value is expected to increase with increasing importance of suboxic and anoxic respiration pathways, and subsequent accumulation of reduced compounds such as pyrite, as respiration along these pathways is releasing DIC without consuming oxygen. Alternatively, dissolution of carbonates may also lead to an increased RQ due to the release of 1 mole DIC per 1 mole CaCO3 dissolved. Equally, the RQ value is expected to decrease when a stock of reduced compounds is being re-oxidized, i.e., the annihilation of a previously accumulated oxygen debt.

Our results show that North Sea sediments can be a substantial source of AT, with sediment effluxes ranging from 0 to 28.7 mmol m-2 d-1 (Fig. b). The AT fluxes reported here were obtained by monitoring the temporal evolution of AT in the overlying water of enclosed sediment incubations. In the same sediment incubations, measured the efflux of the short-lived radium isotopes (224Ra and 223Ra) and subsequently estimated AT fluxes. To this end, these authors used the Ra-isotope data to estimate the overall water exchange rate between the pore water and the overlying water column. This water exchange rate was then multiplied by the excess concentration of AT in pore water, which was estimated from pore-water analysis. This indirect estimate procedure resulted in benthic AT fluxes between 4.7 and 22.1 mmol m-2 d-1, and is in good agreement with the AT flux values obtained here (Fig. ). The disagreement (e.g., at Station 38) between observed and calculated AT flux may result from applying the same AT pore-water concentration for all stations in calculating the AT fluxes based on Ra-isotope data. This assumption might be wrong and could lead to an overestimation of the AT flux at Station 38. For a more detailed discussion on the correlation between both methods, see .

The benthic AT release in the North Sea follows the same spatial pattern as the benthic oxygen uptake and the DIC efflux, suggesting that organic matter mineralization is the primary driver for the AT release from sediments. We observed the highest AT release in the SNS (mean: 6.6 mmol m-2 d-1 in 2011 and 5.7 mmol m-2 d-1 in 2012), closely followed by the SKNT (mean: 4.3 mmol m-2 d-1). The lowest AT fluxes were obtained in the NNS (mean: 1.7 mmol m-2 d-1). As for the TOU and DIC efflux, we found a weak positive correlation between AT efflux and water depth (Spearman; p=0.06), though no significant correlation between AT efflux and porosity (Spearman; p=0.4) and a significant negative correlation between the AT efflux and median grain size (Spearman; p<0.01) Fig. . The lack of negative correlation with porosity is mainly due to SKNT stations, which showed a high AT efflux at high porosity. Furthermore, the increase in AT fluxes with increasing TOU rates (Fig. ) is an indication of metabolically driven dissolution of CaCO3, as carbonate dissolution is fueled by CO2 released during organic matter mineralization, or alternatively, of the increase in the relative importance of anoxic respiration pathways at high TOU. Metabolic dissolution would induce a AT / DIC flux ratio of 1, which is indeed confirmed in some stations, though not for all (Fig. ). A positive correlation is obtained between AT and DIC fluxes (Spearman; p=0.02), but there are marked differences in the AT / DIC flux ratio between different stations, which suggests that a different type of biogeochemistry is acting at different locations. Below we will discuss in more detail how different biogeochemical processes in the sediment are releasing AT and DIC in distinct ratios.

To our knowledge, the AT fluxes presented here are the first field observations of such fluxes in the North Sea area. estimated sedimentary AT fluxes in the southeastern bight of the North Sea and the Wadden Sea during summer and autumn based exclusively on water column data. In this approach, the sedimentary AT release appeared as the unknown closure term in an AT budget based on water column data, which provided AT flux estimates of about 8.0 mmol m-2 d-1. These values are similar to the benthic AT fluxes reported here for the SNS.

AT release from sediments has also been the subject of a number of studies on coastal systems other than the North Sea. Based on basin-wide budget calculations, estimated that sediments of the Baltic Sea generate AT with a mean rate of 2.4 mmol m-2 d-1, which is on the same order as the AT input by rivers in that basin. As main sources of AT, these authors propose denitrification together with sulfate reduction and/or silicate weathering. estimated that sediments in the East China Sea generate between 2.9 and 4.9 mmol m-2 d-1 AT. Over 80 % of this AT flux was thereby attributed to iron and sulfate reduction, with no contribution from carbonate dissolution.

used a one-dimensional reactive transport model to estimate AT fluxes from coastal sediments. For that study, the authors divided the global shelf into four different environments, each with distinctive particular organic and inorganic carbon fluxes. Ultimately, they identified non-carbonate shelves (e.g., the North Sea) as an AT source of 2.7 mmol m-2 d-1.

On a global scale, estimated an AT flux for the coastal ocean of 29 Tmol yr-1. Whereas this estimate agrees well with an independent estimation by (16–31 Tmol yr-1), obtained a much smaller flux of 4–6 Tmol yr-1. Note that all these three studies are based on different assumptions about the underlying processes that are generating AT. Whereas the AT flux estimated by is mainly generated by sulfate reduction with zero contribution of carbonate dissolution, acknowledged carbonate dissolution as a benthic AT source. The estimate of is based on anaerobic AT generation alone; hence, possible contributions by carbonate dissolution are not taken into account. Furthermore, these latter authors treat the coastal sediment–water column system as a single system, while the other two papers only consider the sediment.

To put these different assumptions into perspective, and to verify their consequences, we now develop an alkalinity budget for the SNS in which we first provide a sediment budget, and subsequently we extend this argument to arrive at a combined sediment and water column alkalinity budget. We restrict this discussion to the SNS, as we have shown in Sect.  that AT fluxes in the SNS significantly exceed those of the other two zones. High primary production, riverine input of terrestrial organic matter, and the shallow water depth favor a high benthic respiration rate in the SNS. Furthermore, in the context of atmospheric CO2 uptake, a direct link between alkalinity release from the sediments and the atmosphere is needed, which requires a non-stratified water column as encountered in most parts of the SNS.

The high benthic AT effluxes observed in the SNS invoke the question of which processes generate AT in the sediment. In the following we will discuss how different biogeochemical pathways contribute to the overall AT release from sediments, and in this way, we will try to assemble a closed AT budget for the SNS seafloor. To this end, we developed a simplified biogeochemical model of the SNS sediment, which accounts for carbonate dissolution (CD), aerobic respiration (AR), nitrification (NI), denitrification (DNF), iron reduction (IR), sulfate reduction (SR), free sulfide oxidation (SO), and pyrite formation (PF) as biogeochemical pathways. The input parameters, diagenetic relations, and output variables are summarized in Table . The AT / DIC flux ratio for each of these reactions is indicated as a straight line in Fig. . In the next paragraphs, we discuss how the rates of the individual reactions can be constrained based on our flux measurements and literature data.

In theory, the dissolution of 1 mole of CaCO3 releases 2 mole of AT, assuming no other processes are acting. CaCO3→CO32-+Ca2+

The undersaturation of the pore water with respect to the phases of calcium carbonate present (e.g., high Mg calcite, aragonite or calcite) determines the rate at which carbonate will dissolve. In general the undersaturation of the pore water can have two causes : (1) undersaturation of the overlying water and (2) additional undersaturation of the pore water due to metabolic respiration. In the North Sea, the overlying water is always oversaturated with respect to all common carbonate phases , and so, dissolution of CaCO3 in the sediment must be exclusively metabolically driven. Both the CO2 production during organic matter mineralization as well as the oxidation of reduced compounds produced during anaerobic respiration processes can make the pore water undersaturated with respect to carbonates and, therefore, fuel metabolic driven dissolution . The increase in the AT fluxes with increasing TOU rates (Fig. b) indeed forms an indication of metabolic driven dissolution of CaCO3 .

From the flux data set available here, it is not possible to constrain the actual CaCO3 dissolution rate in each station. However, it is still possible to provide an upper limit for the CaCO3 dissolution rate on a larger geographic scale. Gazeau et al. (personal communication, 2014) recently estimated carbonate production in the southern North Sea to be 2.7 mmol C m-2 d-1. Net carbon burial does not occur in the North Sea, except for small amounts in the Skagerrak and the Norwegian Channel . Therefore, we consider carbonate production (in both water column and sediment) and dissolution (in the sediment) to be in balance with each other. Accordingly, carbonate dissolution in sediments can release up to 5.4 mmol m-2 d-1 of AT.

In addition to carbonate dissolution, AT generation can be linked to various organic matter degradation pathways and secondary re-oxidation reactions. In the presence of oxygen, aerobic respiration is used to break down organic matter: OM+15H++106O2→106CO2+16NH4++H2PO4-+148.4H2O, where we assume organic matter (OM) to be of Redfield elemental composition and follow the classical expression for the composition of organic matter (CH2O)106 (NH3)16 (H3PO4) after and . This reaction essentially generates AT by the consumption of protons linked to ammonium release, and to a lesser extent, it consumes AT by the release of phosphate. have estimated that aerobic respiration accounts for between 5 and 25 % of the total benthic mineralization in shelf and coastal sediments. If we adopt a value of the benthic mineralization rate Rmin= 10.3 mmol C m-2 d-1, which is the mean TOU value recorded for the SNS in September 2011 (Table ; see also the discussion below), we can estimate the aerobic respiration rate to range between 0.52 and 2.58 mmol C m-2 d-1 for the SNS (mean: 1.55 mmol C m-2 d-1). Accounting for stoichiometry (AT/O2 ratio = 15/106), the corresponding AT release ranges from 0.09 to 0.37 mmol m-2 d-1, with a mean of 0.22 mmol m-2 d-1. A theoretical doubling of Rmin would lead to a ≈ 7 % increase in the benthic AT flux, and to a ≈ 115 % net AT increase in the water column (Fig. ), that is, if Rmin is the only parameter changed and thus the relative importance between different diagenetic processes remains the same.

Nitrification of ammonium released during mineralization consumes AT and can be represented by the reaction equation NH4++2O2→NO3-+H2O+2H+.

Per mole of ammonium that is released during mineralization, exactly 2 mole of AT are consumed. If we use the TOU as a proxy for Rmin (10.3 mmol C m-2 d-1), the rate of ammonification (i.e., 16/106Rmin) becomes 1.55 mmol N m-2 d-1 for the SNS. If all the ammonium is re-oxidized, and hence no ammonium escapes the sediment, the nitrification rate should match the ammonification rate, and so the associated AT consumption becomes 3.10 mmol m-2 d-1. Also note that complete aerobic respiration, i.e., the combination of aerobic respiration Eq. () with nitrification Eq. (), consumes 17 mole AT per 106 mole of DIC released.

Denitrification is a second mineralization pathway via which organic matter degradation proceeds in the sediment, and can be represented as OM+84.8NO3-+99.8H+→106CO2+42.4N2+16NH4++H2PO4-+106H2O.

Per mole C in POC (particulate organic carbon) denitrified, the AT increases by 0.94=99.8/106 mole. To discuss the impact of denitrification on sedimentary AT release, we need to consider the origin of the nitrate that is used for denitrification. A portion of the nitrate is internally generated in the sediment through coupled nitrification–denitrification, while another part of the nitrate is derived externally (i.e., from the overlying water column). It has been estimated that these coupled nitrification–denitrification reactions account for 80 % of the total denitrification rates in coastal environments , which thus implies that the total denitrification rate should scale as 1/0.8 times the nitrification rate (1.55 mmol N m-2 d-1). This way, we obtain a denitrification rate of 1.94 mmol N m-2 d-1 or 2.43 mmol C m-2 d-1, with an associated AT release of 2.29 mmol m-2 d-1. In the North Sea, a tight coupling of nitrification and denitrification has indeed been reported by and , and observed benthic denitrification rates in the SNS range from 1.6 to 4.6 mmol C m-2 d-1 , a range that centrally embraces our estimate here. Note that the coupling of ammonification, nitrification and subsequently denitrification does not lead to net production of AT. Only if nitrate is derived from the overlying water column does one obtain a net production of AT in the pore water .

Once nitrate and oxygen are fully consumed, dissimilatory iron reduction and sulfate reduction are the prevailing respiration pathways (in the SNS we assume manganese oxides to be a minor electron acceptor, and so we ignore this pathway).

OM+53SO42-+121H+→53H2S+106CO2+16NH4++H2PO4-+106H2OOM+424FeOOH+863H+→424Fe2++106CO2+16NH4++H2PO4-+742H2O

Both reactions consume protons, and hence produce AT. Per mole C in POC oxidized, sulfate and iron reduction release 1.14 and 8.14 mole of AT, respectively. However, to assess the net AT generation in the sediment, we need to take the fate of the reduced species into account. The free sulfide that is generated by sulfate reduction can follow two major pathways. Some of the free sulfide (fraction p) will react with the reduced iron liberated by iron reduction to form pyrite. Due to its thermodynamic stability, pyrite is considered to be the primary sink for both iron and sulfur on the timescales of early diagenesis . Dissolved iron fluxes between the sediment and the overlying water column are considered to be low throughout the southern North Sea and are thus ignored for the following discussion . The overall process of pyrite formation can be represented by the reaction equation 12Fe2++14O2+H2S→12FeS2+H++12H2O.

First, reduced iron reacts with HS- to form FeS, and in a second step the produced FeS reacts with H2S to form pyrite (FeS2) . The remaining part of the free sulfide generated by sulfate reduction (fraction 1-p) is transported upwards towards the oxic zone and re-oxidized with oxygen: H2S+2O2→SO42-+2H+.

Both pyrite formation and free sulfide reoxidation consume alkalinity. However, when combined with the alkalinity impact of sulfate reduction, pyrite formation results in a net generation of alkalinity, while the alkalinity consumed in free sulfide reoxidation exactly compensates for the alkalinity generated during sulfate reduction. calculated for sediments of Aarhus Bay that the burial of reduced sulfur accounts for only 20 % of the total sulfate reduction rate, and similar values are reported in other studies . The sediments of SNS are generally coarser and more permeable than those of Aarhus Bay, which leads to more advection and reoxygenation, which favors sulfide reoxidation as opposed to pyrite formation. Therefore, here we adopt a reduced degree of pyrite formation (10 %; Table ).

Adopting steady state, our simplified diagenetic model provides a set of linear relations (Table ), which generate a particular division of the total organic matter mineralization into aerobic respiration (15 %), denitrification (24 %), iron reduction (0.4 %), and sulfate reduction (61 %). These estimates are comparable to the model analysis of , who estimated that sulfate reduction is the main respiratory pathway on continental shelves, accounting for 77–85 % of the organic matter mineralization, followed by aerobic respiration (16 %), denitrification (2–4 %) and iron reduction (0.3–0.7 %). Similarly, estimated that sulfate reduction accounted for 10–53 % of the total organic matter mineralization in the SNS, while a coupled benthic–pelagic model for the North Sea suggests that up to 30 % of the organic matter is mineralized using sulfate as the electron acceptor . The TOU rate predicted by the model (10.4 mmol O2 m-2 d-1) agrees well with the mean TOU observed in the SNS in September 2011 (10.3 mmol O2 m-2 d-1), but the predicted respiratory coefficient (RQ =1.25) is higher than the observed one (RQ =0.95). This higher simulated RQ could be due to an overestimate of carbonate dissolution rate. However, such a reduction of the carbonate dissolution rate would at the same time lead to a substantial underprediction of the sedimentary alkalinity release, and so we stick to the carbonate dissolution rate CD = 2.7 mmol C m-2 d-1.

Accounting for the contributions of all biogeochemical reactions as discussed in this section above, the total alkalinity generation in the sediment Rsed becomes (Fig. ) Rsed=2rCD+16-1106rAR+84.8+16-1106rDNF+848+16-1106rIR+106+16-1106rSR-2rNIT-2rSO-rPF.

Note that all the calculations are based on measurements carried out in June and September, and thus care should be taken when transferring these rates to an annual scale. As we assume no alkalinity flux to deeper sediment layers, the total alkalinity generation in the sediment should match the efflux of alkalinity across the sediment–water interface. The model predicts a total alkalinity production Rsed= 6.32 mmol m-2 d-1, which is indeed in good agreement with the mean AT flux of 6.60 mmol m-2 d-1 as measured in our flux incubations within the SNS. Positive contributions to the AT efflux are due to carbonate dissolution, aerobic respiration, denitrification, sulfate reduction and iron reduction, whereas AT is consumed during nitrification, sulfur oxidation and pyrite formation. The most dominant AT-producing reaction is carbonate dissolution, followed by sulfate reduction and denitrification. The most dominant AT-consuming reactions are sulfide oxidation and nitrification. The AT contribution of iron reduction is small and balances that of pyrite formation.

Above we have documented how the sediments of the SNS can be a source of AT for the water column of the SNS. However, before we can estimate the resulting effect of this benthic AT release on the pCO2 dynamics of the southern North Sea, we first need to assess the AT balance of the water column. An efflux of AT from the sediment does not necessarily result in an increase in AT in the whole system, as in the water column, some biogeochemical processes oppose the AT generation in the sediment. As noted above, the following calculations are based on measurements in June and September, and hence represent a budget for this period (not an annual average). Here, we identify carbonate formation (CF), primary production (PP), aerobic respiration (AR), and nitrogen fixation (NF) as processes that potentially produce or consume alkalinity in the water column.

As shown above, the dissolution of carbonates in the sediment (Eq. ) forms a strong source of AT, but the production of carbonates in the water column has the opposite effect on AT. Ca2++2HCO3-→CaCO3+CO2+H2O

If the annual production of carbonate in the water column matches the carbonate dissolution in the sediments, there will be no net generation of AT on a system-wide scale over a whole seasonal cycle. Note however that when there is a temporal shift in production versus dissolution of carbonates, these processes may still have an impact on the coastal AT balance on shorter timescales. Here we assume that no net burial or export of solid carbonates occurs in the SNS on an annual scale . Accordingly, the annual benthic carbonate dissolution must be fully compensated for by carbonate production in the water column; hence, carbonate production in the water column will consume 5.4 mmol m-2 d-1 of AT.

A second process that consumes alkalinity in the water column is primary production. This is mostly pelagic, but also benthic in some very shallow waters of the SNS. If we assume that nitrate is the main nitrogen source, primary production can be described by the reaction equation 106CO2+16NO3-+H2PO4-+122H2O+17H+→OM+138O2.

estimated primary production in the southern North Sea to be 199 g C m-2 yr-1, or equally, 45.36 mmol C m-2 d-1, thus producing 7.28 mmol m-2 d-1 of AT. Note that estimated annual primary production rate, whereas rates estimated in this study are based on summer measurements. In general, primary production rates are expected to be greater than the annual average in summer. Aerobic respiration in the water column can be described by the opposite reaction equation of primary production (Eq. ). To estimate the total respiration rate, we need to account for two different sources of the organic matter that is being respired. The majority of the organic matter is derived from local primary production, while a smaller part originates from riverine input of terrestrial compounds. It has been estimated that about 80 % of the locally produced organic matter respires in the water column, whereas the remaining 20 % sinks down to the sediments . The value of Rmin= 10.3 mmol m-2 d-1 as derived in the previous section, compares well with this assessment, as it comes down to 23 % of the PP value of 45.36 mmol C m-2 d-1 as estimated above. The pelagic mineralization rate derived from local net primary production hence becomes 35.06 mmol C m-2 d-1. Additionally, organic matter input from land also fuels respiration, either in the water column or in the benthic compartment. The total riverine input of organic matter in the SNS was estimated by to be in the range of 1.0 mmol C m-2 d-1, which is hence small compared to the local net primary production. Accordingly, the total respiration rate in the water column must be 36.11 mmol C m-2 d-1, thus consuming 5.79 mmol m-2 d-1 of AT.

Finally, the loss of fixed nitrogen due to denitrification in the sediment can be compensated for by nitrogen fixation in the water column. However, the salinity of the North Sea is too high for diazotrophic cyanobacteria and too cold for open ocean cyanobacteria . Thus, we consider N2 fixation in the North Sea to be negligible.

Similar to what was done for the sediment, we can define the total alkalinity generation in the water column Rwc as Rwc=17106rPP-17106rAR-2rCF-rNF, which hence leads to a total consumption of 3.92 mmol m-2 d-1 of AT.

Based on the analysis above, we can now write an alkalinity balance for the combined sediment and water column of the SNS:

ddt[∫0LsedAT(z)dz]=-Fsed+Rsed,ddt[∫0LwcAT(z)dz]=Friver+Fsed-Fout+Rwc.

In summation of both mass balances, and assuming steady state, the alkalinity for the complete SNS system comprising both water column and sediment becomes

Friver+Fsed-Rsed+Rwc=0.

This balance shows input and export processes (F terms) as well as processes that cause internal alkalinity generation (R terms). The total net generation of alkalinity in the SNS amounts to Rnet=Rsed+Rwc=6.32-3.92=2.40 mmol m-2 d-1 or 244 Gmol yr-1. Furthermore, the riverine input of alkalinity was estimated by as Friver= 1.5 mmol m-2 d-1, based on a river load compilation for the years 1977 to 2002. Accordingly, the AT export flux from the SNS to the rest of the North Sea (SKNT and NNS) is Fout=Friver+Rsed+Rwc=1.5+6.32-3.92= 3.90 mmol m-2 d-1 or 398 Gmol yr-1. Additionally, AT from the Wadden Sea might increase the AT budget of the SNS. estimated that ≈ 68 % of the yearly AT change in the German Bight is due to AT export from the Wadden Sea. estimated an AT flux from the Wadden Sea into the SNS of 9.6 mmol m-2 d-1. However, the important question is whether the Wadden Sea exports “uncompensated” TA that then drives CO2 uptake in the SNS. Any net AT generation in the Wadden Sea will generally also induce an atmospheric CO2 uptake in that area. If the residence time of the water is sufficiently long, water masses of the Wadden Sea will be in equilibrium with respect to CO2 in the atmosphere. In this scenario, the net AT generation in the Wadden Sea will be fully compensated for by CO2 uptake in the Wadden Sea, and so this will not influence the air–sea CO2 exchange of the SNS. In the absence of reliable estimates, uncompensated AT inputs from the Wadden Sea were ignored in the budget here, and so the Friver terms for DIC and TA must be interpreted as inputs from both rivers and the Wadden Sea to the SNS.

As shown in Table , the different terms that contribute to net generation of alkalinity in the SNS have been attributed to four different types of elemental cycling (carbon, nitrogen, sulfur + iron, phosphate). This analysis reveals that no alkalinity is associated with carbon and phosphate cycling, but that 26 % can be attributed to sulfur and iron cycling (i.e., anoxic mineralization taking place in the sediment), and the remaining 74 % is linked to nitrogen cycling (this part is dominated by denitrification in the sediment). Sedimentary denitrification is hence the most prominent driver of net alkalinity generation in the SNS. Applying our estimated mean denitrification rate (1.9 mmol N m-2 d-1) to the whole SNS, this implies a removal of 194 Gmol N yr-1 of bioavailable nitrogen, which is roughly twice as high as the denitrification estimate of of 119 Gmol N yr-1 for the whole North Sea. However, whereas our estimation covers mainly the summer months, calculated an annual average rate. Accounting for the fact that the sediments of the SNS are the prime locations for denitrification, our estimates of denitrification rates and the associated alkalinity generation rate in the SNS seem to be in line with previous studies.

Figure  shows the associated DIC and AT budget of the water column in the SNS. The direction of a CO2 flux between the surface water and the atmosphere is determined by the pCO2 gradient between water and atmosphere, which is ultimately governed by the ratio of internal DIC over internal AT release . As shown above, the net generation of alkalinity amounts to 2.40 mmol m-2 d-1. Similarly the net generation of DIC amounts to 1.0 mmol m-2 d-1. As we assume that organic matter burial is negligible, there is no net internal net DIC generation due to respiration of locally produced organic matter (the respiration of autochtonous organic matter in both water column and sediment matches the primary production of autochtonous organic matter in the water column). However, as noted above, riverine input of organic matter, and so respiration of this organic matter, releases 1.0 mmol C m-2 d-1 of DIC. At constant atmospheric CO2 and retaining a constant alkalinity, one can calculate that the addition of 1 mole AT to seawater will lead to an addition of 0.85 mole of DIC. This calculation was carried out in R using the AquaEnv package, utilizing the values of for the first and second dissociation constants of carbonic acid in seawater , and assuming mean parameters of the SNS for the time of sampling (salinity: 34 PSU; temperature: 16.3 ∘C; AT: 2270 µmol kg-1) and an atmospheric CO2 concentration of 392 ppm). Accordingly, a net internal alkalinity release of 2.40 mmol m-2 d-1 will hence lead to an associated DIC increase of 1.96 mmol C m-2 d-1. Of this, 1.0 mmol C m-2 d-1 is already supplied by internal DIC generation, and so, the remainder of 0.96 mmol m-2 d-1 will be supplied by CO2 uptake from the atmosphere. The CO2 uptake for the SNS as a whole amounts to 98 Gmol C yr-1, which is around 14 % of the total CO2 uptake of the entire North Sea . This value is slightly lower than the estimate by , who calculated that sedimentary AT can potentially facilitate up to 25 % of the total CO2 uptake of the North Sea. This latter estimate was not based on direct flux measurements, but was derived from an AT budget for the North Sea, where AT generation in Wadden Sea sediments was introduced as a closure term.

The idea that internal alkalinity generation in the SNS can drive atmospheric CO2 uptake is further supported by previous CO2 studies, although significant seasonality has been observed in the atmospheric CO2 uptake. reported an uptake of CO2 for the months of February to August (-0.52 mmol C m-2 d-1), while the SNS was reported to be a source of CO2 between September and January (+0.60 mmol C m-2 d-1). described the SNS as a strong source of CO2 with fluxes of 0.8 to 1.7 mmol C m-2 d-1 in late summer, while on an annual scale, the SNS was characterized as a sink of CO2 with an air–sea flux of -0.55 mmol C m-2 d-1. Due to the temporal variability of CO2 fluxes, and the seasonal cycles of primary production and respiration, it is hard to compare these numbers with our calculated CO2 air–sea flux based on internal AT generation only, but the estimated impact on the pCO2 dynamics of the SNS by benthic AT release appears to be on the same order of magnitude as determined in previous studies of the pCO2 dynamics of the SNS. Overall, the AT release from sediments, driven by denitrification and anoxic respiration of organic matter, seems to play an important role in the atmospheric CO2 uptake of the North Sea.