Drivers of summer oxygen depletion in the central North Sea

In stratified shelf seas, oxygen depletion beneath the thermocline is a result of a greater rate of biological oxygen demand than the rate of supply of oxygenated water. Suitably equipped gliders are uniquely placed to observe both the supply through the thermocline and the consumption of oxygen in the bottom layers. A Seaglider was deployed in the shallow (≈ 100 m) stratified North Sea in a region of known low oxygen during August 2011 to investigate the processes regulating supply and consumption of dissolved oxygen below the pycnocline. The first deployment of such a device in this area, it provided extremely high-resolution observations, 316 profiles (every 16 min, vertical resolution of 1 m) of conductivity, temperature, and depth (CTD), dissolved oxygen concentrations, backscatter, and fluorescence during a 3-day deployment. The high temporal resolution observations revealed occasional small-scale events (< 200 m or 6 h) that supply oxygenated water to the bottom layer at a rate of 2± 1 μmol dm day. Benthic and pelagic oxygen sinks, quantified through glider observations and past studies, indicate more gradual background consumption rates of 2.5± 1 μmol dm day. This budget revealed that the balance of oxygen supply and demand is in agreement with previous studies of the North Sea. However, the glider data show a net oxygen consumption rate of 2.8± 0.3 μmol dm day, indicating a localized or shortlived (< 200 m or 6 h) increase in oxygen consumption rates. This high rate of oxygen consumption is indicative of an unidentified oxygen sink. We propose that this elevated oxygen consumption is linked to localized depocentres and rapid remineralization of resuspended organic matter. The glider proved to be an excellent tool for monitoring shelf sea processes despite challenges to glider flight posed by high tidal velocities, shallow bathymetry, and very strong density gradients. The direct observation of these processes allows more up to date rates to be used in the development of ecosystem models.


Introduction
Mooring observations from 2007 and 2008 (Greenwood et al., 2010), historical data and a hydrographic survey of the North Sea in August 2010 (Queste et al., 2013) revealed repeated incidents of seasonal oxygen depletion in offshore waters of the central North Sea. In August 2011, a Seaglider was deployed in the region with the 5 lowest recorded bottom mixed layer (BML) oxygen saturation from the 2010 survey to further investigate the mechanisms causing this seasonal oxygen depletion.
The North Sea is a relatively shallow (15-200 m) shelf sea situated between the British Isles and northwestern continental Europe. It gradually deepens from south to north with the exception of the shallow Dogger Bank (Fig. 1). Dogger Bank effectively 10 separates the North Sea into two regions of different physical, chemical and biological properties (Otto et al., 1990;Lenhart et al., 1997). Water properties in the northern half of the North Sea are largely dominated by North Atlantic inflow. For their study of carbon dynamics, Thomas et al. (2005) determine the waters of the North Sea to be 90 % sourced from the North Atlantic (9000 km 3 yr −1 via the Pentland Firth and Fair Isle, 15 42 000 km 3 yr −1 via the Shetlands), 8 % from the English Channel (4900 km 3 yr −1 ), and 2 % from a combination of Baltic (500 km 3 yr −1 ) and riverine flow (200 km 3 yr −1 ). Water entering the North Sea from the northern boundary generally circulates anticlockwise, following the Scottish coast southward before turning east and crossing the North Sea north of Dogger Bank (Brown et al., 1999;Turrell et al., 1992;Hill et al., 2008). The 20 strength of this anticlockwise circulation is strongly correlated with the North Atlantic Oscillation Index (NAOI); positive NAOI is associated with strong anticlockwise circulation while negative NAOI is associated with greatly reduced anticlockwise circulation in the northern North Sea (Lenhart et al., 1995(Lenhart et al., , 2004Rodwell et al., 1999). The northern half of the North Sea is seasonally stratified through surface heating. This stratification Introduction as under specific conditions of circulation or surface wind stress, the fresh water run-off issuing from the Baltic, Kattegat and Skagerrak can be advected into the central region (Otto et al., 1990). This not only affects stratification by increasing the salinity difference between surface mixed layer (SML) and BML but also provides supplementary nutrients to the SML. OG regions. The historical data also showed a sharp decline in summer oxygen saturations at these two sites since 1990. Queste et al. (2013) suggested that the same mechanisms likely lead to the depletion of oxygen at OG and ND. It is thought that the replenishment of oxygenated waters through advective processes is limited by local topography. ND is situated away from the fronts associated with the north side of the 15 shallow Dogger Bank; it is characterised by variable, weak wind-driven currents (Otto et al., 1990). This slow circulation limits the horizontal supply of DO while weak tidal currents also promote settling of organic matter. Weak winds and strong surface heat fluxes promote stratification and keep the BML isolated from air-sea exchange. There is also weak inflow of oxygenated waters as tides and topography lead to the forma- 20 tion of bottom fronts to the south and east of the OG site (Van Raaphorst et al., 1998;Weston et al., 2008). Organic matter produced at the pycnocline along the deep chlorophyll maximum (DCM), particularly at ND, is exported directly into the BML (Thomas et al., 2004;Weston et al., 2008;Fernand et al., 2013). This organic matter, when remineralised, leads to the consumption of DO in the BML. At both the OG and ND sites, it 25 has been suggested that mixing events may cause resuspension of bottom sediment. Transfer of this organically rich sediment into an oxic water column could cause rapid increases in pelagic DO consumption (Van Raaphorst et al., 1998;Greenwood et al., 2010). The intensity and duration of oxygen depletion depend on the relative magnitudes of oxygen consumption and oxygen supply. To accurately predict potential seasonal oxygen consumption under future climate scenarios, it is necessary to identify the relative magnitude of sources of organic matter that lead to the consumption of DO in the BML and the amount of DO supply through the pycnocline on a finer scale. There is grow-5 ing interest in the issue of low oxygen in coastal waters around the world (Diaz and Rosenberg, 2008) and further research is required to effectively manage the impact of hypoxia in coastal waters. This study aims to improve our understanding of the key short term mixing processes that regulate oxygen supply at high temporal frequency by quantifying through Seaglider observations the major inputs and sinks of dissolved 10 oxygen. in a sawtooth pattern between the surface and the seabed (or 1000 m in deeper water; Eriksen et al., 2001). Typically, the glider recorded data between 2 m from the surface and 7 m off the seabed; sample points beyond this range were discarded because of poor flow conditions due to the glider turning around. Seagliders sample every 5 s using an on-board sensor suite composed of a sis remained in temperature dependent variables (salinity, density, oxygen) relating to the thermal inertia of the different sensors. As the region contained two fairly uniform mixed layers, composite profiles were created using upcasts between the seabed and the thermocline and downcasts between the surface and the thermocline. This eliminated the bias due to the thermal inertia of the conductivity cell and oxygen optode.

5
The composite profiles were then gridded over time and pressure by taking the mean value in each 45 min by 1 dbar grid square to provide regularly spaced data. Figures 3 and 4 present the data collected by SG510 during its three day mission. The glider is able to resolve significant and systematic changes in all the measured parameters on this timescale. The glider observed strong stratification (> 1.5 kg m −3 ) throughout the three days (Figs. 3 and 4) separating two well mixed layers: the SML and the BML. The SML showed warming throughout the survey with a temperature difference of 1 • C between beginning and end, and a decrease in surface salinity (0.35) as the glider entered a region of fresher surface water on the third day likely originating from Baltic outflow (Fig. 5). The BML exhibits a small and opposite pattern with 15 temperatures decreasing by less than 0.1 • C on the third day and salinity increasing by 0.15. As the glider crossed a surface feature and observed a significant change in temperature and salinity at the end of the survey, data from the last 15 h of the survey are excluded when calculating rates of oxygen depletion. The strong pycnocline maintains a sub-surface phytoplankton community, as evi-20 denced by the DCM (Figs. 3d and 4d). Variations in chlorophyll a concentration (as estimated by fluorescence intensity) occur throughout the survey; peaks in DCM fluorescence coincide with decreases at the surface. The surface variation of chlorophyll a fluorescence is likely caused by quenching as it occurs daily at noon, rather than a loss of biomass or pigment. Fluctuations in chlorophyll a concentration are visible 25 throughout the BML but do not coincide with the diurnal signal described above. These BML fluctuations likely relate to sinking organic matter and are also observed by the optical backscatter sensor; pulses in optical backscatter at 650 nm can be seen in the BML repeatedly throughout the three day survey ( Fig. 3d and e). Apparent oxygen utilisation (AOU, defined as the difference between the solubility of dissolved oxygen and observed oxygen concentration; Fig. 3f) shows a pattern typical of oxygen change within the central stratified North Sea (Queste et al., 2013). The SML is homogeneously saturated. At the pycnocline, supersaturation (negative AOU) is visible and correlated with elevated chlorophyll a concentration (Fig. 4f); this supersat-5 uration is evidence of strong primary production occurring at the DCM (Weston et al., 2005). The BML exhibits increasing AOU throughout the survey; this oxygen depletion is caused by the remineralisation of organic matter within the BML and isolation from air-sea exchange by the strong density gradient. 10 To constrain the balance of DO consumption and supply, we quantify the various sources and sinks of DO and compare these with the observed change in DO over the survey period. Pena et al. (2010) review the processes that affect DO (Eq. 1). Advective-dispersive transport, air-sea exchange, and biological production/consumption are the three dominant processes which can be described mathe-15 matically as:

Dissolved oxygen budget
where O 2 is the concentration of DO, (u,v ,w ) is the water velocity, z is depth, K Z and K H are vertical and horizontal eddy diffusivity respectively, ase is air-sea exchange and bio biological processes (photosynthesis, respiration, bacterial remineralisation of 20 organic matter).
In the context of a strongly stratified two-layer system, air-sea exchange (ase) does not affect the BML and can be neglected. In this region and on short time scales (3 days), the u and v components (horizontal advection by currents), K Z (vertical eddy  et al., 2013). Therefore, for the BML in this region, the w component (vertical crossthermocline transport) and biological processes (bio) dominate (Greenwood et al., 2010).

5
As evidenced by the chlorophyll a and supersaturation signatures (Fig. 3), the majority of the production occurs within the DCM and SML. Consequently, the dominant biological processes within the BML are (i) benthic oxygen demand and (ii) the remineralisation of organic matter which may derive from (ii.a) sinking from the DCM and (ii.b) organic matter still in suspension issuing from the spring bloom. We can therefore 10 simplify the equation describing supply and consumption of DO within the BML to: where R is respiration of (i) Benthic organic matter and originating from (ii.a) the DCM and from (ii.b) past production (Bloom).

Observed change in dissolved oxygen in the BML
15 Figure 6 illustrates AOU at different depths throughout the BML. AOU is shown at depths beginning several metres below the pycnocline to account for tidal vertical displacement. There is a clear trend of uniformly increasing AOU over time, indicating oxygen depletion, throughout the entire BML. A linear regression of mean AOU throughout the BML indicates a rate of 2.8 µmol dm −3 day −1 of DO consumption with a standard 20 error of the regression of 0.3. Three sharp decreases in AOU are indicated by the vertical dotted line. These relate to vertical mixing events identified in Fig. 7 and will be discussed in the following section. An oxygen consumption rate of 2.8 ± 0.3 µmol dm −3 day −1 is very large; to reach the observed AOU of 76 µmol dm −3 from saturation would require a month. However in this  et al., 2006), more than three months before the survey occurred. If this rate were representative of consumption rates throughout the summer season, the North Sea would suffer from severe hypoxia. Greenwood et al. (2010) observed a lower rate of DO consumption, with an average of 0.4 µmol dm −3 day −1 across the entire summer season at ND. It is therefore 5 likely that strong reoxygenation events occurred or the mean rate of oxygen consumption was lower during the first half of the stratified season and that the glider observed a temporal local maximum in consumption rates. Despite this, these observations are representative of specific conditions which occur in the central North Sea, likely a recurring maximum of depletion rates within the summer season. It therefore follows that 10 consumption of organic matter in the sediment and pelagically (from the DCM and spring bloom) surpasses the supply of oxygenated water from cross-thermocline exchange by 2.8 ± 0.3 µmol dm −3 day −1 (Eq. 3).

µmol dm
The following sections aim to quantify the remaining terms in the equation using previ- 15 ous studies and the glider observations, in order to close the budget.

Supply of oxygenated water
To observe supply of DO to the BML through time, ideally the glider would act as a Lagrangian platform moving with the water. This requires no horizontal displacement of the glider relative to the water.  Fig. 3b where the Seaglider observed a freshening by 0.3 of the SML at the end of the survey. This change was not reflected in the BML, where the glider observed a weak but opposing trend. In the BML, the Seaglider observed a 0.1 • C cooling and 0.03 salinifi-5 cation (together equivalent to a densification of 0.035 kg m −3 ), likely due to increasing Atlantic water influence.
Although both 2010 and 2011 show an overall similar spatial pattern with depleted oxygen in the northern half, areas of lowest oxygen saturation are located in different regions (Fig. 2). Given the potentially very high consumption rates, these patterns of 10 oxygen saturation are mainly a snapshot set of observations of late summer and may not be representative of the entire season or the previous month. This illustrates the interannual variability and importance of horizontal transport processes on broad scales in the North Sea, and also potentially in other shelf seas, linked to other dominating climate modes. Horizontal transport processes likely affect seasonal oxygen depletion by 15 influencing BML and SML temperatures thereby affecting stratification while also providing nutrients to sustain productivity and potentially supplying labile organic matter to the BML.
On time scales similar to this survey, it is likely that advective processes play a limited role in regulating seasonal oxygen depletion; instead, vertical mixing is assumed to be 20 the largest potential source of oxygen input into the BML. Seaglider observations reveal a highly stratified water column with a strong pycnocline spanning approximately 3 m with a density difference of 1.5-2 kg m −3 (Fig. 3c). Evidence of mixing can be found in changing properties of the BML as observed by the glider. Here, temperature is used as a tracer to assess mixing across the pycnocline on short time scales through the 25 large temperature difference between the SML and BML (Fig. 7). Temperature data from below the thermocline are used to derive the vertical temperature gradient in the BML (Fig. 7,  in vertical temperature gradients on the order of a few hours (or hundreds of metres) horizontally. The BML temperature gradient increases simultaneously to mean BML temperature. This indicates warming of BML water by mixing across the thermocline and injection of warm SML water into the cooler BML. Referring back to Fig. 6, we observe similar fluctuations in AOU. The three principal 5 mixing events, illustrated by the grey vertical lines, correspond to sudden decreases in AOU indicating sudden and rapid reoxygenation of the BML. After these reoxygenation events, net oxygen consumption rates increase within a few hours. The three events over the three days show increases in BML DO (decreases in AOU) of approximately 1 to 3 µmol dm −3 per event (2 ± 1 µmol dm −3 day −1 ). Figure 8 shows wind and tidal velocities along with bathymetry at the glider's position. There is no obvious correlation between the mixing events identified in Fig. 7 and the speed or direction of the tidal velocities. The majority of mixing events tend to occur during periods of peak wind velocities, but the duration of mixing events does not seem to match the time scale of wind velocity changes; furthermore, two mixing events occur 15 in periods of low wind velocities. Mixing events do not seem to correlate with bathymetry in the region either (Fig. 8).
These mixing events are short scale (a few hours or a couple of kilometres) processes that the Seaglider is able to reveal through its very high resolution observations. Since wind events do not seem to be responsible, these events may be linked to 20 a variety of physical processes occurring in the region. Van Haren et al. (1999) highlighted the importance of internal waves as a potential source of mixing around the Dogger Bank. Another possibility is the generation of shear spikes as described by Burchard and Rippeth (2009) due to the interaction of wind and tides, explaining the lack of correlation with either one in particular. 12,2015 Drivers of summer oxygen depletion in the central North Sea We now assess the relative magnitude of the different oxygen sinks within the BML. As the North Sea is a shallow region, vertical transport of organic matter from the SML and BML to the benthic layer is fairly rapid. On a whole North Sea scale, Van Raaphorst et al. (1998) report that 17 to 45 % of primary production is remineralised in the sediments, although the bulk of this occurs in the Skagerrak and Norwegian Trench 10 area (50 to 70 %) due to transport processes. On a more local scale, tidal currents resuspend and transport this organic matter to temporary depocentres. This makes estimating benthic respiration across the North Sea difficult due to the consequent spatial heterogeneity. Van Raaphorst et al. (1998) report that particulate matter from the East Anglian plume and the southern North Sea may settle in the OG and the 15 southern flank of the Dogger Bank. Neubacher (2009) directly measured nutrient and oxygen dynamics at the benthicpelagic interface at the ND site. DO uptake by the sediment was determined to be approximately 250 µmol m −2 h −1 with variations linked to organic matter export to the benthic layer. Assuming a uniformly mixed BML with a height of 38 m (Fig. 3), this amounts 20 to a decline in oxygen concentration of 1.6 µmol dm −3 day −1 consumed throughout the BML from respiration in the sediment.

Remineralisation of DCM-originating organic matter
Primary production estimates for the North Sea range from 40 to 300 g C m −2 yr −1 depending on area and technique (modelling or empirical), with maxima in the central North Sea area (north of Dogger Bank; Weston et al., 2005). A transect by Weston et al. (2005) north of Dogger Bank showed whole water column-integrated primary pro-5 duction values of 167, 370 and 270 g C m −2 yr −1 for areas classified as Dogger Bank, the front (located along the northern edge of Dogger Bank) and the stratified area (the ND site) respectively. More recently, the importance of the DCM has been reviewed. The DCM contribution to total primary production for the stratified North Sea during the summer has been 10 estimated at between 58 and 60 % (Fernand et al., 2013;Weston et al., 2005). Values for new production range from 37 to 66 % produced at the DCM for the whole water column (Fernand et al., 2013;Weston et al., 2005). As the DCM relies on small scale mixing across the pycnocline to provide nutrients from below, it is necessarily located very near to the pycnocline. Consequently, any mixing also injects highly labile organic 15 matter into the BML. Furthermore, sinking organic matter produced at the DCM is exported to the BML immediately; therefore any remineralisation of organic matter occurs in the BML. This means that nearly all export production from the DCM is consumed within the BML.
The Seaglider observed DCM chlorophyll a concentrations of 2 to 3 mg m −3 in 2011 20 (Fig. 3). These values are similar to those found by Fernand et al. (2013) but lower than concentrations observed by Weston et al. (2005) (ca. 5-10 mg m −3 ). This is likely explained by the systematic variation in DCM chlorophyll a concentrations across the North Sea. The site considered here is located further north than the site investigated in Weston et al. (2005) study and is not subject to as much mixing from internal waves 25 fuelling the DCM in the ND region (Ducrotoy et al., 2000;Weston et al., 2005). Chlorophyll a concentrations at the DCM fluctuate on a near diurnal cycle (Fig. 3). This may be linked to either fluctuations in light availability, or a diurnal change in wind velocities. Increased wind speed could lead to an increase in chlorophyll a fluorescence a few hours later by encouraging nutrient supply to the DCM. Fluctuations in chlorophyll a at the DCM (Fig. 3) are not due to quenching as periods of high DCM chlorophyll a concentrations are present when surface chlorophyll a fluorescence is affected by high light intensity. Instead, it is likely that these fluctuations in chlorophyll a fluorescence at 5 the DCM are caused by diurnal increases in nutrient supply linked to stronger winds in the mornings (Fig. 8).
We now investigate the export of DCM organic matter to the BML. Figure 3d show a similar pattern for BML chlorophyll a concentration and DCM chlorophyll a concentrations. We observe a delay between peaks in DCM chlorophyll and BML chlorophyll with BML chlorophyll a concentrations peaking a few hours later. As we have already observed a delay between peaks in wind velocities and DCM chlorophyll a concentrations, we can assume that wind-driven mixing is not responsible for injecting these large amounts of organic matter into the BML. Neither are these increases in BML chlorophyll a correlated to tidal velocities or changes in BML salinity, thus ruling out the 15 influence of horizontal transport processes or tidal mixing on export of DCM organic matter to the BML. It seems likely that the wind drives production at the DCM but export mechanisms are driven by biological processes.
The Seaglider data show pulses, or short-lived increases, in BML chlorophyll a concentrations of 0.15 mg m −3 occurring on daily cycles. We assume these are events 20 where DCM organic matter sinks to the BML. With an approximate BML height of 38 m, these pulses equate to 5.7 mg m −2 day −1 of chlorophyll a export to the BML. Using a particulate organic carbon (POC) to chlorophyll a ratio of 50 : Oxygen consumption arising in the water column from recently exported matter (0.9 µmol dm −3 day −1 ) and benthic processes (1.6 µmol dm −3 day −1 ) provides a potential for oxygen consumption of 2.5 µmol dm −3 day −1 . This is described as "potential" as the value is likely an overestimate of oxygen consumption rates because it is unlikely that all of the sinking organic matter is remineralised pelagically. Much of this sink-10 ing organic matter is deposited before being remineralised and therefore a portion of the DCM originating consumption (0.9 µmol dm −3 day −1 ) is already accounted for in the benthic consumption (1.6 µmol dm −3 day −1 ). Furthermore, no POC : chlorophyll a ratio could be obtained from North Sea DCM phytoplankton communities in this region. It is likely that in reality the ratio is in fact lower guished; horizontal advection does occur throughout the bottom mixed layer. Such an event occurs at the end of the record and is excluded from our study. Even by accounting for additional organic matter issuing from the DCM and SML (faecal pellets, lysed cells, zooplankton) and doubling the potential for oxygen consumption from the DCM, the sum of oxygen consumption potential from the DCM and 5 the benthic compartment remain only marginally larger than the total net observed rate of dissolved oxygen consumption within the BML (2.8 µmol dm −3 day −1 ). This either implies oxygen resupply rates much lower than the 2 µmol dm −3 day −1 identified in Fig. 7, or more likely, the existence of an additional oxygen sink relevant to the time and location of the survey which has not been accounted for.
In the results of this study, observed net consumption is significantly greater than previously reported values in the literature. As previously stated, the observed rate of apparent oxygen utilisation would lead to lower oxygen saturations than observed if 15 maintained from the start of the stratified season until the glider deployment. The implication here is that the Seaglider surveyed the region during a short-lived or localised increase in apparent oxygen utilisation.
If we look at the budget in a wider seasonal context and ignore the glider-observed net consumption rate (2.8 µmol dm −3 day −1 ), the estimates of oxygen supply and con-20 sumption derived above (−2 + 2.5 µmol dm −3 day −1 ) are in agreement with past studies of the North Sea where net seasonal consumption rates of 0.4 µmol dm −3 day −1 were observed (Greenwood et al., 2010). This also points to the unknown sink being either short-lived or very localised and explains why the observed net consumption rate does not correlate with previous studies.

Potential dissolved oxygen sinks
Several processes could account for this unidentified oxygen sink. Bacterial recycling within the BML is rarely accounted for and could contribute to this unaccounted for oxygen depletion. In addition, recent work has begun focusing more on the importance of resuspension events in seasonal oxygen depletion (van der Molen et al., 2012;Cou-5 ceiro et al., 2013;Thompson et al., 2011;Greenwood et al., 2010). Background benthic DO consumption is limited by oxygen penetration depth and Red-Ox levels (Neubacher, 2009). When deep benthic organic matter is resuspended into an oxic water column, bioavailable surface area is greatly increased and oxygen is readily available leading to very rapid and short-lived oxidation of reductants such 10 as nitrite, ammonium and sulphide. Organic matter is likely also briefly degraded more rapidly when resuspended and more surface area is exposed to the oxic BML.
Three particular causes have been identified as providing enough turbulent energy to the seafloor to resuspend organic matter: tidal resuspension, large wind mixing events and trawling (van der Molen et al., 2012;Couceiro et al., 2013;Thompson et al., 2011;15 Jago et al., 200615 Jago et al., , 2002Van Raaphorst et al., 1998;Greenwood et al., 2010). Current speeds and storm events sufficient to cause resuspension have been recorded at both the OG and ND sites (de Jonge et al., 1996;Kröncke and Knust, 1995;Greenwood et al., 2010;Weston et al., 2008). The influence of trawling has only recently been investigated but has been shown to potentially lead to a small (0.5 %) decrease in BML 20 DO (van der Molen et al., 2012) in modelling studies. A strong resuspension event leading to an even greater decrease in DO was documented by Greenwood et al. (2010) at the Oyster Grounds.
In the 2011 survey data, glider observations of optical backscatter at 650 nm in the BML show the existence of resuspension events (Fig. 3e). Glider optical backscat-  (Fig. 4e). The peaks in BML optical backscatter do not correlate with observed cross-pycnocline mixing (Fig. 7) but seem to occur on a diurnal cycle, about 4 to 6 h after daily peak wind speeds (Fig. 8).

5
Correctly understanding and quantifying resuspension becomes increasingly important when one considers temporary depocentres. These sites accumulate organic matter originating from much wider regions through slowing of currents. These sites have the capacity to rapidly sequester large amounts of organic matter, not only during peak production but also throughout the winter. This creates a reservoir of highly labile organic matter which, when resuspended, leads to very rapid consumption of DO. Depocentres are not exclusive to the North Sea; they are driven by local hydrography and topography, acting as hotspots of oxygen depletion in shelf sea environments around the world.

15
This study, through the use of an autonomous underwater ocean glider, has given new insights into the time scales of processes that regulate the supply and consumption of oxygen across thermoclines and into the BML. The average observed rate of oxygen consumption rate of 2.8 ± 0.3 µmol dm −3 day −1 is high; this would have resulted in the observed AOU of 76 µmol dm −3 within only a month of the onset of stratifica-20 tion. The glider observations' high temporal and vertical resolution reveals the periods when resupply of oxygenated water occurs, and enables these vertical processes to be distinguished from horizontal processes. Long term trends in net oxygen consumption hide a more complex short term variability reflecting variations in oxygen supply and consumption. During a period when the BML was isolated, the AOU was shown 25 to increase significantly over a short 3 day period. The observed vertical mixing events have no apparent correlation with winds or tides, and remain a process to be further explored. Optical backscatter readings from the Seaglider in the BML showed varying suspended sediment, which generally increased near the bed and varied on diurnal time scales. Greenwood et al. (2010) recorded average (since stratification) oxygen depletion rates in 2007 of 0.43 µmol dm −3 day −1 at ND and 0.75 µmol dm −3 day −1 at the OG.

5
We observed export of organic matter from the DCM with the potential of consuming 0.9 µmol dm −3 day −1 , and the study by Neubacher (2009) found rates of 1.6 µmol dm −3 day −1 for the sediment-water oxygen flux in incubation experiments.
These rates highlight two particular aspects of the mechanisms governing seasonal oxygen depletion in the North Sea. The first is that consumption along the seafloor plays a predominant role in the consumption of BML oxygen. This supports the idea that depocentres, such as the OG, are particularly prone to seasonal oxygen depletion. Advected organic matter will add to the consumption budget, increasing the ratio of consumption to supply, whilst the reduced water flow that promotes this deposition also promotes stronger stratification. The second is that supply and consumption of 15 organic matter are tightly coupled. Net oxygen depletion rates averaged over a season are an order of magnitude smaller than gross oxygen consumption and supply rates. A small shift in either gross DO consumption or supply would have a large impact on net oxygen consumption rates. This study has shown the potential of new AUVs in regions such as the North Sea. 20 This deployment was cut short and only 5 % of battery capacity was utilised; a full duration glider deployment within the North Sea could last upwards of two months. Despite the deployment lasting only three days, the data revealed small scale mixing mechanisms that would not be identifiable with traditional oceanographic methods. A longer glider deployment would provide more confidence in these estimates of rates 25 and would reveal the temporal variability of biological processes across an entire season. A Seaglider performing repeat transects in and out of known depocentres could provide critical high resolution observations of subsurface primary production and organic matter transport in regions of both low and high biomass accumulation. A longer survey would also increase the likelihood of observing larger mixing events (i.e. storm events) and the impact these have on resuspension of organic matter and subsequent oxygen drawdown. High resolution observations, such as those obtained by a glider, are critical to improving shelf sea ecosystem models. There remain limitations to such models; many processes are simplified and they do not resolve finer 5 processes due to lack of understanding and observations, particularly feedback and subsurface processes. van der Molen et al. (2012) highlight that, in the case of the North Sea, GETM-BFM's representation of advection, particulate organic carbon transport and remineralisation requires further knowledge to adequately represent these carbon and oxygen dynamics. These are gaps that autonomous underwater vehicles, 10 by providing season-long observations, could fill. These models could then, in turn, fill the gaps in glider observations to provide a comprehensive view of the processes occurring in these highly dynamic and heterogeneous environments. As it is now well established that many shelf seas undergo seasonal oxygen depletion (Diaz and Rosenberg, 2008), it is critical to increase the presence of persistent observation systems to Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | of ecosystem health, Biogeosciences, 7, 1357Biogeosciences, 7, -1373Biogeosciences, 7, , doi:10.5194/bg-7-1357Biogeosciences, 7, -2010Biogeosciences, 7, , 2010 Hill, A., Brown, J., Fernand, L., Holt, J., Horsburgh, K., Proctor, R., Raine, R., and Turrell, W.: Thermohaline circulation of shallow tidal seas, Geophys.     depth (m, c) at the Seaglider's location during the survey. Tidal data was obtained from the TMD tide toolbox and OTIS European Shelf Model (Egbert et al., 2010). Wind data was obtained from ECMWF ERA-Interim reanalysis data. Bathymetry was gathered by the Seaglider's on-board altimeter.