Sediment oxygen consumption: Role in the global marine carbon cycle

The seabed plays a key role in the marine carbon cycle as a) the terminal location of aerobic oxidation of organic matter, b) the greatest anaerobic bioreactor, and c) the greatest repository for reactive organic carbon on Earth. We compiled data on the oxygen uptake of marine sediments with the objective to understand the constraints on mineralization rates of deposited organic matter and their relation to key environmental parameters. The compiled database includes nearly 4000 O 2 uptake data and is available as supplementary material. It includes also information on bottom water O 2 concentration, O 2 penetration depth, geographic position, water depth, and full information on the data sources. We present the different in situ and ex situ approaches to measure the total oxygen uptake (TOU) and the diffusive oxygen uptake (DOU) of sediments and discuss their robustness towards methodological errors and statistical uncertainty. We discuss O 2 transport through the benthic and diffusive boundary layers, the diffusion- and fauna-mediated O 2 uptake, and the coupling of aerobic respiration to anaerobic processes. Five regional examples are presented to illustrate the diversity of the seabed: Eutrophic seas, oxygen minimum zones, abyssal plains, mid-oceanic gyres, and hadal trenches. A multiple correlation analysis shows that seabed O 2 uptake is primarily controlled by ocean depth and sea surface primary productivity. The O 2 penetration depth scales with the DOU according to a power law that breaks down under the abyssal ocean gyres. The developed multiple correlation model was used to draw a global map of seabed O 2 uptake rates. Respiratory coefficients, differentiated for depth regions of the ocean, were used to convert the global O 2 uptake to organic carbon oxidation. The resulting global budget shows an oxidation of 212 Tmol C yr (cid:0) 1 in marine sediments with a 5-95% confidence interval of 175-260 Tmol C yr (cid:0) 1 . A comparison with the global flux of particulate organic carbon (POC) from photic surface waters to the deep sea, determined from multiple sediment trap studies, suggests a deficit in the sedimentation flux at 2000 m water depth of about 70% relative to the carbon turnover in the underlying seabed. At the ocean margins, the flux of organic carbon from rivers and from vegetated coastal ecosystems contributes greatly to the budget and may even exceed the phytoplankton production on the inner continental shelf.


Introduction
The seabed is the largest reservoir of organic matter on Earth and is a long-term sink for biomass produced in the ocean or supplied from rivers. A fourth of the particulate organic carbon (POC) exported from the photic zone of the global ocean reaches the seafloor (Dunne et al., 2007). Along the continents, there is also an important lateral transport of POC from terrestrial and coastal ecosystems to the sea (Regnier et al., 2013). Less than a tenth of the globally deposited POC is buried and stored in the seabed over thousands to millions of years (Wallmann et al., 2012). By far the most is mineralized back to CO 2 and inorganic nutrients that are subsequently released from the sediments into the water column.
CO 2 from the atmosphere equilibrates with the surface ocean and is assimilated by phytoplankton. A fraction of this fixed, particulate organic carbon sinks out of the photic zone and down into the dark ocean. This generates a biological pump by which carbon is sequestered in the deep ocean and in the seabed. By the net burial of organic carbon, the seabed provides an important long-term sink in the global CO 2 balance. It also affects the global O 2 balance in the atmosphere, which, due to the 500-fold larger concentration of O 2 than of CO 2 , is regulated on a much longer time scale of many thousands of years. Our current understanding of the deposition and mineralization of organic matter in the seabed is based on many types of data and analyses, ranging from remote sensing of the chlorophyll distribution and experimental productivity studies in the photic zone to sediment trap studies, measurements of the sediment community oxygen consumption, and modeling. There are several well-established approaches to determine the Total Oxygen Uptake (TOU) or the Diffusive Oxygen Uptake (DOU) of marine sediments, including a) the enclosure of sediment to measure the rate of oxygen depletion in the overlying seawater, b) the measurement of oxygen penetration into the sediment and calculation of the diffusion flux, and c) the measurement of oxygen and turbulence of water flow at a point above the sediment and calculation of the oxygen flux by eddy covariance technique. The measurements by a) and b) may be done either directly on the seafloor using flux chambers and profiling instruments or shipboard in retrieved sediment cores.
In order to understand the role of seabed respiration in the global marine carbon cycle, we established a comprehensive database of sediment O 2 uptake rates and accompanying environmental information from the literature and from public databases. From these data we developed algorithms for a multiple correlation of TOU or DOU to key parameters. The algorithms are used to map the global distribution of oxygen uptake of the seafloor and determine the environmental controls on benthic oxygen uptake and organic carbon turnover. A main objective was thereby to establish a global budget for the oxygen and carbon cycles of the seabed, including analyses of their robustness towards methodological errors and statistical uncertainty.

What is in this review
Our data compilation and analysis includes only marine sediments. Our objective is to relate pelagic marine primary productivity to export from the photic zone, deposition on the seafloor, and seabed mineralization processes. The sediments receive POC both from the overlying water column and from lateral transport by ocean currents and from land. The sediments also receive material that moves along the seabed through resuspension-resettling cycles or through down-slope mass wasting. In shallow, coastal regions there is a large organic production directly on the seafloor in the form of microbenthic photosynthesis and growth of seagrasses that contribute to the organic source. In our analysis, we do not include vegetated coastal ecosystems, such as salt marshes, mangroves, macroalgae or seagrass beds. Due to their great heterogeneity, intertidal mud flats or rocky bottoms are also not included in the dataset.
We start this review with a presentation of the TOU and DOU database and our rationale for selecting only in situ data for the depth range 10-6500 m by the global modeling. This is followed by an overview of the mechanisms of seabed oxygen uptake, the role of diffusion and faunal activity, and the coupling to anaerobic processes in the sediment. We then discuss five examples of sediment environments that demonstrate the great diversity of the seabed. The global distribution of in situ sediment oxygen uptake is mapped, using a multivariate correlation analysis of the database, and the controls on seabed oxygen uptake are discussed. This provides the basis for a global budget of seabed oxygen uptake and a new perspective on its role in the marine organic carbon cycle. The constraints on the budget are discussed in relation to other available information on organic carbon fluxes to the seafloor and the mineralization or burial of organic matter.

Different approaches
There are generally three basic approaches to quantify the benthic O 2 consumption rates: by sediment incubations, by flux calculations from oxygen profiles, and by eddy covariance technique. Below we will briefly describe these approaches and the extent to which they provide accurate data for the in situ O 2 consumption or possibly interfere with the solute exchange of the seabed.

Total O 2 uptake by sediment incubations
The most widely used approach for quantifying the benthic O 2 consumption rate is to enclose a sediment area and monitor the O 2 decline in the well-mixed overlying water phase (e.g., Pamatmat and Fenton, 1968). This can be achieved by incubating sediment cores in the laboratory (ex situ) or by placing benthic chambers directly on the seafloor (in situ). When knowing the enclosed sediment area and water volume, the O 2 decline can be converted into a total oxygen uptake rate of the sediment. A commonly used unit, which will also be applied here, is mmol O 2 m − 2 day − 1 . In order to determine the O 2 consumption rate of the undisturbed seabed, it is important not to change critical environmental parameters. Numerous designs of benthic chambers and sediment core incubation systems have been developed and applied over the years (reviewed by Tengberg et al., 1995), but given that the incubation approach is invasive, it is difficult not to affect the benthic system. It should always be considered how best to ensure that temperature, O 2 concentration, flow velocity and faunal activity during the incubation remain as close to natural conditions as possible.
Temperature affects metabolic rates and the solubility and transport of O 2 . It is therefore critical to incubate sediments at in situ temperature. If the sediment has been transiently heated during recovery, sufficient time should be given for the system to revert to the earlier steady state O 2 distribution before starting measurements. It can be more challenging to ensure unaltered O 2 availability, especially in environments with low O 2 concentration. By laboratory incubations, the correct initial O 2 concentration of the overlying water can be established using a manual or digital gas-mixer (Rasmussen and Jørgensen, 1992). Given that O 2 is gradually consumed during incubation, the decreasing O 2 concentration of the overlying water will ultimately affect the O 2 uptake rate (Hall et al., 1989;Kononets et al., 2021), and the initial, linear O 2 decline should therefore be determined (Bender et al., 1989;Glud, 2008).
The flow velocity of seawater affects the benthic O 2 uptake, both in permeable sand and in non-permeable sediments with high O 2 consumption rates where the thickness of the Diffusive Boundary Layer (DBL) affects the O 2 exchange (Jørgensen and Revsbech, 1985;Boudreau, 2001;Glud et al., 2007;Chapter 3.2). Stirred benthic chambers or laboratory core incubation systems have been designed to establish a rather uniform flow velocity and DBL thickness across most of the sediment surface (Berelson and Hammond, 1986;Buchholtz-ten Brink et al., 1989;Glud et al., 1995a;Khalili et al., 2008). In permeable sediments, advective porewater transport is induced by wave action, nearbed flow gradients, and bioirrigation (Huettel et al., 2014;Jahnke et al., 2000;Reimers et al., 2004). The near-bed hydrodynamics are critical for the solute exchange but are difficult to establish during enclosure. The strategy has therefore been to establish well-defined pressure gradients over the sediment and impose known rates and patterns of porewater advection to assess the TOU under reproducible and realistic conditions (Huettel and Gust, 1992b;Janssen et al., 2005).
Fauna respiration and bioirrigation contribute to the benthic O 2 uptake (Aller and Aller, 1998;Kristensen et al., 2012;Snelgrove et al., 2018), yet it is difficult to assess whether an enclosure maintains the natural activity of the fauna. The best representation of the natural benthic community is to apply relatively large chambers (Tengberg et al., 1995), which increase the probability that also the larger fauna is included and that no, or only few, animals are damaged during insertion of the chamber (Glud and Blackburn, 2002). However, sampling instruments designed to recover sediment cores with an intact surface generally use rather small core liners of ≤10 cm diameter (Barnett et al., 1984). They tend to exclude larger fauna and to affect faunal behavior. The TOU of laboratory incubations therefore generally underestimate the in situ TOU measured by larger benthic chambers (Archer and Devol, 1992;Glud et al., 1998Glud et al., , 2003Ferrón et al., 2008). Changes in light conditions, flow velocities, or concentrations of suspended material can also lead to alterations of faunal behavior. This affects the bioirrigation and ventilation of anoxic sediment layers. Passive flushing of relict burrows in the seabed also depends on flow conditions (Munksby et al., 2002).

Small-scale benthic heterogeneity and diffusive O 2 uptake
The fine-scale O 2 distribution across the sediment-water interface can be resolved by Clark-type microelectrodes (Revsbech et al., 1980;Revsbech, 1989) or by microoptodes (Klimant et al., 1995). These enable the quantification of the Diffusive Oxygen Uptake (DOU) of the sediment. The DOU can be determined in different ways, but the simplest and most widely used approach is to apply Fick's first law of diffusion. If O 2 is measured in the DBL, the flux is simply the product of the molecular diffusion coefficient and the O 2 gradient (Jørgensen and Revsbech, 1985). If the O 2 gradient is measured just beneath the sediment surface, also the sediment porosity and the tortuosity must be taken into account (e.g., Andrews and Bennett, 1981;Ullman and Aller, 1982;Glud et al., 1995;Revsbech et al., 1998; Chapter 3.1 and 3.2). Finally, the DOU can be derived from the whole porewater O 2 profile by modeling the volume-specific O 2 consumption rate, which also requires knowledge of porosity and the tortuosity in the sediment (Berg et al., 1998). Oxygen microprofiles provide detailed insight into the distribution and turnover of O 2 at a given position in the sediment. However, sediments are often characterized by small-scale heterogeneity, and a number of O 2 microprofiles may be required to reach a representative value for the DOU of a seafloor area Glud et al., 2009a).
In continental shelf and slope sediments, the Oxygen Penetration Depth (OPD) can be determined directly from measured O 2 profiles, which are usually restricted to <10 cm depth. In the oligotrophic deep sea, oxygen penetration often exceeds the measurement depth achieved by vertically inserted microsensors. An oxygen penetration depth of several meters, as found in mid-oceanic gyre sediments, can then be measured by retrieving whole sediment cores and inserting microsensors into the side through small holes in the core liner (Røy et al., 2012;Fischer et al., 2009). It is necessary to use microsensors for such measurements as these cause minimal disturbance and have minimal oxygen consumption (micro-optodes do not consume O 2 ).
The difference between TOU and DOU measured concurrently in non-permeable sediments reflects the faunal contribution to the benthic O 2 uptake (Rasmussen and Jørgensen, 1992;Glud et al., 2003). The difference includes the faunal respiration and the O 2 uptake along burrow walls surrounded by otherwise anoxic sediment. More detailed insights into the spatial benthic O 2 distribution can be obtained by planar optodes fixed to the wall of a benthic enclosure or to an "inverted periscope" (Glud et al., 1996(Glud et al., , 2001. The O 2 distribution in the sediment can thereby be imaged over an area of several cm 2 in real time and provide important information on the spatio-temporal O 2 dynamics around burrow structures or seagrass roots (Wenzhöfer and Glud, 2004;Volkenborn et al., 2012;Santner et al., 2015). The approach is invasive, however, and conditions along the wall may not correctly represent conditions of the undisturbed seabed (Polerecky et al., 2006;Meysman et al., 2010).

In situ versus ex situ measurements of DOU and TOU
Both DOU and TOU can be measured ex situ in recovered sediment cores or in situ directly on the seabed, either by divers in shallow water or by benthic landers in deeper waters (Fig. 1).
In situ technologies for measuring benthic fluxes have been developed for a wide depth range, from shallow waters to the deep sea, usually with 6000 m as the maximal depth rating (e.g., Pamatmat and Fenton, 1968;Smith Jr. et al., 1976;Reimers, 1987;Tengberg et al., 1995;Reimers and Glud, 2000;Reimers et al., 2001). Recently, lander systems have been developed that cover the depth range down to the bottom of the deepest hadal trenches (11,000 m) Luo et al., 2018).
Parallel in situ and ex situ O 2 microprofile measurements in deep sea sediments have revealed that core recovery can induce significant alterations in the consumption and distribution of O 2 in the sediment (Glud et al., 1994). The difference between ex situ and in situ data increases with increasing water depth because sample retrieval alters chemical, physical and biological properties. Oxygen microprofiles measured on board a ship generally exhibit shallower O 2 penetration depths and higher O 2 consumption rates than measured in situ (Fig. 2). This difference increases with increasing water depth (Chapter 5.2.5). The effect is ascribed to strong decompression and transient heating during core recovery, both of which may cause damage and cell lysis to sensitive organism and induce enhanced metabolic activity of resilient microbes that benefit from released, labile organic compounds . This effect is reflected in elevated concentrations of dissolved organic matter and ammonium in the surface sediment of cores recovered from the deep sea (Hall et al., 2007). When incubating the recovered sediment cores at in situ temperature, the O 2 microprofile may gradually revert to the original shape. Yet, given that O 2 diffusion is a slow process, this may take days to months if the O 2 penetrates several cm or dm into the sediment. Therefore, ex situ measurements tend to overestimate the benthic O 2 consumption rate and underestimate the O 2 penetration depth in the deep sea (Glud, 2008).
Much of our knowledge about O 2 exchange and carbon mineralization at the vast ocean floor originates from single point ("snapshot") measurements by stationary lander systems. However, marine sediments often exhibit strong heterogeneity and lateral variability at different scales. At the small scale, this can be visualized by multiple O 2 microsensor measurements in recovered sediment cores  or directly in situ (Glud et al., 2009a). Both approaches have demonstrated the existence of mm to cm-scale microbial hotspots with intensified O 2 consumption, presumably associated with local deposition of POC or fauna activity. On an intermediate dm-scale, such a mapping of DOU and OPD requires a moving, transecting microprofiler. Deep sea "crawlers" (Fig. 1D;Sherman and Smith Jr., 2009, Wenzhöfer et al., 2016a, Lemburg et al., 2018 can move autonomously over the seabed while repeatedly inserting chambers and microsensors during long-term deployments. This has documented lateral heterogeneity and temporal variability, such as large organic aggregates or abrupt deposition events (Smith Jr. et al., 2008von Appen et al., 2021). Future studies should focus on such benthic dynamics by more extensive use of long-term observatories and integrative technologies to quantify benthic carbon mineralization (Smith Jr. et al., 2013;Soltwedel et al., 2016;Toussaint et al., 2014).

Benthic O 2 exchange determined by Eddy Covariance
In recent years, the Aquatic Eddy Covariance (AEC, also called eddy correlation) approach has been increasingly applied to quantify benthic O 2 exchange. The technique was introduced to aquatic applications by Berg et al. (2003). The in situ benthic O 2 exchange rate is determined from concurrent measurement of fluctuations in the vertical flow component (w') and the O 2 concentration (C ′ ) in the same minute water volume within the turbulent boundary layer. Provided that w' and C ′ are measured continuously over tens of minutes, and the water flow is parallel to the seabed, the vertical O 2 flux can be calculated from the time-integrated product of w' and C ′ (Berg et al., 2003). The two parameters are typically recorded at 64 or 32 Hz by an Acoustic Doppler Velocimeter (ADV) and a fast O 2 microsensor, respectively.
Measurements are done 10-20 cm above the seabed (Fig. 1), the height depending on the topography and rugosity of the sediment surface .
The eddy covariance measures the benthic O 2 exchange rates of a 10-100 m 2 large ellipsoid footprint upstream from the measuring point . The technique can, thus, provide non-invasive measurements of the benthic O 2 exchange rate across a relatively large sediment area without affecting the in situ conditions (Berg and Huettel, 2008). In this respect it is superior to chamber incubations or O 2 profile measurements for assessing benthic O 2 exchange rates. Due to the rather advanced technology required and the complex data handling procedures, the approach is still only used by relatively few research groups. This situation will change as standardized instruments, software packages, and data treatment procedures are developed (McGinnis et al., 2011;Long, 2021;Berg et al., 2022). Currently, the number of AECderived O 2 uptake data is rapidly increasing. The approach has particularly been applied in complex settings such as seagrass beds, permeable sands, and coral reefs (Long et al., 2013;Rheuban et al., 2014;McGinnis et al., 2014;Reimers et al., 2016). The approach also has great potential in deep sea settings (Berg et al., 2009;Donis et al., 2016).

Data analysis
Benthic oxygen fluxes compiled in our database were derived from previously published data compilations Glud, 2008;Stratmann et al., 2019), supplemented by data from other databases or sources. We searched the literature using combinations of relevant search terms (such as benthic, sediment, O 2 , oxygen, flux, consumption, respiration, and uptake). The benthic oxygen consumption rates were measured using different techniques, including core or chamber incubations, O 2 profile modeling, and eddy covariance. Data are reported accordingly in the database as total or diffusive oxygen uptake. Since our goal was to establish a database on seafloor respiration for the modeling of global organic matter mineralization, a few quality criteria were needed to select the published data: 1) Ex situ data from >1000 m water depth are expectedly compromised and were not included in the database (see discussion in Chapter 2.1; Fig. 2). 2) Shallow coastal sites are influenced by benthic photosynthesis, which produces O 2 within the sediment and adds fresh, labile organic matter at the seafloor. By shallow water oxygen uptake rates, we therefore used only those data that were clearly measured under dark conditions. 3) Low O 2 concentrations affect the efficiency of aerobic respiration, including the re-oxidation of products from anaerobic mineralization, especially in dynamic settings (Middelburg and Levin, 2009;Fig. 1. Instruments for measuring oxygen uptake rates in situ, directly on the seabed. A) Free-falling autonomous benthic lander system carrying three incubation chambers (inset B) and an array of microsensors (inset C) on a transecting micro-profiler. D) Deepsea crawler with two benthic chambers and three multiple sensor profiling units for longterm studies. E) Eddy covariance system deployed at 28 m water depth to measure the benthic O 2 flux from 20 cm above seafloor (photo by Lucas Zañartu). F) Acoustic Doppler velocimeter (ADV) and fast O 2 microsensors used on the eddy covariance instrument (photo by Karl Attard).

Fig. 2.
Effect of core retrieval on sediment oxygen uptake. Two oxygen profiles measured ex situ and in situ, respectively, at the same temperature on the same station at 5500 m water depth in the Atacama Trench off Chile (data from Glud et al., 2021). Levin et al., 2009a). We therefore excluded data from locations with <25 μM O 2 in the bottom water in our global modeling. Sediments underlying such oxygen minimum zones constitute only 4% of the outer shelf and upper slope and play no significant role in the global sediment O 2 uptake budget (Chapter 4.2). 4) Extreme environments, such as fish farms or sites affected by sewage outlets or other local, anthropogenic influence, do not represent natural marine sediments, as benthic communities are modified and benthic consumption rates are enhanced (Sweetman et al., 2014). Data from such sites were therefore excluded.
Using the above criteria, we identified a total of 3904 data on benthic O 2 uptake rates, of which 1780 were measured in situ within the full ocean depth range of 0-11,000 m and 2124 were measured ex situ at <1000 m water depth.
These data on TOU and DOU are compiled in an Excel file and are available as supplementary material (Supplementary Table S1). Data are presented in the unit mmol O 2 m − 2 d − 1 . If originally published in another unit, fluxes have been recalculated to this unit. If multiple measurements were done at the same site, these were averaged before use for our correlation analysis. Such multiple measurements are indicated by "0", while single measurements are indicated by "1" and the mean values of multiple measurements are indicated by "2" in Suppl. Table S1. If available for DOU data, we also included the Oxygen Penetration Depth (OPD). Additionally, we included the station number (if available), bottom water O 2 concentration (in μM or % saturation), geographic position, water depth, and full reference to the data source. We labeled the data according to depth region: Inner shelf 0-10 m, Inner shelf 10-50 m, Outer shelf 50-200 m, Slope 200-1000 m, Slope 1000-2000 m, Rise 2000-4000 m, Abyss 4000-6500 m, and Hadal (>6500 m). Data were also categorized according to ocean (Arctic, Atlantic, Indian, Pacific and Southern Ocean) and according to region (e.g., Black Sea).
We used a multiple correlation analysis to model the trends of our data relative to the controlling environmental parameters and to calculate the global distribution of oxygen uptake in the seabed (Chapter 5). For this purpose, we made a critical quality control of all data in order to ensure that the data included in our analysis fulfilled certain criteria. All data selected for our model analysis are supplied with the identifier "1" or "2" in Suppl. Table S1. The selection criteria for data used for our multiple regression analysis and model were the following: A) We used only measurements done in situ directly on the seafloor for the depth range 10-6500 m in our global regression analysis, for reasons explained above. Oxygen uptake rates at water depths of 0-10 m are treated separately in our assessment (see Chapter 5.2.1). Given that there are only few in situ data from the coastal zone and that recovery artifacts for this depth range are considered minor, also ex situ data from 0 to 10 m were included in our assessment. Date from water depths <10 m are supplied with the identifier "3" in Suppl. Table S1. B) Pore water advection in permeable sediments is an important factor for the benthic O 2 exchange rates (Huettel and Webster, 2001). Pore water advection fluctuates in nature due to varying bottom water currents, wave dynamics and changes in seafloor topography (e.g., by ripple movements). Benthic oxygen consumption measurements in permeable sediments are therefore often strongly biased by these environmental conditions and are consequently excluded. C) For the multiple regression analysis, we also excluded fluxes from hadal trenches, as these are characterized by enhanced accumulation of organic material and relatively high mineralization rates and cannot be quantitatively assessed by extrapolating findings from deep-sea trends in general see Chapter 4.5).
In order to correlate the seafloor O 2 uptake rates to the organic carbon produced in surface waters for the same geographic position, we extracted the net primary production (NPP; mg C m − 2 d − 1 ) from monthly averaged data computed over a 10-year period (1998-2007; available at www.science.oregonstate.edu/ocean.productivity/custom. php). These NPP data are also provided in Suppl. Table S1.
In Chapter 6.5 we discuss the quantitative importance of benthic environments not included in the regression analysis and global extrapolation.

The benthic boundary layer
The benthic boundary layer (BBL) is the lowest part of the water column, in which the flow is distinctly affected by the seabed, and in which shear stress dampens the motion of seawater (Dade et al., 2001). Vertical gradients of solutes that are exchanged between seawater and sediment, such as oxygen and inorganic nutrients, become apparent in the logarithmic layer that comprises the lower meters to decimeters of the BBL Trowbridge and Lentz, 2018). When such gradients are combined with in situ determinations of the turbulent diffusion coefficients, solute fluxes across the sediment-water interface can be calculated (Inoue et al., 2007;Holtappels and Lorke, 2011).
Within a few dm or cm of the seafloor, vertical motion and turbulence fade out and viscous transport of momentum becomes predominant (Boudreau and Jørgensen, 2001). At even smaller scale, in the viscous sublayer within a few cm from the seabed, there is a steep drop in flow velocity towards the stagnant sediment-water interface. In the thin water film adjacent to the sediment surface, with a thickness of a mm or less, water motion is so strongly reduced that the vertical transport of solutes is dominated by molecular diffusion. This is the diffusive boundary layer (DBL). Moving into the sediment, the seawater flow over the topography of coastal sands and other permeable sediments causes pressure gradients that drive a slow advection of pore water within the uppermost centimeters of the sediment. These flow and transport properties of the BBL are critical to consider when performing oxygen uptake measurements because the approaches used may cut off the natural flow and thereby compromise the oxygen flux.
To avoid interference with the benthic boundary layer, the eddy covariance technique was developed for measurements of the oxygen flux to the seabed. A comparison of the eddy covariance data with oxygen flux determinations using benthic chamber incubations or in situ O 2 microprofile measurements have shown rough agreement without a clear systematic difference. Thus, somewhat contrary to expectations, the eddy covariance data were statistically not higher than the other techniques for sediments in which the oxygen uptake was dominated by microbial respiration (Reimers et al., 2012;Attard et al., 2014;Donis et al., 2016). In a sediment with abundant fauna, the eddy covariance showed ca. 30% higher oxygen flux than flux chambers, which again showed about 30% higher oxygen flux than calculated from O 2 microprofiles (Berg et al., 2003).
As described, the diffusive boundary layer drapes the sediment surface as a sub-mm thin, unstirred layer through which solute exchange is dominated by molecular diffusion. The role of this DBL for the oxygen flux across the sediment-water interface has been demonstrated by the use of microsensors in laboratory-incubated sediments (Jørgensen and Revsbech, 1985) and directly on the seafloor (e.g., Reimers et al., 1986;Archer et al., 1989;Gundersen and Jørgensen, 1990;Glud et al., 1994). The thickness of the DBL generally varies between a tenth of a mm and a few mm. It is thinner by higher current velocities of the overlying seawater and thicker by increasing roughness of the sediment surface (Dade, 1993;Røy et al., 2002;Han et al., 2018). Fig. 3A shows an O 2 microprofile measured in situ in a coastal sediment at 10 • C. It illustrates the abrupt shift from the well-mixed water flow above the sediment to the steep gradient in the 0.5-mm thick DBL. Since the O 2 flux is continuous from the mixed layer and into the DBL, the steepness of the vertical gradient above and within the DBL is inversely proportional to the transport coefficient for oxygen and other solutes.
The stability of the DBL depends on currents and turbulence in the overlying seawater and is constantly changing. Dynamic modeling has shown that the benthic O 2 uptake of a coastal sediment may vary by 30% or more within hours or days because of changes in the DBL thickness . The DBL may impede the O 2 uptake of sediments that have a high respiration rate (Kelly-Gerreyn et al., 2005). However, the impedance is generally small, less than 5-20%, even in sediments of the eutrophic coastal zone (Rasmussen and Jørgensen, 1992;Glud et al., 2007).
High-resolution microsensor measurements of O 2 in the DBL enable calculations of the diffusive oxygen flux to the sediment-water interface. Since diffusion through the stagnant DBL is not impeded by particles, it can be calculated by a simple diffusion equation: where D is the molecular diffusion coefficient of O 2 at the relevant temperature and salinity (Schulz and Zabel, 2006), while dC/dz. is the vertical O 2 gradient. A similar flux calculation from microsensordetermined O 2 gradients within the sediment must take both porosity and tortuosity into account: where φ is the porosity of the surface sediment and Ds is the whole sediment diffusion coefficient. Ds is equal to the molecular diffusion coefficient corrected for tortuosity, for example using the equation: Ds = D/(1 + 3(1φ)) (Rasmussen and Jørgensen, 1992;Iversen and Jørgensen, 1993). It is apparent that the measurement of O 2 microgradients within the DBL provides a simpler approach to calculate the diffusive O 2 flux to the sediment surface than measurements within the sediment. The latter require high-resolution determinations of either the porosity or the diffusivity or both, e.g. by the use of a diffusivity microsensor (Revsbech et al., 1998). The importance of porosity and tortuosity is evident from the O 2 microgradient shown in Fig. 3A, where the slope of the gradient increases abruptly by 1.65-fold right at the sediment-water interface. Since the O 2 flux must be the same just above and beneath this interface, the ratio D/(φ × D s ) is also 1.65. The molecular diffusion coefficient at 10 • C is 1.46 × 10 − 5 cm 2 s − 1 (Schulz and Zabel, 2006). The sediment porosity in the top few tenths of a mm is then: φ = 0.79. The diffusive oxygen uptake of the sediment according to Eq. 3.1 is 13 mmol O 2 m − 2 d − 1 , which is within a typical range for a coastal marine sediment.
Some sediments have an extremely high oxygen uptake, and nearly the entire drop in O 2 concentration occurs in the DBL. Fig. 3B shows such an example, where oxic seawater with 265 μM O 2 at 20 • C flowed over a highly sulfidic sediment with a 0.5-mm thick DBL. The diffusive O 2 uptake was 85 mmol m − 2 d − 1 (Jørgensen and Revsbech, 1983). This approaches the theoretical maximum, about 100 mmol O 2 m − 2 d − 1 , of diffusive flux of across the sediment-water interface. An even higher diffusive O 2 flux would require a higher flow velocity and, thus, a thinner DBL to enable a steeper O 2 gradient.
The DBL closely follows the topography of the sediment surface, even down to the mm-scale. It is smoothed out over sediment roughness elements smaller than half the thickness of the DBL (Jørgensen and Des Marais, 1990;Røy et al., 2005). Fig. 4 shows the topographic surfaces of the DBL, the sediment surface, and the base of the oxic zone in the sediment. The DBL was 0.6 mm thick and the mean depth of O 2 penetration was 1.2 mm, i.e. rather similar to Fig. 3A. Due to this minitopography of the DBL, the diffusion of oxygen is not simply a vertical flux but rather a complex, 3-dimensional flux, where the direction of net diffusion is always perpendicular to the DBL and sediment topography.  This 3-dimensional DBL-flux was found to exceed the vertical DBL-flux by about 10% in estuarine mud of low roughness and up to about 25% in nearshore sand of high roughness (Glud et al., 2003;Røy et al., 2005). None of the diffusive oxygen uptake data in our database have been corrected for this 3-dimensional enhancement and are, therefore, generally too low by a small percentage.
Due to currents and wave action, half of the inner continental shelf (<50 m depth) consists of permeable sands (Hall, 2002), often with distinct bedforms such as ripples and dunes. The water flow across these topographic bedforms generates pressure gradients that drive pore water advection through the upper 2-15 cm. The depth of flow penetration depends on the sediment topography and permeability. The forced advection enhances the solute exchange across the sedimentwater interface (Huettel and Gust, 1992a;Huettel et al., 2014). Sandy sediments thereby trap phytoplankton and other fine organic particles from the water column and often have a high oxygen respiration in spite of a low organic carbon content (Huettel and Rusch, 2000;Cook et al., 2007). As the oxygen uptake depends on the overlying seawater flow and internal pore water advection, the proper measurement of oxygen consumption in permeable sediments poses methodological challenges. The eddy covariance technique is the most useful approach for this purpose at moderate current velocities (McCann-Grosvenor et al., 2014). Such measurements at water depths of 30-70 m, e.g. in the North Sea or off the US west coast Reimers and Fogaren, 2021), indicate that the O 2 consumption is similar to or slightly higher than non-permeable sediments at similar water depths and with similar primary productivity in the overlying water column.
Currents are sometimes so strong that they cause bedform migration and severely complicate even eddy covariance measurements (Ahmerkamp et al., 2015). Strong currents across the seabed combined with a resulting high shear stress drive particle resuspension from the sediment surface into the turbulent benthic boundary layer (e.g., Tengberg et al., 2003). A dynamic balance between active resuspension and resettling of particles thereby establishes a nepheloid layer of elevated particle density in the lowest few m above the seabed. The particles typically have high O 2 respiration and nutrient regeneration rates and constitute a highly mobile, particulate component of the boundary layer. Particle resuspension is a function of current velocity and possible wave action, while the settling velocity is also affected by the degree of particle flocculation. Flocs have been found to be larger and have higher settling velocity close to the seafloor in summer, possibly because the production of transparent extracellular polymers (TEP's) by microorganisms cause particle agglutination during the warm summer period (Fettweis and Baeye, 2014). Very high flocculation may generate a fluid mud layer flowing over the sediment in river estuaries with strong tides (Becker et al., 2013).

Diffusion-mediated O 2 uptake
The relative contribution of diffusive O 2 uptake to total O 2 uptake was found to increase with depth in the ocean, from half on the eutrophic inner continental shelf with dense benthic communities to being completely dominant in the oligotrophic deep sea with a more sparse benthos (Glud, 2008;Stratmann et al., 2020). It is interesting that, at both the high and the low extremes of oxygen uptake, molecular diffusion becomes the predominant mechanism of oxygen transport. At the high end, microbial biofilms of filamentous sulfur bacteria, Beggiatoa spp., have an O 2 penetration depth of only 0.1-0.2 mm, and the entire O 2 pool in the sediment turns over in about one second ( Fig. 3B; Jørgensen and Revsbech, 1983). At such a high flux with short diffusion distance and fast turnover, molecular diffusion outruns advection. At the low end, oxygen diffuses tens of meters into sediments of the central ocean gyres, even down to the basaltic crust, and O 2 in this deep oxic zone turns over extremely slowly in tens of thousands of years (Chapter 4.4;D'Hondt et al., 2015).
An example of a sediment O 2 profile with an intermediate penetration depth of 3 cm is shown in Fig. 5A. The profile was measured at 0.1 mm depth resolution directly on the seafloor in one of the deep oceanic trenches by the autonomous free-fall "Hadal-Profiler Lander". The 400 data points form a smooth curve that even resolves the O 2 gradient within the 0.5-mm thick DBL. The volume-specific O 2 consumption rates were calculated from the O 2 concentration profile, using the modeling software PROFILE of Berg et al. (1998) (Fig. 5B). The model showed two intervals of high O 2 consumption rates. One was the top 0.5 cm where freshly deposited organic matter was available for aerobic respiration by the microbial community. The other was the bottom of the oxic zone, near the oxic-anoxic interface, driven by the oxidation of reduced inorganic products from anaerobic degradation. Thus, out of the total diffusive oxygen uptake of 1.63 mmol O 2 m − 2 d − 1 , ca. 15% occurred in the lower peak at 2-3 cm depth. Compared to the extreme examples described above, the total O 2 pool in this sediment had an intermediate turnover time of ca. 1 day. Such a turnover time shows that the O 2 distribution and flux may still be compromised during the time it takes to retrieve sediment cores from the deep sea and perform O 2 analyses on board a ship .

Fauna-mediated O 2 uptake
Faunal activity plays an important role for the biogeochemical function of benthic habitats. The benthic fauna ingests and decomposes organic material, reworks sediment and ventilates burrow structures. These latter activities are commonly referred to as bioturbation (Kristensen et al., 2012). The fauna-mediated O 2 uptake encompasses fauna respiration and fauna-stimulated microbial respiration. It also includes enhanced re-oxidation of reduced products from the microbially driven anaerobic respiration, which can contribute significantly to the benthic O 2 uptake (e.g., Aller, 1994;Kristensen and Kostka, 2006;Middelburg, 2018). The way that fauna activity affects diagenetic pathways and enhances the benthic O 2 uptake depends on the type of fauna and its feeding mode.
Filter feeding epifauna, such as mussels, oysters or cold-water corals, acts as bioengineers that form beds and reefs, which effectively entrap suspended material and host a diverse biota. Recent investigations using eddy covariance technique have shown that the benthic O 2 consumption in such beds and reefs exceeds the O 2 uptake of adjacent sediments by 5to 20-fold and that they act as hotspots for organic carbon mineralization (Rovelli et al., 2015;Attard et al., 2020;Volaric et al., 2020). Deposit-feeding epifauna such as brittle stars, sea stars and sea cucumbers, move along the sediment surface, ingesting detrital material. They are particularly prominent along the deeper margins and in the abyss (Gorska et al., 2020). Their feeding mixes the upper sediment layers and their digestion hydrolyses complex organic material, while their defecation concentrates organic material and forms hots-spots with intensified microbial oxygen consumption (Jørgensen, 1977b;Huffard et al., 2016;Durden et al., 2020;www.youtube.com/watch?v=MZ mF8-Wudu4).
Burrowing macrofauna (in-fauna) strongly affects the benthic redox zonation and is particularly important for the biogeochemical processes in coastal sediments (e.g., Aller, 1994;Kristensen et al., 2012). Depending on their feeding mode, burrowing deposit feeders rework and mix the sediment matrix. Biodiffusers typically induce a random local mixing over short distances, yet the material transport can be considerable. As an example, it was estimated that a natural population of the sea urchin, Echinocardium sp., with 40 individuals per m 2 , translocated a volume of 20 L of sediment per m 2 per day (Lohrer et al., 2005). Natural coastal populations of biodiffusers typically support a particle biodiffusion coefficient of 1-2 cm 2 yr − 1 (Dupont et al., 2006;Gilbert et al., 2007;Quintana et al., 2007). Upward and downward conveyors continuously displace material vertically from deeper layers to the sediment surface or vice versa. A natural population of 50 lugworms, Arenicola marina, per m 2 , living head-down in J-shaped burrows, has been estimated to yearly ingest and completely mix the upper 30 cm of sediment (Riisgård and Banta, 1998). Regenerators, such as fiddler crabs, live in permanent burrow systems and infills that are constantly excavated. The material is deposited on the sediment surface, inducing an effective and continuous overturn of the sediment (Huang et al., 2007). Thus, deposit feeding fauna directly contributes to the degradation of organic material via ingestion, while sediment reworking exposes otherwise buried recalcitrant organic material to frequent redox shifts, which stimulates the microbially driven degradation of the material (Aller, 1994(Aller, , 2004Kristensen and Holmer, 2001). Overall, sediment reworking enhances the efficiency of the oxidative processes, maintains a high buffer capacity against sulfide release, and enhances the phosphate binding capacity of coastal sediments (Aller, 2004).
Filter feeding and ventilating in-fauna establishes burrow structures and mediates a rapid solute exchange between the oxic bottom water and the subsurface sediment. The irrigated burrows enlarge the oxicanoxic interface and form a complex, 3-dimensional mosaic of microhabitats with highly heterogeneous redox zonations as compared to nonirrigated sediments (Wenzhöfer and Glud, 2004;Zhu et al., 2006;Volkenborn et al., 2012;Camillini et al., 2019). Bioirrigation significantly enhances the benthic O 2 uptake due to stimulated aerobic microbial respiration and enhanced oxidation of reduced products along the walls of ventilated burrows and galleries (Fig. 6). Furthermore, irrigation facilitates a close coupling between oxic and anoxic microbial processes, such as nitrification and denitrification, or enhanced formation of iron and manganese oxides, thus stimulating microbial metal respiration (Aller, 1990;Pelegri and Blackburn, 1994;Munksby et al., 2002;Webb and Eyre, 2004). The irrigation is typically intermittent, creating highly dynamic oxic environments in the subsurface sediment, and results in small-scale variation in diagenetic processes. These dynamic conditions support a diverse and physiologically versatile microbial community (Papaspyrou et al., 2006;Bertics and Ziebis, 2009;Woulds et al., 2016).
Densities of fauna and empirical allometric relations for fauna respiration have been used to assess that fauna respiration in coastal settings may account for 25 to 40% of the TOU (Piepenburg et al., 1995;Heip et al., 2001;Rodil et al., 2019). The fauna-mediated O 2 uptake generally exceeds the calculated faunal respiration, however, mainly due to fauna-enhanced, microbial respiration and re-oxidation of reduced products from anaerobic respiration (Forster and Graf, 1995;Hansen and Kristensen, 1997;Vopel et al., 2003;Glud et al., 2003).
To correctly assess the in situ contribution of fauna, it is essential to maintain the natural behavior and respiration of the communities. This is difficult in small, confined sediment cores or in laboratory enclosures.
Natural faunal assemblies consist of numerous species of very different sizes that are difficult to transfer and maintain in the laboratory. Therefore, in situ assessments of TOU using large benthic chambers generally provide a more correct representation of the fauna-mediated O 2 uptake, with higher rates than obtained from incubations of smaller, retrieved sediment cores (Glud and Blackburn, 2002;Glud et al., 2003) (Chapter 2.1). Furthermore, the benthos in coastal sediments may respond to changes in light intensity, water flow and food availability, exhibiting diel or tidally induced temporal patterns in activity (Loo et al., 1996;Wenzhöfer and Glud, 2004;Thouzeau et al., 2007). Thus, chamber enclosures might not capture environmentally induced dynamics in oxygen flux or may underestimate the contribution of filter feeding by communities that gage their filtration activity to the   6. The effect of bioirrigation. Permeable sediment from the German Wadden Sea was placed in a large flume with a transparent wall. The photographs show the zonation before (A) and one week after (B) the introduction of benthic in-fauna (lugworms and soft-shell clams). The fauna induced subsurface irrigation with oxic water and gradually oxidized the otherwise reduced sediment. Oxidized sediment is brown due to iron oxides, reduced sediment is darkgray due to iron sulfides. (Photos by Ronnie N. Glud).
food availability (Riisgård, 1994;Riisgård et al., 2003). Under such circumstances, combined measurements of in situ O 2 microprofiles and non-invasive eddy covariance can best reveal the importance of faunamediated O 2 uptake .
Only a limited number of in situ studies have quantified the TOU and DOU in parallel. Studies in coastal settings generally show that the fauna-mediated O 2 uptake accounts for about half of the TOU (Archer and Devol, 1992;Forster et al., 1999;Meile and Van Cappellen, 2003;Glud, 2008, and references therein). The fauna-mediated O 2 uptake increases with the faunal abundance (Clough et al., 2005;Norkko et al., 2015;Belley and Snelgrove, 2016). However, in many heterogeneous settings macrofauna biomass correlates poorly with the fauna-mediated O 2 uptake (Moodley et al., 1998;Glud, 2008;Snelgrove et al., 2018). This is presumably due to the variability in feeding modes of the benthos, the behavioral responses of natural communities, and the fact that mass-specific respiration rates are size dependent (Ghedini et al., 2018). Functional traits, and the extent to which a given fauna activity enhances the benthic O 2 demand, are also context dependent. For instance, the enhancement of benthic O 2 uptake by a given irrigation activity depends on the O 2 penetration depth.
Studies of the importance of fauna for benthic solute exchange mostly focus on macrofauna, but in sediments with very shallow O 2 penetration, meiofauna (metazoans of 0.04-1 mm size) can add significantly to benthic solute transport and to O 2 consumption (Aller and Aller, 1992;Rysgaard et al., 2000;Nascimento et al., 2012;Bonaglia et al., 2014). It was recently demonstrated that meiofaunal activity enhanced reoxygenation of surface sediments after hypoxia (Bonaglia et al., 2020). The contribution of meiofauna to the metazoan biomass generally increases with water depth (Stratmann et al., 2020). Allometric trends suggest that meiofauna respiration is responsible for 8-19% of the benthic O 2 consumption at abyssal and hadal water depths in the southwest Pacific (Leduc et al., 2016). Yet, the importance of meiofauna for the biogeochemical function and the oxygen demand of marine sediments is still poorly understood.

Time scales
Suspended POC in the open ocean originates mainly from primary production in the surface waters according to its young 14 C-age and isotopically light δ 13 C-composition (McNichol and Aluwihare, 2007). The flux of POC from rivers and estuaries to coastal seas, together with the resuspension of sediment from the shelf and upper continental slope, also contribute by a lateral transport of POC from the continental shelf to the deep sea (Bauer and Druffel, 1998;Raymond and Bauer, 2001). The total pool of suspended POC in the ocean has a mean residence time of 5-10 years according to Thorium isotope studies (Bacon and Anderson, 1982). The settling speed of marine snow varies strongly depending on the degree of flocculation and the sediment load of particles, and the settling time from the photic zone to the seafloor at several thousand m depth may vary from days for a flocculating phytoplankton bloom to years for the finest suspended particles (e.g., Smith Jr. et al., 2018).
The rate of POC mineralization drops continuously with its increasing age as it sinks down through the water column, settles on the seafloor, and is gradually buried into the sediment. Middelburg (2019) compiled data on the degradation rate of POC from sediment traps, from laboratory experiments, and from sediment cores. His combined plot shows that the degradation rate constant drops by 9 orders of magnitude as its age increases by 9 orders of magnitude, from less than a day to a million years. The log-log plot in Fig. 7 shows a linear fit, i.e. a power law function, between POC reactivity and POC age, whereby the firstorder kinetic coefficient, k (in yr − 1 ), is equal to: 0.21 × t -0.985 , where t is the time in years. The exponent is very close to − 1, which shows that the POC reactivity drops in inverse proportion to its age. This relationship has important implications for whether the POC is degraded in the water column or in the sediment, and where CO 2 and inorganic nutrients are released. It also has implications for which fraction of the deposited organic matter is buried on a long time scale and is thereby withdrawn from the modern oceanic carbon cycle.
After the POC settles on the seafloor, the aerobic degradation continues. The organic matter is here orders of magnitude more concentrated than in the water column, and a very different community of benthic organisms participate in the degradation. The macrobenthos may drive particle mixing down to 5-15 cm in the sediment (Boudreau, 1998). In ocean margin sediments, this ensures that aerobic mineralization continues to play a role until the organic matter is completely buried beneath the bioturbated zone. The anaerobic microorganisms then take over and continue the mineralization at ever decreasing rates. Direct measurements of oxygen respiration and sulfate reduction rates in a coastal marine sediment show how aerobic and anaerobic degradation follow a power law function with depth and, thus, with age of the Fig. 7. Degradation rate constants ("reactivity") of organic matter as a function of its age. Data sources are from sediment trap material, laboratory experiments, and sediment cores. The data plot is double logarithmic and is fitted by a power law function (after Middelburg, 2019; data courtesy of Jack J. Middelburg). organic matter. An example is presented in Fig. 8 in a semi-log plot in which data at 3-50 cm depth are fitted by the power law function: Mineralization rate = 2200*z -2.2 nmol C org cm − 3 d − 1 , where z is the depth in cm.
The global distribution and budget of oxygen uptake in the seabed, as discussed in Chapters 5 and 6, represent processes that occur today and have occurred through the later part of the Holocene. However, on a time scale of tens of thousands of years, the ocean productivity and the seafloor respiration have changed greatly. Marine sedimentation on the continental shelves has occurred only since the last 10,000 years, after the shelves became flooded following the last glaciation and sea-level low-stand. Today, the coastal zone and inner shelf region have the highest rates of sediment accumulation and organic matter mineralization. Thus, an estimated 88% of the global accumulation of organic carbon in marine sediments currently occurs within 500 km from the nearest coast or out to 2000 m water depth (Wallmann et al., 2012;Xie et al., 2019). Beyond this ocean margin, the seafloor has much lower sedimentation rates, and the degradation of organic matter in the deep sea is affected by geochemical processes that take place on a much longer time scale. An extreme example of this are the 10-to 100-m deep, oxic sediments of the mid-oceanic gyres where the calculated turnover time of O 2 is 20,000-40,000 years, i.e. much longer than the current interglacial period (D'Hondt et al., 2015).

Coupling to anaerobic processes
A fraction of the organic matter deposited on the seafloor escapes the uptake and respiration by aerobic organisms and is buried beneath the bioturbated zone. Here, it continues to be attacked by extracellular hydrolytic enzymes from the microbial community and is gradually degraded and fermented to small organic molecules ( Fig. 9) (Jørgensen, 2006). These molecules are substrates for bacteria and archaea with an anaerobic respiration. The dominant terminal electron acceptors by this respiration change down through the anoxic sediment, largely in accordance with a decreasing energy yield of the respiratory process: nitrate, manganese oxide, iron oxide, sulfate and carbon dioxide (Froelich et al., 1979).
The anaerobic respiration processes result in a corresponding sequence of reduced products: N 2 /NH 4 + , Mn 2+ , Fe 2+ , H 2 S and CH 4 . In addition, CO 2 , NH 4 + , HPO 4 2− and other inorganic ions are released by the organic matter degradation. The reduced products accumulate in the sediment where a small fraction is buried. Next to organic matter, pyrite (FeS 2 ) is the most important sink for reducing power. Most of the dissolved, reduced products diffuses upwards along concentration gradients. As they reach into overlying biogeochemical zones with higher redox potential, they are oxidized through a redox cascade of processes as illustrated in Fig. 9. Very simplified, the CH 4 is thereby oxidized by SO 4 2− , H 2 S is oxidized by Fe(III), Fe 2+ is oxidized by Mn(IV), etc. In reality, the zonation of processes is not that clear-cut. The different processes overlap, and the sediment is temporally and spatially heterogeneous due to faunal activity and other factors. The ultimate oxidant at the end of the redox cascade is O 2 , which is consumed by the chemical or biological oxidation of the inorganic species. The quantitative role of each electron acceptor differs strongly between sediments, depending on water depth, sedimentation rate, organic carbon content and other parameters. In coastal waters and inner shelf sediments, sulfate reduction is the most important anaerobic process (Jørgensen, 1982;Canfield et al., 2005). Studies of sulfate reduction in these sediments using 35 S radiotracer show that about 90% of the produced H 2 S is ultimately oxidized back to sulfate . If the sediment is highly sulfidic, the oxidation may take place near the sediment surface by aerobic or nitrate-reducing sulfide oxidizers, such as filamentous sulfur bacteria or cable bacteria (Jørgensen, 1977a;Pfeffer et al., 2012). If the sediment is less sulfidic, the coupling to oxygen is mostly indirect via iron, which is then the main oxidant for sulfide (Findlay et al., 2020). The resulting Fe 2+ and iron sulfides may subsequently be oxidized by O 2 as a result of bioturbation and burrow ventilation by benthic invertebrates. The iron oxidation may also be indirect via manganese oxides that oxidize both solid phase and dissolved Fe(II) (Aller and Rude, 1988;Schippers and Jørgensen, 2001).
With increasing water depth and distance from land, iron reduction and manganese reduction take over the role as the main electron acceptors by anaerobic mineralization (Canfield et al., 2005). Abyssal sediments under the ocean gyres constitute the extreme end-members of this trend, with oxygen penetration all the way down to the basaltic crust (D'Hondt et al., 2015). The anaerobic mineralization processes and their contribution to sediment oxygen consumption are, thus, biased towards the ocean margins, even more biased than the oxygen consumption . Whereas the coastal zone and continental shelf out to 200 m water depth comprise only 7% of the ocean area, they include about 75% of the global marine sulfate reduction (Canfield et al., 2005) and about 60% of the global sediment oxygen Fig. 9. Organic matter mineralization in marine sediment with a vertical zonation of the main biogeochemical processes and electron acceptors. The scheme emphasizes the degradation pathways via hydrolysis and fermentation that provide substrates for anaerobic respiration with an energetic depth sequence of oxidants. The reduced respiration products diffuse upwards and are re-oxidized through a redox cascade that ultimately transfers most of the reducing power from the anoxic sediment to O 2 . (From Jørgensen, 2021). consumption (Chapter 6.3). Whereas oxygen uptake in the seabed is mostly driven by the respiration of heterotrophic organisms, including benthic macrofauna, meiofauna and microorganisms, oxygen is also consumed by the oxidation of reduced, inorganic products from the redox cascade. In shelf sediments, these products are dominated by sulfide and iron sulfides, and 25-50% of the entire oxygen consumption is used for the re-oxidation of these inorganic species .
Since about 90% of the produced sulfide is re-oxidized to sulfate by oxygen, either directly or indirectly, most of this anaerobic mineralization is included in the total oxygen uptake (TOU). The TOU measurement thereby includes also most of the organic matter mineralization that takes place beneath the oxic and bioturbated zone. It may not include the deep mineralization to methane and CO 2 . However, the methane production and its oxidation at the bottom of the sulfatic zone generally contribute only 1-10% to the total mineralization by sulfate reduction and is thus a small component in the overall organic carbon turnover (Flury et al., 2016;Jørgensen, 2021).
In conclusion, half of the oxygen consumption in inner shelf sediments is involved in the re-oxidation of sulfide and only half of the oxygen consumption is available for the benthic fauna and for aerobic, heterotrophic microorganisms. This oxygen consumption by sulfide oxidation drops with ocean depth to about a tenth in the continental rise and fades out in the abyss (Canfield et al., 2005). The coupling of oxygen uptake to sulfide and iron oxidation in coastal sediments may impose a seasonal delay in the sediment oxygen consumption. During summer and fall the surface sediment builds up a pool of iron sulfides that hold back the sulfide from immediate reaction with oxygen and gradually becomes oxidized during winter. This "iron courtain" can function as a barrier against release of sulfide from the sediment in summer and delay the oxygen debt of the sediment (Seitaj et al., 2015).
An important aspect of the re-oxidation of sulfide and other reduced inorganic species is that the processes are microbially catalyzed and provide energy for a chemoautotrophic growth of microorganisms. This has been studied by dark incubation experiments with marine sediment amended with 14 C-labeled HCO 3 − . Based on such experiments, Vasquez-Cardenas et al. (2020) calculated the autotrophic growth in coastal marine sediment to be equivalent to 1-11% of the total oxygen uptake with a mean value of 4%. They also made a literature survey of such measurements and found a range of 1-22%, with a mean of 7%. This mean corresponds to the 7% estimated as a maximum value by Jørgensen and Nelson (2004), based on a growth yield by chemoautotrophic sulfide oxidizers of 15% and assuming that half of the oxygen uptake was allocated to sulfide oxidation. The anaerobic oxidation of organic matter has a lower energy yield than the aerobic oxidation, but this is compensated by the oxidation of reduced respiration products through chemoautotrophic microorganisms. Their dark CO 2 fixation and growth generate easily degradable microbial necromass that can be used by aerobic organisms (Middelburg, 2011). Yet, this accounts for only a small fraction relative to the total influx of organic matter and is probably not detectable in budgets of carbon cycling in marine sediments (Bradley et al., 2018).

Regional examples
The global map of seabed oxygen uptake rates and their correlation with environmental parameters presented in Chapter 5 are compiled from highly diverse marine systems, ranging from eutrophic seas to oxygen minimum zones, abyssal plains, mid-oceanic gyres, and deep-sea trenches. We here provide examples of the environmental conditions of these systems and discuss how the oxygen and carbon cycles of their sediments are controlled.

Eutrophic seas: Baltic Sea
The Baltic Sea represents an example of a eutrophic coastal sea with strong terrestrial influence and a close benthic-pelagic coupling. With an area of about 400,000 km 2 , it is one of the largest brackish water bodies in the world and arguably has one of the largest anthropogenically induced oxygen-deficient bottom waters in the world (Meier et al., 2018). The water column is stratified by a seasonal thermocline and a permanent halocline at 60-80 m depth, between a brackish upper water mass and more saline bottom water originating from the North Sea. The deglaciation of the entire basin some 13,000 years ago, and its transition from freshwater to brackish-marine conditions about 8000 years ago, explains why the relatively young sea still has a low diversity of genuine brackish-water species of animals and plants (Björck, 1995;Hällfors et al., 1981).
Starting in the 1950s, the Baltic Sea has become increasingly eutrophic due to a high nutrient load of nitrogen and phosphorous from the large catchment area of predominantly forestland to the north and farmland to the south. This has accelerated the carbon cycling and oxygen consumption in the Baltic seabed. Jonsson and Carman (1994) estimated a > 1.7-fold increase in organic matter deposition in the northern Baltic Sea proper between the late 1920's and the late 1980's. This trend will continue, enhanced by a predicted increase of up to 20-50% in POC flux from land due to increasing runoff (Omstedt et al., 2012). Reduction in nitrogen and phosphorous loads since the 1980's has led to a gradual alleviation of eutrophication (Andersen et al., 2017) but not yet to an alleviation of hypoxia caused by high oxygen consumption in the lower water column (Meier et al., 2018).
Oxygen concentrations are low in the deep basins, and hypoxia may even occur during the summer months in shallow coastal areas. A third of the central Baltic Sea had <2 mL O 2 L − 1 during the past two decades, while a fifth was totally anoxic and sulfidic. Major inflows of saline and oxygenated bottom water occur irregularly and provide a transient oxygen supply to the deep basins. However, this ventilation with saline bottom water also increases the stability of stratification and reduces the vertical mixing of oxygen from above until the stratification slowly weakens again (Conley et al., 2009;Carstensen et al., 2014).
Long periods of extensive hypoxia have occurred in many highlatitude shelf basins during the Holocene due to the combined effect of global sea level rise and local glacio-isostatic rebound. In the central Baltic Sea, such periods in the past are apparent from organic-rich, laminated sediment intervals that have now returned, starting in the 1950's, due to the combined effect of eutrophication and continued warming (Kotilainen et al., 2014;Jilbert et al., 2015). Thus, remote sensing data from 1990 to 2008 suggest an accelerated temperature increase of up to 1 • C per decade, mostly in the northern part of the Baltic Sea (International Council for the Exploration of the Sea, 2018).
The net primary production in the Baltic Sea is about 3.8 Tmol C yr − 1 (Leipe et al., 2011), of which a third is exported beneath the photic zone (Heiskanen and Leppänen, 1995). Leipe et al. (2011) estimated a total organic carbon (C org ) burial rate in the Baltic Sea of 0.30 ± 0.24 Tmol C yr − 1 , corresponding to 8% of the net primary production. This C org burial is a relatively high fraction of the deposited POC, in particular in the hypoxic basins where the buried fraction ranges from one to two thirds (Omstedt et al., 2014). A large burial efficiency was suggested to be due to a large contribution of poorly degradable terrestrial POC . Recent estimates by Nilsson et al. (2019) conclude that 1.9 ± 0.7 Tmol C yr − 1 is deposited in total on the Baltic seabed, with a mean burial efficiency of 4%, but with a large variability between the different basins (2.5-16%).
Data from sediment trap studies and measurements of the total oxygen uptake of the sediments are rather scattered in the Baltic Sea. Rates of TOU from the muddy seabed at water depths from 15 to 130 m are mostly in the range of 7-15 mmol O 2 m − 2 d − 1 (Koop et al., 1990;Rasmussen and Jørgensen, 1992;Conley et al., 1997;Glud et al., 2003). The contribution by benthic macrofauna respiration is high in some areas, up to 50% of the TOU and occasionally even higher (Glud et al., 2003).
Climate change during the 21st century will impact the phytoplankton production in many regional seas (Holt et al., 2016). An increase of the net primary productivity is predicted for the Baltic Sea, whereby a primary driver is an increased wind-driven ventilation and upwelling in the central Baltic Sea and reduced sea ice cover in the northern Baltic Sea. Massive cyanobacterial blooms are currently observed by satellite nearly every summer. The cyanobacteria provide a positive feed-back on eutrophication by extensive N 2 -fixation. The sedimentation and decay of senescent cyanobacteria stimulate hypoxia in the deep bottom water, which in turn enhances phosphate release from the sediment and thereby supports cyanobacterial blooms (Funkey et al., 2014). The fast, annual deposition of a diatom spring bloom may also cause a peak in sediment O 2 uptake and even create a sulfidic surface sediment (Moeslund et al., 1994).

Oxygen minimum zones
Oxygen minimum zones (OMZs) along the open ocean margins are mainly driven by strong coastal upwelling and a high export rate of plankton-derived organic matter into the underlying water column. Hypoxia develops when the O 2 consumption during degradation of the sinking detritus exceeds the O 2 renewal by mixing with oxic water masses. There is not a strict definition of hypoxia or of the resulting OMZs in the ocean, but <0.5 mL O 2 L − 1 is often used as an upper threshold (e.g., Helly and Levin, 2004;Paulmier and Ruiz-Pino, 2009). This corresponds to <22 μM O 2 or < 7% of sea surface air saturation at 5 • C. The largest OMZs occur near the eastern ocean margins where predominant wind-systems drive surface water off shore and draws up nutrient-rich deeper water that supports a high primary productivity. The main OMZs occur in the Eastern South Pacific and Eastern Tropical North Pacific plus the Arabian Sea and Bay of Bengal. In some of these OMZs, the O 2 concentration reaches zero (<2 nM O 2 ; Revsbech et al., 2009). The anoxic or near-anoxic conditions in the OMZs cause massive losses of combined nitrogen to N 2 in the ocean by supporting the microbial processes of anammox and heterotrophic denitrification (Lam and Kuypers, 2011).
Paulmier and Ruiz-Pino (2009) estimated that the total areal extent of the oceanic OMZs is 30 × 10 6 km 2 , or 8% of the total oceanic area. Over most of this area, the O 2 minimum occurs in midwater of the open ocean at depths of 50-1000 m. Within that depth range, the OMZs impinge on the seafloor mainly on the outer continental shelf and the upper slope. Helly and Levin (2004) estimated that hypoxic OMZ bottom water with <0.5 mL O 2 L − 1 covers a global sediment area of 1.15 × 10 6 km 2 , corresponding to 4% of the outer shelf and upper slope together (50-1000 m water depth) and only 3% of the total OMZ area. An estimated 0.76 × 10 6 km 2 of this seabed has anoxic, or nearly anoxic, bottom water with <0.2 mL O 2 L − 1 .
The areal extent of benthic hypoxia has expanded over the past halfcentury and will expectedly continue to expand over the coming decades (Schmidtko et al., 2017;Breitburg et al., 2018). This is primarily due to the diverse effects of ocean warming, which include decreasing O 2 solubility by increasing temperature, reduced ventilation by stronger thermal stratification, stronger winds that drive upwelling, increased nutrient supply, and increased respiration rates of sinking POC in the upper ocean (Levin, 2018). A concurrent shoaling of the upper boundary of OMZs now spreads hypoxic bottom water further onto the continental shelves, which will affect their benthic fauna and microbial processes (Gilly et al., 2013). These trends have similarities to coastal hypoxia in silled basins and fjords, which is also enhanced by global warming but is increasingly affected by anthropogenic processes such as eutrophication (Middelburg and Levin, 2009;Breitburg et al., 2018). The Baltic Sea is an example of periodic hypoxia (Chapter 4.1), while the Black Sea is an example of permanent anoxia beneath a water depth of 50-150 m (e.g., Stanev et al., 2018).
Benthic flux chambers have been used for in situ measurements of the O 2 uptake at very low bottom water O 2 concentrations in OMZ regions. They have also been used to detect the out-flux of reduced solutes or DIC produced from anaerobic mineralization in the sediment. Chamber incubations at low ambient O 2 concentration are challenged by a relatively rapid exhaustion of oxygen in the chamber, which might affect exchange rates of O 2 and other solutes. To ensure reliable benthic exchange rates at low O 2 concentration, sophisticated gill systems have been developed that maintain the O 2 level of the enclosed water by a continuous O 2 supply (Sommer et al., 2008). Yet, to determine the benthic organic carbon mineralization rates in such settings, DIC exchange rates may be a better option than O 2 consumption rates (e.g., Rassmann et al., 2020).
Due to such in situ measurements, data are available for the O 2 uptake of ocean margin sediments underlying the major OMZs with different degrees of hypoxia in the bottom water. We extracted the relevant data from our database using the following criteria: a) only data from the open ocean margins were selected, not data from coastal basins, silled fjords and other enclosed systems (such as the Black Sea or the Baltic Sea); b) only data from the outer shelf (50-200 m) and the upper slope (200-1000 m) were selected, i.e. from the ocean depths where OMZs may impinge on the seabed; c) only data with concurrent measurements of TOU and bottom water O 2 concentration were selected.
This selection left 199 data for the outer shelf and 88 data for the upper slope. A plot of TOU vs. bottom water O 2 concentration is shown for the entire dataset in Fig. 10A. In order to better resolve data for the most hypoxic OMZ regions, the same data are shown with expanded bottom water O 2 scale in Fig. 10B. Fig. 10 shows clearly that the measured in situ O 2 uptake rates are strongly reduced under low O 2 concentrations relative to rates under high O 2 for the same ocean depth of 50-1000 m. Thus, TOU rates in OMZ sediments with <25 μM O 2 (<0.5 mL O 2 L − 1 ) appear to be an order of magnitude lower than in sediments with much higher bottom water O 2 concentrations (Archer and Devol, 1992). Nearly all these data are from the Pacific margin of South and North America, notably off Peru, Mexico, California and Oregon. By the lowest O 2 concentrations, nitrate may be an important alternative oxidant. For example, on the Pacific shelf off Mexico at 100-200 m depth there was only 0-6 μM O 2 in the bottom water but 20-30 μM NO 3 − (Harnett and Devol, 2003). This could explain why the respiratory quotient, RQ, calculated from the DIC release relative to the O 2 uptake was mostly very high, in the range of 2-10 ( Berelson et al., 1987Berelson et al., , 2019Jahnke et al., 1990Jahnke et al., , 1997Sommer et al., 2016). As discussed in Chapter 6.2, a mean RQ for this depth range of the seabed outside of OMZs is DIC:O 2 = 0.80-0.85. The low O 2 uptake relative to the organic carbon mineralization may be compensated by NO 3 − uptake (see below). This shows that the DIC flux can be a more useful measure of organic matter mineralization than the O 2 flux under low bottom water O 2 in OMZ sediments.
The very low TOU rates occur in the absence of fauna and faunal respiration. They are therefore also related to a shift in O 2 transport across the sediment surface from partly bioirrigation to purely molecular diffusion (cf. Chapter 3). On the other hand, OMZ sediments underlie regions with high primary productivity and, expectedly, have high POC deposition rates with high mineralization potential. Although a relatively large fraction of the deposited POC may be buried in OMZ sediments (Baroni et al., 2020), it is not obvious whether the opposite directed effects of enhanced POC flux and low O 2 concentration will result in higher or lower O 2 flux. Apparently, the balance is a lower O 2 flux.
A large part of the mineralization in OMZ sediments must be anaerobic and lead to reduced products that diffuse to the sediment surface and require an oxidant. Nitrate in the bottom water or iron(III) in the sediment may contribute as oxidants. It should thereby be noted that each TOU data point in Fig. 10 represents only a momentary situation, at the time of measurement, while the long-term data may be more complex. The ability of the sediments to retain sulfide depends on their periodic re-oxidation as the bottom water O 2 and NO 3 − concentrations vary seasonally or occasionally due to mesoscale variability in ocean circulation, intensity of upwelling, and primary productivity (e.g., Vergara et al., 2016). Such a re-oxidation establishes a pool of oxidized iron minerals near the sediment surface that oxidizes and precipitates the diffusive sulfide flux from below (Seitaj et al., 2015).
In some hypoxic shelf regions, such as the shelf off northern Chile and Peru, benthic exchange rates are affected by dense communities of filamentous, sulfide-oxidizing bacteria, Thioploca spp. These use their high concentrations of intracellular nitrate for sulfide oxidation and may function as an effective barrier against H 2 S release as long as nitrate is available in the anoxic bottom water (Fossing et al., 1995). The bacteria perform a dissimilatory nitrate reduction to ammonium, which results in a high NH 4 + flux from the sediment . Sediments under hypoxic bottom water are thus important sources of NH 4 + to the water column, also in other OMZs, such as off SW and NW Africa Kalvelage et al., 2013). The NH 4 + is oxidized to NO 2 − and NO 3 − by nitrifying bacteria in the hypoxic OMZ water, or is combined with NO 2 − to produce N 2 by anammox bacteria under anoxic conditions (Lam and Kuypers, 2011). A direct H 2 S flux from the sediment was observed by flux chamber measurements to occur only when the bottom water became anoxic and nitrate dropped to <5 μM .
Extensive release of H 2 S from OMZ sediments appears to be an exception, even under fully anoxic bottom water. It has been observed in the most productive upwelling systems, such as on the shelf off Namibia where the oxidation of sulfide to elemental sulfur after H 2 S emission into the water column generates large areas of milky-white sulfur turbidity that can even be observed from satellite (Ohde and Dadou, 2018). This emission of H 2 S from the diatom ooze on the Namibian shelf is not only due to anoxia but is also driven by eruptions of methane gas that accumulates in the mud (Brüchert et al., 2003).
Organic carbon preservation in the seabed is generally enhanced under hypoxic or anoxic conditions (Canfield, 1994), and burial is therefore more efficient under OMZs compared to more oxic regions of the continental margin (Harnett et al., 1998;Baroni et al., 2020). Yet, the significance of bottom water O 2 on organic carbon burial efficiency is complex and may depend on several confounding factors, such as oxygen exposure time, bioturbation by burrowing macrofauna, sediment deposition rate, and age or degradation state of the deposited POC (e.g., Canfield, 1994;Dale et al., 2015).
There is a strong selectivity in the benthic faunal composition where the OMZs impinge on ocean margin sediments. Hypoxia leads to a reduction in species richness of all benthic faunal groups and in some cases selects for species that are specially adapted to thrive under very low O 2 (Gooday et al., 2010). At the same time, OMZs may enhance regional diversity in some taxa, such as Foraminifera, many of which are particularly tolerant of anoxia and may even respire by dissimilatory nitrate reduction (Piña-Ochoa et al., 2010;Orsi et al., 2020). There is generally a reduced activity of burrowing macrofauna in OMZ sediments, where bioturbation was observed to fade out and where laminated sediments became predominant at <5 μM O 2 (Levin et al., 2009b).

Abyssal plains
The abyssal plains cover 85% of the ocean bed and represent the largest marine ecosystem on Earth. The term "plain" might be misleading as the abyssal realm is characterized by an undulating bathymetry with reliefs of 300-1000 m tall formations (Harris et al., 2014). The spatial distribution of benthic fauna and of the O 2 consumption rate at the abyssal seafloor is ultimately controlled by the export of organic material from the photic zone (Rowe, 1973(Rowe, , 2013. Thus, the benthic activity is generally decreasing with increasing distance to the ocean margins and with increasing water depth (Jahnke et al., 1990;Wenzhöfer and Glud, 2002). However, this pattern may be confounded by the regional or local bathymetric relief focusing material in depressions and throughs of the abyss (Chen et al., 2013).
Beside the increasing awareness of the spatial variation in deposition rates at abyssal depths, there is increasing evidence that the benthic carbon requirement of abyssal communities as inferred from O 2 uptake measurements exceeds the vertical supply of organic carbon calculated from sediment trap studies (Smith Jr. and Kaufmann, 1999;Wiedmann et al., 2020). This is partly attributed to lateral transport of POC from ocean margins or mounts and the vertical transport of larger algal aggregates and carrion that circumvent sediment traps (Jahnke and Jackson, 1992;Smith Jr., 1987). Indeed, large food falls of whales, salps and jellyfish have been identified as locally important sources of organic matter input to the deep seafloor (Smith, 2006;Billett et al., 2006;Treude et al., 2009;Smith Jr. et al., 2014) (Fig. 11).
Aggregation enhances the sinking velocity of POC (Deuser and Ross, 1980;Nowald et al., 2015) and governs the rapid transport of fresh organic matter to the seabed. Large pulses of fresh phytodetrital matter have on several occasions been observed on abyssal sediments (Billett et al., 1983;Smith and Baldwin, 1984;Lampitt, 1985). One striking example was the observation of large algal aggregates released from melting sea ice settling at the seafloor at 4500 m water depth in the central Arctic Ocean (Fig. 11). The settling aggregates locally increased the benthic oxygen uptake by more than ten-fold . The deposition in deep sea sediments is thus very dynamic, and sedimentation events that are quantitatively important for the pelagicbenthic coupling may not be recorded. To our knowledge, there are today only three long-term observatories that monitor the seasonal variation of carbon export, benthic community structure and benthic oxygen consumption in abyssal settings. These are the North-East Pacific Station M (Smith Jr. et al., 2018), the North Atlantic PAP (Lampitt et al., 2010), and the Arctic HAUSGARTEN (Soltwedel et al., 2016).
The most comprehensive data set of seasonal variations in carbon export and benthic oxygen consumption in the deep sea has been realized by Station M in the deep sea off central California (Fig. 12) (Smith Jr. et al., 2001, 2013. The time series has revealed distinct seasonal pulses of elevated export fluxes during or shortly after surface blooms that transiently enhanced the supply of labile organic material at the seabed . Similar observations have been made for PAP and HAUSGARTEN (Lampitt et al., 2010;Bauerfeind et al., 2009;Lalande et al., 2013). These deposition events fuel the benthic ecosystem and trigger a strong growth response by the benthic community (Witte et al., 2003). Deposition events also enhance the sediment oxygen consumption rate, at Station M by a factor of two to three (Fig. 12;Smith Jr. et al., 2016) and at HAUSGARTEN by a factor of two (von Appen et al., 2021). Such events can easily be missed by single lander deployments during deep sea expeditions. The available database does not reflect the dynamics of the deep sea and may therefore also not correctly represent the average carbon mineralization rates. More high-resolution and long-term studies by autonomous instruments (Fig. 1) are required to meet this challenge (Wenzhöfer et al., 2016a;Lemburg et al., 2018;Sherman and Smith Jr., 2009) and to supplement deep ocean observation data (Levin et al., 2019;MBARI, 2009).

Mid-oceanic gyres
Ocean gyres are large systems of circular ocean currents driven by global wind patterns and by forces caused by Earth's rotation. Due to these Coriolis forces, the circular currents run clockwise in the northern hemisphere and counter-clockwise in the southern hemisphere. There are five major gyre systems, located in the North and South Pacific, the North and South Atlantic, and the Indian Ocean. The gyres isolate nutrient-poor, central parts of the upper ocean and are distinct by remote sensing due to their low chlorophyll concentration (see Fig. 17b). Gyres have very low phytoplankton productivity and a low export of sinking particulate material, either of POC derived from the photic zone or of clastic material derived from the continents. The underlying seabed therefore has extremely low sedimentation rates, low rates of organic matter degradation, and deep oxygen penetration.
While measurements of total oxygen uptake (TOU) and oxygen penetration depth (OPD) are abundant along the ocean margins, measurements are scarce in the central, low-productive parts of the oceans. This may be due to their remoteness, to less scientific interest in these regions, and to the difficulty of measuring very low O 2 consumption rates and very deep O 2 penetration. Chamber landers for measuring in situ flux can barely detect the extremely slow depletion of O 2 in the chamber within useful incubation periods. Most in situ O 2 measurements with microsensors using benthic landers have had a depth range limited to the top 5-10 cm (Chapter 2).
In order to overcome the latter limitation, Wenzhöfer et al. (2001) constructed a benthic lander with O 2 optodes that could penetrate half a meter into the sediment. In the western Atlantic they recorded O 2 down to 8-26 cm depth. In large areas of the ocean, however, O 2 reaches much further down and it is necessary to retrieve cores and perform measurements on board ship (e.g., Grundmanis and Murray, 1982;Fischer et al., 2009;Røy et al., 2012;D'Hondt et al., 2015). This can be done by drilling small holes in the side of the core liner and pushing microsensors into the sediment core to let them equilibrate at in situ temperature. Although pressure drop and transient heating of the cores tend to affect the O 2 consumption rates, these rates are so low in the deep-sea cores that the O 2 distribution hardly has time to change before measurements are done. The deep diffusive oxygen uptake (DOU) can therefore be calculated by diffusion-reaction modeling of the measured O 2 distribution.
An example from the North Atlantic gyre of such a deep O 2 profile is shown in Fig. 13A. The data reaching 25 m into the seabed were obtained during RV Knorr Expedition KN223. We calculated the diffusive O 2 flux using the O 2 gradient in the top 0.4 m and assuming a porosity of 0.75, a temperature of 2 • C, a salinity of 35, and a molecular diffusion coefficient of 1.05 × 10 − 9 m 2 s − 1 (Schulz and Zabel, 2006). The resulting flux is 0.0045 mmol O 2 m − 2 d − 1 . This is ca. 100-fold lower than the globally mean DOU in abyssal sediments (0.6 mmol O 2 m − 2 d − 1 ; Table 4). The linear gradient below 5 m depth shows even a 30-fold lower O 2 flux of 0.00014 mmol O 2 m − 2 d − 1 . However, neither of these fluxes represents the O 2 uptake of the seabed, as the following example shows.
The South Pacific Gyre is a global extreme with respect to low sedimentation rates and low O 2 consumption. In the central gyre, oxygen and aerobic respiration persist through the entire sediment column, even reaching down to the crustal basement, tens of meters below D'Hondt et al., 2009). Fig. 13B shows an example of O 2 data from a 5-m long piston core with a distinct drop in concentration in the upper 0-1 m. The O 2 flux modeled from these data was 0.014 mmol m − 2 d − 1 in the top 10 cm and 0.00018 mmol m − 2 d − 1 across 1.5 m depth. However, the main zone of oxygen consumption is not revealed at  Together with several ex situ measurements, these data from Fischer et al. (2009) reveal that nearly the entire O 2 consumption in the South Pacific gyre sediments takes place within a few-cm thick surface layer and only 1% takes place in the 10-20-m deep sediment column below. Fig. 14 shows that this is apparently a general phenomenon in the ocean gyres. The high-resolution DOU at the sediment surface is roughly 100fold higher than the O 2 flux calculated from measurements in retrieved sediment cores. The much higher oxygen consumption in the surface layer (0.12-1.32 mmol O 2 m − 2 d − 1 ) corresponds quite well to the multiple regression of DOU in the global seabed for the same water depths and sea-surface primary productivity (ca. 0.4 mmol O 2 m − 2 d − 1 ) (see  Table 5), this corresponds to a potential oxygen consumption rate of 0.14 mmol O 2 m − 2 d − 1 .
The sedimentation rates in the central gyre are extremely low, 0.1-1 mm per 1000 years, bioturbation is negligible, and the thin surface layer of recently deposited organic matter is therefore the main oxygen sink. Oxygen that is not consumed in this layer, on the other hand, may diffuse far down into the underlying deposits that are many millions of years old and in which the buried organic matter is extremely resistant to further degradation . If such an extreme displacement of oxygen consumption towards the sediment surface is generally characteristic of abyssal sediments from low-productivity regions, it means that O 2 data measured at several-cm resolution in retrieved sediment cores may grossly underestimate the benthic oxygen uptake. There is clearly a need for further in situ microsensor data to understand how oxygen consumption is controlled in ocean gyre sediments. D'Hondt et al. (2015) used data on total sediment thickness and mean sediment accumulation rates to develop a global map of regions where O 2 is likely to penetrate through the entire sediment column and thereby exclude anaerobic microorganisms and anaerobic processes (Fig. 15). These regions coincide to a large degree with the geographic extent of the ocean gyres. D'Hondt and co-authors estimated that O 2 penetrates all the way to the basaltic crust in 15-44% of the Pacific seabed and in 9-37% of the global seabed.
Modeling of sediment O 2 profiles in the central South Pacific Gyre yielded calculated O 2 consumption rates per sediment volume that ranged from 6 × 10 − 11 mol O 2 cm − 3 yr − 1 in the upper few meters to a minimum of 3 × 10 − 13 mol O 2 cm − 3 yr − 1 at 50 m depth (Site U1370; D' Hondt et al., 2015). At that specific site, the residual O 2 concentration was 8 μM at 50 m depth. This means that the in situ turnover time of the small O 2 pool at that depth was 20,000 years. A similar calculation for a deep sediment core from the North Pacific Gyre indicated an O 2 turnover time of 40,000 years at 25 m depth (Røy et al., 2012). This extremely slow aerobic respiration is driven by deeply buried organic matter that becomes increasingly persistent with depth and age, but continues to be  slowly degraded and to feed the deep sub-seafloor biosphere . The slow respiration is also driven by H 2 and oxidants produced in situ by radiolysis of water due to high-energy radiation by the natural, long-lived radionuclides 40 K, 232 Th and 238 U (Blair et al., 2007;Sauvage et al., 2021). However, this radiolysis-fueled respiration may not contribute to net O 2 reduction, since oxidants are produced in stoichiometric balance with the H 2 (Sauvage et al., 2021). In the fully oxic sediments underlying the South Pacific Gyre, the extremely low energy availability maintains viable communities of aerobic microorganisms in deep sediments of up to 100 million years age (Morono et al., 2020).
It was a striking observation in the periphery of the South Pacific Gyre (Site U1371) that O 2 penetrated 0.9 m down into the sediment, but that O 2 was also supplied from the underlying ocean crust and penetrated 15 m up into the sediment from below (D'Hondt et al., 2015). A similar observation of a deep O 2 and nitrate flux from the underlying crust was observed in the tropical eastern Pacific (D'Hondt et al., 2004). It was explained by a slow, thermohaline circulation of deep, oxic seawater that entered and vented the crust through seamounts. This long-distance circulation caused a lateral seepage of seawater through the crustal aquifer that slowly leaked O 2 up into the overlaying sediment Bekins et al., 2007). A deep O 2 source was also observed on the western flank of the Mid-Atlantic Ridge, driven by a hydrothermal circulation through the ridge crust (Orcutt et al., 2013). While these inverted O 2 fluxes are interesting and geochemically and hydrologically important, they are insignificant for the global seabed O 2 budget.

Hadal trenches
The hadal zone is the deepest part of the global ocean, stretching from 6 km to almost 11 km water depth at the bottom of the Challenger Deep in the Mariana trench. The hadal zone represents 45% of the oceanic depth range and covers an area of 3.44 × 10 6 km 2 , corresponding to 1.1% of the global seabed (Harris et al., 2014). It includes 47 individual trench systems and is dominated by 27 large trenches that extend along tectonic subduction zones, mainly in the Pacific Ocean (Stewart and Jamieson, 2018).
It is a general observation that biological activity declines with increasing ocean depth. The hadal trenches were therefore originally believed to be largely deprived of life, but it has now been documented that the trench systems act as depo-centers for organic material. Hadal sediments often harbor surprisingly high abundances of meiofauna and microorganisms that thrive at the extremely high hydrostatic pressure of 10 3 bar (Danovaro et al., 2002(Danovaro et al., , 2003Itoh et al., 2011;Schauberger et al., 2021). The biogeochemical function of hadal trenches therefore cannot be assessed by simply extrapolating relations to water depth, which otherwise appears to be a robust procedure for most other oceanic settings.
Trenches have in cross-section a distinct V-shaped bathymetry and are situated in regions of high seismic activity. Thus, material deposited along the trench slopes is frequently translocated to the bottom of the trench (the trench axis) during mass wasting, typically triggered by earthquakes (Itoh et al., 2000;Bao et al., 2018). Using bathymetric mapping, Parasound profiling, and sediment TOC measurements it was estimated that 1 Tg of organic carbon was re-deposited to the interior of the Japan Trench during the Tohoku-oki earthquake in 2011 (Kioka et al., 2019). This corresponds to 1‰ of the total annual TOC deposition on the deep global seabed (>4000 m). Based on preliminary assessments using specific biomarkers, signatures of carbon stable isotopes, and thermochemical analysis of deposited TOC, it appears that trench sediments represent a globally significant marine storage of resilient terrigenous organic material (Luo et al., 2017;Xu et al., 2020Xu et al., , 2021. Hadal trenches may thus play an important role for the long-term burial of organic material in the deep sea, much beyond their areal contribution to the seafloor . In situ measurements of DOU document 2-5 times higher benthic O 2 consumption along hadal trench axes as compared to the adjacent abyssal plain ( Fig. 16; Glud et al., 2013;Wenzhöfer et al., 2016b;Luo et al., 2018). Together with elevated cell abundances, this indicates that it is not just relict organic carbon that accumulates in the hadal trenches, but deposition also includes labile organic material that can sustain a relatively high level of microbial activity. Generally, the average benthic O 2 consumption in the trenches correlates well to the regional surface primary production, but measurements within individual trenches also show extensive variations, indicating that the linkage to surface production is modulated by other factors .
Sediment mass wasting is a catastrophic event for mega-and macrofauna within the trenches, and their carcasses tend to be deposited along the trench axis, representing highly labile nutrition for the hadal communities . Four months after the Tohoku-oki earthquake and the associated tsunami, the bottom water in the Japan Trench was still turbid and contained high levels of suspended detrital material . Recovered sediment cores from the trench axis revealed recent, but highly variable, deposition of labile organic material, even within relatively short distances of a few kilometres. Surprisingly, a high level of the short-lived 134 Cs isotope, which could only originate from the flooded Fukushima nuclear power plant, was found at the trench axis . Thus, large seismic events can efficiently transport terrigenous and labile coastal material to the trench interior. Variability in benthic O 2 uptake within the trench might thus reflect the deposited amount and the time since the last mass deposition.
The distinct V-shaped bathymetry of subducting trenches facilitates tidally induced internal waves (Van Haren, 2020), which cause local erosion of the trench slopes. Such waves may also cause focusing or winnowing of sediment material along the trench axis and enhance transport of surface derived material towards the trench bottom (Turnewitsch et al., 2014). This further contributes to the relatively high benthic deposition and large spatio-temporal variations in benthic O 2 consumption in hadal trenches .
The benthic O 2 uptake is generally a robust proxy for total benthic carbon mineralization, which includes the efficient re-oxidation of reduced products of anaerobic mineralization. However, trench sediments with frequent mass wasting and little bioturbation driven by macrofauna might be the exception to this rule. Trench sediments may not show a simple decline in TOC availability with increasing sediment depth Wenzhöfer et al., 2016b). They may even have an inverted distribution of sediment layers with old and sparse TOC overlaying layers with younger and more abundant TOC (Xu et al., 2021). Thus, mass wasting may facilitate burial of labile organic material well below the oxic surface layer. A recent study at 7000 m water depth in the Atacama Trench showed high levels of iron(III) reduction and sulfate reduction, well below the sediment surface, and deep layers of reduced iron sulfides . Given frequent masswasting, as apparent by distinct sediment layering and lack of efficient bioturbation, the reduced products from anaerobic degradation may be buried and never be re-oxidized in such settings. In that case, the benthic O 2 consumption includes only a fraction of the total carbon mineralization ). Yet, the importance of anaerobic carbon mineralization in hadal trenches remains largely unknown.
Hadal trenches also act as depositional centers for persistent pollutants. The hadal biota appear to hold high levels of anthropogenic organic pollutants and heavy metals, presumably transported to the trench interior by sinking carrions (Jamieson et al., 2017;Blum et al., 2020;Sun et al., 2020). Recent studies from the Atacama and Kermadec trench regions show benthic mercury accumulation rates that are 30-50 times higher than the average for the deep ocean . Hadal trenches are thus important sites for oceanic deposition, possibly even of global importance. At the same time, trenches function as deepsea hot spots for early diagenesis, mediated by largely unexplored microbial communities.

Global distribution of seabed O 2 uptake
In this chapter, the extensive database described in Chapter 2.2 is used to analyze the multiple correlations to environmental parameters and to develop a global map of the seabed O 2 uptake rates. We also discuss, which environmental factors play the most important role for the O 2 uptake.

From database to global map
We analyzed data for total oxygen uptake (TOU) and diffusive oxygen uptake (DOU) collected from a total of 3904 stations. Due to potential problems with ex situ measurements, described in Chapter 2, we used only data from in situ measurements for depths >10 m by our multiple regression analysis. We also excluded measurements according to other defined criteria, as explained in Chapter 2. Multiple measurements from the same station were averaged before our multiple regression analyses.
This selection removed 70% of all data from the original database and resulted in a total of 798 data from stations distributed throughout the world oceans. These in situ data were used for a global quantification of seabed O 2 uptake. Fig. 17a shows the geographic distribution of available data, plotted on a map of ocean depths. It is apparent that the highest data density is strongly skewed towards the continental shelves and slopes along the ocean margins, and towards the northern hemisphere. The data density is low in the open Pacific Ocean and Atlantic Ocean and data are largely missing from the Indian Ocean. Table 1 shows the geographic area of seven depth zones in the ocean, the number of total observations of TOU and DOU in each zone, and the density of observations per million km 2 . The shelf region out to 200 m   -averaged (1998-2007) sea-surface net primary production (NPP) distribution (for reference, see text). Note that color scales are non-linear. water depth covers 7% of the global ocean and has a data density of 8-18 TOU and 1.3-3.0 DOU measurements per 10 6 km 2 . The abyssal plains at >4000 m depth cover 54% of the global ocean and have a mean data density of 0.3 TOU and 0.2 DOU per 10 6 km 2 . That corresponds to 3 and 2 data points, respectively, per area the size of the USA or Europe. There is clearly a need for more in situ data, in particular from the deep sea, to reach a representative data coverage and reduce the large statistical uncertainty of global estimates. Fig. 17b shows the distribution of net primary production (NPP) in the surface ocean monitored by remote sensing. We used the NPP data available through the Oregon State University (www.science.oregon state.edu/ocean.productivity/custom.php) on monthly climatologically averaged physical data computed over a 10-yr period (1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007). The distribution based on calculations from remote sensing data was processed with the vertically generalized production model, VGPM, of Behrenfeld and Falkowski (1997). The map shows that the highest phytoplankton primary production is strongly skewed towards the higher latitudes and towards the ocean margins with westward extensions along the Equator in the Pacific and the Atlantic Ocean.
This coincidence between the highest data density, the shallowest ocean depths, and the highest sea-surface NPP means that deep and oligotrophic regions of the ocean are statistically poorly represented by the data distribution. This is illustrated in Fig. 18, where the white bars correspond to the frequency distribution of observations in our database relative to (a) the ocean depths and (b) the sea-surface NPP. The blue bars show the actual, global distribution of (a) ocean depths and (b) seasurface NPP. Fig. 18 shows that the frequency of data is strongly skewed towards shallow depths and high sea-surface NPP. This confirms that the seabed O 2 uptake remains poorly constrained in large, central parts of the ocean. Yet, correlation plots of TOU and DOU versus ocean depth and versus sea surface primary production indicate that the available data from abyssal plains in low productivity regions fall along the general trend of the entire seafloor dataset (see Chapter 5.2).
For each depth region, as defined in Table 1 (excluding data from depths <10 m) we calculated the mean value of log transformed TOU and DOU data versus log depth and versus log NPP (Fig. 21) and determined the standard error of the mean. We then performed a multicomponent linear regression analysis of the log-log-transformed mean values and derived the following empirical relationships for TOU and DOU: where z is ocean depth and NPP is net primary productivity. The model parameters A, B, and C for TOU and DOU, as well as the 5% and 95% confidence interval (CI), the coefficients of determination (R 2 ), and the P values are shown in Table 2. A confidence interval was calculated for each depth interval from the mean ± 2 × standard error. The CI's for each depth interval were used to run a multiple regression and calculate the 5% and 95% values for the model fit. The model parameters show, as expected, that the log(TOU) and log(DOU) are negatively correlated with ocean depth and positively correlated with sea-surface NPP. The coefficient of determination, R 2 , is high, 0.995 and 0.998 for TOU and DOU, respectively, with a significance level, P < 0.001. When using Eq. 5.1 and 5.2 for geographic extrapolation and mapping, a correction factor was applied to account for the skewness bias inherent in the back conversion from a log-log transformed linear regression model to arithmetic units. As described by Middelburg et al.

Table 1
Geographic area of different depth zones in the global ocean. The number of in situ data for total oxygen uptake (TOU) and diffusive oxygen uptake (DOU) from each zone is shown and the number of observations per 10 6 km 2 is calculated.   (1997), high values lose significance relative to low values in a log-log linear regression, which therefore results in a bias towards lower values. The correction factor (CF) takes this bias into account and can be calculated using the variance of the data, CF = e 2.65×variance (Middelburg et al., 1997). The variance was calculated for each ocean depth interval and used to correct the mean value for that interval. The correction factors for TOU ranged from 1.22 to 1.61, lowest on the inner shelf and highest on the upper slope.
Using the power law equations derived from the multiple linear regression of log-log transformed mean values per ocean depth (Eq. 5.1 and 5.2), we predict the modeled TOU and DOU for each observation site based on the corresponding water depths and sea-surface NPP values. To test the validity of this model, the predicted (modeled) values are compared to the actual value for each observation site. In general, the predicted oxygen uptake rates are in good agreement with the observed values (Fig. 19a, c), with the predicted oxygen uptake rates generally deviating much less than one order of magnitude from the observations and with the median values matching 1:1 (Fig. 19 b,d).
To create a global map of sediment oxygen uptake on a geographically highly resolved 0.1 • × 0.1 • grid, the predictive relationships for TOU and DOU were combined with a gridded global dataset of water depth (Amante and Eakings, 2009) and annual mean sea-surface NPP (see text above). For each 0.1 • × 0.1 • grid cell of the world ocean, we predicted TOU and DOU by applying Eq. 5.1 and 5.2, respectively. We also applied the correction factors mentioned above to account for the skewness bias inherent in the back conversion from a log-log transformed linear regression model to arithmetic units. TOU and DOU for coastal sediments at 0-10 m water depth were not calculated from this multiple regression but were calculated separately using only the arithmetic mean of data from that depth interval.
The resulting map of global TOU distribution is shown in Fig. 20. A separate map of global DOU distribution is shown in Supplementary Fig.  S1. The DOU map is graphically very similar to the TOU map. The color scale is arithmetic and covers a TOU range from 0 to 119 mmol O 2 m − 2 d − 1 . The scale is designed to provide high color sensitivity by the low TOU values in the open ocean and low color sensitivity by the high TOU values along the ocean margins. The map shows a very strong focus of seabed O 2 uptake towards the continental shelves and upper slopes. In the deep sea, the O 2 uptake reflects the distribution of sea surface net primary production (cf. Fig. 17B).

Water depth and sea-surface net primary production
Water depth is the most important factor controlling the flux of organic matter to the seabed and, thereby, its oxygen uptake rate. Fig. 21a and b show the total and the diffusive O 2 uptake as a function of water depth, ranging from 1 m to the abyss. The TOU and DOU data have been fitted by a log-log linear correlation for mean values per corresponding depth region as defined in Table 1, and excluding water depths <10 m. The R 2 of these correlations is high (>0.99; Table 2). In shallow water of <10 m there is no apparent depth dependence of TOU or DOU. We have therefore fitted a constant mean rate for each parameter (dashed horizontal lines in Fig. 21a and b) and used this for the global budget in Chapter 6.3. Sea-surface primary production is the secondmost important factor controlling benthic O 2 uptake ( Fig. 21c and d). All data in Fig. 21 are from in situ measurements, except for the DOU from 0 to 10 m where very few in situ data exist and where also ex situ data were included.
The large scatter of data in Fig. 21 is partly due to the fact that these data are snapshots of processes that vary strongly in both time and space. As described in Chapter 4.2, the deposition of fresh phytoplankton detritus is often a brief event lasting only a few days or weeks of the year, even in the deep sea. During such deposition events, the O 2 consumption rate can be very high right at the sediment surface, while most of the year the rate is moderate and more constant. Depending on whether in situ measurements catch such events, the measured rates of TOU and DOU can show strong variability and conceal the importance of sea-surface NPP in data scatter. Furthermore, the degradation rate of deposited organic matter depends on a complex interplay of biological, geochemical and physical properties of the sediment (e.g., LaRowe et al., 2020).

Oxygen penetration depth (OPD)
The oxygen penetration depth (OPD) into the seabed is determined as the depth where O 2 profiles appear to reach zero concentration. The OPD and the diffusive oxygen uptake rate are mutually coupled, although not in a simple manner. The relationship depends on the depth distribution of volumetric O 2 consumption rates, especially when O 2 penetrates deeper than a cm. Such deeper O 2 profiles show highest  respiration rate at the sediment surface, where fresh detritus settles, and at the bottom of the oxic zone, where reduced products of anaerobic degradation become oxidized (Fig. 5). The relationship also depends on the porosity of the sediment and on the molecular diffusion coefficient of O 2 , which is a function of temperature. Revsbech et al. (1980) tested a simple relationship for a depth-independent O 2 respiration in a coastal marine sediment with 1-5 mm O 2 penetration: where D s is the molecular diffusion coefficient in the sediment and φ is sediment porosity, both considered to be constant within the depth interval. They found a relatively deeper OPD than predicted by Eq. 5.3 and explained this as a result of bioturbation, perhaps by meio-and microfauna. Others have tested the same relationship on sediment O 2 profiles from different stations, e.g. off the SW coast of Japan (Glud et al., 2009a), in the South Atlantic (Hensen et al., 2006), and off the east and west coast of the USA (Cai and Sayles, 1996). Cai and Sayles (1996) found that the relationship holds well for O 2 profiles along the continental margin but that it underestimates the OPD in the open ocean. Seitaj et al. (2016) tested the relationship in a saline coastal reservoir and found that the pronounced oxidation of products from anaerobic mineralization at the bottom of the oxic zone tended to reduce the oxygen penetration to half of the penetration calculated from Eq. 5.3. Our database indicates a log-log linear correlation between DOU (in mmol O 2 m − 2 d − 1 ) and OPD (in mm) (Fig. 22a): with an R 2 of 0.64 and a P value of <0.001. A similar relationship has been observed in earlier analyses Glud, 2008). One single data point, which was not included in the log-log linear regression, shows extremely deep O 2 penetration relative to the DOU (Fig. 22A). This data point by Fischer et al. (2009) originates from the central gyre in the South Pacific and is discussed in Chapter 4.4. The general goodness of the fit is illustrated in Fig. 22B, which shows the ratio between predicted and observed OPD. The peak of this frequency distribution nearly coincides with the perfect 1:1 fit (log = 0) between OPD pred and OPD obs . As discussed in Chapter 4.4, the relationship in Eq. 5.4 breaks down in the deep sea, in particular in the mid-oceanic gyres where bioturbation is practically absent and the O 2 consumption rate in the uppermost few cm of the sediment may be orders of magnitude higher than in the subsurface. An example from the South Pacific Gyre presented in Fig. 13B shows O 2 penetration to the ocean crust (OPD > 10 m), i.e. log OPD (mm) >4, while the DOU was 0.23 mmol O 2 m − 2 d − 1 , i.e. log DOU = − 0.63 . This illustrates how the extremely deep O 2 penetration in oceanic gyre sediments gives a misleading expectation that the DOU must also be extremely low. Data in Fig. 13, 14 and 22 suggest that this is not the case.  Fig. 22. A) Log-log plot of observed oxygen penetration depth (OPD) versus observed diffusive oxygen uptake (DOU). Oxygen profiles visually affected by bioirrigation were excluded from this analysis. The line represents the log-log linear regression (Eq. 5.4). A strongly deviating data point is marked by a red arrow (see text). B) Histogram of predicted OPD (OPD pred ) versus observed OPD (OPD obs ) relative to the log-log linear regression in (A). The red line indicates a perfect model fit (predicted/observed OPD = 1).

Temperature
Temperature has a fundamental impact on the metabolic rates of benthic microorganisms and fauna. Temperature is also a major environmental variable that selects for the predominant biotic communities. The temperature of deep-sea sediments is low, 0-3 • C, and generally very stable for each ocean region. Short-term experiments with the oxygen consumption rate of marine sediments from the temperate inner shelf show maximum rates at 20 • C both summer and winter but relatively higher rates at low temperature at the end of winter (Fig. 23). Similar experiments with Arctic sediments showed maximum rates at 10 • C (Thamdrup and Fleischer, 1998). However, the true adaptation of sediment bacteria to permanently low temperature is not clearly detected by such short-term respiration measurements but rather from the temperature dependence of the growth yield of the microbial communities, which may be highest at the low in situ temperature (Scholze et al., 2021).
Sediments from tropical, temperate and polar latitudes are exposed to temperatures that may differ by 20-30 • C on the inner shelf. It is not clear to which extent this large temperature difference controls the rate of oxygen uptake or the efficiency of organic matter mineralization. Experiments on the degradation of kelp across a wide range of geographic latitudes and temperatures of coastal sediments showed enhanced mineralization rates and shorter retention times of degrading biomass in warmer climates (Filbee-Dexter et al., 2021). The accumulation rate of POC is relatively high in Arctic sediments that receive a large contribution of terrestrial organic matter (Wallmann et al., 2012). Yet, we are not aware of data that indicate that the fraction of deposited, plankton-derived organic matter, which becomes buried, is controlled by long-term temperature. Benthic microbial and faunal communities in different regimes are well adapted to the ambient temperature, and bacteria in polar sediments have relatively high cell-specific respiration rates at the low in situ temperature (Knoblauch et al., 1999).
The situation is different in temperate coastal sediments with a strong seasonal temperature variation. Here, the sediment oxygen uptake tends to vary over the year together with the ambient temperature. The variation is a combined effect of accelerated respiration rate of the benthic organisms at higher temperature and accelerated influx of fresh organic matter to feed their respiration during spring and summer. In many coastal waters this potentially enhanced oxygen consumption is counteracted by hypoxia during summer, which limits the oxygen flux and may result in a rather constant oxygen uptake throughout the year (Rasmussen and Jørgensen, 1992;Glud et al., 2003). The sediment may instead build up an oxygen debt in the form of iron sulfide minerals that are gradually oxidized during fall and winter (Seitaj et al., 2015).

O 2 concentration in bottom water
Sediment oxygen uptake in oxygen minimum zones along the ocean margins is discussed in Chapter 4.2. Since benthic O 2 uptake is a poor proxy for mineralization rates during permanent or transient hypoxia, we decided to exclude data from our multiple correlation analysis when the bottom water O 2 concentration was below 25 μM. Benthic mineralization of organic carbon in oxygen minimum zones was therefore assumed to be similar to oxic regions with similar surface productivity and similar water depth. Sediments underlying such oxygen minimum zones constitute only 4% of the outer shelf and upper slope.

Fauna
Benthic animals play important roles for the oxygen uptake in marine sediments, as discussed in Chapter 3.3. Apart from their own respiration, they increase the extent of the sediment-water interface and of the diffusive oxygen flux through their burrow structures. They also increase the advective oxygen flux by ventilating their burrows and enhance the degradation of organic matter by particle fragmentation and sediment reworking, so that a smaller fraction of the deposited organic carbon is buried.
The importance of fauna for the benthic O 2 exchange is often assessed by the difference between TOU and DOU (see Chapter 3.3). Previous data compilations of in situ TOU and DOU, have demonstrated that, while TOU exceeded DOU by a factor of 2-4 in coastal settings, values converge beyond the shelf and become similar at abyssal water depths (Archer and Devol, 1992;Glud, 2008).
This may partly reflect that the O 2 penetration depth increases with water depth from mm in the coastal areas to cm or dm in abyssal settings and faunal activity in the upper sediment layers therefore has little impact on the benthic O 2 uptake in the deep sea (Glud et al., 1994. Furthermore, the biomass of benthic macrofauna generally drops relatively faster with water depth than the biomass of microorganisms (Rex et al., 2006;Wei et al., 2010), implying that the relative importance of microbially driven O 2 consumption increases with water depth. Exceptions to this include sponge grounds and cold-water coral reefs, which act as deep-sea hotspots for organic carbon mineralization (Beazley et al., 2013;Rovelli et al., 2015).
This general depth trend was not observed in the extended database of the current study. This could imply a depth-independent contribution of fauna to the benthic O 2 exchange rate. However, the scatter is extensive by inclusion of all the non-coordinated data (Fig. 21). Indeed, when selecting only the in situ studies where both TOU and DOU were measured in parallel at the same site at the same time, the fauna contribution clearly declines with increasing water depth. This is apparent from the calculated decrease in mean TOU:DOU ratio from >2.2:1 on the inner shelf to 1.2:1 in the abyss (Fig. 24).

Physical factors
A range of physical factors affect the mechanisms of sediment oxygen uptake and potentially also of the resulting total oxygen uptake, as described in Chapter 3.1. Sediment topography enhances the effective area of the sediment-water interface at different scales and thereby the surface area through which oxygen may diffuse. In very shallow waters, this should be taken into account by oxygen flux calculations based on microsensor measurements of vertical O 2 gradients in the diffusive boundary layer or in the sediments. That is rarely done, which may cause an underestimation of the true diffusive oxygen uptake when the topography is distinct and the O 2 penetration depth is shallow. Examples from coastal sediments show that the underestimation is generally only about 10% or less, but it can also be higher (Gundersen and Jørgensen, 1990;Røy et al., 2002;Glud et al., 2003). Bottom currents and wave motion on the inner shelf affect the sediment oxygen uptake. Higher current speeds cause a thinner diffusive boundary layer, which may transiently enhance the O 2 diffusive flux into sediments with high potential oxygen consumption rate ; Chapter 3.1). Currents combined with sediment topography generate lateral pressure gradients that induce pore water advection in sand and in other porous sediments. Such an advective flow within the sediment has been documented in experimental flumes and is also pronounced in coastal sands with, e.g. ripple marks or faunal casts and feeding pits (Huettel and Webster, 2001;McGinnis et al., 2014). By increasing current speeds, sediment resuspension plays an increasing role by lifting freshly deposited organic particles up into the benthic boundary layer where oxic mineralization rates are enhanced (Holtappels and Lorke, 2011;Camillini et al., 2021).
In conclusion, the oxygen uptake of the seafloor is controlled by multiple parameters that complicate the accurate measurement of TOU and DOU. Some of these parameters affect how and where the oxygen consumption takes place but do not affect the resulting total oxygen uptake. Yet, the ultimate control on benthic TOU is the flux of particulate organic matter, of which most is completely mineralized in the sediment at the expense of O 2 and only 6% is buried, as an estimated global mean (Stolpovsky et al., 2015).

Role of the seabed in the oceanic carbon cycle
The map of benthic O 2 consumption rates presented in Chapter 5.1 enables us to assess the quantitative importance of the seabed for the global marine carbon cycle. This requires that O 2 consumption rates are converted to benthic carbon mineralization rates assuming a respiratory quotient for early diagenesis in marine sediments. However, the nature of organic matter reaching the seabed depends on the sources and age of the material, which differ substantially with water depth and distance to terrestrial sources. The relative amount of O 2 required to oxidize deposited organic matter therefore differs across the seascape (Williams and del Giorgio, 2004). In addition, the extent to which anaerobic mineralization is included in the O 2 consumption depends on the degree of recycling of the anaerobic pathways and the burial rate of reduced products from anaerobic diagenesis (Canfield, 1993;Chapter 3.5). In this chapter, we discuss sources of organic matter and qualify the applied benthic respiratory quotient used to assess the benthic carbon mineralization from our TOU data base. The calculated budgets for benthic carbon mineralization of different oceanic depth horizons is discussed and compared with previous assessments as derived from sediment incubations and sediment trap studies. Finally, the quantitative importance of the seabed for the global marine carbon cycle is assessed and discussed.

Sources of organic carbon in marine sediments
Rivers deliver large quantities of terrigenous organic material to the coastal ocean, where it is retained, mineralized or transported onwards to the open ocean (Burdige, 2005;Kandasamy and Nath, 2016). Current estimates have considerable uncertainties, but are in the order of 35 Tmol C org yr − 1 of terrigenous organic material entering the coastal zone where 40% is mineralized, 50% is sequestered in the sediment while 10% is exported to the open ocean (Schlesinger and Melack, 1981;Regnier et al., 2013). The relatively low estimated mineralization efficiency in the coastal zone reflects the complex and recalcitrant nature of terrigenous organic material with high content of nitrogen-poor humic substances. Vegetated estuarine habitats such as salt marshes, mangroves, seagrass beds and macroalgae represent other important marine sources of organic carbon. Also here, assessments are highly uncertain. It has been estimated that between 95 and 280 Tmol C org yr − 1 is exported from vegetated estuarine habitats to the coastal ocean and beyond (Duarte et al., 2005). Most of this material is presumably mineralized and deposited in coastal and shelf sediments, but recent studies based on e-DNA sequencing from deep sea sediments suggest a significant export of organic material from coastal vegetated habitats to the deep ocean (Ortega et al., 2019). Yet, the above estimate of organic carbon export appears to be exceedingly high compared to the total oxygen consumption of continental shelf sediments (Table 5).
More specific, it has been estimated that approximately 40% of the organic material produced in macroalgal beds is exported, corresponding to 50 Tmol C org yr − 1 (Krause-Jensen and Duarte, 2016). Most of this material is retained and mineralized in shelf settings, but 3 Tmol C org yr − 1 reaches deep sea sediments as particulate organic material (Krause-Jensen and Duarte, 2016). A similar export has been proposed for the estimated 40 Tmol C org yr − 1 of organic material produced in seagrass meadows (Duarte and Krause-Jensen, 2017). Thus, the benthic supply of organic material from rivers and from coastal marine vegetated habitats is substantial, but despite its high C:N ratio and refractory nature, a considerable fraction of these organic sources is degraded in coastal, shelf and even deep-sea sediments (Burdige, 2005). The benthic imprint of refractory organic material from vegetated habitats clearly attenuates with distance from the continents (Hedges and Keil, 1995;Regnier et al., 2013;Li et al., 2017).
Benthic microphytes also contribute to the primary production of shallow settings (0-10 m) (Cahoon, 1999;Glud et al., 2009b) and represent a quantitatively important and nutritious food source for coastal food webs (Miller et al., 1996;Christiansen et al., 2017). Benthic primary production may be detected even down to ~200 m water depth, but in most coastal settings light attenuates more steeply and benthic primary production rarely contributes to the benthic carbon supply beyond 20-30 m water depth (Gattuso et al., 2006;Rovelli et al., 2019;Attard et al., 2014).
The distribution of phytoplankton in the ocean is very heterogeneous, with high concentrations along the ocean margins, in tropical and subtropical upwelling regions, and at high latitudes, and with extremely low concentrations in the central gyres of the Pacific, Atlantic and Indian Oceans (Fig. 17b). Estimates of the depth-integrated net primary productivity have been made from remote sensing of ocean color coupled to general circulation models and biogeochemical models. Global estimates vary by a factor of two between models, with 4300 Tmol C org yr − 1 as a recent, mean value (Field et al., 1998;Carr et al., 2006;Westberry and Behrenfeld, 2014;Silsbe et al., 2016).
The export of organic matter from the photic surface waters, which feeds the flux of particulate organic carbon (POC) sinking towards the seafloor, has high spatial and temporal variability. Different data and models have been used, and estimates of the export efficiency differ considerably among studies (e.g., Laufkötter et al., 2016). As examples, Najjar et al. (2007) estimated a global POC export flux of 1400 ± 500 (±s.d.) Tmol C org yr − 1 across a 75 m depth boundary, while Dunne et al. (2007) estimated a global export flux of 800 ± 300 Tmol C org yr − 1 with a flexible lower boundary of the photic surface water.
The nutritious value of sinking POC generally declines with increasing ocean depth leading to a relatively higher fraction of recalcitrant organic material reaching the deeper realm (Loh and Bauer, 2000). Yet, pulses of relatively fresh phytodetritus are observed at great water depth and on abyssal sediments (Billett et al., 1983;Beaulieu, 2002). The quantity and quality of organic material reaching the seabed therefore vary extensively across the global seascape. Inter-annual and seasonal variation in ocean productivity and riverine run-off combined with extensive horizontal transport complicate the linkage between the estimated deposition of organic material and the measured benthic O 2 consumption at any given location.

Conversion of O 2 consumption to organic carbon oxidationthe Respiratory Quotient
The conversion of seabed oxygen consumption to organic carbon mineralization is not straightforward. The export production from the photic surface ocean has an elemental composition close to the originally proposed Redfield ratio of P:N:C = 1:16:106. Different authors have estimated that the respiratory quotient (RQ), linking the O 2 required for the complete oxidation of this organic carbon and nitrogen to CO 2 and NO 3 − , is P:N:C:-O 2 = 1:16:103:-172 (Takahashi et al., 1985) or 1:16:117:-170 (Anderson and Sarmiento, 1994). The stoichiometry between organic C oxidized and O 2 consumed by the respiration of phytoplankton biomass may vary, but is mostly in the range of 0.67-0.78, depending on the growth phase and composition of the algae and their content of protein, carbohydrate, lipid and nucleic acid (Finkel et al., 2016;Tanioka and Matsumoto, 2020). Based on a mean composition of phytoplankton, Anderson (1995) derived the following equation for a complete mineralization: This implies that the oxidation of organic N to nitrate consumes 13% of the O 2 . If the organic N were released as ammonium and not being oxidized to nitrate, less O 2 would be required and the RQ would increase to 0.82. Körtzinger et al. (2001) estimated an RQ of 0.75 for the Northeast Atlantic, while Tanioka and Matsumoto (2020) compiled data from the global ocean and determined a mean RQ of 0.7 for the complete mineralization of export production in the upper ocean. They noted that the C:N ratio of sinking POC increases systematically with depth in the ocean, but so does the H:C ratio, thus maintaining a rather constant RQ of 0.7 into the abyss.
As stated in Chapter 6.1, the coastal ocean receives a large contribution of organic matter from terrestrial runoff and from vegetated coastal ecosystems. These POC sources have a higher C:N ratio than the Redfield ratio of 7:1 and are much richer in structural carbohydrates than phytoplankton. As carbohydrate has the highest mean oxidation state among the main cellular components, these POC sources require relatively less O 2 to oxidize the organic carbon and therefore have a higher C org :O 2 ratio by mineralization than the phytoplankton RQ of 0.7. This implies that the C org :O 2 ratio of freshly deposited organic matter is relatively high in the coastal zone and on the inner shelf and declines offshore.
The fate of NH 4 + released from organic matter degradation in marine sediments complicates this C org :O 2 stoichiometry (e.g., Li and Katsev, 2014). Deeply oxygenated offshore sediments tend to be nitrate sources, as ammonium released from the mineralized organic nitrogen is quantitatively oxidized to nitrate, so that the C org :O 2 ratio approaches that of complete aerobic mineralization. The oxygen consumption by nitrification may thereby account for up to 12% of the total oxygen uptake (Li and Katsev, 2014). Coastal sediments, in contrast, tend to be nitrate sinks, where ammonium from organic nitrogen mineralization is oxidized to nitrate, which is recycled within the sediment by reduction to N 2 during denitrification. Nitrate from the overlying water diffuses into the sediment and is also denitrified to N 2 . Ammonium and nitrite may be used by anammox bacteria near the sediment surface and converted to N 2 . It is not important for the C org :O 2 stoichiometry whether the conversion to N 2 takes place via anammox or via nitrificationdenitrification. By anammox the stoichiometry is: By nitrification-denitrification, more O 2 is consumed per ammonium than by anammox but, at the same time, additional organic carbon is oxidized during denitrification. Seitzinger et al. (2006) estimated that 80% of the global denitrification in shelf sediments is due to coupled nitrification-denitrification (or anammox) and 20% is due to nitrate uptake from the water column. As a result, the C org :O 2 ratio of shelf sediments is relatively high and approaches a stoichiometry without NH 4 + oxidation. This reinforces the above mentioned effect of terrestrially derived POC towards a higher RQ in shelf sediments. Another complication of the C org :O 2 stoichiometry is the fact that much of the degradation of sediment organic matter takes place below the oxic zone and involves fermentation processes and anaerobic respiration with alternative electron acceptors, as described in Chapter 3.5. This anaerobic respiration produces NH 4 + , Mn 2+ , Fe 2+ and H 2 S as well as CH 4 and a range of solid-phase minerals such as pyrite (FeS 2 ). Bioturbation and upwards diffusion of the dissolved species drive a redox cascade of oxidation processes by which most of these reduced products becomes oxidized with O 2 as the ultimate oxidant. As an example, for coastal sediments, 50% of the oxidation of the organic matter may be due to sulfate reduction from which 90% of the produced sulfide is re-oxidized at the expense of O 2 (Jørgensen, 1982). This means that, in such a case, only 5% of the total organic carbon oxidation is not included in the O 2 uptake. If an RQ of 0.85 were assumed for the aerobic mineralization, then a C org :O 2 stoichiometry closer to 0.90 should be chosen for the complete mineralization in order to convert sediment oxygen uptake to total organic C oxidation in the seabed.
An alternative approach to determine the C org :O 2 ratio is to directly measure the fluxes of O 2 into the sediment and dissolved inorganic carbon (DIC, total CO 2 ) out of the sediment using benthic in situ incubation chambers. Table 3 compiles data from such concurrent DIC and  Hammond et al. (1996) O 2 flux measurements. The resulting RQ shows large variations between 0.69 and 1.31 with no clear trend with ocean depth. A reason for this large variation is probably that carbonate dissolution in the sediment enhances the DIC flux and requires concurrent analysis of alkalinity production or modeling of the Ca 2+ and Mg 2+ fluxes to enable a correction of this source (e.g., Berelson et al., 1994;Hammond et al., 1996). Another reason may be that the DIC and O 2 fluxes are not closely coupled in time due to episodic events in the POC influx (Therkildsen and Lomstein, 1993;Smith Jr. et al., 2018). It is striking how different authors have chosen different RQ values for local or global conversions from oxygen data to carbon budgets. For example , Smith Jr. (1987) assumed an RQ of 0.85 for the deep sea and used this number in later publications (Smith Jr. et al., 2018). Jahnke (1996) assumed a rather low RQ of 0.6 based on a Redfield stoichiometry determined by Takahashi et al. (1985). Dunne et al. (2007) cited the RQ of 0.6 of Jahnke (1996) and used this value in their calculation of the global marine carbon budget. Other RQ values used for the global seabed include 0.77 (Seiter et al. (2005) and 1.0 (Snelgrove et al., 2018).
Based on the above discussion, we conclude that the RQ on the continental shelf is expectedly higher than the 0.82 estimate for a mineralization that does not include organic N oxidation, while in the deep sea it is lower. As our best estimate, we will in the following use an RQ of 0.90 for the inner shelf, 0.85 for the outer shelf, 0.80 for the slopes, and 0.75 for the deep sea. As seen from Table 3, these values are all within the relatively large standard deviations of estimates from in situ flux chamber measurements.

A global budget of O 2 uptake and organic carbon mineralization
To quantify the global oxygen uptake of marine sediments, we calculated the areal depth-integrated TOU and DOU values for every 0.1 • × 0.1 • grid cell of the seafloor and summed these for specific water depths. Table 1 shows the seafloor area covered by each of the seven depth regions used. The total seafloor area covered equals 354 × 10 6 km 2 , corresponding to ~98% of the total ocean. Table 4 shows the mean areal TOU and DOU for each depth region, as predicted from the multicomponent regression analysis described in Chapter 5.1. These data are compared to the mean TOU and DOU for the same depth regions calculated from the observed in situ data in the database. A plot of the two sets of data, predicted versus observed, was shown in Fig. 19.
The mean, observed areal TOU drops with ocean depth from about 45 mmol O 2 m − 2 d − 1 on the inner shelf to 0.8 mmol O 2 m − 2 d − 1 on the abyssal seafloor (Table 4). The similar drop in mean, observed areal DOU is from 20 to 0.6 mmol O 2 m − 2 d − 1 . The standard deviations are calculated for each depth region for both predicted and observed TOU and DOU. Table 4 shows model predictions for both the entire global seabed and for the specific sites of observation. There is a good agreement between the predicted and the mean observed values of TOU and DOU of the measurement sites within each depth interval. The relative standard deviations are generally 20-50% of the mean for the predicted values and 50-90% of the mean for the observed values. The difference between predicted and observed values is generally much less than the standard deviation for each depth region. The mean predicted values for the global seabed are on average 75% of the mean observed values for TOU and 100% for DOU. This means that a simple arithmetic mean of the observed values for each depth interval would slightly overestimate the global seabed O 2 uptake by ca. 30% for the TOU but would correspond to the DOU calculated from the modeled values. The lower predicted values for the global ocean compared to the sites of observation is due to the bias of sampling sites towards regions with high net primary production (Fig. 18b).
It should be noted that the multicomponent linear regression models for TOU and DOU were only applied for water depths greater than 10 m. For the coastal waters of the inner shelf (0-10 m water depth), the data showed no significant correlation with water depth or sea-surface POC. We therefore used the arithmetic mean of all observations from this shallow depth range in Table 4, corresponding to 45 mmol O 2 m − 2 d − 1 for TOU and 20 mmol O 2 m − 2 d − 1 for DOU.

Table 5
Global budget of oxygen uptake rates in marine sediments calculated by adding up all the modeled TOU values in each 0.1 • -by -0.1 • grid cell within the defined depth regions. Based on the TOU data, the rate of organic carbon oxidation to CO 2 has been calculated, using the specified respiratory quotients (RQ) for each depth region. The 5%-95% confidence intervals (CI) are shown in parenthesis. Using the specified RQ values, the TOU was converted to benthic carbon mineralization rates for each of the defined depth regions. The integrated value for all regions expressing the global benthic carbon mineralization is 212 Tmol C yr − 1 ( Table 5). The regional and global integrals have a 5-95% confidence interval of ca. ±20%. This underscores that, despite all efforts during the past decades and despite an ever-growing data base of high-quality in situ assessments of benthic carbon mineralization, the global and regional budgets are still moderately constrained. This mainly reflects the extensive spatial heterogeneity and temporal variability in diagenetic activity at the seabed. To further constrain the budget, more measurements are needed, in particular in regions that are poorly represented in the data base, such as the open oceanic gyres where there are still very few available in situ measurements. Regions with extensive spatio-temporal dynamics are also poorly represented in the database, and slope and abyssal sediments are under-sampled.
The shelf region is clearly highly important for the global oceanic carbon budget. Although the inner shelf (0-50 m) covers only 3% of the global seabed, it accounts for an estimated 45% of the global benthic carbon mineralization (Table 5), not including sediments of vegetated ecosystems. This activity is to a large extent sustained by benthic primary production and supply of organic material exported from shallow vegetated ecosystems and from terrigenous run off (Chapter 6.1). The inner shelf region is characterized by a mosaic of diverse habitats and high seasonal variability, so despite a relatively good spatial data coverage, we still have only a moderate quantitative assessment of carbon turnover (and productivity) in shelf sediments.
To constrain the organic carbon export from the surface ocean, several studies have assessed the benthic carbon requirements from benthic O 2 consumption rates in the deep sea (>1000 m). Values range over a factor of three, from 33 Tmol C yr − 1 (Jahnke, 1996) to 97 Tmol C yr − 1 (Andersson et al., 2004) bracketing our own current estimate of 67 Tmol C yr − 1 ( Table 6). The bottom-up assessments of benthic mineralization rates of the entire global seabed range from 99 Tmol C yr − 1 to the 212 Tmol C yr − 1 of the current study. Different regression analyses have been applied and, considering the extreme variation of data on benthic O 2 consumption, the assessments are remarkably similar. In fact, the difference between the deep-sea estimates is to some extent explained by the range of RQ values used. This, however, does not mean that the global benthic mineralization of organic material is well constrained. Not all studies acknowledge the large uncertainty or confidence interval of the average (or median) values derived. Further, the studies largely use the same -albeit gradually increasingdatabase and apply similar relations to water depth and surface production for areal extrapolation.

Constraints on the oceanic carbon budget
As stated before, particulate organic carbon that sinks down through the water column from the photic surface waters or is transported laterally from the coastal zone and into the ocean exerts the primary control on seabed oxygen uptake. While a net primary production of about 4300 Tmol C yr − 1 has been estimated from satellite measurements of ocean color and modeling, the export production and the flux into the deep ocean are determined from the POC captured by multiple sediment traps. The fraction of produced POC that sinks out of the photic zone corresponds to a global export efficiency of 20-30%. As the sinking organic particles are gradually broken down and mineralized, the POC flux drops with depth. Based on multiple data from bottom-tethered time-series sediment traps at 134 globally distributed stations, Honjo et al. (2008)  This estimate is compared in Fig. 25 to published data on the POC flux in the ocean determined from multiple free-floating sediment traps. Data from different authors and different ocean regions show large scatter but a clear depth trend of steeply decreasing flux with depth in the upper ocean. This drop in organic particle flux indicates which fraction of the sinking POC is mineralized in the water column. The sediment trap data were fitted by different models that all assume an exponential decrease in POC flux with depth (Fig. 25). The model of Antia et al. (2001) also took the sea surface net primary production into account. It is striking how closely the estimate by Honjo et al. (2008) at 2000 m water depth matches these models, although they are based on different data sets from different sediment trap systems.
Since these sediment trap data describe the vertical POC flux that feeds the O 2 consumption of the global seabed, we integrated the sediment TOU beneath different ocean depths according to our multiple regression analysis. This was done by adding up the TOU calculated for all the 0.1 • × 0.1 • grid cells of the world ocean beneath each ocean depth. The TOU data were recalculated to organic carbon using the RQ values stated in Chapter 6.2. The results are plotted in Fig. 25 and show Table 6 Compiled values of benthic carbon mineralization of deep sea sediments (>1000 m) and of global ocean sediments, respectively. All values have been derived by a bottom-up approach using assessment of the benthic O 2 consumption rates but using different RQ values as indicated.  Honjo et al. (2008) based on bottom-tethered time-series traps is shown as a vertical black line. The total oxygen uptake of the global seabed below different water depths, modeled from our study, was converted to C org as described in the text and is shown as a red, dotted curve. that a POC flux higher than that determined by modeled sediment trap data is required to account for the calculated organic carbon oxidation in the seabed. Stolpovsky et al. (2018) also used a global database of benthic O 2 and NO 3 − fluxes to estimate POC rain rates at different depths in the ocean using a multi-G model. Although the calculated depth distribution was comparable to the Martin et al. (1987) curve, they found a similar mismatch between POC trap flux data and higher organic matter mineralization rates in the sediment as shown in Fig. 25. Several other studies have noted that trap-derived POC sedimentation rates were insufficient to account for the TOU-based organic carbon demand Smith Jr. and Kaufmann, 1999;Smith Jr. et al., 2016;Witte and Pfannkuche, 2000). The mean fluxes per m 2 shown in Fig. 25 can be extrapolated to the global ocean to calculate the total organic carbon flux across a depth of 2000 m. The ocean area deeper than 2000 m is 301 × 10 6 km 2 (Honjo et al., 2008; Table 1). Honjo et al. (2008) calculated a mean flux of 0.33 mmol C m − 2 d − 1 at an ocean depth of 2000 m, which corresponds to a global flux of 36 Tmol C org yr − 1 across that depth. Henson et al. (2012) combined models of export and transfer efficiencies with remote sensing data to estimate a global POC flux at 2000 m water depth of 55 Tmol C org yr − 1 . Our estimated mean organic carbon demand needed to cover the global seabed oxygen consumption below 2000 m is 61 Tmol C org yr − 1 (Table 5). An additional 1-8 Tmol C org yr − 1 is buried in the seabed beneath 2000 m ocean depth (Muller-Karger et al., 2004;Dunne et al., 2007), which requires a global carbon flux to the seabed beneath 2000 m of 62-69 Tmol C org yr − 1 . As only 36 Tmol C org yr − 1 was estimated to sink down across 2000 m depth, ca. 30 Tmol C org yr − 1 , corresponding to half of the seabed C org mineralization, remain unaccounted for by the sediment trap data.
Why is there such an apparent discrepancy between the measured organic carbon flux in the water column and the organic carbon oxidation and burial in the global seafloor? There are several possibilities: A) the sediment trap data underestimate the POC flux, B) the TOU data overestimate the mineralization rates, C) the estimated RQ values are too high, D) there is lateral transport down along the continental shelf, E) there is an additional flux of dissolved organic carbon (DOC) to the deep sea, and F) the discrepancy is within the uncertainty of the different estimates and is therefore statistically not significant.
A) Sediment traps may be biased relative to the total flux of POC due to factors related to hydrodynamics, lateral particle advection, zooplankton swimmers, and POC solubilization (Buesseler et al., 2007). Approaches to calibrate sediment traps include corrections using natural radionuclides (e.g., 234 Th/ 238 U, 210 Po/ 210 Pb) or mass balance of stable elements. Such calibrations have indicated a significant under-trapping, in particular in the mesopelagic zone (e.g., Buesseler et al., 2007;Burd et al., 2010;Wiedmann et al., 2020). Long time series with concurrent measurements of sedimentation rates and sediment oxygen consumption at specific deep-sea locations have also shown that sediment traps tend to underestimate the total POC flux (Chapter 4.3).
B) It is well documented that the retrieval of sediment cores from the deep sea and the measurement of O 2 uptake on board ships tends to overestimate TOU and DOU relative to measurements in situ (Glud, 2008; Chapter 2.1). We have therefore used only in situ data in our model analysis for all depths >10 m.
C) The RQ values used here decrease from 0.90 on the inner shelf to 0.75 in the deep sea and fall within the published range of 0.6-1.0 (Chapter 6.2). A recalculation to an extreme value of 0.6 for the deep sea would reduce the C org oxidation of deep-sea sediments by 20%, which is still far from the apparent discrepancy.
D) The lateral transport of POC is well documented for many regions, in particular for the continental slopes where it was suggested to account for an 80% deficit between the vertical POC flux and the sediment oxygen consumption (Lampitt et al., 1995). E) There is a huge reservoir of DOC in the ocean, which is partly refractory material derived from rivers and coastal ecosystems and from old DOC originally excreted from plankton communities. There is also labile DOC that is continuously released by plankton organisms. This labile DOC may globally account for 17% of the net primary productivity, but it is mostly turned over in the upper ocean (Hansell and Carlson, 1998). DOC is also released from sinking POC in the ocean, but this fraction is largely included in the POC flux determined by sediment traps. In contrast, there is a flux of DOC from the seabed and into the overlying water column, which would add to the POC flux required to balance the sediment TOU and would exacerbate the discrepancy between sediment trap and TOU data. Burdige et al. (1999) estimated that such a DOC flux may globally correspond to up to 10% of the DIC flux from the sediments.
F) The 5-95% confidence interval of our 61 Tmol C org yr − 1 estimate is 55-to 71 Tmol C org yr − 1 . We do not know the confidence interval of the sediment trap estimates, but it is possibly larger. It is therefore not clear whether the two estimates are statistically different.
A potential bias between POC flux data from sediment traps and sediment C org oxidation data from TOU measurements could possibly arise because the two data sets are obtained from different ocean regions with different water depths and different sea-surface productivity. Fig. 26 shows the geographic distribution of stations occupied by freefloating sediment traps and by bottom-tethered time-series traps. All stations included in Fig. 25 and in Honjo et al. (2008) are presented in Fig. 26. The world map shows that the sediment trap positions tend to be skewed towards the ocean margins.
In order to compare this geographic distribution of sediment trap data with the distribution of sediment TOU data (Fig. 17a) we have analyzed the frequency distribution of both data sets relative to the distribution of ocean depths and sea-surface net primary productivity. The sediment trap distributions are shown in Fig. 27 (white bars). Their geographic distribution matches the global distribution of ocean depths very well. They also match the global distribution of net primary productivity fairly well, with some bias towards highly productive regions. A comparison of Fig. 27 with Fig. 18 shows that the global TOU data are much more biased towards shallow water and high productivity regions than are the sediment trap data. We have not been able to distinguish the role of such a selectivity in depth and NPP in our analysis of sediment TOU and DOU data, but it remains a potential bias.

Role of other benthic habitats
In our global budget of seabed O 2 consumption we have generally excluded vegetated coastal ecosystems such as salt marshes, mangroves and seagrass beds. We have also excluded advective systems, such as cold seeps and hot vents, and we excluded the hadal trenches. Here we will briefly touch upon their potential role in the global marine carbon cycle and how they compare to the role of non-vegetated marine sediments.
The export of organic carbon from vegetated coastal ecosystems was discussed in Chapter 6.1. Canfield et al. (2005) estimated the average O 2 uptake rates of sediments within these ecosystems. Considering that the source of organic matter here is rich in structural carbohydrates, we will assume a respiratory quotient of 1.0. The total organic carbon oxidation calculated from Canfield et al. (2005) is then 34 Tmol C yr − 1 in salt marshes, 6 Tmol C yr − 1 in mangroves, and 16 Tmol C yr − 1 in seagrass beds. Intertidal flats add only 0.7 Tmol C yr − 1 to this. Crossland et al. (1991) estimated the export production from coral reefs to be 1.7 Tmol C yr − 1 . This all adds up to 58 Tmol C yr − 1 , which exceeds our estimate of 34 Tmol C yr − 1 for all other sediments on the inner shelf at 0-10 m water depth (Table 5). The fluxes cited in Chapter 6.1 of 35 Tmol C yr − 1 from rivers and 95-280 Tmol C yr − 1 from vegetated estuarine habitats to the coastal ocean exceeds the total organic carbon mineralization of 97 Tmol C yr − 1 , calculated for continental shelf sediments out to 50 m depth (Table 5). Although these estimates have wide error margins, they illustrate the extremely high productivity of ecosystems along the coastal margin (Najjar et al., 2018). They also illustrate the need to better constrain current estimates of sources and sinks of organic carbon along the ocean margins.  reviewed the tectonically driven cold seeps on the continental slopes that leak methane-rich fluids from subsurface reservoirs to the sea floor. They compiled data from seeps at 700 to 3000 m water depth. They estimated that a few tens of thousands of active cold seep systems exist on continental slopes worldwide and that seep-fueled benthic communities consume on average two orders of magnitude more O 2 per area than the surrounding seabed. They estimated a methane flux of up to 3 Tmol C yr − 1 , of which two thirds are emitted into the overlying seawater while one third, up to 1 Tmol C yr − 1 , is oxidized by the benthic community of free-living and symbiotic microorganisms, thus consuming 2 Tmol O 2 yr − 1 (CH 4 + 2O 2 → CO 2 + 2H 2 O). While this is only a few percent of the organic carbon oxidation on the lower slope and rise, it exceeds the estimated global methane production of 0.5 Tmol CH 4 yr − 1 in marine sediments at that depth range (Egger et al., 2018).  suggested that some of the methane may originate from dissolving gas hydrate or from the deep, hot subsurface at >1 km depth. Bach and Edwards (2003) quantified the low-temperature alterations of upper ocean crust in the flanks of mid-ocean ridges and estimated an annual oxidation of 1.7 Tmol Fe and 0.1 Tmol S. These reactions could, directly or indirectly, consume up to 0.6 Tmol O 2 yr − 1 and support a chemolithoautotrophic biomass production of 0.1 Tmol C yr − 1 . Elderfield and Schultz (1996) reviewed the hydrothermal fluid flux along the mid-ocean ridges whereby a complete thermochemical reduction of seawater sulfate takes place. From this fluid flux they estimated a hydrothermal venting of 0.1-1 Tmol H 2 S yr − 1 , which by oxidation to sulfate in cold deep-sea water would consume 0.2-2 Tmol O 2 yr − 1 .
Together, these oxygen sinks by seawater circulation through cold or hot ocean crust are of the order of 2-4 Tmol O 2 yr − 1 in the deep sea. For comparison, the estimated global seabed oxygen uptake below 1000 m depth is 90 Tmol O 2 yr − 1 (Table 5).
Hadal trenches cover about 0.8 × 10 6 km 2 (Stewart and Jamieson, 2018), corresponding to 0.4% of the seabed area beneath 4000 m depth (Table 1). Their O 2 uptake rate is 4-5-fold higher than that of the surrounding abyssal seabed (Chapter 4.5), which implies that their contribution to the global deep sea carbon budget is about 2%.

Conclusions and future perspectives
The present global budget of oxygen uptake and organic matter mineralization in the seabed started out with a database with nearly 4000 entries on benthic O 2 uptake rates. The budget was then calculated from a multiple regression analysis of data relative to key environmental parameters, of which ocean depth and sea surface primary productivity explained most of the variation. The aim was to understand the controls on in situ sediment O 2 uptake rates of the seafloor. We therefore needed to perform a critical selection of the data according to well-defined criteria, which discarded 70% of all data. The remaining data showed mean TOU rates that dropped with ocean depth from about 45 mmol O 2 m − 2 d − 1 on the inner shelf to 0.8 mmol O 2 m − 2 d − 1 on the abyssal seafloor. The relative standard deviations of the corresponding modeled rates were generally 20-50% of the mean. The calculated organic carbon oxidation in the global seabed was 212 Tmol C yr − 1 with a 5-95% confidence interval of 175-260 Tmol C yr − 1 . This confidence interval is probably of similar magnitude as by other published estimates and is important to keep in mind when comparing global budgets in the

literature.
A potential quality criterion for published TOU data, which would be hard to quantify and use, is how the measurement of sediment O 2 uptake was actually performed, i.e. how carefully the sediment was treated and the experiment was done. Each measurement of total seabed O 2 uptake is indeed a sensitive experiment on a biologically, chemically, and physically complex system. The correct in situ data therefore depend strongly on minimum disturbance and use of the best techniques. Unfortunately, the optimal techniques tend to be technically complex and costly, and there is a need for simpler and more robust instruments that are easier to use.
One such candidate is the non-invasive eddy covariance technique, which has the potential to become much more widely used, once standardized instruments, software packages, and data treatment procedures are developed (Berg et al., 2003(Berg et al., , 2022. The technique is noninvasive and does not disturb the natural flow in the benthic boundary layer or the natural bioirrigation of benthic fauna. It also enables measurements of O 2 flux across the surface of permeable sandy sediments, whereby it integrates a relatively large area of the seabed, but it provides only limited insight into the spatial variability within a habitat. Benthic crawlers are much more complex instruments but have the unique capability to perform repeated measurements and generate time-series data for a range of parameters within an area of the seafloor (Lemburg et al., 2018;Smith Jr. et al., 2021). An increased capacity to combine "snapshot" measurements and long-term observatories will strongly improve our understanding of seabed dynamics.
The combination of several techniques is important because each approach tends to reveal only one aspect of the complex and dynamic processes. As an example, from the South Pacific gyre, where O 2 penetrates tens of meters down into the sediment, the combination of in situ O 2 microprofiles from a free-falling lander and ex situ O 2 measurements in retrieved sediment cores revealed that 99% of the oxygen respiration took place in the top few cm, which are generally missed by the coring technique . As another example, sediment O 2 uptake and sediment trap deployments on the same stations have shown a several-fold discrepancy. It is therefore important to combine timeseries of both in order to reconcile bottom-up and top-down measures of POC flux to the seafloor (Smith Jr. et al., 2018).
The seafloor constitutes a global interface between the oxic ocean and the anoxic subsurface biosphere. Both aerobic and anaerobic processes in the seabed are strongly focused towards the ocean margins and the organic-rich sediments of the coastal zone. If vegetated coastal ecosystems are included in the global budgets, about 75% of both the oxygen respiration and the sulfate reduction takes place in continental shelf sediments at 0-200 m water depth, covering about 7% of the total ocean area. In these shelf sediments, benthic fauna plays an important role for the O 2 respiration, and the re-oxidation of sulfide produced during sulfate reduction consumes 25-50% of the total O 2 uptake. In our model, we correlated the rates of seafloor O 2 uptake to sea-surface primary productivity. However, in estuaries and on the inner shelf there is a large contribution of organic matter from continental run-off and from the vegetated coastal zone (Regnier et al., 2013;Duarte et al., 2005). Recent estimates suggest that these sources may exceed the plankton productivity on the inner shelf and that they also export POC down the continental slopes, yet their contributions are poorly constrained.
This balance between aerobic and anaerobic processes is very sensitive to the organic loading of the sediment and to the availability of O 2 in the bottom water. The large oxygen minimum zones in the ocean are currently expanding, driven by ocean warming, which reduces ventilation of deeper waters due to stronger stratification in the upper ocean and decreases O 2 solubility (Levin, 2018). This causes an impoverished bottom fauna, a shift from bioirrigation towards diffusive O 2 uptake, and a relative enhancement of anaerobic degradation processes. Coastal seas are today the most strongly affected by this change due to the combination of anthropogenic eutrophication and the general ocean warming (e.g., Meier et al., 2018;Ni et al., 2019). Polar regions have the fastest warming, which results in reduced sea ice cover and enhanced pelagic and benthic photosynthesis over the year (e.g., Arrigo et al., 2008;Gattuso et al., 2020). Thus, about 20% of the global shelf area is in the Arctic, with a considerable potential for benthic primary production as the open water period is expanding. It is critical to understand the extent and the time scale of these changes that are a result of global warming and anthropocene CO 2 emission. Predicted changes will affect the biological pump and the sequestering of organic carbon in the seabed, with significant effects on the benthic fauna and the mechanisms of O 2 consumption, both in the coastal seas and in the deep sea.

Data availability
Data are available in Supplementary Table S1, which includes all data from in situ and ex situ measurements. Data were filtered according to selection criteria outlined in Chapter 2.2 and specified in the table.
The globally gridded TOU model data will be uploaded to Figshare and thereby made publicly available upon acceptance of this review.

Declaration of Competing Interest
None.