Remote and local drivers of oxygen and nitrate variability in the shallow oxygen minimum zone off Mauritania in June 2014

. Upwelling systems play a key role in the global carbon and nitrogen cycles and are also of local relevance due to their high productivity and ﬁsh resources. To capture and understand the high spatial and temporal variability of physical and biogeochemical parameters found in these regions novel measurement technics have to be combined in an interdisciplinary manner. Here we use high-resolution glider-based physical-biogeochemical observations in combination with ship-based underwater vision proﬁler, sensor and bottle data to investigate the drivers of oxygen and nitrate variability across the shelf break off Mau- 5 ritania in June 2014. Distinct oxygen and nitrate variability shows up in our glider data. High oxygen and low nitrate anomalies were clearly related to water mass variability and probably linked to ocean transport. Low oxygen and high nitrate patches co-occurred with enhanced turbidity signals close to the seabed, which suggests locally high microbial respiration of resuspended organic matter near the sea ﬂoor. This interpretation is supported by high particle abundance observed by the underwater vision proﬁler and enhanced particle-based respiration rate estimates close to the seabed. Discrete in-situ measurements of dissolved 10 organic carbon and amino acids suggest the formation of dissolved organic carbon due to particle dissolution near the seabed fueling additional microbial respiration. Our high-resolution interdisciplinary observations highlight the complex interplay of remote and local physical-biogeochemical drivers of oxygen and nitrate variability off Mauritania, which cannot be captured by classical shipboard observations alone. properties and the oxygen and nitrate concentrations due to enhanced local remineralization processes near the seabed. Our glider-based observations reveal negative O 2 anomalies of about 10 - 20 µ mol kg − 1 . This variability is huge when compared to the mean oxygen concentration of about 50 µ mol kg − 1 within the shallow OMZ. These low O 2 anomalies cannot be explained by water mass variability but estimated local respiration rates are high enough to create these 15 anomalies within time scales of a few weeks. These low O 2 anomalies are very pronounced close to the seabed were enhanced local particle-associated O 2 respiration associated with the resuspension of organic matter seems to occur. Iversen et al. (2010) Author contributions. ST designed and carried out the experiment on board of Meteor in cooperation with MD. ST carried out the data analysis and wrote the main manuscript. JK motivated the water mass analyses methods applied in this study and contributed to the writing of the manuscript. GK carried out the CTD and glider sensor data processing and calibration. RK estimated the UVP5-based particle-related oxygen respiration rate and assisted with the interpretation of the UVP5 dataset. AE contributed to the interpretation of the organic matter datasets and wrote parts of the manuscript. All co-authors reviewed the manuscript and contributed to the scientiﬁc interpretation and 5 discussion. -


15
The Mauritanian upwelling region is located in the shadow zone of the eastern tropical North Atlantic (ETNA), an area characterized by sluggish mean circulation (Luyten et al., 1983). The local balance between oxygen supply (ventilation) and respiration creates a vertical oxygen structure with two minima indicating a shallow and a deep Oxygen Minimum Zone (OMZ) (Karstensen et al., 2008;Brandt et al., 2015). The deep OMZ below has a core depth of about 400 m and minimum oxygen concentrations around 40 µmol kg −1 (Karstensen et al., 2008;Brandt et al., 2015). Here the focus lies on the shallow OMZ 20 with a core depth of about 100 m and oxygen concentrations between 40 and 60 µmol kg −1 (Karstensen et al., 2008;Brandt et al., 2015;Klenz et al., 2018). Close to the coast near the seabed high oxygen variability and even oxygen concentrations well below 35 µmol kg −1 have been reported for the shallow OMZ (Yücel et al., 2015;Gier et al., 2017). Despite the potential im-and associated lability of DOM in the ETNA, and in EBUS in general, is however scarce. DOC is a heterogeneous pool of organic compounds, often categorized by its turnover time into labile (hours-days), semi-labile (weeks-months) and refractory (years-centuries) components. Semi-labile DOC is mainly represented by high molecular weight DOM, i.e. biopolymers, such as combined carbohydrates and hydrolysable amino acids (Benner, 2002). Semi-labile and refractory DOC thus reside long enough in seawater to be transported away from their source of production by ocean currents. For the ETNA upwelling regions 5 off North-Africa, offshore transport of DOC has been hypothesized to support microbial respiration in the more oligotrophic open Atlantic regions (Alvarez-Salgado, 2007). Understanding the chemical composition, offshore transport and microbial respiration of fresh DOC in the highly productive ETNA is thus important to understand the remote and local drivers of oxygen and nitrate variability.
As it is typical for an EBUS, the Mauritanian upwelling region has been found highly variable in its local physical and 10 biogeochemical characteristics, both in time and space (Schafstall et al., 2010;Peña-Izquierdo et al., 2012;Yücel et al., 2015).
In this study we investigate the oxygen and nutrient (here the sum of nitrate and nitrite) distribution along 18 • N and in the depth range of the upper OMZ at the end of the upwelling season in June 2014, during the transition phase from strong to weaker upwelling. We aim to better understand the drivers of oxygen and nitrate variability to evaluate and possibly improve regional model simulations and predictions of the future state off the Mauritanian oxygen minimum zone and the associated 15 ecosystem. Our study is mainly based on a data set that include physical (temperature, salinity) and biogeochemical (oxygen, nitrate, turbidity and chlorophyll fluoresecne) parameters measured by sensors attached to an autonomous underwater glider.
Significant amount of data were acquired which allows to analyze the high variability of physical and biogeochemical parameters in in the Mauritanian upwelling system in a very detailed view and with much better statistics then obtained from ship based observations. Our study also devote attention to the role of the pelagic processes close to the benthos in contributing 20 to the local oxygen and nitrate structure. Benthic lander observations with novel Lab-on-Chip technology reported high nirate and nitrite variability on timescales of less than 40 hours for our working area (Yücel et al., 2015). As underwater glider for the water column, the high resolution lander based time series observations highlight the advantage of new technology in revealing variability that can not be captured by traditional observing methods. However, ship-based profile and bottle data allowed us to carry out high precision reference data as well as collection a suite of parameters so far not accessible from autonomous 25 instrumentation.
Based on the observational data at hand we aim to decompose and identify drivers for remote (via transport) and local (biogeochemical cycling) variability in oxygen and nitrate. This includes estimating local microbial particles-associated oxygen respiration as well as the possible role of DOC as microbial substrate. The paper is structured as follows: In section 2 the observational datasets, including data processing and calibration procedures are described. In section 3 we introduce two  MODIS Aqua and Terra measurements (both day and night) at 8 days between 12 and 22 June 2014. Multiple SST measurements during a single day were averaged prior to the overall temporal average. The water depth (black contours), the glider tracks (black dots), offshore reference profiles (see section: 3.1, red circles), as well as conductivity, temperature and depth (CTD) stations (blue dots) are shown on the right. The wind direction and strength is shown with black arrows.
to remote (transport) and local respiration and remineralization processes. The local remineralization signals are compared with oxygen respiration rates derived from particle abundance measurements with the underwater vision profiler UVP5 in section 4.3. In subsection (4.4) we investigate the possible role of DOM, including DOC and dissolved hydrolysable amino acids (DHAA), for the local respiration and remineralization processes. We discuss and conclude our results in section 5 and 6 respectively. ( Fig. 1, e.g. (Yücel et al., 2015;Gier et al., 2017). In this study we analyze and interpret physical and biogeochemical gliderbased measurements of one Slocum glider (IFM13) deployment in combination with ship-based profiles and bottle data.

Ship-based sensor and nutrient measurements
The ship-based instrumentation included a pumped Seabird SBE 9-plus CTD system, equipped with double sensor packages for temperature, conductivity (salinity), and oxygen, and with single sensors for chlorophyll fluorescence and turbidity. In total 5 62 CTD stations were carried out along the 18 • N transect between June 8 and 27, 2014. The CTD was mounted on a 24 bottle General Oceanic rosette system used to collect discrete water samples for analysis of salinity and various biogeochemical parameters. From the observational data we calculated conservative temperature (Θ), absolute salinity (S A ), and potential density anomaly (σ Θ ) using the Thermodynamic Equation of Seawater-2010 Matlab Toolbox Version 3.04 (McDougall and Barker, 2011). The salinity of bottle samples were measured on board RV Meteor with a Guildline Autosal 8 model 8400B 10 salinometer and used to calibrate the CTD conductivity sensor. The calibrated salinity measurements have an accuracy of 0.002 g/kg. The CTD oxygen sensor (SBE 43) was calibrated with Winkler titration oxygen measurements of discrete water samples obtained from the rosette (Winkler, 1888;Grasshoff et al., 1983). Considering the scatter of the calibrated oxygen measurements an accuracy of 2.5 µmol kg −1 from the oxygen concentrations was determined. In total 835 bottle samples along 18 • N were analysed for this study for nitrate and nitrite using a QuAAtro autoanalyzer (Seal Analytical) on board. The 15 nutrient data was used for sensor intercalibration (section 2.5) and to characterize the upper shelf nutrient distribution where not glider measurements were available. See also (Yücel et al., 2015) and  where parts of the nutrients dataset have already been published.

Underwater Vision Profiler
An Underwater Vision Profile 5 (UVP5; Picheral et al. (2010)) was mounted on the General Oceanic rosette and operated 20 during downcasts to obtain full depth particle size spectra (0.06 mm to 26.8 mm equivalent spherical diameter, ESD). Microbial particle-associated respiration rates (PARR) were calculated according to Kalvelage et al. (2015) for particle sizes between 0.06 and 10.64 mm ESD as a function of particle size (ESD, mm), ambient temperature and oxygen concentration. The empiric relationship of particle size and oxygen respiration is based on a dataset close to the study area in the northern Mauritanian upwelling systems (Iversen et al., 2010). The PARR of single size classes were multiplied by particle abundances in the different 25 size classes and summed up for all size classes to obtain PARR.

Dissolved organic carbon
For DOC, samples (20 ml) were collected in duplicate, filtered through combusted (8 hrs, 500 • C) GF/F filters and filled into combusted (8 hrs, 500 • C) glass ampoules. Samples were acidified with 80 µL of 85% phosphoric acid, heat sealed immediately, and stored at 4 • C in the dark until analysis. DOC samples were analyzed by high-temperature catalytic oxidation (TOC-VCSH, 30 Shimadzu) modified from Sugimura and Suzuki (1988) and as described in more detail in Engel and Galgani (2016).
For DHAA, samples were first filtered through 0.45 µm Millipore Acrodisc syringe filters. A measure for the diagenetic state 5 of organic matter is the amino acid-based degradation index (DI) (Dauwe and Middelburg, 1998;Dauwe et al., 1999). For the calculation of DI from THAA in this study, mole percentages of amino acid were standardized using averages, and standard deviations and multiplied with factor coefficients as given in Dauwe et al. (1999) based on Principal Component Analysis. DI values often range between +2 and -2, with lower values indicating more degraded, higher values more fresh organic material.

Glider-based measurements
The SLOCUM G2 underwater electric glider IFM13 was operated from June 12 to the June 27 2014 and did 6 sections of roughly 68 km each. Data processing and calibration of salinity and oxygen measurements followed the procedures described and cited in Thomsen et al. (2016). The glider was equipped with a Satlantic Deep SUNA nutrient sensor measuring nitrate 15 and nitrite, thereafter named N O x following Yücel et al. (2015). The configuration and data processing of the SUNA was done as described in Karstensen et al. (2017).

Methods
Given the long integration time for remineralization and respiration processes in OMZ regions the background signal in nutrients enrichment and oxygen loss is large. Thus locally generated signals can be difficult to detect. Here we use two different 25 approaches to separate locally and remotely forced oxygen and nitrate variability. The first one is based on a determination of the local Apparent Oxygen Utilization (AOU) ("AOU method") and the second approach makes use of a water mass mixing analysis ("OMP method"). Both methods will be introduced in the following.

Local Apparent Oxygen Utilization (AOU) method
AOU is defined as the difference between oxygen saturation calculated from an empirical relationship based on temperature and salinity (Weiss, 1970), and the observed oxygen. The "AOU method" is based on a simple model for an upwelling region that is controlled by along-isopycnal (lateral) spreading of the water masses from the offshore regions (remote signal) towards the coast and respiration/remineralization within the area from the reference to the coast (local signal). By identifying a "suitable" 5 mean AOU profile in an offshore region the two components could easily be separated. In practice we fitted a 4th order polynomial function to a group of 18 offshore AOU reference profiles using density as the independent variable ( Fig. 1). The so obtained AOU(σ) function was applied to the observed density field to reconstruct the respective remote signal AOU field, which in turn was subtracted from the observed oxygen concentrations to obtain the local AOU anomaly. To minimize the impact of possible local ventilation we limit the analysis to the interior ocean below the mixed layer but also above the depth where NACW and SACW T/S characteristic are difficult to distinct. Moreover a smooth offshore AOU (σ) profile was needed that allow for fitting a simple linear fit. Thus our final solution space covers all waters between 26.1< σ <26.7 kg m −3 . We applied the same strategy to estimate a N O x from the σ field in order to estimate local remineralization on total nitrate and ultimately the local stoichiometry AOU / N O x . It is important to note that as we limit our analysis to waters well below the mixed layer, only water layers below 100 m can be investigated near the coast where the isopycnals are descending.

15
Nevertheless the along-isopycnal spreading of subsurface waters between the offshore region and the seabed is well captured by our local AOU approach. For simplicity and a more intuitive comparison with the OMP method, we multiply our results of the local AOU method with -1 and just refer to local oxygen anomalies.

Extended Optimum Multiparameter (OMP) method
A second approach, the extended Optimum Multiparameter (OMP) method, a water mass mixing analysis that also consid-20 ers the bulk remineralization / respiration of nutrients and oxygen is following Karstensen and Tomczak (1998); Hupe and Karstensen (2000), is applied to separate remote and local processes. In brief, the extended OMP analysis decomposes observed conservative (Θ, S A ) and non-conservative (nutrients, oxygen) parameters into water mass fractions of predefined source water types (see table 1). The decomposition is done by applying a non-negative least square fit in a multidimensional space (spanned by all parameters). For the conservative parameters only mixing fractions are resolved while for the non-conservative 25 parameters mixing fractions and a bulk remineralization / respiration, controlled by a predefined set of stoichiometric ratios (here: 8.625 = 138 O 2 / 16 N O x ), is considered. As we are interested in the local remineralization (and respiration) we used source water types based on the data set at hand (see table 1) and considering potentially youngest NACW and SACW guided by maximum oxygen and minimum N O x concentration within the nearby Θ(+/−0.04 2). In this way the OMP method will provide us the local respiration and remineralization signals in the observational data.

30
Note the main differences between the AOU method and the OMP method: The OMP method does not provide a remote signal "field" as the AOU method does via the fitted parameter curves. On contrary provides the OMP method a quantification of the water mass fractions but the local remineralization/respiration signal still is a "bulk estimate" based on the misfit relative  to a "conservative parameter" only solution. However, in regions with high contribution of SACW or NACW the local remineralization / respiration should be assumed unbiased by mixing signals. We apply both methods to test the robustness of the estimated oxygen and N O x anomaly patterns.  around 20 m over the shelf break (not shown). In the offshore part of the section a rapid temperature drop is seen e.g. at 115 km offshore from 23 • C in 25 m depth to 18 • C in 40 m depth. Approaching the coast the vertical temperature gradient is less pronounced and below 40 m depth the slope of the isotherms even reverses i.e. the depth of the isopycnals decreases towards the coast. Furthermore the vertical spreading of the isopycnals results in a less stratified water body along the continental slope ( Fig. 3a). In contrast to chlorophyll we also observe higher turbidity (0.2 NTU) near the seabed at the shelf break (Fig. 3f). It is important to note that this higher turbidity signal is collocated with reduced oxygen concentrations ( Fig. 3c). This collocation will be investigated in more detail in the single transect analysis (subsection 4.2.1).
In summary we find in June 2014 a hydrographic structure along 18 • N that shows typical upwelling pattern with cold and

Oxygen and nitrate variability within the shallow oxygen minimum zone
In this section at first single glider transects will be used to investigate the synoptic oxygen and nitrate variability along 18 • N in June 2014 (section 4.2.1). Secondly the low frequency temporal change during the whole deployment will be investigated in section 4.2.2. The oxygen and nitrate variability will be described based on the observed bulk oxygen and nitrate distribution (Fig. 4). Additionally we apply the two methods (AOU and OMP method) to the glider section data as introduced in section 5 3.1 and 3.2 respectively in detail. In short, with these methods we aim to investigate remote and local drivers of the observed variability (Fig. 5). The AOU method reveals local oxygen anomalies which either point to a local excess of oxygen (positive values) relative to the offshore reference profile or local oxygen loss (negative values) (Fig. 5b, g, l). The former are interpreted as a result of a remote ventilation while the latter are signals for local oxygen consumption. The OMP method also resolves explicitly water mass composition and bulk oxygen and nitrate variability relative to the predefined source water masses. As we 10 defined the source water masses to carry the maximum oxygen concentrations the OPM method only reveals negative oxygen anomalies pointing to local oxygen loss (Fig. 5d, i, n). For more details on the methods the reader is referred to section 3.

Synoptic oxygen and nitrate variability along single glider transects
The glider transects reveal high spatial and temporal variability in all observed parameters (Fig. 4, 5). Here we focus on three representative transects to investigate the observed oxygen and N O x variability within the depth range of the shallow A co-location of high oxygen anomalies and low salinity patches is clearly visible e.g. along the first transect (June 13 -15) at 100 km offshore (Fig. 4b, c). The OMP method shows that the low salinity is accompanied by by high (> 75%) SACW 25 fractions (Fig. 4b, 5a) and point to the importance of SACW in supplying oxygen to the region. The local AOU method reveals that these high oxygen patches are associated with oxygen anomalies of the order of 10 µmol kg −1 (Fig. 5b, g, l). We conclude that these positive oxygen patches are caused by physical transport of SACW into the region.
The second type of oxygen anomalies are low oxygen patches which exhibit a clear co-existence with enhanced turbidity signals. They are predominantly found near the seabed e.g. along the first two transects (June 13-15, June 20 -22, Fig. 4c, h)

30
where oxygen concentrations down to 29 µmol kg −1 are observed (Fig. 4c). In transect 2 (June 20 -22) also an elongated low oxygen (45 µmol kg −1 ) patch, aligned with enhanced turbidity, is found between 100 and 250 m depth and reaching more than 100 km offshore (Fig. 4h, j). The high turbidity signal may indicate offshore transport of resuspended organic matter which in turn may indicate locally enhanced microbial respiration near the sea floor. This hypothesis is further evaluated using a combination of UVP5-based particle abundance profiles (section 4.3) and CTD bottle-based DOM measurements (section 4.4). The oxygen concentrations associated with this anomaly seem to increase with time when comparing the section from mid June with the section from end of June (Fig. 4c, h, m) and goes along with a reduction of the turbidity signal. This low frequency changes become particularly visible in OMP method results (Fig. 5d, i, n) and will be discussed in more detail in formation are identified: One is mainly found at the shelf break close to the seabed and is associated with enhanced turbidity suggesting the signal may originate from the benthic boundary layer and was resuspended into the water column. The second type of low oxygen anomalies are found further offshore on similar isopycnal layers than the ones close to the shore. However, 20 no enhanced turbidity signals are seen. This suggest that these offshore anomalies may originated from the shelf break but much longer ago and all turbidity signal is removed due to gravitational settling of particles. Both applied methods (AOU and OMP) gave similar oxygen respiration/nutrient remineralization patterns pointing to the robustness of these results. Our data does not allow to identify when the oxygen anomalies have been formed. However, we will use local particle-associated oxygen respiration rate estimates in section 4.3 to further evaluate the timescales that could explain the observed signals.

Temporal changes of the oxygen and nitrate during June 2014
Beside the high spatial variability of oxygen and nitrate along the single transects we also observe temporal variability during the 2 weeks of the glider observations. This temporal variability of oxygen and nitrate is particular pronounced near the seabed on the shelf break and will be investigated in more detail in the following.
The temporal change in the oxygen concentrations during our glider deployment is clearly visible also in the single transect drivers we extract the oxygen and nitrate anomaly estimates from the AOU and the OMP method including the water mass fractions between the isopycnals of 26.27 and 26.29 close to the seabed (Fig. 6).
It can be see that the observed changes cannot be explained by variability in water mass composition. The SACW fraction stays relatively constant near the seabed during the time of the observations (Fig. 6c) and even decreased slightly from 85% to 80%. The continues advection of SACW from the south via the boundary current system is an oxygen source for the density This analysis suggests that a strong modification of typical water mass signals near the shelf such as typically high oxygen 10 concentrations associated with SACW. This points to the importance of enhanced local oxygen respiration rates which act on faster time scales than the water mass renewal rate. In the following section we will investigate these high turbidity waters close to the seabed in more detail using UVP5, CTD sensor and bottle data to investigate the distribution and composition of the organic matter. In particular we aim to learn more about the importance and associated time scales of local organic matter remineralization processes and oxygen respiration rates.

Particle abundance and particle-associated oxygen respiration rate estimates
In this section we investigate the particle abundance along the transect using UVP5 observations to estimate particle associated respiration rates. These respiration rate estimates are then used to investigate the possible role of local particle-associated oxygen consumption for the observed low oxygen patches described in section 4.2.1. Oxygen is continuously consumed within the interior ocean by microbial respiration processes but the magnitude of oxygen respiration varies regionally.
We distinguish here between two different particle size classes, where the diameter of small particles ranges from 0.14 to 0.53 mm (Fig. 7a) and for large particles from 0.53 to 16.88 mm (Fig. 7b). The oxygen respiration rate estimates can be used to learn about the approximate formation timescale of the observed low oxygen anomalies with high turbidity signals close to the seabed described in section 4.2.1. For this we combine the observed magnitudes O(10 -20 µmol kg −1 ) of these anomalies with the estimated respiration rates and assume an alongshore 15 advection of an enclosed water mass. This gives a formation time scale of about 12.5 to 25 days for oxygen anomalies of 10 to 20 µmol kg −1 when using the rates near the seabed (0.8 µmol kg −1 day −1 ). Contrary the interior offshore rates of 0.06 to 0.2 µmol kg −1 day −1 ) result in formation timescales of 100 to 333 days. Although we investigate the transect due to the availability of the data just in a two dimensional way we are aware about the fact that the observed water masses are advected through the transect. In the following we assume a typical advection speed along the coast of about 0.1 m/s within the undercurrent 20 close to the seabed and a constant alongshore respiration rates. This implies that the water masses move about 100 km during 12.5 days, which was the lower end of the formation time scale of these observed oxygen anomalies. Given the remote supply path of the SACW this is a relatively short distance and can still be referred to as local. Of cause these are very crude first order estimates. However these results suggest that particle associated oxygen respiration close to the seabed is indeed able to change the oxygen concentrations within the study area during relatively short time periods. Thus beside advection of water 25 mass properties via physical transport also fast local respiration might impact on the observed oxygen variability. In section 5 we further discuss the possible role of resuspension processes occurring within the bottom boundary layer.
In section 4.2.2 we describe an increase of oxygen concentrations close to the seabed despite relatively constant water mass properties. The transect has been sampled at relatively high temporal and spatial resolution including the repetition of various stations with the shipboard CTD/UVP5 measurements. Nevertheless there are not enough measurements available to make 30 reliable estimates of temporal variability of respiration. However, the temporal change of the turbidity signal, i.e. a decrease of the turbidity signal near the seabed with time (Fig. 4e, j, o), suggests a reduction of particle-abundance and possibly the magnitude of particle-associated respiration during the study period. Our a reduction in respiration might be relevant for the low frequency temporal change in oxygen concentrations, when assuming a relative steady water mass renewal rate. This assumption is supported by relatively constant water mass composition observed during the oxygen increase as described in section 4.2.2.
In summary high particle related respiration rates are estimated near the surface above the OMZ and at depth close to the seabed. Near the bottom we also find pronounced low oxygen patches in combination with high SACW fractions. As SACW brings oxygen into the region high oxygen concentrations might be expected. However the fact that we observe low oxygen 5 patches points to a strong modification of typical water mass characteristics especially close to the coast driven by high local respiration.

Dissolved organic matter and amino acid distribution
As described in the previous section our UVP5 observations show enhanced particle abundance within the OMZ on the shelf break near the seabed. In this section we use CTD bottle-based measurements of DOC to investigate the potential role of DOC 10 remineralization for the formation of the low oxygen anomalies close to the seabed as described in section 4.2.1. In particular we want to address the question whether we also find higher DOC within the OMZ associated with the high turbidity and low oxygen signals. We furthermore quantify the DHAA concentrations and estimate a DHAA based degradation index (DI).
Highest DOC concentration of up to 100 µmol l −1 are found within the upper 50 m of the water column above the OMZ (Fig. 7d, 8b). DOC decreases rapidly to average values of around 60 µmol l −1 below 50 m within the oxycline (Fig. 8b). In the 15 upper 50 m of the waver column no significant difference in the average DOC concentrations was determined between offshore (red, > 70 km to the coast) and onshore (red, < 70 km to the coast) stations (Fig. 8b). Below 50 m depth DOC concentrations of up to 70 µmol l −1 are found within the OMZ close to the coast, which are absent further offshore (Fig. 7d). Due to the patchiness of the DOC concentrations this becomes more clear when averaging all data points close to the coast and offshore data (Fig. 8b). Indeed this analysis reveals an increase of DOC of about 5 µmol l −1 within the OMZ close to the coast.

20
As seen for DOC also the highest DHAA concentrations of up to 1000 nmol l −1 are found in the surface layers ( Fig. 7,8). in various studies (Peña-Izquierdo et al., 2012Klenz et al., 2018). Our analysis support these findings as we find highest 5 oxygen concentrations in combination with high SACW fractions. An early high-resolution modeling study by Glessmer et al. (2009) highlighted the role of the equatorial current system for the supply of water into the upwelling region off Mauritania.
The good representation of this remote supply pathway in models might be crucial to capture remotely forced oxygen and nitrate variability off Mauritania. On contrary, (Kounta et al., 2018) did not find a direct pathway between the North Equatorial Counter Current and the eastern boundary region of the ETNA. 10 Despite this important role of remote oxygen supply via SACW into the region we see also a strong decoupling between the physical water mass properties and the oxygen and nitrate concentrations due to enhanced local remineralization processes near the seabed. Our glider-based observations reveal negative O 2 anomalies of about 10 -20 µmol kg −1 . This variability is huge when compared to the mean oxygen concentration of about 50 µmol kg −1 within the shallow OMZ. These low O 2 anomalies cannot be explained by water mass variability but estimated local respiration rates are high enough to create these 15 anomalies within time scales of a few weeks. These low O 2 anomalies are very pronounced close to the seabed were enhanced local particle-associated O 2 respiration associated with the resuspension of organic matter seems to occur. Iversen et al. (2010)  observed particle size spectra further north of our study region near Cape Blance. The authors also found a dominance of small particles and an increase of particle numbers with depth as also described by . This increase might be caused by lateral transport as suggested by modeling studies (Karakaş et al., 2006;Lovecchio et al., 2017). Here we focus just on the upper 250 m of the water column but our observations also suggest the resuspension of organic matter and the consecutive offshore transport of this material.

5
Additionally to the enhanced particle abundance we also find enhanced DOC concentrations close to the shelf (Fig. 7,8).
So far the DOC distribution has not been described for this region. However, further north in the Canary upwelling system DOC was identified to play a major role in mesopelagic respiration (Santana-Falcón et al., 2017). Our observations suggest that DOC plays also an important role for the remineralization process within the study area. Our analysis of the quality of the DOC revealed that the DOC at depths is mainly formed due to resuspension of organic matter and the dissolution of particles 10 during microbial decomposition near the seabed. The focus in this study lies on the organic matter dynamics close to the seabed.
Nevertheless it is important to note that we observe high concentrations of DOC at the outer edge of the transect pointing the important of offshore DOC transport as already suggested by (Alvarez-Salgado, 2007) for the northern part of the upwelling area.
Beside pelagic oxygen respiration also benthic oxygen uptake is driven by the export of organic matter from the surface.  The Underwater Vision Profiler (UVP) data can be accessed at https://doi.pangaea.de/10.1594/PANGAEA.885759. According to the SFB 754 data policy (https://www.sfb754.de/de/data), all remaining data (organic matter dataset) associated with this publication will be published at a world data center (www.pangaea.de, search projects:sfb754) when the paper is accepted and published. Upon publication all code necessary for the data analysis and preparation of the figures of this manuscript will be freely available at: https://github.com/soerenthomsen.