Diapycnal dissolved organic matter supply into the upper Peruvian oxycline

The Eastern Tropical South Pacific (ETSP) hosts the Peruvian upwelling system, which represents one of the most productive areas in the world ocean. High primary production followed by rapid heterotrophic utilization of organic matter supports the formation of one of the most intense oxygen minimum zones (OMZ) in the world ocean, where dissolved oxygen (O2) concentrations reach less than 1 μmol kg. The high productivity leads to an accumulation of dissolved organic 10 matter (DOM) in the surface layers that may serve as a substrate for heterotrophic respiration. However, the importance of DOM utilization for O2 respiration in the Peruvian upwelling system in general and for shaping the upper oxycline in particular remains unclear so far. This study reports the first estimates of diapycnal fluxes and supply of O2, dissolved organic carbon (DOC), dissolved organic nitrogen, dissolved hydrolysable amino acids (DHAA) and dissolved combined carbohydrates (DCCHO) for the ETSP off Peru. Diapycnal flux and supply estimates were obtained by combining measured 15 vertical diffusivities and solute concentration gradients. They were analysed together with the molecular composition of DCCHO and DHAA to infer the transport of labile DOM into the upper OMZ and the potential role of DOM utilization for the attenuation of the diapycnal O2 flux that ventilates the OMZ. The observed diapycnal O2 flux (50 mmol O2 m day at max) was limited to the upper 80 m of the water column, the O2 supply of ~1 μmol kg day, was comparable to previously published O2 consumption rates for the North and South Pacific OMZs. The diapycnal DOM flux (31mmol C m day at 20 max) was limited to ~30 m water depth, suggesting that the labile DOM is extensively consumed within the upper part of the shallow oxycline off Peru. The analyses of DCCHO and DHAA composition support this finding, suggesting that DOM undergoes comprehensive remineralization within the upper part of the oxycline, as the DOM within the core of the OMZ was found to be largely altered. Estimated by a simple equation for carbon combustion, aerobic respiration of DCCHO and DHAA, supplied by diapycnal mixing (0.46 μmol kg day at max), could account for up to 38% of the diapycnal O2 supply 25 in the upper oxycline, which suggests that DOM utilization plays a significant role for shaping the upper oxycline in the ETSP.

organisms are conducted by oxidation with O2 (e.g. Bender and Heggie 1984). The eastern tropical South Pacific (ETSP) embodies one of the largest oxygen minimum zones (OMZ) in the world ocean (Karstensen et al., 2008;Paulmier and Ruiz-Pino, 2009). The core of the Peruvian OMZ is considered fully anoxic (e.g. Ulloa et al. 2012), as O2 concentrations below the detection limit (DL) of ~0.01 µmol kg -1 were observed between 20 and 400 m depth by high precision STOX sensor measurements (Revsbech et al., 2009;Kalvelage et al., 2013;Thomsen et al., 2016a). Those low O2 concentrations are due to 5 a sluggish ventilation by ocean currents, carrying low-O2 waters to the ETSP and microbial respiration attributed to utilization of organic matter (OM) originating from the upper water column (e.g. Czeschel et al., 2011;Brandt et al., 2015;Kalvelage et al., 2015).
Elevated primary production in the Peruvian upwelling region above the OMZ (Pennington et al., 2006) leads to an accumulation of both particulate (POM) (Franz et al., 2012a) and dissolved (DOM) organic matter (Romankevich and other solutes, may be reduced or even predominated by upwelling fluxes due to Ekman divergence in the coastal upwelling region (e.g. Steinfeldt et al., 2015). Machadevan (2014) suggested that transport of OM (via eddy fluxes) into the OMZ should be accompanied by O2 in amount that is sufficient for full remineralization of the subducted OM. Therefore, this physical transport of OM and O2 should stimulate heterotrophic aerobic respiration in the OMZ, which was suggested to be the main pathway of OM remineralization 5 in the upper OMZs by Kalvelage et al. (2015). However, so far, no direct O2 and DOM supply estimates exist for the Peruvian OMZ.
Here, we investigated the possible importance of diapycnal DOM supply by turbulent mixing processes for the O2 utilization off Peru using combined physical and biogeochemical observational data that were collected during the R/V METEOR "M93" (M93) research cruise to the ETSP off Peru in February-March 2013. Specifically, we directly estimated the diapycnal O2 and 10 DOM supply into the upper oxycline off Peru. Additionally, we analyzed diapycnal fluxes and the composition of dissolved combined carbohydrates (DCCHO) and dissolved hydrolysable amino acids (DHAA) to learn, whether DOM and its labile and semi-labile constituents may be supplied to the upper OMZ and the potential contribution of DOM based respiration to O2 flux attenuation.

Study area
The observational data were acquired during the research cruise "M93" which took place from 7th of February to 9th of March 2013 between 12°S and 14°S and 76°W and 79°W off Peru (Fig. 1). During the measurement program, the study area was affected by moderate southeasterly winds (1-9 m/s) (Thomsen et al., 2016a). The water column was highly stratified during the cruise (Fig. 2a,b). High concentrations of inorganic nutrients (~30 µmol L -1 (NO3 -), ~3 µmo L -1 (PO4 3-)) just below the 20 surface (Thomsen et al., 2016a) collocated with highest chlorophyll a (chl a) concentrations near the surface Fig. 2c) (Loginova et al., 2016). The oxycline was located at upper 5-80 m depth, here oxygen concentrations dropped from >200 µmol kg -1 to <1µmol kg -1 (Fig. 2d) (Thomsen et al., 2016a). In summary, our observations were carried out during a period which corresponds to typical summer conditions off Peru.

25
Seawater was sampled with a rosette (GO; General Oceanics, USA) equipped with a conductivity, temperature and depth profiler (CTD; Sea-Bird (SBE) 9-plus, Sea-Bird Electronics Inc., USA), an O2 optode (SBE43, Sea-Bird Electronics Inc., USA), a WETStar chl a fluorometer (WET Labs, USA) and 24 x 10 L Niskin bottles. Additional water samples were taken with a PUMP-CTD-System (an integrated measurement device, which was developed in collaboration between Leibniz Institute for Baltic Research (IOW) and the Max Planck Institute for Marine Microbiology (MPI) Bremen: PUMP-CTD; Strady offshore) and from 2 to 200 m at stations offshore (~100 km offshore). DOC/DON analyses were performed for 50 GO rosette stations, and for 8 PUMP-CTD stations. DHAA and DCCHO analyses were performed only for samples from the GO rosette. CTD, O2 and chl a recordings were taken at 172 profiles (Fig. 1a).
Apparent oxygen utilization (AOU) was then calculated as a difference of measured O2 concentrations and its equilibrium saturation using Gibbs-Sea Water Oceanographic Toolbox (McDougall and Barker, 2011) for MatLab (MathWorks, USA) for 10 analyses of potential relationship between DOM reworking and the utilization of O2. DOC/DON duplicate samples (20 mL) were collected into combusted glass ampoules (8 h, 450° C) after filtration with combusted GF/F filters (5 h, 450°C). Samples were acidified (80mL of 85% H3PO4), sealed with flame and stored at 4°C in the dark until analysis. DOC samples were analysed by the high-temperature catalytic oxidation method (TOC -VCSH, Shimadzu) modified from Sugimura and Suzuki (1988). The detection limit (DL) was 1 µmol L -1 . Total dissolved nitrogen (TDN) was determined simultaneously to DOC with DL of 2 µmol L -1 using the TNM-1 detector of a Shimadzu analyser [Dickson et al., 2007]. DON concentrations were calculated by subtracting inorganic nitrogen concentrations from 20 concentrations of TDN. The description of the instrument calibration and measurements may be found in Loginova et al. (2015).

Diapycnal flux calculations
To estimate the diapycnal fluxes of various solutes, CTD sensor (O2) and bottle data (DOC, DON, DCCHO and DHAA) were combined with near-simultaneous measurements of turbulence in the water column. The turbulence measurements were performed with a microstructure profiling system (MSS) from the rear of the vessel. The loosely-tethered profiler (MSS90-D,

15
S/N 32, Sea & Sun Technology) was optimized to sink at a rate of 0.55 m s -1 and was equipped with three shear sensors and a fast-response temperature recorder, as well as an acceleration sensor, two tilt sensors and CTD, sampling with lower response time. At each CTD station, 3-6 microstructure profiles were collected. Standard processing procedures were used to determine the rate of kinetic energy dissipation of turbulence in the water column ( , m 2 s -3 ), as given in Schafstall et al. (2010).
Diapycnal diffusivities ( , m 2 s -1 ) were determined at 14 m depth intervals, following Osborn (1980): where is stratification (in s -1 ) and Γ is the mixing efficiency, for a which value of 0.2 was used. The diapycnal diffusivity of the solutes (O2, DOC, DON, DCCHO, and DHAA) --was assumed to be equivalent to the diapycnal diffusivity of the mass (e.g. Schafstall et al., 2010;Fischer et al., 2013).
where ∇ is the vertical gradient of the molar concentration of the solutes (mmol m -4 ).
The mean diapycnal supply (−∇Φ ̅̅̅̅̅̅ , µmol kg -1 day -1 ) of a solute was determined at 28 m depth intervals as an attenuation of the diapycnal solute flux profile over depth, according to the Eq. 3: where -is the in-situ density of the seawater (kg m -3 ), z -is depth (m) and Φ ̅̅̅̅ (mmol m -2 day -1 ) -is the estimated mean diapycnal flux profile of a solute. The mean diapycnal solute supply was interpreted to balance the amount of a solute that is lost per unit of time over a specific depth interval of the water column due to the microbial utilization of the solute. This interpretation assumes that sources other than turbulent mixing or sinks other than microbial consumption are negligible.

5
For DCCHO and DHAA the diapycnal flux estimates were based on 14 combined CTD/MSS stations, while for DOC and DON fluxes 22 stations were available (Fig. 1b). The diapycnal O2 flux was determined from 50 combined stations. All combined data sets include stations from the continental slope, as well as stations in deeper waters, where bottom depth was larger than 4000m.
For each combined CTD/MSS station a mean was estimated based on a 2 profile (CTD) and mean dissipation profile 10 (turbulence probe) averaged over all MSS profiles conducted at the CTD station. In combination with the vertical solute gradient, a mean flux profile for each station was estimated. Only measurements below the mixed layer, which was defined by a threshold criterion of a 0.2°C temperature decrease below the maximum and a minimum depth of 10 m, were used.
Measurements from different sensors and instruments were averaged in density space to reduce the impact of internal waves.
The mean diapycnal flux (Ф ̅ ) was determined by arithmetically averaging all fluxes from individual stations in 14m depth 15 intervals. The diapycnal solute supply was then determined from the divergence of the mean diapycnal flux (∇Ф ̅̅̅̅ ).
The 95% confidence interval of the diapycnal flux was calculated following the procedure described by Schafstall et al. (2010).
From this error estimate the uncertainty of the supply was derived by error propagation.
A simple equation of carbon combustion: parameter was considered significant, when module of the coordinate of the parameter exceeded 0.5 on the "variables factor map". The PCA was performed using "FactorMineR" package (Husson et al., 2010) for "R" (R Core Team, 2013).

Distribution of O2 and DOM
In this section the horizontal and vertical distribution of O2 and the different DOM components including DOC, DON and 5 their labile and semi-labile constituents, DCCHO and DHAA are described. The vertical gradients of the different solutes are crucial for estimating the associated diapycnal fluxes, as described in section 3.2. Near surface O2 concentrations were observed ranging between 100 µmol kg -1 at the coast and 240 µmol kg -1 further offshore (Fig. 2d). These values dropped to less than 1 µmol kg -1 at ~20m depth at the coast and ~80 m depth offshore (Fig. 2d).
Patches of isolated DOC maxima (up to 120 µmol L -1 ) were measured at a depth range from 20 to 120 m ( Fig. 3a). DOC concentrations of >100 µmol L -1 had been reported previously for the water column off Peru (Romankevich and Ljutsarev, 1990;Franz et al., 2012a). However, since concentrations >100 µmol L -1 were observed only sporadically, we cannot exclude a possible contamination of these samples. The main decrease of DOC occurred between 5 and 30 m. Thus, the main vertical DOC gradient was found at shallow depth, compared to the oxycline. This becomes even more apparent, when comparing the  (Table   1), while Glc-H was detected only sporadically. Overall, S-N and S-H comprised 0.04±0.03 µmol L -1 and 0.02±0.02 µmol L -1 , contributing 6±3 % and 3±2 % to DCCHO, respectively. Thus, the major part of DCCHO was represented by nS (Table 1).
DHAA concentrations varied from 0.075 µmol L -1 to 1.39 µmol L -1 (Fig. 3d). Like for DCCHO, the highest DHAA concentrations were found above the oxycline, where C contained in DHAA represented 2±1 % DOC (max. 4 %) and nitrogen 25 (N) contained in DHAA represented 15±14 % DON. Lowest DHAA concentrations were mainly found below 80 m depth and equivalent to ~1 %DOC and 6-8 %DON ( Table 1). The major part of DHAA was represented by α-amino acids. The concentrations of GABA, which is commonly used as a signature of microbial activity (Davis et al., 2009), was very low in all samples and represented generally <1% of DHAA. In summary, the concentrations of all the DOM compounds were highest above the oxycline, and the mean concentration gradients of the DOM compounds were restricted to a shallower depth,

Diapycnal fluxes and supply
As outlined in the previous section vertical gradients of O2, DOC, DON and their constituents were observed at 30 to 80 m depth in the study area. In this section we combine these vertical gradients with turbulence measurements to estimate the associated diapycnal fluxes and supply i.e. the diapycnal flux divergences.

15
The diapycnal C supply via DCCHO and DHAA at its maximum comprised ~38% of ∇Ф ̅̅̅̅ 2 , when estimated by Eq. (4). In summary, our diapycnal flux and supply calculation revealed that the diapycnal O2 supply reaches deeper into the oxycline than the diapycnal DOM supply. This is especially true for DCCHO and DHAA, representing the labile and semi-labile parts of DOM.

20
To understand, whether low-O2 conditions of the OMZ may cause changes in DOM composition, we complement our quantitative estimates of the DOM and O2 supply with the analyses of DOM quality. For this, the composition of neutral DCCHO and DHAA via PCA was compared to environmental factors, i.e. temperature, AOU and salinity, and to OM composition from the well oxygenated water column as described in Kaiser and Benner (2009) Table 1). Gly, Thr and Glc mol% were increasing along with increase in AOU ( The differences on the second dimension of PCA (Dim.2) were driven likely by regional differences in the DOM composition, i.e. by mol% of Ala, Arb, and Fuc, and distributions of mol% Asp, Phe, Val and Leu over depth (Fig. 5, Table 1, Kaiser and Benner, 2009).

Discussion
The observed distributions of O2 and of DOC and DON components are the result of sinks and sources in the water column, 5 mainly due to microbial processes and isopycnal and diapycnal supply (i.e. flux divergences) controlled by physical processes.
A quantification of each of those individual processes is essential for understanding of important mechanisms, controlling O2 and OM cycling off Peru and, therefore, formation and maintenance of the Peruvian OMZ.
Previous studies have shown that turbulent mixing processes in the eastern boundary upwelling systems (EBUS) are strongly enhanced and that the resulting diapycnal supply is often a leading term in the flux divergence balances of O2, nutrients and 10 other solutes in the upper ocean (e.g. Schafstall et al., 2010;Kock et al., 2012;Brandt et al., 2015;Steinfeldt et al., 2015).
The diapycnal O2 and DOM fluxes and supply determined in this study represent average values for the continental margin ranging from the shelf to about 100 km offshore. This spatial averaging is likely responsible for a lower near-surface diapycnal

15
In their study, the diapycnal O2 flux was able to sustain benthic respiration on the continental shelf down to a bottom depth of 100 m. Herewith, the diapycnal O2 supply, found in our study, was of similar magnitude as the rates of O2 consumption (~1

20
Other terms of the O2 transport budget, such as isopycnal supply by meso- (Thomsen et al., 2016a)  Herewith, our data suggest that the diapycnal DOC flux in the upper 20m of the water column off Peru is in the same order of magnitude as the diapycnal O2 flux ( Table 2). The annual diapycnal DOC flux (2.7 mol C m -2 yr -1 ) into the upper OMZ, estimated from our results by averaging Ф ̅ above the mean depth of the oxycline (from below the mixed layer to 80 m depth) and integrating over a year, is in the same order of magnitude as previously reported data for the North Pacific

25
Subtropical Gyre, where DOC export was estimated by a mass balance approach (1.6-2.7 molC m -2 yr -1 ; Emerson et al., 1997) and by fitting an exponential decay function over depth (0.5±0.1 molC m -2 yr -1 ; Kaiser and Benner 2012). Furthermore, in the upper water column (from below the mixed layer to 38 m water depth), the diapycnal DOC supply was higher, than the diapycnal O2 supply, suggesting that DOC respiration could exhaust all O2. However, the vanishing of DOC flux above the upper oxycline suggests that the bioavailable fraction of DOM is respired well before entering the upper OMZ. This is even 30 more apparent, when considering diapycnal DHAA and DCCHO fluxes, which decayed more rapidly compared to the diapycnal DOC flux, suggesting preferential uptake of DHAA and DCCHO in the water column. The diapycnal supply of DHAA and DCCHO could not fully explain the diapycnal supply of DOC, as those were responsible for only ~26% of ∇Ф ̅̅̅̅ when summed up together. This may hint to a presence of an additional bioavailable DOM component that was respired in the water column, and/or to other DOM removal mechanisms in the near-surface waters. For instance, DOM may form marine microgels and hence POM (Chin et al., 1998;Engel et al., 2004, Verdugo et al., 2004 or be trapped in the pore space of already existing particles (e.g. Benner, 2002 (2007), that implies that carbon yields of DHAA above 1.6 %DOC and 1.09 %DOC are corresponding to labile and semi-labile DOM, respectively, to our data suggests that the labile and semi-labile DOM off Peru 10 was restricted upper 50 m of the water column.
The compositional analyses of DHAA and DCCHO suggested preferential microbial uptake of Glu, Phe, Ser, Leu and Rha, Gal, Fuc, Ara in the near surface waters, as below 50 m depth, the composition of DHAA and DCCHO were dominated by Gly and Glc, respectively (Fig. 5, Table 1). Glc was previously suggested to be less susceptible to microbial degradation compared to preferentially removed Fuc, Gal, and Ara (Ittekot et al., 1981;Sempere et al., 2008;Goldberg et al., 2010;Engel et al., 2012). Enrichment in Gly with depth has also been proposed to reflect the low nutritional value of Gly in anoxic sediments off Chile (Pantoja and Lee, 2003) and in sediments of the North Sea (Dauwe and Middelburg, 1998

20
"compress" labile and semi-labile DOM towards the surface, while the rapid microbial utilization of DOM shall prevent labile and semi-labile DOM export into the OMZ, and also would imply a pronounced heterotrophic respiration. The latter was suggested by our PCA analyses, as DOM composition was highly interrelated to AOU. Herewith, the diapycnal supply of DHAA and DCCHO could explain up to 38% of ∇Ф ̅̅̅̅ 2 . This suggest, that despite the diapycnal fluxes of labile and semi-labile fractions of DOM may not reach deep into the core of the OMZ, DOM based microbial respiration above the OMZ may 25 substantially attenuate the diapycnal O2 flux that ventilates the upper oxycline. In other words, DOM may alter the shape of the upper oxycline, and, therefore, contribute to the formation and maintenance of the OMZ.

Conclusions
Our results suggest that DOM, i.e. DCCHO and DHAA, is significantly consumed and altered above the upper oxycline in the ETSP off Peru. Thus, despite the presence of high DOC concentrations in the euphotic zone, DOM may enter the OMZ in an 30 already highly reworked stage. Herewith, DOM respiration may contribute substantially (~38%) to O2 reduction in the upper water column, potentially controlling the shape of the upper oxycline of the OMZ. The elevated diapycnal supply of DOC to the upper oxycline, which cannot be explained by microbial processes solely, hint to the presence of an additional DOM removal mechanism, such as microgel formation or absorption onto particles.

Competing interests
The authors of this manuscript are not aware of any real or perceived financial conflicts of interests for other authors or authors that may be perceived as having a conflict of interest with respect to the results of this paper.

10
This study was supported by the Deutsche Forschungsgemeinschaft (dfg.de) trough SFB754 "Climate-Biogeochemical Interactions in the Tropical Ocean" (subproject B9) and CP1403 "Transfer and remineralization of biogenic elements in the tropical oxygen minimum zones".
We thank the chief scientists of the M93 cruise G. Lavik and T. Kanzow for station planning and support during sampling, as well as the crew and scientists onboard. We are also grateful to G. Krahmann for processing the CTD data, to C. Mages and Jackson, G.A. and Williams, P.M.: Importance of dissolved organic nitrogen and phosphorus to biological nutrient cycling, Deep-Sea Res II, 3, 223-235, 1985. Kaiser, K. and Benner, R.: Biochemical composition and size distribution of organic matter at the Pacific and Atlantic timeseries station, Mar. Chem., 113, 63-77, doi:10.1016/j.marchem.2008.12.004, 2009.    The data from all transects and stations were averaged over intervals of 10 km on "Distance from the coast" axis and over 1 m on "Depth" axis. Isolines represent potential density, averaged over intervals of 10 km on "Distance from the coast" axis and over 1 m on "Depth" axis.