On the role of circulation and mixing in the ventilation of oxygen minimum zones with a focus on the eastern tropical North Atlantic

13 Ocean observations are analysed in the framework of the Collaborative Research Center 754 14 (SFB 754) "Climate-Biogeochemistry Interactions in the Tropical Ocean" to study 1) the 15 structure of tropical oxygen minimum zones (OMZs), 2) the processes that contribute to the 16 oxygen budget, and 3) long-term changes in the oxygen distribution. The OMZ of the eastern 17 tropical North Atlantic (ETNA), located between the well-ventilated subtropical gyre and the 18 equatorial oxygen maximum, is composed of a deep OMZ at about 400 m depth with its core 19 region centred at about 20° W, 10° N and a shallow OMZ at about 100 m depth with lowest 20 oxygen concentrations in proximity to the coastal upwelling region off Mauritania and 21 Senegal. The oxygen budget of the deep OMZ is given by oxygen consumption mainly 22 balanced by the oxygen supply due to meridional eddy fluxes (about 60 %) and vertical 23 mixing (about 20 %, locally up to 30 %). Advection by zonal jets is crucial for the 24 establishment of the equatorial oxygen maximum. In the latitude range of the deep OMZ, it 25 dominates the oxygen supply in the upper 300 to 400 m and generates the intermediate 26 oxygen maximum between deep and shallow OMZs. Water mass ages from transient tracers 27 indicate substantially older water masses in the core of the deep OMZ (about 120-180 years) 28 compared to regions north and south of it. The deoxygenation of the ETNA OMZ during 29 recent decades suggests a substantial imbalance in the oxygen budget: about 10 % of the 30 Peter Brandt 28.10.2014 18:59 Gelöscht: carried out 31 Peter Brandt 28.10.2014 18:59 Gelöscht: are used 32

We expect that there is a seasonal variation of the shallow OMZ. Unfortunately we don't have a good seasonal coverage of oxygen data in the eastern boundary upwelling region of Mauretania and Senegal and cannot give a clear statement.
2. Page 12077, line 6: I could not wrap my head around this first sentence of this paragraph. Is there a simpler way to write this?
We split the sentence into two: "The 23° W section (Fig. 6) cuts through the ETNA OMZ, which can be identified by low oxygen levels as well as by the high age of the water masses. The gradual change of salinity on density surfaces along this section defines the transition between low-and high-saline water masses of southern and northern origin, respectively." 3. Page 12081: How good is the assumption that meridional advection is negligible? What about the possible significance of cross-equatorial exchanges via thermocline convergence, upwelling and Ekman divergence?
The thermocline convergence, upwelling and Ekman divergence describes the flow within the subtropical cell (STC). The water masses subducted in the eastern subtropics have to follow equatorward and westward pathways without a mean meridional flow into the OMZs, which is described by the ventilated thermocline theory (Luyten et al., 1983b) and observed geostrophic water mass pathways (Zhang et al., 2003). This is described in Sect The apparent "spike" that is seen for the density range 26.1 to 26.2 originates from the comparably young reservoir age of this density range (related to enhanced subduction rates). We added a statement to the text (also with a reference to the Figure 9 in Karstensen et al. 2008 which nicely shows the enhanced subduction within the given density range).
a suboxic system, i.e. the ETSP OMZ. Particularly the observed deoxygenation trend in the hypoxic ETNA OMZ might lead to a shift of the ETNA OMZ to suboxic conditions and hence the comparison of the two systems will lead to a better understanding of differences and similarities of both systems finally to investigate possible consequences of such a possible regime shift in the future. This was not made clear in the earlier text and will be clarified in the revised manuscript.
We included in the introduction: "The Atlantic and Pacific OMZs have many similarities particularly regarding OMZ shape and circulation pattern. The ETNA and the eastern tropical South Pacific (ETSP) OMZs (Figs. 1, 2) are both located in the shadow zones of the ventilated thermocline and are ventilated by lateral and vertical mixing as well as by zonal advection in the equatorial band. However, the striking difference between both OMZs is that the ETNA OMZ is hypoxic (oxygen below ~60 to 120 µmol kg -1 ) and the ETSP is suboxic (oxygen below about 10 µmol kg -1 ). Karstensen et al. (2008) concluded that this difference is the result of reduced oxygen levels in the eastward current bands of the Pacific OMZs compared to the Atlantic OMZs, which they argue can be traced back to the larger ratio of the total volume of OMZ layer to the renewal or subduction rate in the Pacific compared to the Atlantic." and "The ETSP OMZ has been studied as well using a reduced observational program.
However, the comparison between the hypoxic ETNA and the suboxic ETSP is of particular interest here, as the observed deoxygenation in the ETNA, or future climate change, might lead to a shift from hypoxic to suboxic conditions." At beginning of section 8 we included: "A continuation of the observed deoxygenation in the ETNA would turn the ETNA OMZ suboxic within a century, hence it is worth to look at differences and similarities of the ETNA and the ETSP with regard to a possible shift of a hypoxic system to a suboxic system." In the summary and discussion we included: "The relative importance of the different terms affecting the oxygen budgets of the ETNA und ETSP OMZs appear to be similar. For both OMZs the eastward advection of oxygen-rich waters from the wellventilated western boundary was found to be a dominant ventilation process. As the zonal currents are of similar strength in the tropical Pacific and Atlantic, the difference in the basin width of both oceans consequently results in lower oxygen concentrations and larger water mass ages in the eastern tropical Pacific (Fig. 20) compared to the eastern tropical Atlantic (Fig. 6)." Reviewer #2: Much modelling and observational work has been undertaken on the role of planetary wave systems and dynamics to explain O2 variability and trends in the tropical OMZs but this is not really reflected in this study. Given that these dynamics appear to explain a significant part of the variability in the ETSP and the ETSA it seem that the study should explain why these are under-represented in the ETNA. Reviewer #2: Given that one of the major scientific benefits of such a synthesis is a better understanding of the climate sensitivities of the ETNA, it would have been useful to see some discussion on where models may look to improve the way they reflect the climate sensitivity of the OMZ.
Answer to reviewer #2: In the summary and discussion, we suggest directions for model improvements: "The increase in resolution of ocean circulation models improves the tropical circulation and associated oxygen distribution in the Atlantic (Duteil et al., 2014) and the Pacific OMZs (Montes et al., 2014), suggesting that model physics largely contribute to the oxygen bias in coarser-resolution models. However, particularly the intermediate circulation (below 250 m) is still underestimated by these high-resolution simulations in realistic settings." "Such a regional pattern is most likely due to changes in the circulation pattern associated with forced ocean dynamics as well as with internal ocean dynamics. […] Improvements of model ventilation physics by increased resolution and/or improved parameterizations will reduce errors in the simulated mean oxygen distribution and its variability, but at the same time will help to better understand the climate sensitivity of OMZs with regard to anthropogenic climate change." Reviewer #2: The meridional negative anomaly of the oxygen trend (Fig. 18) between 10 -30N and 100 -500m would seem a good basis to examine where the imbalance may be emerging in the proposed budgets Fig. 13 and 14.
Answer to reviewer #2: We are so far not able to conclude from the budget calculation about the trend pattern (Fig. 18). However, we included in Sect. 7: "Changes in the strength and location of the wind-driven gyres are a possible explanation for the longterm oxygen trends observed between 15° and 30° N in Fig. 18." Reviewer #2: Finally, the summary is again too long and much of the discussion points are repeating the text. Overall, an effort to clarify the objectives and context of the study as well as removal of non critical parts will help further highlight the strengths of this otherwise comprehensive excellent study.
Answer to reviewer #2: We removed repeating parts of the discussion and also streamlined the text in many places in the main body of the manuscript, which hopefully helps to clarify its main points. indicate substantially older water masses in the core of the deep OMZ (about 120-180 years) 28 compared to regions north and south of it. The deoxygenation of the ETNA OMZ during 29 recent decades suggests a substantial imbalance in the oxygen budget: about 10 % of the explains the existence of non-ventilated, near-stagnant shadow zones in the eastern tropics. 67 The remaining slow ventilation of such shadow zones, which under the assumption of steady 68 state is required to balance oxygen consumption, is expected to be the consequence of lateral 69 fluxes of oxygen from oxygen-rich water masses of the subtropics as well as due to diapycnal 70 oxygen fluxes from oxygen-rich layers above and below the thermocline of the OMZs. 71 The near-surface layers (upper ~250 m) of the tropical oceans are characterized by the 72 presence of energetic zonal current bands. In the Atlantic below that layer, substantial mean 73 zonal currents are also found particularly in the depth range of the OMZs (Fig. 1) to multidecadal changes in the strengths of these jets might play a significant role in 87 modulating long-term oxygen changes in the ETNA OMZ (Brandt et al., 2010). 88 The Atlantic and Pacific OMZs have many similarities particularly regarding OMZ shape and 89 circulation pattern. The ETNA and eastern tropical South Pacific (ETSP) OMZs (Figs. 1, 2) 90 are both located in the shadow zones of the ventilated thermocline and are ventilated by 91 lateral and vertical mixing as well as by zonal advection in the equatorial band. However, the 92 striking difference between both OMZs is that the ETNA OMZ is hypoxic (oxygen below ~60 93 to 120 µmol kg -1 ) and the ETSP is suboxic (oxygen below about 10 µmol kg -1 ). Karstensen Bourles et al. (2008)) and at a subsurface 141 mooring at 23° W, 2° N (Fig. 1). For the analysis of hydrographic and velocity data acquired 142 along 23° W, we used the measurements given in Table 1. Besides the 23° W section, we  143 shall present here also data acquired along 18° N at the northern boundary of the ETNA OMZ 144 ( Fig. 1 The subtropical gyre circulation of the northern hemisphere is, to first order, determined by 165 the negative wind stress curl associated with mid-latitude westerlies and northeast trade 166 winds. The resulting Ekman pumping drives subduction of oxygen-rich surface water masses 167 in the subtropics. According to theory, equatorward and westward propagation of subducted 168 water masses forms the northern boundary of the shadow zone of the ventilated thermocline 169 (Luyten et al., 1983b). Within the shadow zone, which is characterized by a weak mean 170 circulation, the ETNA OMZ with a core depth at about 400 m is found. Lowest oxygen 171 concentrations at the core depth are found away from the continental margin at about 20° W, 6 southwestward along the Cape Verde Frontal Zone. It transports oxygen-rich Central Water 188 Equatorial Current (SEC) to the subduction region in the eastern subtropical gyre (Tsuchiya, 208 1986;Stramma and England, 1999). The CW also includes water from the Indian Ocean that 209 are brought into the Atlantic by eddy shedding from the Agulhas retroflection. There is generally higher oxygen variance at 300 m depth close to the oxycline above the 292 deep OMZ core compared to 500 m depth (cf. Figs. 8, 9). Time scales of processes driving the 293 variance in moored time series cover a wide range from those associated with internal waves 294 and tides, inertial oscillations, the mesoscale eddy field to seasonal and interannual variability, 295 including planetary waves (Hahn et al., 2014). Using repeat ship sections, the effect of 296 vertical motion of isopycnals can be removed by calculating oxygen variance on potential 297 density surfaces and projecting back onto depth space (Fig. 10). The remaining oxygen 298 variance in regions of weak mean flow surrounding the ETNA OMZ might be associated with 299 processes responsible for vertical and lateral mixing that is discussed in the following 300 subsections. 301

Vertical mixing 302
Vertical mixing acts on the vertical oxygen gradients and leads to an oxygen supply to the 303 diffusivity of (1.2±0.2)×10 −5 m 2 s -1 was derived (Banyte et al., 2012). The tracer was injected 312 on the isopycnal =26.88 kg m −3 (about 330 m), corresponding to the oxycline above the 313 deep OMZ. GUTRE was accompanied by extensive microstructure and finescale shear 314 measurements that delivered an estimate of (1.0±0.2)×10 −5 m 2 s -1 for ! for the depth range 315 between 150 and 500 m . The value inferred from microstructure 316 measurements only considers diapycnal mixing due to small-scale turbulence. However, 317 double diffusive enhancement was found to be small (~0.1×10 -5 m 2 s -1 ) in this depth interval 318 , so the total diffusivities estimated by the two independent methods 319 agree within the error bars. This estimate of diapycnal mixing is considerably larger than the 320 expected background mixing at this latitude (e.g. Gregg et al. (2003)), probably due to the 321 presence of rough topography (e.g. the Sierra Leone Rise) in the southern part of the OMZ. 322 Combining ! with simultaneous profiles of the vertical oxygen gradient allows 323 determination of the profile of the diapycnal oxygen flux. Its divergence represents the 324 oxygen supply to the OMZ and amounted to about 1 µmol kg -1 yr -1 in the OMZ core, with the 325 required oxygen transported downwards from the upper CW . 326 Deeper ranging microstructure profiles acquired during the most recent cruises to the ETNA 327 OMZ (Table 1) allowed us to extend the analysis into the deeper water column down to 800 m 328 depth; i.e. allowed us to estimate the diapycnal oxygen flux from the AAIW below as well. In 329 total 200 microstructure profiles, 40 of them down to 800 m, were about equally partitioned to 330 three subregions of the OMZ: a seamount subregion (7 % of OMZ area), an abyssal plain 331 subregion (80 % of OMZ area), and a transition subregion (13 % of OMZ area). They served 332 to estimate subregional mean profiles of the turbulent part of diapycnal diffusivity (Fig. 11). 333 Double diffusive enhancement of ! from simultaneous CTD profiles for each subregion 334 following St Laurent and Schmitt (1999) was accounted for to obtain subregional total ! 335 profiles ( Fig. 11) and an area-weighted mean total ! profile (Fig. 12). The mean diapycnal 336 supply ( Fig. 13) that, in the following, will be used in the oxygen budget was then derived as 337 the divergence of the low-pass filtered mean diapycnal flux. The mean flux profile was 338 calculated as the area-weighted mean of the three flux profiles from the three subregions, 339 which in turn were obtained by combining mean ! with vertical oxygen gradient profiles 340 from the three subregions. Error estimates are reported as 95% confidence limits and are 341 based on standard errors of the mean of individual ! and oxygen gradient profiles for each 342 subregion. Subsequent error estimates for the mean total ! profile ( Fig. 12 The oxygen that is meridionally supplied to the ETNA OMZ regime by lateral mixing can 414 then be derived as the divergence of the eddy-driven meridional oxygen flux. The average 415 profile of this eddy-driven meridional oxygen supply (6°-14° N, 23° W) obtained using the 416 diffusive flux parameterization shows a substantial gain of oxygen at the depth of the OMZ 417 and a loss of oxygen above (Fig. 14). The corresponding error was derived both from the error 418 of the curvature of the meridional oxygen distribution (95% confidence) and the error of the 419 eddy diffusivity (factor 2 assumed following (Ferrari and Polzin, 2005)). 420 The tropical and subtropical oceans are generally assumed to be associated with an

Advection 428
We now turn to the remaining ventilation term in the budget; that is, the term associated with 429 zonal advection (meridional advection is assumed to be negligible). We are only able to 430 quantify this term as a residual. A rigorous determination of the advection term would require 431 mean sections around a closed box to fulfil mass balance within the box. This cannot be 432 achieved with the present observing system. However, our measurements along 23° W 433 confirm that the advection term is a major player in the ventilation of the OMZ, especially 434 above 400 m depth. 435 The key factor for carrying the relatively oxygen-rich waters eastwards from the western 436 boundary is the presence of a series of latitudinally stacked zonal jets that are now known to 437 be an ubiquitous feature of the tropical oceans (e.g. Maximenko et al. (2005), Qiu et al. 438 (2013)). Near the equator in the Atlantic, these jets are confined below the Equatorial 439 Undercurrent (EUC), but away from the equator they extend to the surface, and at all latitudes 440 they tend to have a strong depth-independent (barotropic) structure (Fig. 6). Brandt et al. 441 (2010) suggested that a reduction in the strength of these jets north of the equator was a factor 12 in the recent reduction in oxygen within the OMZ. The influence these jets have on the 447 meridional oxygen distribution can clearly be seen in Figure 13  1938; Wyrtki, 1962) and, although being a prominent part of the local oxygen budget of the 481 OMZs, it is among the poorest constrained ones. We will consider here only the net 482 consumption that is the combined effect of removal and production of oxygen. Removal of 483 oxygen is related to the metabolism of marine life as well as to elementary chemical reactions, 484 whereas production of oxygen is related to photosynthesis and as such confined to the 485 euphotic zone (e.g. Martz et al. (2008)). We will focus in this section on pelagic oxygen 486 consumption; removal of oxygen from the water column by uptake at the sediment-water 487 interface will be discussed in Sect. 6. 488 Direct observations of oxygen in-situ respiration are rare, primarily due to technical 489 difficulties (e.g. Holtappels et al. (2014)). The most commonly applied approach to quantify 490 time and space integrated oxygen removal and production processes is through an apparent 491 oxygen utilization rate (AOUR; e.g. Riley (1951); Jenkins (1982Jenkins ( , 1998 2012)). The AOUR is calculated as the ratio between 493 the apparent oxygen utilization (AOU) and age (τ). Hereby AOU is determined as the 494 difference between the air saturation value of dissolved oxygen (e.g. Weiss (1970)) at a given 495 temperature and salinity (surface water saturation is commonly assumed to be 100 %) and the 496 observed in-situ oxygen concentration. The aging of the water starts when a water parcel 497 leaves the surface mixed layer (τ=0) and enters the ocean interior. As the aging is closely 498 linked to the ventilation process the age is also called ventilation age. In many cases the The basic concept behind a ratio of along pathway oxygen removal and along pathway age 511 e.g. it is assumed that as age increases so does AOU (Thiele and Sarmiento, 1990). While this 513 seems to be a reasonable assumption for the ventilated gyre it is questionable for the shadow 514 zone where the OMZs are located. Here diapycnal mixing (Fischer et  influence on the local oxygen transport and water parcels from multiple source regions with 517 different ventilation ages and along-path integrated oxygen consumption meet and mix. 518 Water mass composition and water ages can also be considered in a TTD approach (Haine 519 and Hall, 2002), but limitations exist for non-steady state tracers (such as transient tracers). 520 The TTD concept acknowledges the shortcomings in age calculations, which assign a single 521 tracer age to a water parcel, and provides a framework to more realistically characterize the 522 ventilation age (e.g. Waugh et al. (Δ/Γ=1), where Γ is the mean age and Δ defines the width of the TTD. In the limit of Δ/Γ=0, 526 the mean age of the TTD equals the single tracer age. 527 Here an extended set of CFC-12, SF 6 and oxygen data collected in the ETNA OMZ is used to 528 apply the TTD approach for exploring the oxygen consumption rate. Using CFC-12 and SF 6 529 data (SF 6 preferentially used if available and CFC-12 if CFC-12>450 ppt, i.e. corresponding 530 to atmospheric mixing ratios at about the end of the near-linear atmospheric increase) the 531 AOUR is calculated using two different Δ/Γ ratios (Fig. 16). Note that the AOUR for Δ/Γ=0 is 532 larger than values reported previously (Fig. 16) that were obtained by using a single tracer age 533 concept applied to data collected in the ventilated gyre (e.g. Karstensen et al. (2008)). 534 The two estimates for Δ/Γ=0 and Δ/Γ=1 represent an upper and lower limit of the AOUR 535 within the ETNA OMZ, respectively. A shortcoming of the TTD concept in this region is its 536 one-dimensionality (single water mass), i.e. it only considers the along-isopycnal mixing of 537 parcels of a single source water mass, which might have encountered different advection and 538 diffusion pathways and thus differ in age and AOU. The influence of diapycnal mixing 539   Δ/Γ=1 on tritium ( 3 H) and 3 He measurements. They derived AOUR values close to 5 µmol kg -1 yr -1 for the potential density level of 27.0 kg m -3 that were similar to AOUR values obtained 546 by Karstensen et al. (2008) using CFC-11 ages. For the same density level that is close to the 547 OMZ core depth at roughly 400 m, we derived AOUR values of only about 1.5 µmol kg -1 yr -1 548 using the TTD approach with Δ/Γ=1. The main differences are that the waters off Bermuda 549 are much better represented by a single water mass and that they are significantly younger 550 with a TTD derived mean age of a few tens of years. Waters in the ETNA OMZ instead are a 551 mixture of water masses from multiple sources, some of which might be rather old resulting 552 in a mean TTD age of 120-180 years (Fig. 7). 553 Another approach to estimate the large scale AOUR is based on the reservoir age (Bolin and 554 Rodhe, 1973), which is derived as the ratio of the total volume of the reservoir for an 555 isopycnal range and the corresponding ventilating flux (that is the subduction rate). The 556 AOUR based on the reservoir age is then given by the ratio of the mean AOU of the isopycnal 557 volume and the corresponding reservoir age. For the ETNA OMZ, the AOUR obtained using 558  2008)) it is closer to the AOUR from the tracer age 561 approach (Fig. 17). 562

Processes at the continental margin 563
Processes contributing to the ventilation of the OMZ at the continental margin are advective 564 oxygen transport within the eastern boundary current system, upper-ocean diapycnal oxygen 565 supply due to increased turbulent mixing on the continental slope and shelf, and eddy-driven 566 isopycnal oxygen transport. In comparison to the open ocean OMZs, the consumption of 567 oxygen at the continental margin is generally enhanced due to high pelagic primary 568 production, which in turn results in an increased respiration associated with sinking particles 569 in the water column and at the sediment-water interface. These processes are largely 570 responsible for the regional oxygen distribution particularly defining the shape of the shallow 571 OMZ. Along the eastern boundary, oxygen concentrations within the shallow OMZ decrease 572 towards the north reaching a minimum at about 20° N (Fig. 4). For the deep OMZ, minimum 573 oxygen levels at the continental margin are found south of 16° N (Machin and Pelegri, 2009). 574

Upwelling and circulation 575
The continental margin off Mauritania and Senegal is part of the Canary eastern boundary 576 upwelling system that extends from the northern tip of the Iberia peninsula at 43° N to south 577 of Dakar at about 10° N (e.g. Mittelstaedt (1991)). Due to changes in wind forcing associated with the migration of the Intertropical Convergence Zone, coastal upwelling off Mauritania 596 and Senegal exhibits a pronounced seasonality. Here winds favorable to upwelling prevail 597 primarily from December to April. The seasonality in upwelling and associated primary 598 production must be reflected in oxygen consumption and thus in water-column oxygen 599 concentrations at the continental margin. 600 The ventilation of the waters above the continental margin occurs primarily through the 601 Mauritania Current in the surface layer and the Poleward Undercurrent below. Both currents 602 transport relatively oxygen-rich South Atlantic Central Water, which is supplied by the 603 eastward flowing NECC and NEUC (Figs. 1, 4) Previous studies showed that the Mauritania Current exhibits a seasonal behavior 612 (Mittelstaedt, 1991), which was found to be associated with the seasonality of the NECC, 613 suggesting that the ventilation of the water masses above the continental margin also varies 614 seasonally. In boreal winter and early boreal spring, when the NECC is weak, the Mauritania 615 quasi-exponentially to about 3 mmol m -2 d -1 in a depth of 1000 m. To compare TOU rates to 655 pelagic oxygen consumption, we have to apply the TOU to a water volume with a given in-656 situ density: the consumption within a 1 m thick layer above the bottom due to TOU is three 657 orders of magnitudes larger when compared to pelagic oxygen consumption occurring at 658 similar depths. This is due to the volume-specific production and degradation of organic 659 material in surface sediments, which supports high densities of microbes and metazoans 660 (Glud, 2008). In shelf areas, it is estimated that 10 to 50 % of the pelagic primary production 661 reaches the sediment (Canfield, 1993;Wollast, 1998) and benthic remineralization plays a key 662 role in this region for the recycling of nutrients and burial of carbon. 663 Although the benthic oxygen consumption due to TOU at the shelf strongly exceeds pelagic 664 oxygen consumption, benthic processes play a minor role for oxygen depletion within larger 665 volumes as that of the deep OMZ. To illustrate this, we assume that oxygen depleted water 666 masses are laterally exchanged between the shelf and the open ocean. Between 300 and 600 m 667 depth the continental margin has a typical average topographic slope of about 4 % corresponding to 25 m shelf width per 1 m depth change. Assuming a TOU of 5 mmol m -2 d -1 670 results in an oxygen depletion by the sediments of 125 mmol d -1 per 1 m depth range and 1 m 671 along-shelf distance. Using the range of pelagic oxygen consumption determined in Sect. 5 (1 672 to 5 µmol kg -1 yr -1 ) and corresponding in-situ density, the equivalent water volume resulting 673 in an oxygen depletion of 125 mmol d -1 would be 44×10 3 m 3 to 9×10 3 m 3 , corresponding to a 674 distance from the shelf, where both processes have comparable influence, of 44 km to 9 km. 675 In other words, pelagic oxygen consumption within the deep OMZ, typically extending about 676 1000 km offshore, is 1 to 2 orders of magnitude larger than benthic oxygen consumption due 677 to oxygen fluxes into the continental slope sediments. Reduced topographic slopes at 678 shallower depths suggest a more important role of benthic oxygen uptake for the shallow 679 OMZ, which is characterized by minimum oxygen concentration close to the continental 680 margin and is not as widespread as its deeper counterpart (cf. Figs. 3, 4). reduced to less than 10 mmol m -2 d -1 at a depth of 60 m below the mixed layer, which has an average thickness of about 20 m. Diapycnal mixing is thus able to fully supply the oxygen 717 that is required by the benthic oxygen uptake for water depths shallower than about 80 m. At 718 about 150 m depth, however, the diapycnal flux changes sign due to the presence of the 719 shallow OMZ and oxygen here is essentially fluxed upward, although at small rates. Thus, 720 oxygen from the sea surface cannot contribute to ventilating the deeper water column via 721 diapycnal mixing. 722 It should be noted that the diapycnal oxygen flux divergence from the mixed layer to 60 m 723 below the mixed layer yields a diapycnal oxygen supply of about 400 µmol kg -1 yr -1 . In steady 724 state other oxygen transport processes and consumption are required to balance this 725 substantial oxygen supply. While vertical advection during the upwelling season might 726 contribute to the balance, the oxygen supply due to other transport processes should be at least 727 an order of magnitude lower in this region. The diapycnal oxygen supply to the upper 728 thermocline can thus be used to define an upper limit of the oxygen consumption below the 729 mixed layer. Such a consumption rate is, however, two orders of magnitude larger than the 730 one estimated for the deep ocean as discussed above. 731 The results suggest that the high oxygen demand of the water column and the sediments 732 within the upwelling region at shallow depths above the shallow OMZ may well be supplied 733 from the surface via diapycnal mixing. At larger depths however, the continental slope must Similar to the hypoxic ETNA OMZ, the suboxic ETSP OMZ is located in the shadow zone 819 equatorward of the subtropical gyre with lowest oxygen levels near the shelf-break. The most 820 prominent difference between both OMZs is that the ETSP OMZ covers a much wider region 821 and that oxygen values in its core region are close to zero (Karstensen et al., 2008) while the ETNA would turn the ETNA OMZ suboxic within a century, hence it is worth to look at 825 differences and similarities of the ETNA and the ETSP with regard to a possible shift of a 826 hypoxic system to a suboxic system. 827

The large scale distribution 828
Different to the ETNA with its Guinea Dome and the eastern tropical South Atlantic and 829 eastern tropical North Pacific with similar domes there is no dome in the ETSP (Kessler, 830 2006). Similar to the equatorial Atlantic, the equatorial Pacific is characterized by a local 831 oxygen maximum and a system of eastward and westward currents (Figs. 2, 20). Near the 832 equator, the EUC, the NICC and SICC all carry water richer in oxygen than the adjacent 833 The eastward mass transport associated with the EUC, SCC's and ICCs was estimated to be 892 about 30×10 9 kg s -1 . It was assumed that this mass transport is returned by the adjacent 893 westward currents with a typical relative oxygen difference between eastward and westward 894 currents of about 20 µmol kg -1 . The resulting net advective molar oxygen supply across 125° 895 W is 0.6×10 6 mol s -1 (Stramma et al., 2010a). The diffusive supply was estimated through the 896 climatological 60 µmol kg -1 surface surrounding the tropical Pacific OMZ. Vertical and lateral 897 oxygen gradients were evaluated at this surface and multiplied with a diapycnal diffusivity of characteristic for the off-equatorial regions (Davis, 2005), respectively. Integrating these 901 products over the surface area resulted in a vertical diffusive molar oxygen supply of 0.4×10 6 902 mol s -1 mostly through the upper surface, where the gradients are large, and in a lateral 903 diffusive molar oxygen supply of 0.8×10 6 mol s -1 (Stramma et al., 2010a). The mass of the 904 tropical Pacific OMZs between 30° N and 30° S with oxygen concentrations <60 µmol kg -1 is 905 about 16×10 18 kg. Dividing the estimates of molar supply by the mass leads to an advective 906 oxygen supply of about 1.2 µmol kg -1 yr -1 , a lateral diffusive oxygen supply of 1.6 µmol kg -1 907 yr -1 and a vertical diffusive oxygen supply of 0.8 µmol kg -1 yr -1 . The oxygen utilization rate 908 calculated to balance the net oxygen supply resulted in about 3.6 µmol kg -1 yr -1 . These rough

Trends in oxygen 918
As the ETSP OMZ is extremely low in oxygen a decreasing trend is much more difficult to 919 determine. Furthermore, data are sparse to investigate the trend. However, for the eastern 920 oxygen supply is strongest slightly above the deep OMZ core, where it accounts for about one third of the oxygen supply required to balance consumption. There are, however, indications 972 of regional variations in the diapycnal eddy diffusivity with higher values over the seamount 973 region (up to one order of magnitude) compared to the abyssal plains (Fig. 11) resulting also 974 in a general increase of the diapycnal eddy diffusivity with depth (Fig. 12). 975 The contribution of the mean advection to the oxygen budget of the OMZ cannot yet be 976 quantified from observational data. Instead, idealized advection-diffusion models were used 977 to estimate this contribution (Brandt et  sinking rate, and mineral ballast effect, they therefore suggested to use the sum of two 999 exponential functions with different decay. Processes that would also contribute to a deviation 1000 from a single exponential profile include respiration associated with the daily vertical 1001 migration cycle of zooplankton  or oxygen consumption at the sediment-1002 ocean interface and associated lateral spreading of low oxygen waters. To tackle the problem 1003 of regional and temporal consumption variability new targeted data/model approaches are oxygen consumption.
To identify the physical mechanism responsible for the mean and variable zonal jets, idealized high-resolution models have been employed (Ménesguen et