Present past and future of the OMZ in the northern Indian Ocean

Abstract. Decreasing concentrations of dissolved oxygen and the resulting expansion of anaerobic ecosystems is a major threat to marine ecosystem services because it favors the formation of greenhouse gases such as methane, endangers the growth of economically important species, and increases the loss nitrate. Nitrate is one of the potential primary nutrients, which availability controls the marine productivity. The Arabian Sea and the Bay of Bengal are home to ~ 59 % of the Earth's marine sediments exposed to severe oxygen depletion and approximately 21 % of the total volume of oxygen-depleted waters (oxygen minimum zones, OMZs). The balance between physical oxygen supply and the biological oxygen consumption controlled the oxygen concentrations. In the Arabian Sea and most likely also in the Bay of Bengal the supply of oxygen sustained by mixing and advection associated with mesoscale eddies compensated the biological oxygen consumption. These steady states maintain low (hypoxic) oxygen concentrations allowing the competition between anaerobic and aerobic processes. However, due to slightly higher oxygen concentrations, the aerobic nitrite oxidization inhibits the anaerobic nitrite reduction and thus denitrification (the reduction of nitrate to N2) to become significant in the Bay of Bengal. A feedback mechanism caused by the negative influence of decreasing oxygen concentrations on the biological oxygen demand helped to maintain these steady states. Furthermore, it might have also counteracted a reduced physical oxygen supply into the Arabian Sea caused by climate-driven changes in the ocean's circulation during the last 6000 years. However, due to human-induced global changes, the OMZs in Arabian Sea and the Bay of Bengal intensified and expanded, which included also the occurrence of anoxic events on the Indian shelf. This affects benthic ecosystems, and in the Arabian Sea it seems to have initiated a regime shift within the pelagic ecosystem structure. Consequences for biogeochemical cycles are unknown, which, in addition to the poor representation of mesoscale features reduces the reliability of predictions of the future OMZ development in the northern Indian Ocean.



Introduction 24
The rise of oxygen at about 600 Million years ago initiated a revolution by facilitating aerobic live 25 forms to displace anaerobic ecosystems from the Earth's surface (Canfield, 2014 ;Lenton et al., 2011;26 Lyons et al., 2014). Albeit it seems that today anaerobic microorganisms do not emerge from their 27 shadow existence in guts of animals, wetlands and marine sediments, yet they strongly influence the 28 productivity of aerobic ecosystems and the Earth's climate as they reduce nitrate to N 2 and produce 29 methane. Nitrate limits the productivity in many of the aerobic ecosystems and methane is the most 30 important greenhouse gas in the Earths' atmosphere after water vapor and CO 2 (Gruber et al., 2008;31 Kirschke et al., 2013;Myhre et al., 2013;Nisbet et al., 2016). 32 The transition from solely aerobic to anaerobic ecosystems occurs in steps at which microorganisms 33 utilize oxygen bound to nitrogen (e.g. nitrate and nitrite) as well as to sulfur (e.g. sulfate) to decompose 34 organic matter. Heterotrophic organisms use the resulting energy for running their metabolism whereas 35 autotrophic life forms oxidize reduced metabolites to gain energy which additionally sustains the build-36 up of new biomass (e.g. Middelburg, 2011). The absence of elementary oxygen (anoxia) and oxygen 37 bound to nitrogen and sulfur inhibits this chemosynthesis, and organic matter is decomposed to carbon 38 dioxide, methane, ammonia, and hydrogen sulfide. At anoxic conditions and in the presence oxygen 39 oxygen consumption reflects uncertainties caused by the poorly constrained physical oxygen supply 237 and export production rates (Rixen et al., 2019b;. Nevertheless, these two 238 processes are linked to each other if the seasonal thermocline is hypoxic. 239

Interplay between the intensity of the OMZ and export production 240
The seasonal thermocline is the subsurface layer from which water is introduced into the euphotic zone 241 via physical processes such as upwelling and vertical mixing on a seasonal timescale. Nutrient supplied 242 by these mechanisms largely sustain the productivity of pelagic ecosystems and the associated export 243 production (Eppley et al., 1979). Hence, the seasonal thermocline is the main nutrient reservoir of 244 pelagic ecosystems and to fulfill this role the vast majority of the exported organic matter must be 245 respired within the seasonal thermocline. Accordingly, the season thermoclines represents the main 246 zone of respiration and similar to soils on land, accommodates the nutrient recycling machinery of the 247 pelagic ecosystem. Nutrient losses from the seasonal thermocline via particle fluxes into the deep sea, 248 denitrification, and lateral advection must be compensated by nutrient inputs in order to maintain the 249 productivity (Rixen et al., 2019a). Nitrogen fixation, river discharges, and atmospheric deposition can 250 be important nutrient sources but in the Arabian Sea lateral inflow of water masses from the south are 251 the main source balancing nutrient losses from the season thermocline (Bange et al., 2000;Gaye et al., 252 2013). In contrast to the Bay of Bengal nitrite accumulates in the seasonal thermocline of the Arabian 253 Sea (Fig. 3). 254 The accumulation of nitrite in the upper part seasonal thermocline, which was first described during 255 John Murray expedition of 1933 -34 (Gilson, 1937), is assumed to indicate active denitrification and is 256 called the secondary nitrite maximum (SNM, Naqvi, 1991). The role of the SNM as indicator of active 257 denitrification is further supported by stabile isotopic ratio of nitrogen in nitrate (δ 15 N NO3 ) and nitrate 258 (NO 3 -) concentration profiles (Gaye et al., 2013;Rixen et al., 2014). Since denitrification increases 259 δ 15 N NO3 in the water column due to the preferential uptake of the lighter 14 NO 3 - (Cline et al., 1975;260 Mariotti et al., 1981) low nitrate concentrations correspond to high δ 15 N NO3 within the SNM. 261 In the Arabian Sea the SNM indicates the core of the OMZ (Fig. 1 c) and during the last decades it 262 expanded towards the south and west due to the decreasing oxygen concentrations in these regions 263 (Banse et al., 2014;Rixen et al., 2014). The SNM occurs at water depths between 200 and 400 m in the 264 central Arabian Sea (Fig. 3a) and as deep as 500 m in the eastern Arabian Sea. It divides the main 265 respiration zone in an aerobic upper part at water depths between ~40 and 200 m and an anaerobic 266 lower part down to the base of SNM (Fig. 3a). The base of the SNM is still located in the hypoxic 267 OMZ but, in contrast to the SNM, associated with increasing nitrate concentrations. Therefore 268 anaerobic processes including also the sulfate/nitrate based respiration (Canfield et al., 2010) are 269 assumed to be negligible so that the base of the SNN seems to represent also the base of the main 270 respiration zone. 271 Even though nitrate concentrations decrease within the SNM they remain above 10 µM, which suggest 272 that not nitrate but the supply of decomposable organic matter limits denitrification. A substrate 273 limitation at a water depth of 400 to 500 m and the arrival of organic matter at sediment traps deployed 274 in the deep sea at a water depth of 3000 m support the concept of export production that is divided 275 between free (reactive) and protected (low contribution of reactive) organic matter (Armstrong et al.,276 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. 2002). This partition is based on the assumption that ballast-associated, protected organic matter is 277 preferentially exported to deeper waters as fast sinking particles, whereas the slow sinking free organic 278 matter is preferentially respired with the main respiration zone (Fig. 3). Therewith the ballast-effect is a 279 prime factor controlling the nutrient supply to the seasonal thermocline and therewith the export 280 production and the intensity of the OMZ. 281 Although the ballast effect is not specifically addressed in numerical models used to study the OMZ in 282 the northern Indian Ocean, there are models that account to some extent for the concept of protected 283 and free organic matter by considering the formation of fast and slow sinking particles (Aumont et al., 284 2015;Lachkar et al., 2019;Lachkar et al., 2016;Resplandy et al., 2012). These 285 models indicate that organic matter is mostly remineralized within in the upper 300 m of the water 286 column (Resplandy et al., 2012) which nearly encompasses the depth range of the SNM. It also covers 287 approximately the depth range of vertical migrating zooplankton during the large summer bloom in the 288 Arabian Sea (Smith, 2001), and roughly matches the water depth range from where subsurface water is 289 introduced via upwelling into the euphotic zone in the western Arabian Sea (Brock et al., 1992;Rixen 290 et al., 2000). 291 In addition to the ballast-effect also concentrations of dissolved oxygen influence the organic matter 292 export into the deep sea as decreasing oxygen concentrations are assumed to slow down the respiration 293 (Aumont et al., 2015;Laufkötter et al., 2017;Thamdrup et al., 2012;Van Mooy et al., 2002). 294 Consequences of a reduced respiration within the seasonal thermocline are enhanced fluxes of organic 295 matter into the deep sea and a deepening of the respiration zone. Data presented by Acharya and 296 Panigrahi (2016) support the hypothesis by showing that decreasing oxygen concentrations within the 297 OMZ correlate with a deepening of the OMZ on a seasonal time scale (Fig. 4). On the other hand an 298 increased export of organic matter and nutrients out of the seasonal thermocline lowers the productivity 299 and the associated export production, which in turn reduces the oxygen consumption within the OMZ. 300

Implications 301
If decreasing oxygen concentrations within the seasonal thermocline lower the export production, the 302 resulting lower biological oxygen demand could mitigate or even prevent an intensification of the 303 OMZ caused by weaker ballast-effect and or a reduced physical oxygen supply. This feedback 304 mechanism might have played an important role in maintaining the hypoxic conditions within the 305 Arabian Sea and Bay of Bengal OMZ by preventing the development of anoxic conditions. As 306 discussed in the following chapters this also agrees with model and paleoceanographic results 307 suggesting that variations of the physical oxygen supply rather than changes in the biological oxygen 308 demand are drivers controlling the intensity of the OMZ in both the Arabian Sea and the Bay of Bengal 309 (Gaye et al., 2018;McCreary Jr et al., 2013;Resplandy et al., 2012). 310

The role of mesoscale eddy activity as a driver of OMZ ventilation 311
Mesoscale eddies, in the form of coherent vortices and filaments, are ubiquitous in the ocean. They 312 develop from baroclinic and barotrophic instabilities related to the shear of horizontal currents. As they 313 transport heat, salt, nutrients and oxygen across large distances in the ocean, eddies affect both climate 314 and large-scale marine biogeochemistry. Previous studies have also shown that eddies generally 315 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. enhance biological production in oligotrophic environments through nutrient pumping (e.g., Oschlies et 316 al., 1998) and suppress productivity in biologically active eastern boundary upwelling systems as they 317 cause subduction of incompletely consumed nutrients offshore below the euphotic zone (e.g., Gruber et 318 al., 2011). More recent work has highlighted the role of eddies in enhancing ocean mixing in regions of 319 sluggish circulation in the eastern tropical Atlantic and Pacific, thus contributing to the ventilation of 320 the OMZ located there (Bettencourt et al., 2015;Brandt et al., 2015;Gnanadesikan et al., 2013). In 321 particular, stirring of oxygen by eddies along isopycnal surfaces has been suggested to modulate the 322 intensity and distribution of low-oxygen waters in the ocean (Fig. 5, Gnanadesikan et al., 2012). 323

Effects of eddies on the Arabian Sea OMZ 324
In the Arabian Sea, numerical model studies have shown that eddies play an important role in the 325 transport of nutrients and oxygen (Lachkar et al., 2016;McCreary Jr et al., 2013;Resplandy et al., 326 2012;Resplandy et al., 2011). For instance, Resplandy et al., (2011) emphasized the role of mesoscale 327 eddies in spreading nutrients vertically and horizontally in the Arabian Sea (Fig. 5). Furthermore, 328 mesoscale eddies and filaments were shown to dominate the supply of oxygen to the OMZ in the 329 Arabian Sea on an annual timescale due to the semiannual reversal of the mean circulation and a 330 resulting reduced oxygen supply (Resplandy et al., 2012). This study also showed that eddy-driven 331 advection enhances the vertical supply of oxygen along the western coast of the Arabian Sea and 332 contributes to the lateral transport of ventilated waters offshore into the central Arabian Sea. In a 333 process study aiming to explore the dynamics of the Indian Ocean OMZs, McCreary et al (2013) 334 highlighted the important role of vertical eddy mixing in the ventilation of the western Arabian Sea in 335 addition to the inflow of ICW. Their work suggests that this mechanism strongly contributes to the 336 eastward shift of the upper OMZ relative to the region of highest productivity located along the western 337 part of the Arabian Sea. 338 Using a suite of regional model simulations with increasing horizontal resolution, Lachkar et al (2016) 339 found that isopycnal eddy transport of oxygen to the Arabian Sea OMZ strongly limits the extent of its 340 suboxic core. Within the model this leads to a suppression of denitrification and in turn to an increase 341 in biological productivity driven by an increase of nitrate availability in the subsurface. As more 342 organic matter is produced near the surface, more remineralization and oxygen consumption occur at 343 depth. This in turn results in an expansion of the volume of hypoxia and a compression of habitats of 344 O 2 -sensitive species. Thus, eddies affect the Arabian Sea marine biogeochemistry and living organisms 345 both at lower (e.g., plankton) and higher trophic levels (e.g., fish). 346 Finally, eddies have also been shown to control the transport and the spreading of the Persian Gulf 347 Water (PGW) into the Gulf of Oman (Queste et al., 2018;Vic et al., 2015). These dense waters, 348 relatively rich in O 2 , subduct in the northern Arabian Sea and strongly contribute to the ventilation of 349 the upper OMZ there (Lachkar et al., 2019;Schmidt et al., 2020). Using a 350 series of computer simulations, it could be shown that a warming driven decrease in the sinking of 351 oxygen-saturated dense waters formed in the Persian Gulf contributes to a drop in oxygen at depths 352 between 200 and 300 m in the northern Arabian Sea (Lachkar et al., 2019). 353 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.

Eddies and the ventilation of the Bay of Bengal 354
In the Bay of Bengal, previous studies have highlighted the role of eddy pumping of nutrients in 355 enhancing biological productivity during all seasons (Kumar et al., 2007;Prasanna Kumar et al., 2004;356 Singh et al., 2015). Eddies have also been shown to affect the ventilation of the Bay of Bengal and 357 subsequently the intensity of its OMZ. For instance, Sarma et al (2018a) showed that while cyclonic 358 eddies inject nutrients into the euphotic zone, thus enhancing productivity and oxygen consumption at 359 depth, anticyclonic eddies supply oxygen to the subsurface layer and hence weaken the OMZ. Sarma 360 and Baskhar (2018b) focused on anticyclonic eddies sampled by bio-Argo floats between 2012 and 361 2016 in the Bay of Bengal and found these to form in the eastern side of the basin and propagate 362 westward, ventilating the layer between 150 and 300 m and weakening the OMZ. The frequent 363 episodic injection of oxygen, likely by mesoscale eddies, could be the physical oxygen supply 364 mechanism that inhibited denitrification and/or prevent it from becoming significant. 365

Implications 366
The variability of eddy activity in space and time can modulate the intensity of OMZs between 367 different regions and across time, thus contributing to the observed variability of dissolved O 2 . In this 368 context, previous work has linked long-term changes in oxygen to changes in the intensity of eddy 369 activity. For instance, Brandt et al (2010) have shown that a reduction in filamentation and the strength 370 of alternating zonal jets associated with mesoscale eddies between the periods 1972-1985 and 1999-371 2008 in the tropical north Atlantic has contributed to a reduction in the ventilation of the OMZ located 372 there and hence contributed to its deoxygenation. In the Bay of Bengal, strong interannual variations in 373 the intensity of the eddy activity have been reported (Chen et al., 2012). These are expected to cause 374 strong variations in the subsurface ventilation that may eventually lead to deoxygenation and onset of 375 denitrification at the core of the OMZ (Johnson et al., 2019). 376 The fact that eddies affect both the supply of O 2 (through ventilation) and its consumption (through 377 biological productivity) in a non-trivial manner can explain the fundamental difficulty to adequately 378 parameterize the effects of eddies on dissolved oxygen in coarse resolution models. An additional 379 potential source of error in the currently used parameterizations is their underlying assumption that the 380 eddy-driven isopycnal tracer mixing and isopycnal flattening occur at similar rates (Griffies, 1998). 381 Yet, recent studies (e.g., Gnanadesikan et al., 2013) suggest that the two can be substantially different. 382 In the Arabian Sea, Lachkar et al (2016) show that the eddy driven transport of O 2 is mostly driven by 383 enhanced mixing along the isopycnal surfaces with very little change in the slope of the isopycnals. 384 However in the Arabian Sea, both Resplandy et al. (2012) and  found that the 385 biological oxygen consumption is counterbalanced by the supply of oxygen sustained by mixing and 386 advection associated with mesoscale eddies and filaments. This in turn agrees with paleoceanographic 387 studies, implying that remotely-forced changes in physical oxygen supply cause long-term changes to 388 the intensity of the OMZ. 389 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.

The δ 15 N as an indicator of OMZ strength in sediments 391
Changes in OMZ oxygenation were shown to be reflected by the δ 15 N of nitrogen in sediments (Altabet 392 et al., 1995;Ganeshram et al., 1995). The average δ 15 N NO3 value of oceanic deep water is ~5‰ 393 (Sigman et al., 2005) but δ 15 N NO3 in OMZs can be much higher during periods of denitrification as this 394 process has an isotopic effect of 20-30 ‰ and resulting δ 15 N NO3 can exceed 20‰ (Altabet et al., 1999;395 Brandes et al., 1998). Convective mixing and especially upwelling force nitrate-deficient water masses 396 to the surface, so that the nitrate with high δ 15 N values is transported into the euphotic zone. After 397 assimilation into biomass by phytoplankton, 15 N-enriched particulate matter sinks through the water 398 column to the seafloor where the signal of denitrification and OMZ intensity is preserved in sediments 399 (Altabet et al., 1995;Gaye-Haake et al., 2005;Naqvi et al., 1998;Suthhof et al., 2001). Early 400 diagenesis may raise sedimentary δ 15 N values by 2-5 ‰ and the diagenetic effect increases with water 401 depth (Altabet, 2006;Tesdal et al., 2013). Nevertheless, the relative changes of δ 15 N in deep-sea 402 sediments record variations in the OMZ intensity while records from the continental slopes are 403 subjected to negligible diagenetic enrichments so that they retain the signal of the nitrogen source 404 (Altabet et al., 1999;Gaye et al., 2018). 405

OMZ Fluctuations in the Holocene 406
A core from the northern Bay of Bengal, which at present has the lowest oxygen concentrations of the 407 basin shows a range of δ 15 N between 4.4 and 5.0 ‰ during the Holocene and even slightly lower δ 15 N 408 during the last glacial maximum so that denitrification can be ruled out from a paleoceanographic 409 perspective (Contreras-Rosales et al., 2016). The δ 15 N values at the core top of 4.6 ‰ were similar to 410 values in sediment trap materials of 3.7-4.5 ‰, and were explained by a mixture of nutrients or 411 suspended matter from the Ganges-Brahmaputra-Meghna river system with nitrate from subsurface 412 water (Contreras-Rosales et al., 2016;Gaye-Haake et al., 2005;Unger et al., 2006). Enhanced δ 15 N 413 values in the early Holocene to 6000 years BP (BP = before present, whereas present means 1950) 414 coincide with a stronger monsoon and were attributed to enhanced supply of nitrate from the 415 subsurface which has elevated δ 15 N compared to the depleted values of the riverine endmember (Sarkar 416 et al., 2009). Nevertheless, to our knowledge there is only one published record from the Bay of Bengal 417 spanning the entire Holocene (Contreras-Rosales et al., 2016) so that we know nothing about the 418 spatial variability within the basin. However, the available data imply so far that results presented by 419 Bristow et al. (2017) and discussed before indicate a recent onset of denitrification within the Bay of 420

Bengal. 421
In contrast to the Bay of Bengal denitrification has prevailed in the Arabian Sea during the warm 422 interstadials of the Pleistocene and during the entire Holocene as can be discerned from δ 15 N >6 ‰ 423 with maxima of >11 ‰ (Agnihotri et al., 2003;Higginson et al., 2004;Kessarkar et al., 2018;Möbius 424 et al., 2011;Pichevin et al., 2007). Productivity increased with the onset of the Holocene as the summer 425 monsoon strengthened and monsoonal upwelling off Somalia and Oman commenced and became a 426 permanent feature of the Holocene Arabian Sea (Böning et al., 2009;Gaye et al., 2018). A rise of δ 15 N 427 by at least 2 ‰ shows that onset of upwelling immediately strengthened the OMZ and led to 428 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. denitrification in the entire basin (Böll et al., 2015;Gaye et al., 2018). Furthermore, production of the 429 oxygen-enriched ICW was reduced by the southward retreat of the Antarctic Sea Ice, so that 430 ventilation of the Arabian Sea OMZ from the south was in turn reduced (Böning & Bard, 2009;Naidu 431 et al., 2010). 432 A decline in δ 15 N by about 1 ‰ is found in the early Holocene until 6000 years BP in high-resolution 433 sediment cores from the western, northern and eastern Arabian Sea and indicates that the OMZ 434 weakened and became less persistent during this period (Fig. 6a). A possible explanation may be the 435 enhanced input of surface derived and therefore oxygen enriched water from the Red Sea and Persian 436 Gulf due to prolonged sea level rise until about 6000 years BP (Siddall et al., 2003). More vigorous 437 upwelling during this period, discernible from benthic foraminifera, also led to a better ventilation of 438 the basin by Indian Central Water (ICW) from south during this period (Das et al., 2017). After 6000 439 years BP increasing δ 15 N values indicate a strengthening of the OMZ across the entire basin, which is 440 still ongoing (Fig 6a). It is assumed that a weaker ventilation is responsible for decreasing oxygen 441 concentrations and it could be due to reduced inflow of ICW as it is blocked by the enhanced inflow of 442 PGW and Red Sea Water (RSW) since the sea level high stand at 6000 years BP (Pichevin et al., 443 2007). Ventilation of the eastern Arabian Sea by the West Indian Coastal Current also declined and 444 was shifted southward (Mahesh et al., 2014). Climate Model (KCM, Park et al., 2009) and the marine biogeochemistry model PISCES (Aumont et 454 al., 2003). In a first step, KCM was forced with transient orbital parameters and greenhouse gas 455 concentrations from 9500 years BP to present. In a second step, the PISCES model was forced with the 456 ocean physical fields from the above KCM experiment in so-called off-line mode (see

Implications 484
The δ 15 N records from the Arabian Sea and Bay of Bengal reveal the difference in late Pleistocene and 485 Holocene history of denitrification. Oxygen concentrations in the Bay of Bengal never declined below 486 the threshold of denitrification whereas denitrification prevailed in the Arabian Sea during the warm 487 interstadials and the entire Holocene. A data-model comparison shows that the age of the OMZ water 488 mass increased after 6000 years BP in both basins coinciding with a strengthening of the OMZ and 489 denitrification in the Arabian Sea which is still ongoing. It is assumed that a reduced ventilation is 490 responsible for decreasing oxygen concentrations and it could be due to less inflow of ICW as it is 491 blocked by the enhanced inflow of PGW and RSW since the sea level high stand at 6000 years BP 492 (Pichevin et al., 2007). Ventilation of the eastern Arabian Sea by the West Indian Coastal Current and 493 the associated counter current also declined and was shifted southward (Mahesh & Banakar, 2014). 494 The similar temporal evolution of observed OMZ intensity and modelled O 2 concentration in the 495 Arabian Sea thus indicates that the mid-to late Holocene OMZ intensification may be related to 496 oceanic circulation rather than to local processes in the Northern Indian Ocean. The progressive 497 oxygen loss may thus be the result of orbital and greenhouse gas forcing. 498 6. Model predictions 499

Global models 500
For future climate predictions we rely on earth system models (ESM). Although these models 501 reproduce large-scale features and global trends they suffer from considerable mismatches between 502 measured and model oxygen concentrations in the ocean Cabré et al., 2015;503 Oschlies et al., 2018;Oschlies et al., 2008). In comparison to observational data, they underestimate 504 oxygen losses significantly (e.g. Oschlies et al., 2018 and references therein) and simulated volumes of 505 OMZs differ considerably. Unresolved physical oxygen supply mechanisms, poorly constrained 506 biological oxygen consumption rates and their hardly known responses to global change cause these 507 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.
uncertainties (e.g., Oschlies et al., 2018;Segschneider et al., 2013). Furthermore, feedbacks caused by 508 the strong coupling of the marine oxygen and nitrogen cycles complicate long-term predictions (Fu et 509 al., 2018;Oschlies et al., 2019). 510 Especially in the Indian Ocean, global coupled biogeochemical ESMs struggle to represent the OMZs 511 ( Fig. 7, Oschlies et al., 2008). In most ESMs the east -west contrast between the Arabian Sea and Bay 512 of Bengal is backward, with most global models producing lower oxygen values in the Bay of Bengal 513 than in the Arabian Sea. To some degree this problem may be attributed to the fact that ESMs are not 514 tuned for the northern Indian Ocean. In addition, global models generally have coarser resolution to 515 reduce computational costs, thus they are not able to resolve mesoscale processes, which are important 516 for both the ventilation of the OMZ and for resolving upwelling that generates high rates of primary 517 production and biological oxygen demand. These processes are parameterized in the ESMs but the 518 question remains, why do they still fail to represent the OMZs in the northern Indian Ocean? We 519 conclude that more care should be dedicated to the representation of the eddy-driven isopycnal mixing 520 in the global ocean models for a more accurate representation of OMZs and O 2 in general, and an 521 enhanced ability to predict future global oxygen distributions and climate. 522

Future prediction 523
The poor representation of the OMZs in the northern Indian ocean in ESMs reduces the reliability of 524 future predictions of potential changes in the OMZs related to natural and anthropogenic forcing, and 525 thus their ecological impacts and possible feedbacks to climate change. Global models suggest a 526 general decline of oxygen for the entire ocean, but there is no clear trend visible in the Indian Ocean 527 (Oschlies et al., 2017). However, an older set of ESMs analyzed in Cocco et al. (2013) suggest a future 528 decrease in oxygen in the subtropical Indian Ocean in the upper mixed layer and a small increase in the 529 western tropical Indian Ocean. This increasing oxygen concentration is also seen in response to climate 530 change in the RCP8.5 and RCP2.6 scenarios of the 5 th coupled model intercomparison project (CMIP5, 531 Bopp et al., 2013). Specifically, Bopp et al. (2013) showed that a decrease in productivity is 532 consistently simulated across all CMIP5 models and scenarios in the tropical Indian Ocean and that, by 533 2100, all models project an increase in the volume of waters below 80 µM, relative to 1990-1999. 534 This response is more consistent than that of the previous generation of ESMs, i.e., changes varying 535 from −26 to +16% over 1870 to 2099 under the SRES-A2 scenario (Cocco et al., 2013). 536 However, for lower oxygen levels, there is less agreement among the CMIP5 models and also 537 compared to observations regarding the volume of the OMZ . Specifically, for the 538 volume of waters below 50 µM, four models project an expansion of 2 to 16% (both GFDL-ESMs, 539 HadGEM2-ES and CESM1-BGC), whereas two other models project a slight contraction of 2% 540 (NorESM1-ME and MPI-ESMMR). For the volume of waters below 5 µM, only one model (IPSL-541 CM5A-MR) is close to the volume estimated from observations and simulates a large expansion of this 542 volume (+30% in the 2090s). These results for low oxygen waters (5 and 50 µM) agree with those of 543 Cocco et al. (2013), with large model-data and model-model discrepancies and simulated responses 544 varying in sign for the evolution of these volumes under climate change . Thus, 545 future trends in the northern Indian Ocean OMZs derived from the ESMs are highly uncertain, with 546 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.
predicted potential increases or decreases in the volume of low oxygen waters, depending on the model 547 and the oxygen levels under consideration Cocco et al., 2013). 548

Implications 549
The OMZ in the Indian Ocean is the one we know least about but it may also be the OMZ with the 550 most complex dynamics in terms of forcing and variability. Regional modelling studies have been able 551 to reproduce the OMZs and thus they have helped us to understand the interplay between physical and 552 biogeochemical drivers (Lachkar et al., 2019;McCreary Jr et al., 2013;Resplandy et al., 2012;553 Resplandy et al., 2011). However, there is still very little known about the interannual variability of the 554 Indian Ocean OMZs, as there are limited long term observational data and the influence of the remote 555 forcing processes that drive this variability (e.g., IOD and ENSO) is not fully understood. Global 556 models still struggle to reproduce the Indian Ocean OMZ. One explanation for this is the coarse 557 resolution of these models, i.e., they cannot resolve the mesoscale processes that ventilate the 558 subsurface waters and they underestimate coastal upwelling during the monsoon seasons and, 559 therefore, also primary production and biological oxygen demand. As a result, the oxygen trend in the 560 tropical Indian Ocean remains unclear. However, in addition poor representation of mesocale features 561 in global models large uncertainties stem also from largely unknown ecosystem responses to global 562 changes. 563

Pelagic ecosystems 565
Dissolved oxygen concentrations in seawater are crucial for the successful development of many 566 pelagic organisms, particularly marine animals both planktonic vertebrates, and invertebrates whose 567 metabolism, life cycle performance, growth capacity, reproductive success and longevity are intimately 568 linked to oxygen availability (Ekau et al., 2010 and references therein). However, hypoxia tolerance 569 and threshold values vary enormously among species, and even within the same species, and the 570 growth stage of animals, the differences can be very large (Miller et al., 2002). Many fish larvae 571 present in the pelagic realm are incapable of further growth and development at oxygen values <134 572 µM, while organisms such as euphausiids can survive to 4.5 µM. Thus a change in the average or the 573 range of dissolved oxygen concentrations in the water column could have significant impacts on the 574 survival of certain species and consequently the species composition in the ecosystem. As compared to 575 marine vertebrates and invertebrates, the impacts of hypoxia on phytoplankton physiology and growth 576 are less known. What is well known is that large phytoplankton blooms promote oxygen loss following 577 their demise and export into the OMZ. Sea and during the winter bloom in the northern Arabain Sea (Garrison et al., 1998;Garrison et al., 612 2000). During these two periods large diatom dominated blooms occurred. However, despite the 613 emergence of Noctiluca blooms in the northern Arabian Sea, high rates of N 2 fixation occur in the 614 Arabian Sea during spring and fall seasons associated with Trichodesmium blooms in the eastern 615 Arabian Sea (Gandhi et al., 2011;Singh et al., 2019). 616 An on-board experimental study conducted by Gomes et al. (2014)

in the central and western Arabian 617
Sea during the winter monsoons of 2009, 2010 and 2011, provided the first conclusive evidence that 618 the growth of green Noctiluca blooms were being facilitated by hypoxia. Additionally, prior to their 619 appearance as surface blooms, Noctiluca were observed in large numbers at depth in association with 620 the oxycline (Goes & Gomes, 2016). In their study, Piontkovski et al. (2017) were able to show a 621 gradual descent of Noctiluca cells into the water column towards the oxycline following peak blooms 622 at the surface. More recently based on observations that showed that Noctiluca blooms of the eastern 623 Arabian Sea were not associated with hypoxic waters Lotliker et al. (2018) argued that low oxygen 624 waters were not the cause of Noctiluca blooms. Their conclusions were not backed by any 625 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. experimental data and their oxygen data were from Bio-ARGO floats that were located south of where 626 Noctiluca blooms occur. Furthermore, in a recent study, Yan et al. (2019) showed that Noctiluca 627 blooms of the eastern Arabian Sea were largely the result of advection by coherent eddy structures, 628 filaments and streamers and not actively growing as in the western and central Arabian Sea. Gomes et 629 al. (2014) posited that the capacity of the of endosymbionts Protoeuglena noctilucae present within 630 Noctiluca to photosynthesize more efficiently at low oxygen concentrations was probably linked to 631 their primitive origin when oxygen levels in Earth's atmosphere and in the oceans were much lower. 632 The exact role of Noctiluca's endosymbionts is not clearly understood, but preliminary evidence 633 suggests that when the host cell is actively grazing on other phytoplankton, microzooplankton, detritus 634 and fish eggs, the endosymbionts help reduce excessive build-up of ammonia within the central 635 cytoplasm of Noctiluca (Goes & Gomes, 2016). In addition, oxygen produced by the endosymbionts 636 helps to maintain the balance of oxygen within Noctiluca cells. 637 Noctiluca is not a preferred food for most micro-and meso-zooplankton (do Rosário Gomes et al., 638 2014). Instead, its major consumers are salps and jellyfish. Respiration rates of gelatinous zooplankton 639 in gelatinous plankton are rather low and most species belonging to this group are capable of regulating 640 their oxygen consumption allowing them to grow and survive under low-oxygen conditions. 641

Zooplankton migration 642
Knowledge of the concentrations of dissolved oxygen within the water column is also important, 643 because these concentrations can also set limits to horizontal and vertical distribution of zooplankton 644 (Saltzman et al., 1997;Wishner et al., 2008). In general, most zooplankton taxa show minimum 645 abundances in the core of the OMZ, and higher abundances in well-oxygenated waters above or 646 beneath the OMZ (Böttger-Schnack, 1996;Saltzman & Wishner, 1997;Wishner et al., 1995). There 647 are indications that several copepods are highly susceptible to low-oxygen waters that can at times lead 648 to their death (Elliott et al., 2013;Jagadeesan et al., 2013). Certain zooplankton, however, have 649 developed vertical migration strategies that enable them to pass through or even live within the OMZ 650 (Gonzalez et al., 2002;Herring et al., 1998;Longhurst, 1967). The ability to do so has been linked to 651 the presence and activity of lactic dehydrogenase (LDH), an enzyme associated with anaerobic 652 metabolism (Escribano, 2006;Gonzalez & Quiñones, 2002). Gonzalez and Quinones (2002) were also 653 able to show that the specific LDH activity within Euphausia mucronata a species capable of 654 conducting daily vertical migrations through the OMZ in the Humboldt Current upwelling system, was 655 roughly two orders higher than Calanus chilensis, a zooplankton species which restricts itself to the 656 oxygenated waters above the OMZ. In Escribano (2006), bulk zooplankton samples from within the 657 OMZ, were seen to contain very high amounts of LDH. 658 In the Arabian Sea, almost 85% of the epipelagic mesozooplankton biomass are found within the upper 659 100 m within the upper aerobic part of the seasonal thermocline. Nevertheless, the mesozooplankton 660 biomass is roughly only half of that found in areas without a pronounced OMZ (Vinogradov et al., 661 1962). Below 100 m within the anaerobic part of the seasonal thermocline zooplankton concentrations 662 decline sharply (Banse, 1994;Böttger-Schnack, 1996;Smith et al., 2005;Wishner et al., 1998). 663 Comparisons of day versus night hauls revealed that the permanent OMZ of the Arabian Sea does 664 indeed strongly suppress vertical migration of zooplankton (Smith et al., 1998), on account of their 665 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. inability to swim across the OMZ. At locations where the OMZ was forced upwards due to physical 666 processes, mesozooplankton communities were observed as narrow aggregates within the surface layer 667 (Morrison et al., 1999), where they became easily accessible to predators. 668 In the Arabian Sea there appears to be only one species, Pleuromamma indica, that has displayed the 669 ability to survive and thrive in hypoxic waters. This species is not only observed in high numbers in 670 hypoxic waters (Goswami et al., 1992;Haq et al., 1973;Saraswathy et al., 1986;Vinogradov & 671 Voronina, 1962), but is also capable of migrating daily through the well-oxygen surface layer 672 (Saraswathy & Iyer, 1986). There are also indications that the increased abundance of P. indica being 673 witnessed in recent years may be tied to the geographically more widespread oxygen depletion. 674

Implications 675
In comparison to the Arabian Sea, little is known about the plankton dynamics in the Bay of Bengal. 676 Limited data from the Bay of Bengal Process Studies (BOBPS) program suggested a diatom-dominated 677 community that contained more genera compared to the Arabian Sea (Madhupratap et al., 2003). 678 However, the regular occurrence of Noctiluca blooms and the increase in salps and jellyfish being 679 witnessed in the Arabian Sea in recent years is consistent with the idea of an ecosystem shift associated 680 with decreasing concentrations of dissolved oxygen. The emerging trophic structure fundamentally differs 681 from the traditional planktonic food web, with a reduced transfer of biomass to larger size classes and 682 fishes (Mitra et al., 2014). Similar to diatom blooms, senescent salp blooms are also exported efficiently 683 into the deep sea (Martin et al., 2017). Impacts on the export production are still difficult to predict but it 684 is likely that this will have implications for the cycling nutrients and oxygen within the seasonal 685 thermocline and the benthic community (Billett et al., 2006;Lebrato et al., 2012). 686

Benthic communities 688
Hypoxia has major consequences at the sea floor, for benthic communities and for the biogeochemical 689 processes they drive. Benthic communities and processes in the Bay of Bengal have thus far received 690 less study than those of the Arabian Sea. It is however clear that oxygen exerts an important control on 691 benthic communities across the margins of both basins (e.g, Ingole et al., 2010;Raman et al., 2015). 692 There are grain-size related contrasts in communities across the shelves, but also clear oxygen-related 693 patterns across the upper slope depth ranges where mid-water oxygen minima impinge on the sea floor 694 (Fig. 9). In the Arabian Sea, the degree to which this oxygen effect is expressed varies between 695 margins due to differing degrees of bottom-water ventilation. On the Pakistan margin, where 696 ventilation and bottom-water oxygen levels are lowest, hypoxia-resistant foraminifera are the only 697 fauna to persist at the core of the OMZ, and macro-and megafauna are totally absent (Gooday et al., 698 2009). By contrast, on the Indian margin, and even off Oman, where upwelling-driven productivity and 699 delivery of organic matter to sediments are particularly high, macrofauna generally persist across the 700 entire margin, albeit in reduced numbers and diversity at the OMZ core (e.g., Ingole et al., 2010;Levin 701 et al., 2000). Further, across the OMZ boundaries, clear "edge effects" have been observed; sharp 702 changes in community composition and faunal abundance linked to different oxygen thresholds (e.g., 703 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.

Benthic ecosystem function 706
The strong but variable cross-OMZ gradients in bottom-water oxygen and benthic communities 707 translate to contrasts in benthic ecosystem function, which also varies between margins. For example, 708 the numbers, size and depth of faunal burrows, and the extent of bioturbation and bio-irrigation, change 709 across the OMZ boundaries (e.g., Cowie et al., 2009a;Smith et al., 2000). In the extreme case, this 710 leads to total absence of bioturbation and bio-irrigation at the core of the OMZ off Pakistan, and the 711 resulting presence of annually laminated (varved) sediments, which are not observed on the better 712 ventilated margins of the Arabian Sea or in the Bay of Bengal. In the Arabian Sea, there are also clear 713 oxygen-dependent differences in benthic community organic matter processing, as have been revealed 714 by tracer incubation experiments. For example, a threshold oxygen concentration occurs, above which 715 macrofauna dominate short-term OM processing, and below which meiofauna and bacteria dominate. 716 This was illustrated on the Pakistan margin both at sites that spanned the lower OMZ boundary and at a 717 shelf-edge site that underwent strong seasonal change in bottom-water oxygen levels, from fully 718 oxygenated (intermonsoon) to hypoxic (summer monsoon) (e.g., Andersson et al., 2008;Pozzato et al., 719 2013;Woulds et al., 2009;Woulds et al., 2007). 720 Further, the "edge effect" seen in benthic community composition also has been observed in faunal 721 OM processing. At sites in the lower OMZ transition zone, the polychaete Linopherus sp. showed clear 722 morphological adaptation to low oxygen levels, and overwhelmingly dominated both the benthic 723 community and also the uptake and processing of organic matter (Jeffreys et al., 2012). These results, 724 and those of other experiments (e.g., Hunter et al., 2012;White et al., 2019), illustrate that faunal 725 assemblage composition may represent an important factor determining the pattern of seafloor 726 processing, but also the composition, bioavailability and fate of residual organic matter. It is certainly 727 clear that faunal digestive processes are recorded in the composition of organic matter deposited across 728 the margins (e.g., Jeffreys et al., 2009;Smallwood et al., 1999). In summary, oxygen-dependent cross-729 margin variability in benthic communities and ecosystem function (feeding, bioturbation and bio-730 irrigation etc) may be important contributors to the role that oxygen exposure plays in controlling 731 organic carbon distribution and burial across Arabian Sea margins, although other factors, most notably 732 hydrodynamic processes, are also important (e.g., Cowie, 2005;Cowie et al., 2009b;Koho et al., 2013;733 Kurian et al., 2018). 734

Sediment redox conditions and microbial processes 735
Alongside the contrasts in faunal communities, bioturbation and irrigation, there are cross-OMZ 736 differences in sediment redox conditions and microbial processes. Again, these are expressed to 737 varying degree on the different margins of the Arabian Sea (Cowie, 2005), and will be less apparent in 738 the Bay of Bengal due to the less intense oxygen depletion at the OMZ core. In the Arabian Sea, sulfate 739 reduction has generally been shown to be surprisingly limited in near-surface sediments (top ~50 cm) 740 (e.g., Cowie, 2005;Law et al., 2009), and redox conditions overall to be only moderately reducing 741 (e.g., Crusius et al., 1996) relative to rates observed on upwelling/OMZ margins in other basins. 742 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.
Nonetheless, Pakistan margin sediments, and possibly those on other Arabian Sea margins, are home to 743 significant rates of denitrification and anammox (e.g., Schwartz et al., 2009;Sokoll et al., 2012) and 744 authigenic phosphorous (P) burial (e.g., Filippelli GM, 2017;Kraal et al., 2012). These phenomena 745 represent important sink terms in the N and P biogeochemical cycles, and along with sediment-water 746 nutrient fluxes that vary in direction, magnitude and N:P stoichiometry across the OMZ, serve as 747 potential controls on pelagic nutrient inventories. 748 Finally, there is evidence that Pakistan margin sediments (and possibly OMZ sediments on other 749 margins), sequester important amounts of "dark" carbon arising from anammox and possibly other 750 chemoautrophic processes occurring in overlying waters or within the sediments (e.g., Cowie et al., 751 1999;Cowie et al., 2009b;Lengger et al., in press). It is a term that is currently underestimated or 752 ignored in carbon budgets and biogeochemical models. On the Pakistan margin, there are also 753 chemosynthetic bacterial mats associated with methane seeps (Himmler et al., 2018) 754

Implications 755
As mentioned above, the coastal hypoxia on the western Indian shelf can reach the extreme of fully 756 sulfidic conditions in nearshore bottom waters (e.g. Naqvi et al., 2000). Apart from mortality of 757 benthic (as well as pelagic) fauna under extreme conditions, details of the effects of seasonal hypoxia 758 on benthic communities in the shelf and coastal waters of Arabian Sea and Bay of Bengal are not well 759 documented. Thus, while seasonal contrasts in benthic community organic matter processing were 760 reported on the Pakistan shelf (see above), it is not otherwise clear if or how benthic communities have 761 adapted to the recurring, possibly intensifying, hypoxia. What is clear is that wholesale seasonal 762 changes occur in benthic microbial processes and in the magnitudes and directions of sediment-water 763 nutrient fluxes (e.g., Pratihary et al., 2014). 764 Potential benthic ecosystem and biogeochemical consequences of projected intensification and 765 expansion of hypoxia have been the subject of multiple reviews (e.g., Levin et al., 2009a;Middelburg 766 et al., 2009;Stramma et al., 2008). Intensification of hypoxia within the Arabian Sea and Bay of 767 Bengal OMZs would predictably drive distributions in benthic communities, sediment characteristics 768 and biogeochemical processes towards those currently observed off Pakistan. This would result in 769 potentially expanded depth ranges devoid of macro-and megafauna (and thus bioturbation and 770 irrigation), but also shifts in the locations and composition of "edge" populations associated with 771 oxygen gradients at OMZ boundaries. Other hypoxia-related phenomena might also impact on benthic 772 ecosystems. These include the increasing prevalence of Noctiluca and jellyfish and their potential 773 impacts on food webs and organic matter export to depth. Mass deposition of jelly fish on the seafloor 774 off Oman (Billett et al., 2006) have major impacts on seafloor communities and processes (Sweetman 775 et al., 2016). 776 It is not yet clear what the net effect of such changes would be on carbon burial, but changes in faunal 777 populations and transition from hypoxic to fully anoxic conditions could have major impacts on 778 benthic N and P cycling and sediment-water nutrient fluxes (and N:P ratios), as has been observed with 779 expanding hypoxia in the Baltic (Jilbert et al., 2011;Karlson et al., 2007). Intensification of existing 780 seasonal coastal hypoxic zones, or shoaling of upper OMZ boundaries (currently close to shelf edge 781 depth) into shelf waters, could have particularly pronounced impacts on benthic (and pelagic) fauna -782 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. with direct implications in terms of food security for large human populations -and on biogeochemical 783 processes. 784 Intensification or increased duration of coastal hypoxia could lead to increasing occurrence of mass 785 mortality or to reduced ability of faunal populations to recover between hypoxic events. It would also 786 result in expanded areas of reducing sediments and potential changes to carbon sequestration, N and P 787 cycling and N 2 O emissions ). Further, the magnitudes and the dramatic 788 intermonsoon/monsoon (oxic/hypoxic) changes in benthic processes and nutrient fluxes seen at sites on 789 the western Indian shelf (Pratihary et al., 2014), imply that expanded or intensified hypoxia could, 790 through benthic-pelagic coupling, have major influences on nutrient inventories and processes 791 occurring in shallow overlying waters. 792

Conclusion 793
Hypoxic conditions prevail in the Arabian Sea and Bay of Bengal OMZ, which allowed anaerobic 794 microorganisms to thrive and to compete against aerobic organisms. However, in contrast to the 795 Arabian Sea, in the Bay of Bengal the low oxygen concentrations suffice to support nitrite oxidization 796 to a degree that is prevented denitrification to become significant. The in comparison to the Arabian 797 Sea high freshwater fluxes and lithogenic matter supply into the Bay of Bengal might have caused this 798 difference as they influence the balance between biological oxygen consumption and physical oxygen 799 supply. Nevertheless, in the Arabian Sea and probably also the Bay of Bengal the supply of oxygen 800 sustained by mixing and advection associated with mesoscale eddies compensated the biological 801 oxygen consumption. The negative influence of decreasing oxygen concentrations on the respiration of 802 organic matter might have helped to establish these balances and counteracted a reduced oxygen supply 803 in the Arabian Sea during the last 6000 years. This was caused by climate-driven changes in the 804 ocean's circulations. Due to human induced global changes, the OMZ is expanding in the Arabian Sea, 805 and the Bay of Bengal, and hyp-as well as anoxic events occurred on the Indian shelf in both basins. 806 These trends significantly affect benthic and pelagic ecosystems. The regular occurrence Noctiluca is 807 e.g. a new phenomenon, which is assumed to herald a regime shift within the pelagic ecosystem of the 808 Arabian Sea in response to declining concentrations of dissolved oxygen. These recent changes 809 augment the problems to represent the Indian Ocean OMZ in models and thus to predict the impact of 810 the changing monsoon system on productivity and OMZ development under global change scenarios. 811 9 Author contribution 812 The paper was written jointly by all co-authors whereas Tim Rixen coordinated the writing processes 813 and co-authors focused on specific sections as listed in the following: Oxygen concentrations > 20 µM are indicated by white color. The data was obtained from the World Ocean Atlas 2013 (Boyer et al., 2013). The black line indicates the extent of the secondary nitrate maximum (SNM) in 1997 (Rixen et al., 2014). The maps were produced with Generic Mapping Tool.    Table 5 in Acharya and Panigrahi (2016).    (Garcia et al., 2013;bottom right). The model data cover the period from 1900-1999 and are taken from the 'historical' experiment. For more information on the models see Cabré et al. 2015 (Table A1). The maps were produced with MATLALB. https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.  https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. Figure 1 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License. Figure 2 https://doi.org/10.5194/bg-2020-82 Preprint. Discussion started: 6 April 2020 c Author(s) 2020. CC BY 4.0 License.