Oxygen Minimum Zone of the Open Arabian Sea Printer-friendly Version Interactive Discussion Oxygen Minimum Zone of the Open Arabian Sea: Variability of Oxygen and Nitrite from Daily to Decadal Time Scales Oxygen Minimum Zone of the Open Arabian Sea Printer-friendly Version Interactive Discussion We 

Biogeosciences Discussions This discussion paper is/has been under review for the journal Biogeosciences (BG). Please refer to the corresponding final paper in BG if available. Abstract The oxygen minimum zone (OMZ) of the Arabian Sea is the thickest of the three oceanic OMZs, which is of global biogeochemical significance because of denitrifi-cation in the upper part leading to N 2 and N 2 O production. The residence time of the OMZ water is believed to be less than a decade. The upper few hundred meters of 5 this zone are nearly anoxic but non-sulfidic and still support animal (metazoan) pelagic life, possibly as a result of episodic injections of O 2 by physical processes. The very low O 2 values obtained with the new STOX sensor in the eastern tropical South Pacific probably also characterize the Arabian Sea OMZ, but there is no apparent reason as to why the temporal trends of the historic data should not hold. − 2 , besides temperature and salinity, made between 1959 and 2004 well below the tops of the sharp pycno-and oxyclines near 150, 200, 300, 400, and 500 m depth. We assemble nearly all O 2 determinations (originally, 849 values, 695 in the OMZ) by the visual endpoint detection of the iodometric Winkler procedure, which in our data base yields about 0.04 mL L −1 15 (∼ 2 µM) O 2 above the endpoint from modern automated titration methods. We find 632 values acceptable (480 from 150 stations in the OMZ). The data are grouped in zonally-paired boxes of 1 • lat. and 2 • long. centered at 8

We report on discrete measurements of dissolved O 2 and NO Between 1/4 and 1/3 of the total marine denitrification is estimated to occur in the water column, of which up to 1/3-1/2 may take place within the OMZ of the Arabian Sea (Codispoti et al., 2001;Bange et al., 2005). Naqvi et al. (2005 : Table 9) es- 20 timated the contribution to the global marine pelagic denitrification by the Arabian Sea to be between 8 and 21 %. The principal denitrifying zone is located in the upper 1/3 of the OMZ where O 2 concentrations as determined by the visual endpoint detection of the Winkler titration fall below 0.06 mL L −1 (about 2.7 µM; however, see below). The zone is readily identifiable by the secondary NO though, deduced a residence time of a few decades. Accepting a short residence time, temporal O 2 changes should be discernible in historical data from four decades as analyzed here. Because small O 2 shifts at the generally quite low concentrations in the OMZ may suddenly stop or start denitrification, the temporal variability of the intensity and geographical extent of this OMZ are of global biogeochemical significance. 15 After we had largely completed our analysis of the historical O 2 data, we learned of the introduction to the field of the STOX sensor (Switchable Trace amount OXygen, Revsbech et al., 2009). Its detection limit is almost two orders of magnitude lower than that of the automated Winkler titration and will necessitate a re-assessment of the intensity of the oceans' OMZs. Titration was hitherto used worldwide for O 2 studies, 20 including for the calibration of the O 2 probes attached to CTDs. Thamdrup et al. (2012) in February 2007 employed the sensor along a section from 28 • to 5 • S off the continental slope in the eastern tropical Pacific. In the OMZ core between about 100 and 300-350 m depth, O 2 was consistently undetectable on 10 of 12 stations (detection limit 0.01 µmol kg −1 or less). The authors found depletion to < 0.09 µmol kg −1 O 2 25 (∼ 0.002 mL L −1 ) to be normal in the core. Further, NO − 2 in the secondary nitrite maximum occurred only at < 0.05 µmol kg −1 O 2 (∼ 0.001 mL L −1 ), which is lower by at least an order of magnitude than the best automated endpoint detection of the Winkler titration or the colorimetric measurements (Codispoti et al., 2001). Thamdrup et al. (2012)  In answer to the second question, we do not know of a reason why the temporal trends of O 2 to be presented herein should not hold. Were it not so, the ocean-and basin-scale sections of low concentrations of dissolved O 2 currently in the literature would not closely reflect the basin-wide distribution of nutrients. Therefore, we proceed here with using our less accurate data. 5 Below, we describe first the methods and data selection for temperature, salinity, oxygen, and nitrite in subsamples drawn from the same water bottle. Then we review and update the setting of the OMZ to 500 m depth, including the small-scale spatial and temporal (days to weeks) variability in order to provide perspective on the observations. The core of the paper addresses seasonal and four-decadal changes of O 2 and NO − 2 . 10 We conclude with a section on the implications of the results.

Data sources
The principal sources for our discrete observations of O 2 and NO − 2 together with temperature and salinity were the Indian and US national oceanographic data centers  ODC and NODC, respectively). Collections on some additional cruises conducted by India's National Institute of Oceanography but not yet incorporated in the INODC and NODC bases were also utilized. All data had been taken near 150, 200, 300, 400, and 500 m, within boxes of 1 • lat. by 2 • long. The boxes are centered at 8 • , 10 • , 12 • , 15 • , 18 • , 20 • , and 21 • N along 65 • E and 67 • E (from A1 to G1 and B2 to G2, respec-20 tively, in Fig. 1). While the text and the figures refer to rounded nominal depths, the great majority of our data were collected within 5 % and in some cases within 10 % of the nominal horizons. In addition, we include some O 2 observations at exact nominal depths and the very few values from data centers that were already interpolated to them. The measurements used are listed in Supplement We identify the seasons following the US JGOFS mode (Morrison et al., 1998) but the starting dates were lagged by one-half to one month: northeast monsoon (NEM), December-March; spring intermonsoon (SI), April-May; southwest monsoon (SWM), June-September; and fall intermonsoon (FI), October-November. The lag is introduced because the biochemical response at depth depends also on the downward 5 transmission of surface signals by sinking particles.

Temperature and salinity
All temperatures and salinities accompanying the O 2 and/or NO − 2 data were taken at face value except that five hydrographic series (including the O 2 and NO − 2 values) were eliminated as occurring in exceptionally deep eddies (two series) or as clearly due 10 to pre-trips of the entire bottle strings. A very few salinity records were dropped as false by being obvious outliers in T -S diagrams. To fill in large temporal gaps, a few measurements without O 2 and/or NO

Oxygen
Our O 2 data were generated by the iodometric Winkler titration technique with visual end-point detection. This procedure was in general use until it was partially replaced by 20 automated titration, by which also the CTD-attached O 2 probes were calibrated. In turn, the introduction of the STOX sensor by Revsbech et al. (2009) bias as detailed below. Measurements by other methods, i.e., those involving automated (e.g., colorimetric or photometric) endpoint detection of the iodometric titration, were not considered, unless noted, because of systematic differences in the analytical results discussed below. Excluded thus are cruises 118 and 159 of R/V Gaveshani, as well as the observations from cruises 99 and 104 of R/V Sagar Kanya and all those 5 by R/V T. G. Thompson and other recent expeditions. To our knowledge, for avoiding interference by nitrite, sodium azide was not added on any expedition except those by R/V T. G. Thompson (Morrison et al., 1999). Almost all of our O 2 data were recorded in mL L −1 , which we did not convert to µM or µmol kg −1 , except for means and medians, because reporting the exact multiplication 10 would have added a decimal place implying a false precision, whereas rounding off might have introduced errors in means and medians of sets of often only five to ten values. Original observations expressed as µmol kg −1 were retained but also stated as converted v /v or molar units. Most data since the mid-1970s were collected by India's National Institute of Oceanog-15 raphy employing 60 mL bottles, adding 0.5 mL each of the two Winkler reagents, and analyzing with the visual starch-based endpoint detection procedure. The small amount of O 2 carried by the reagents was not subtracted and sodium azide (NaN 3 ) was not added to take care of NO − 2 interference in the Winkler titration. The lower limit of detection is approximately 0.05 mL L −1 (∼ 2 µM). The correction for blanks, if any, for the 20 other (non-Indian) data based on visual endpoints and used by us is unknown.
During the last two to three decades O 2 was analyzed with automated endpoint detection in the Winkler analysis (JGOFS manual; Anon., 1994). By comparing such measurements with visual endpoint detection in the same boxes during the same periods, we find an overestimate by 0.04 mL L −1 O 2 (∼ 2 µM) (Supplement, Sect. 2). With-25 out further evidence we generalize the difference as applicable to our entire material in The accuracy in historical O 2 data may vary for a number of reasons. For quality control we used the accompanying NO − 2 values, all accepted at face value, and eliminated O 2 values > 0.10 mL L −1 when they were accompanied by NO − 2 values > 0.2 µM. As stated above, high NO − 2 is an indicator of denitrification, and if associated with substantial O 2 it implies overestimation of the latter (see Sect. 1; also Supplement, Sect. 2). 5 The 0.2 µM contour was used by Naqvi (1991) to demarcate the boundaries of the suboxic zone in the Arabian Sea. We corrected apparent O 2 measurements < 0.10 mL L −1 accompanied by NO − 2 ≥ 0.2 µM for nitrite interference (Wong 2012) and signified these by italics in Here, all O 2 values > 0.10 mL L −1 from the OMZ were accompanied by NO − 2 > 0.2 µM, suggesting similar overestimates for the remaining determinations. Also dropped at the outset were several single O 2 values from other expeditions that appeared unreasonably high from the context, e.g., as compared to values from adjoining depths, or which came from transition layers with strong gradients. 5 The totals are 632 accepted measurements from 196 stations between 8 • and 21 • N (

Nitrite
Because the O 2 values within the OMZ are often very close to the lower limit of de-10 tection and hence perhaps not as precise as is desirable, we use NO − 2 as a surrogate of near-absence of O 2 . On all cruises nitrite was determined following Bendschneider and Robinson (1952) or variants thereof. In contrast to the O 2 observations, the data from the U.S. JGOFS and WOCE programs were also included in our study. For the five horizons under consideration, 1191 data points from 292 stations (949 from 227 15 stations in the OMZ) were utilized, all analyses having been taken at face value (Table S.1.b). These totals comprise some averages of two or three replicate casts per station, as well as means of ≥ 4 (up to 28) replicated casts within the same day or consecutive days at fixed positions. Most of the high NO − 2 values were < 5 µM except a few that exceeded this concentration (maximum 6.2 µM). One outlier of 10.2 µM is Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | medians or groups of data was assessed by the non-parametric rank test (Wilcoxon T test = Mann-Whitney U test). Because the latter mostly addressed clear differences between two data sets, one-tailed tests were usually applied. In neither approach did those values that were based on means, weight the statistics. Our statements about significance of "differences between medians" are shorthand for "tests for significance 5 of differences between two sets of independent values." The p values reflect the distribution of the variables tested, but also the often low number of samples. Because so many sets are small, reporting even p = 0.2 seems appropriate, but is not intended to serve as proof! 3 Results and discussion 10

Broad setting
The area of study lies, strictly speaking, between 7.5 • and 21.5 • N and 64 • to 68 • E ( Fig. 1), but the southern boxes (A-C, [8][9][10][11][12] • N), which are outside the suboxic OMZ, are treated only in passing. Sketching the general setting of the Arabian Sea, we note that previously the two monsoons were thought to force reversal of surface currents 15 seasonally over the entire basin. Actually, much of the western half of the Arabian Sea is full of cyclonic and anticyclonic, quasi-geostrophic mesoscale eddies and fronts with their associated meandering currents (Flagg and Kim, 1998;Shankar et al., 2002;Artamonov, 2006;Resplandy et al., 2011). The new insight is illustrated by maps of sea level anomalies (SLAs, Kim et al., 2001), sea surface height 20 Weller et al., 2002;Resplandy et al., 2011), and geopotential anomalies (Artamonov, 2006). The eddies and fronts may reach 500 m depth (e.g., Artamonov, 2006: Figs. 3.15C-F, 3.16B;Bobko and Rodionova, 2006: O 2 at 300 m in Fig. 5.6, NO of varying salinity, presumably with varying biochemical histories. The cause probably is dense water from winter convection subducted and advected horizontally, then preserved in the pycnocline (e.g., Banse and Postel, 2009). Between σ t of about 26-27 kg m −3 , similar layering is due to the intrusion from the Persian Gulf (the Persian Gulf Water, PGW, see Supplement Sect. S.3). The overall result of, especially, the hydro-5 graphic processes is substantial variability even within stations replicated during one to three days (Sect. 3.2.4) and is superimposed over marked seasonality even below the permanent thermocline (Sect. 4.1.2). Resplandy et al. (2011) in an eddy-resolving (1/12 • ) model showed the large role of vertical nutrient supply by eddies. However, as noted, e.g., by Shankar et al. (2002), the regularity of the monsoons makes features like  The study region is largely outside the strong physical and biogeochemical activity offshore of the Arabian Peninsula and near the Murray Ridge. Similarly on its eastern side, our meridional band is largely beyond the influences of the sea level changes, currents, and Kelvin waves near the Indian Subcontinent. However, during the SWM period coastal upwelling reigns in the "meso-eastern-boundary-current regime" along 20 the west coast of India, without the large eddies and offshore-drawn filaments as off the western side of the basin. Associated with the upwelling is the north-setting undercurrent, which advects O 2 poleward. Naqvi et al. (2006)  The general distribution of salinity and O 2 along 64 • E as depicted by Olson et al. (1993: Fig. 2) for 1986 is probably valid also for 65 • E. Between 200 and 500 m depth the median temperatures at each depth horizon increase with monotonous slopes by 2-3 • C, with the medians along 67 • E (boxes B2-G2;  Table 1). Naqvi et al. (1993) noted that the transitional zone is observed between these two latitudes, which during 1995 was only one degree wide. The position seems to be fairly stable in time, probably owing to the distribution of wind stress and the resulting quasi-zonal circulation (Warren, 1994; but see the section for the 20 SWM in de Sousa et al., 1996, and Supplement, Sect. S.1). The very steep horizontal O 2 gradients at the southern edge of the upper half of the OMZ in our meridional band, not described previously, are to be treated by Banse and Postel (in preparation). 25 We focus on the OMZ poleward of 12 • N. Our 150 m, uppermost horizon is below the salinity maximum of the non-seasonal pycnocline near a σ t of 24 kg m −3 and below 15468 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the sharp oxycline, which is the upper border of the OMZ. The depths of these clines vary seasonally (Colborn, 1975;Molinari et al., 1986), being deep in the SWM and, in the north, also during the NEM. The climatological depths of the bottom end of the steep vertical gradients of temperature, salinity, and oxygen occur somewhat above the 200 m horizon (see Fig. 2 as an example). Much of the OMZ has only weak vertical gradients of O 2 , but zooplankton species may be layered, perhaps responding to these gradients, as found for copepods by Böttger-Schnack (1997) and Wishner et al. (1998). as we do here.

Oxygen
The intense (suboxic) OMZ extends from north of the C-boxes (12 • N) poleward, although in September 1994 the depths at and slightly above 200 m in our Box C1 were part of this zone (R/V T. G. Thompson cruise TN039; see also the variable latitude of 20 the denitrification zone in Naqvi, 1991). In the OMZ, the median visual Winkler O 2 levels are often below 0.1 or even 0.05 mL L −1 ( north-south differences with the exception of the G-boxes (Table 1). Regarding zonal differences, the means of the annual medians of the depths in the four western boxes are all higher than those of the eastern boxes, but the ranking of the medians in Table 1 shows significant differences only at 200 and 400 m (p = 0.01 and 0.2, respectively). 5 Within the OMZ and in contrast to O 2 , Table 2 shows many significant depth differences for NO  Table 2). For the same reason, the NO − 2 medians at 300 and 400 m in Box G1 decline relative to the adjoining Box F1. The low medians at 150 m in boxes E1 and F2 and at 400 m in the D-boxes, though, hide the fact that several to many high values were present along with zero concentrations (see Table S (2006) from the central and northern Arabian Sea illustrates the role of downwelling from eddies to appreciable depths (cf. Artamonov, 2006). Denitrification to N 2 (dinitrogen) rather than the anammox route was found to be the dominant source of loss of fixed nitrogen in the Arabian Sea by Ward et al. (2009, 20 with earlier references; Bulow et al., 2010). In contrast, Lam et al. (2011) stressed the anammox pathway in the nitrogen cycling leading to N 2 losses in OMZs. Note that while anammox is an autotrophic process, both NO − 2 and NH + 4 are largely derived from the heterotrophic decomposition of organic matter and therefore dependent on the supply of the latter. Ward (2013)

Day-to-day and weekly variability
The highly dynamic nature of the hydrography creates temporal as well as spatial variability even in the central Arabian Sea at the same stations with repeated casts and is especially noticeable on week-long drift stations. For the area treated herein, previous papers emphasized temperature (isopleths, twice for four days in Box E2 (Ramesh  Kumar et al., 2001;Naqvi et al., 2002).
In Table 3 the first two examples present among-days variability for temperature, salinity, O 2 , and NO − 2 on stations near Box D1 and in Box E1, each maintained within 10 1-2 km. The standard deviations (S.D.) for O 2 are absolutely small, but may be as high as the low means. On both stations the variability of NO − 2 at 200-400 m is very large (the variance being larger than the mean), in part because of zero values but in part so even among the high concentrations, as will also be apparent in later sections. Measurements with fewer repeats on seven other locations in our data sets, including 15 US JGOFS cruises, other than the following ones show similar variability. The third set of data in Table 3 come from a drift station near 21 • N, where temporal and spatial variability change cannot be separated, but the spatial range of almost half a degree is relevant to comparisons of stations on sections with one-degree spacing (see also Supplement Fig In conclusion, a number of relatively small water masses with apparently different biochemical history are often present on the same density surfaces. This geographical heterogeneity is also to be expected in regional surveys or in interannual studies at 5 fixed locations. Therefore, data from a few stations must not be over-interpreted.

Oxygen and nitrite co-occur temporally
The observations about hydrographic variability support our assumption set forth in Sect. 1 that O 2 and NO − 2 may co-exist temporally in the OMZ. Of course, they exclude each other spatially below the O 2 threshold for the onset of denitrification of 10 < 0.002 mL L −1 (< 0.09 µM). As stated, 20 % of 646 samples contained ≤ 0.02 µM NO − 2 ; the majority showed zero or 0.01 µM NO − 2 ; the averages include NO − 2 not accompanied by O 2 data (Table S.1.b). In view of the aforementioned salinity spikes reflecting stratification (also Fig. 2), we visualize the dimensions of such patches horizontally to be much larger than vertically. On the average the patches must last many months, if not 15 a year, such that planktonic animals (e.g., copepods) live and persist in an otherwise nearly anoxic milieu of a few tens of nanomoles of O 2 (see also the large "patches" free of NO

Animal life
Metazoan zooplankton, which by its nature requires dissolved O 2 , is found year-round in this OMZ. As remarked above, NO to prevent denitrification and is sufficient to maintain a reduced number of species and specimens. The following is merely to demonstrate that the OMZ harbors resident metazoans (additional reports in Supplement, Sect. S.4). Only nighttime data are considered, which exclude the diel migrators among plankton and nekton entering the top of the 5 OMZ during daytime but which during the night in the mixed layer pay off the O 2 debt incurred at depth. Wishner et al. (1998)  The animal distribution in the upper OMZ of the Arabian Sea is not unique. Similar features, including the minima of biomass and number of species, are found in the 25 OMZ of the Costa Rica Dome of the eastern tropical Pacific, including the secondary maximum of biomass and species occurrence in the lower part of the OMZ (Saltzman and Wishner, 1997). In the same region, Vinogradov et al. (1991), using plankton hauls and visual observations from a submersible, confirmed the similarity of vertical patterns between the Arabian Sea and the Costa Rica Dome but added the large contribution of biomass by gelatinous animals, which are a part of the OMZ fauna. Jelly-like animals accounted for 92 % of wet mass and 16 % of carbon to the meso-and macroplankton in the upper 500 m, omitting very large (> 15 cm) but rare medusae and ctenophores (see also Hamner et al., 1975;Ignatyev, 2006, in Supplement, Sect. S.3). 5 In conclusion of this sub-section, even the upper OMZ of the Arabian Sea is biologically not functionally anoxic, in contrast to the suggestion by Thamdrup et al. (2012) for much of the upper OMZ in the eastern South Pacific.

Nitrate deficit
The maximum of the NO − 3 deficit resulting from denitrification is also observed in the 10 core of the OMZ, although tending to be a few tens of meters deeper than that of NO − 2 (Morrison et al., 1999: Fig. 8; the deficit was calculated by the "NO" approach, which below about 250 m depth might yield slight overestimates, see Naqvi and Sen Gupta, 1985: Fig. 2;Mantoura et al., 1993;also Chang et al., 2012). The maximal values in a season or a cruise range between about 2 and 15 µM. As illustrated by Morrison et 15 al. (1999) for 1995, the deficit declines toward depth and vanishes by 600-900 m, while NO − 2 is not observed below 350 m depth (below 600 m at one of their four stations). However, note that the presence of an NO − 3 deficit does not reflect ongoing anoxia. Rather, the deficit regularly observed in the OMZ well below the deficit maximum, not accompanied by NO − 2 , is apt to be due to mixing downward from the depth of maximum. 20 In the northeastern Arabian Sea, Naqvi et al. (1990) Rad et al., 1995) concluded that the OMZ anoxia of the Arabian Sea started near 7300 BCE, which is supported by Thamban et al. (2007) from a core taken from 500 m depth on the Indian margin at 17 • 45 N. The water below the OMZ has been oxygenated for at least the last 185 000 yr (Schenau et al., 2002). winter monsoon and the spring intermonsoon, respectively).

Temperature-salinity relations
Our T -S diagram (Fig. 3) is based on medians of discrete data and presents the climatology for the OMZ in our meridional swath. The patterns of the two upper depths down to the isopycnals (σ t ) of 26.3-26.4 kg m −3 differ strikingly from those of the 300 to At the 300 to 500 m horizons temperature and salinity increase and decrease in a periodic, isopycnal manner (Fig. 3). Both properties tend to be highest during the SI and SWM periods, as is the case for temperature also at the upper horizons. Note the 20 somewhat elevated salinities at 300 m, relative to the two deeper horizons at a density close to that of the Persian Gulf intrusion from Box C1 (12 • N) poleward. At depth, most temperature differences between the warmest and coolest seasons in the E-to G-boxes are significant, in part highly so (p = 0.01 to 0.05). The increases suggest horizontal advection, with more northerly water tending to be present during the SI 25 period. The salinity increases tend to be more pronounced in the western boxes.
For boxes D1, D2, and G1, zonal seasonal data means and medians of temperature and salinity were also drawn from the averages in the World Ocean Data Base 2001 (Levitus et al., 2002; (2012) regarding artifacts arising from averaging and interpolating of O 2 data as used in the World Ocean Data Base, which in part applies to other variables, we lean toward giving more credence to our own medians, which are simply based on discrete samples. Disregarding the small differences between our and Levitus' data, the principal con-10 clusion from Fig. 3 for the OMZ (Boxes D-G) is the evidence of significant seasonal, periodic advection along isopycnals at 300-500 m depth, most marked during the SI and/or the SWM periods. Because temperature and salinity together increase at that time, the transport direction must be from the northern to the southern quadrants. 15 The general distribution of salinity was described earlier. For the climatological change, Supplement Table S.4 provides the regression slopes for salinity on years for all boxes and depth ranges for 3-4 decades, while

Seasonal variability of oxygen concentrations
Previous studies suggested either slightly better aeration during the SWM than in the NEM seasons in the upper few hundred meters of the OMZ (Naqvi, 1987;Naqvi et al., 1990;de Sousa et al., 1996) or did not find seasonal change (Morrison et al., 1999). 5 Except for Naqvi (1987), the cited papers were based on only two cruises or (Morrison et al.) several cruises during one year and could not distinguish between seasonal and interannual changes.
With often four decades of observations for many boxes in our longitudinal band in hand, which include the above cited measurements by Naqvi and de Sousa, we treat 10 climatology. The 150 m horizon is given short shrift because of the substantial interaction with surface waters (see Fig. 3). A methodological problem in interpreting our O 2 and NO − 2 data, besides the accuracy of the O 2 analyses, is the possible compounding of seasonal with decadal changes, which we disregarded for temperature and salinity. So, for comparisons among seasons we juxtapose similar years and focus on the NEM 15 and SWM (December-March and June-September, respectively) in boxes D2 and F1 from the core of the OMZ. These boxes each offer a pair of regressions for the 200-500 m horizons for similar years in both seasons (Fig. 4, open symbols). Table 5 contains the statistics for all 18 seasonal data with ≥ 5 points (note some differences for periods between Fig. 4 and Table 5). As will be discussed in full in the next subsection, 20 all slopes are negative, i.e., oxygen was decreasing with year and significantly so (at p ≤ 0.2) in 7 out of 9 regression equations in Table 5 for Box D2.
The principal and surprising result about seasonality in the upper OMZ is that at all depths, the NEM and SI periods exhibit the highest O 2 values. All of the comparisons of medians in the 18 pairs in Table 4 are significant (at least at p = 0.2), and only in one 25 is the SWM value the higher one of the pair. We are struck by the relative regularity of our data, although they are strongly biased upward as judged by the STOX observations. The similarity of the patterns (Fig. 4) (Table 1) in spite of the great physical differences acting at the 200 m (diapycnal) and the 300 to 500 m levels (isopycnal) (Fig. 3), is truly noteworthy.
Regarding the causes of the seasonality, the high primary production of the SWM and the presumably enhanced vertical flux of particulate matter (see the sediment trap data at greater depths than our treatment near 15 • N in Ramaswamy and Gaye, 2006) 5 could be expected to lead to higher O 2 consumption and lower O 2 concentrations. Obviously, however, consumption cannot be the cause of the seasonal increase of O 2 during the NEM and SI periods, so the seasonal climatological progression of O 2 concentrations must principally be governed by advection. As noted in Sect. 4.1.2, at the 300 to 500 m levels the T -S diagram (Fig. 3) indicates isopycnal advection, 10 most conspicuously during the FI period, with the transport direction during that time apparently from north to south in view of the increase of salinity to the north. While there is a tendency to a mild increase of O 2 in the far north (G-boxes, Table 1), the meridional gradients of the overall medians are smaller than would be expected from the seasonal ratios in Tables 4 and 5, so that O 2 must also be supplied laterally. SWM periods that the temporal O 2 change from vertical eddy-driven processes was 3-5 times that from biological consumption. The authors, however, estimated an annual amplitude of 6 µM of seasonal change in the upper OMZ to be "of the order of 5 %" of the annual mean O 2 concentrations of 40 µM. The range in the core and the lower OMZ was 1-3 µM of the concentrations of 20-40 µM (op. cit., p. 5105). In contrast, we 25 find the median and mean differences of 0.08 and 0.10 mL L −1 , respectively (∼ 4.0 µM; range, 0.03-0.20 mL L −1 ) of the 14 pairs "> SWM" in Table 4 to amount to 76 and 91 %, respectively, of the 0.105 and 0.11 mL L −1 (∼ 4.5 µM) as median and mean of the 13 median O 2 concentrations for these boxes and depths ( In conclusion, in view of the strong advective processes, the apparently fairly steady maintenance of the annual balance between relatively rapid utilization and advection at the low prevailing oxygen concentrations is truly remarkable. Further, even when 5 neglecting these newly evaluated historical data it is no longer advisable to base O 2 budgets for the central Arabian Sea on data from a single or even several cruises executed during only one season.

Four-decadal changes of O 2 in the upper OMZ
Neglecting the 150 m horizon in Table 5, the slopes of the 29 seasonal regressions -10 half being significant -are overwhelmingly negative, except in the northernmost boxes F2, G1, and G2. Three slopes at depth in F2 (20 • N) are positive with very high significance, which proves that the decrease of O 2 in the bulk of our meridional swath is real rather than an artifact from an accuracy of the Winkler analyses increasing with time. Table S  period, which clearly was not the case because of O 2 advection. Further, Fig. 4 and Tables 5, 6, and S.6 imply the disappearance of O 2 during the current decades from most of the central Arabian Sea. Regrettably, the STOX probe has outdated this implication, which otherwise would have been sensational. In conclusion, the seasonal and decadal O 2 losses at depth appear to be largely replaced by advection, which, however, does not enter the OMZ directly from the south. The same pattern as for the annual regressions of medians holds also for the medians of the within-season slopes on year in the OMZ boxes with ≥ 4 values each after omitting the 1960s, which weight the regressions perhaps unduly (Fig. 4, Table 6). Especially noticeable in the table is the number of statistically significant declines in Box 10 D2, off Goa (for further details, see Supplement Sect. 6).
In summary, as with the O 2 concentrations, the rates of O 2 change on year over four decades in the central OMZ differ significantly between the two principal seasons, but O 2 does decline in both. The faster decreases are observed during the NEM with the higher O 2 concentrations, while the reverse holds for the SWM. For the two intermon-15 soon periods we have too few data to invite comments. The slopes for the two principal seasons change sign in the north.

Nitrite
The geochemical significance of the Arabian Sea principally rests on the presence or absence of denitrification and formation of N 2 (dinitrogen). Our NO − 2 observations 20 for most depths and boxes extend from 1960-1965 to the beginning of the new century. Because we study this parameter as an indicator of active denitrification, the zero or near-zero values are considered only in passing. Note that once the denitrification threshold is crossed, NO − 2 is both produced and consumed. The accumulation is determined by several factors including the availability of substrates and the composition of 25 the microbial community (Naqvi, 1991 In the Arabian Sea, denitrification to N 2 rather than the anammox route was found by Ward et al. (2009, with earlier references; also Bulow et al., 2010) to be dominant. Lam et al. (2011), in contrast, stressed the anammox pathway leading to N 2 losses in OMZs. In this OMZ the highest NO − 3 deficit as well as the excess N 2 tend to coincide with the NO − 2 maximum, so the association of NO − 2 with active denitrification seems 5 hard to argue against. While anammox is an autotrophic process, both NO − 2 and NH + 4 are largely derived from the heterotrophic decomposition of organic matter and are therefore dependent on the supply of the latter. Ward (2013) reviewed new laboratory work showing the importance of the C : N ratio of the substrate, so we believe that the debate about which is the dominant path toward N 2 in the Arabian Sea may be close 10 to solution.

Previous research in the open Arabian Sea had suggested slightly higher NO
− 2 values during the NEM than during the SWM periods (Naqvi et al., 1990;de Sousa et al., 1996). Assuming the NO − 2 concentration roughly reflects the strength of overall sub-15 oxic conditions, a more intense denitrification was inferred for the NEM. In contrast, Morrison et al. (1999) did not find consistent seasonal trends.
To avoid possible decadal bias, we compare seasons of similar intervals of years in Fig. 5 as we did with O 2 . For the observations indicating active denitrification (> 0.5 µM), Table 7 lists the significant differences between seasons in median NO − 2 levels within 20 depth ranges and boxes. As for O 2 , the results may be taken as climatological medians at least for the horizons with more than four or five samples, in contrast to the work by those earlier authors who considered only one or two cruises. Note that we mainly deal with small differences between large numbers. As with O 2 , the pattern is obscured when viewing the significant (Table 7) and non-significant pairs (not shown) together. 25 The occurrences of higher NO − 2 during the NEM suggested here are in line with the cited earlier work. In order to investigate the longer-term (interannual to decadal) changes in the intensity of suboxic conditions, we examined NO − 2 profiles at the location of GEOSECS sta. 416 (19 • 45 N, 64 • 37 E, occupied in December 1977), visited by us nine times between 1992 and 2004. The site, placed in the southwestern corner of Box F1 (see Fig. 1), 5 is close to the periphery of the suboxic zone, which makes it sensitive to changes in the volume and intensity of the reducing zone. Also it is well away from the continental margins and provides high-quality NO − 2 data going back to the 1970s. The NO − 2 data exhibit large variability of the thickness of the secondary NO − 2 maximum, ranging from ∼ 150 to ∼ 500 m, and of the peak concentrations, varying between 10 0 and 4.6 µM (Fig. 6). However, since the observations were made in different seasons it is difficult to distinguish interannual from seasonal changes, or from those caused by smaller-scale spatio-temporal variability, which is fairly large in this region as demonstrated by the aforementioned results of the quasi-Lagrangian time-series study conducted around 21

BGD
• N, 64 • E (Sect. 3.1.d; Nevertheless, it would be reasonable to conclude that interannual changes may be quite substantial as well. Most of the variability at this site seems to result from advection of PGW, which appears in Fig. S.3.1 and is a small but significant source of O 2 to the OMZ in the northwestern Arabian Sea (Codispoti et al., 2001). In the majority of our data from this 20 location, the PGW salinity maximum was associated with a minimum in NO − 2 , whereas NO − 2 maxima often coincided with the salinity minimum overlying the PGW or just below the latter feature (Fig. 6). Moreover, despite the long period covered, the NO − 2 and salinity values within the approximate σ θ range showed a significant inverse correlation (r 2 = 0.36, p = 0.001 for the slope, n = 26; Fig. 7). As is to be expected, the variability 25 of salinity is matched with that of temperature (Fig. 8). For example, the most saline PGW core occurring in July 1995 was also the warmest, when the water column did not contain measurable NO − 2 , indicating the absence of denitrification (Fig. 6) Thus, while the observations demonstrate remarkable variations in the hydrographic structure and consequently in the redox environment, the changes -probably comprising both seasonal and interannual oscillations -are irregular and do not suggest a secular change in the redox environment over the past quarter-century. That is, the GEOSECS profile ( 15 The slopes of NO − 2 concentrations of ≥ 0.5 and ≥ 1.5 µM on years for boxes and depths in the OMZ vary greatly and without an obvious pattern by either sign or value (Table S.7; representative plots in Fig. 5). In the two groupings of NO − 2 levels, positive slopes contribute 21 to a total of 29 in the first and 12 of 24 in the second group, respectively. Thus trends of increases in NO − 2 were more common than decreases. 20 The distribution between seven positive and five negative significant slopes is without any pattern. The ranges (medians) for the positive and negative slopes are 0.0253 to 0.8363 (0.0498) and −0.0168 to −0.1035 (−0.0404) µM a −1 , respectively. The median increases and decreases correspond to 2.0 and 1.6 µM over four decades. Few trends are obvious among the seasonal regressions on years ( To investigate decadal changes of NO − 2 in the OMZ in another way, the number of "zero values" (i.e., ≤ 0.2 µM), of ≥ 0.5 µM, and of ≥ 1.5 µM, each relative to the total number of NO − 2 values for a depth and box for 1985 and earlier, are compared with the data acquired since then (see Supplement, Sect. 7). The result is that denitrification increased after 1985 also as estimated in this manner. 5 In summary, more increases than decreases of NO − 2 were observed on the decadal scale among the regressions on year in spite of the great variability of the parameter, in contrast to the observations from the GEOECS site (Sect. 4.3.2). Similarly, the percentage of "zero values" increased in the eastern boxes in the second period, as did the ≥ 0.5 µM concentrations. For the ≥ 1.5 µM observations, denitrification likewise 10 intensified in the three seasons with a sufficient number of data (NEM, SI, and SWM).

Overview
The introduction into oceanography of the STOX sensor has opened a new era of reassessing all previous measurements of dissolved oxygen in the marine and freshwater 15 bodies experiencing water-column oxygen depletion. We face the possible overthrow of long-standing results, unless a reasonably accurate conversion ratio for the historic data can be found, which we regard as unlikely. In any case, the STOX probe is another example from aquatic sciences of a new technology, rather than concepts, changing the field. Among new concepts in oceanography is the large investment of the last two or 20 three decades into long-term time series observations without a specific hypothesis, which is so different from the canon of proper scientific conduct of hypothesis-driven original research (cf. Church et al., 2013).
We collate historic O 2 measurements -that may no longer be considered accurate enough -and NO probe. In effect, we assume a bias with small enough variability that will not obliterate trends in O 2 . Only time will tell whether our assumption was right, and two to three decades are required for a new, STOX-based time series. The regional and temporal coherence and consistencies of the O 2 distributions presented here, however, argue for the validity of using these seemingly outdated O 2 measurements.

5
Our NO − 2 values are not affected by the new nanomolar procedure for nitrite (Garside, 1982), because 4/5 of our samples are far above the lower limit of the traditional analytical method. In contrast, if the 1/5 of the near-zero records in our collation turn out to be too high by comparison with the new method, it would strengthen our point of the patchy occurrence of O 2 in the OMZ as preventing denitrification and supporting 10 metazoan life.
The NO − 2 levels observed since 1933-1934 by every expedition looking for this species suggest that O 2 concentrations in the OMZ have been broadly nanomolar at least since that time. The presence of metazoans, however, which was noted at least that long, shows that the OMZ of the Arabian Sea not only was not sulfidic but was not 15 functionally anoxic as stated for the tropical Pacific off South America by . Also, high NO − 3 concentrations were observed since the begin of measurement in the late 1950s. The disconnect of concentrations and temporal trends between salinity and O 2 in our meridional swath between the OMZ and the area adjoining to the south (Boxes A-C; Table 1; Sects. 4.1.3 and 4.2.2; Supplement Sect. S.1) shows that 20 O 2 is advected into the OMZ from the southwest rather than directly from the south (cf. McCreary et al., 2013;Resplandy et al., 2012). It also implies that there has not been a general decrease in oxygen in the mesopelagic realm of the Arabian Sea as a whole as reported for the eastern Pacific (Stramma et al., 2010).
In addition, we describe strong seasonality of O 2 between 200 and 500 m depth

Local time change and O 2 consumption
Any time series of O 2 in the sea determines the local (total) time change, which is 5 the sum of the x, y, z components of advection and eddy diffusion plus the biological consumption (utilization). For direct measurement of the latter, enclosing of water and following the concentration changes during incubation are necessary, which is quite difficult because of the small rates of change in OMZ water and bottle effects on the enclosed organisms (for OMZ work, cf. Jayakumar et al., 2009;Stewart et al., 2012).
However, knowing and understanding the regional distribution of seasonal change of biological consumption of O 2 and its variability is one of the holy grails of biogeochemistry of the sea. We know of four computed estimates of O 2 consumption rate for OMZs, all made prior to the appearance of STOX and hence using too high O 2 concentrations, like we 15 do for local time changes. The two earlier ones by Warren (1994) and Sarma (2002) determined the difference between the annual northward transport and the southward export of O 2 across 12 • and 10 • N, respectively, into the OMZ of the Arabian Sea and did not consider vertical or horizontal features. We mention only Warren (1994), who elaborated on a similar model by Olson et al. (1993) and estimated a consumption of these two estimates of O 2 consumption vary from each other, which may not entirely be due to regional differences in primary production and food webs. Importantly, however, Resplandy et al. (2012) reported for the NEM and SWM that the amplitude of seasonal O 2 change from vertical eddy-driven advection was several times that from biological consumption (cf. their Fig. 6, CAS region), which is opposite to the result of Stramma et 15 al. (2010). We cannot judge whether this gross divergence is due to the hydrographic settings of the two regions or the authors' approaches to the problem but note again that the O 2 utilization rate is a key parameter to know.

Decadal change of oxygen?
The origin of this paper during the late 1990s was a question about decadal variability Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | ments prior to the 1960s we cannot rule out the possibility that the decline of O 2 in the OMZ inferred from our data could be a part of the natural secular cycle.

Would global warming expand this OMZ?
Global change has reached the Arabian Sea. Since about 1960 the North Indian Ocean between about 200 and 500 m depth has warmed by up to 0.1 • C, most probably due 5 to advection (Barnett et al., 2005). Based on an updated NODC data set (cf. Levitus, 2002), Harrison and Carson (2007: Fig. 6), using essentially the same database as Barnett et al. for the central open Arabian Sea, found an increase of very approximately 0.5 • C for the period 1950-2000 at 100, 300, and 500 m depth.
In our view the answer to the question posed by the headline may depend on whether 10 the present extent of the OMZ will remain as largely due to bottom-up processes, first of all due to circulation including O 2 advection at depth across the equator. In contrast, would the overall O 2 distribution in a changed climate depend more on the rate of O 2 consumption (utilization) by particulate and dissolved organic matter (POC, DOC), which is supplied from the euphotic zone by the top-down actions in the food web?
Since according to the model by Resplandy et al. (2012) for the OMZ, advection within the basin is several times larger than consumption (see Sect showed a reversal of the processes found for the period between the 1960s and 1987, 25 suggesting a speeding up of the circulation of the subtropical gyre. Thus the sign points to variability rather than unidirectional global change, as stated by Bryden et al. (2003)  for the region near 32 • S, north of the formation region of the Mode Water and the O 2 source for the upper OMZ in the Arabian Sea. If in contrast in the future in parts of the Arabian Sea consumption is to outrank advection (as presently in the eastern equatorial Pacific, Stramma et al., 2010), then the top-down effects of changes in zooplankton on the fraction of primary production 5 that reaches the OMZ from above will become paramount (see Banse, 2013: 2.15.) This major problem is not merely figuratively sitting on top of the dependence on possible climate-related changes of circulation of regional, seasonal, and annual primary production.