What fraction of the Paciﬁc and Indian oceans’ deep water is formed in the Southern Ocean?

. In this contribution we explore constraints on the fractions of deep water present in the Indian and Paciﬁc oceans which originated in the northern Atlantic and in the Southern Ocean. Based on PO ∗ 4 we show that if ventilated Antarctic shelf waters characterize the Southern contribution, then the proportions could be close to 50–50. If instead a Southern Ocean bottom water value is used, the Southern contribution is increased to 75 %. While this larger estimate may best characterize the volume of water entering the Indo-Paciﬁc from the Southern Ocean, it contains a signiﬁcant portion of entrained northern water. We also note that ventilation may be highly tracer dependent: for instance Southern Ocean waters may contribute only 35 % of the deep radiocarbon budget, even if their volumetric contribution is 75 %. In our estimation, the most promising approaches involve using CFC-11 to constrain the amount of deep water formed in the Southern Ocean. Finally, we highlight the broad utility of PO ∗ 4 as a tracer of deep water masses, including descending plumes of Antarctic Bottom Water and large-scale patterns of deep ocean mixing,

Abstract. In this contribution we explore constraints on the fractions of deep water present in the Indian and Pacific oceans which originated in the northern Atlantic and in the Southern Ocean. Based on PO * 4 we show that if ventilated Antarctic shelf waters characterize the Southern contribution, then the proportions could be close to 50-50. If instead a Southern Ocean bottom water value is used, the Southern contribution is increased to 75 %. While this larger estimate may best characterize the volume of water entering the Indo-Pacific from the Southern Ocean, it contains a significant portion of entrained northern water. We also note that ventilation may be highly tracer dependent: for instance Southern Ocean waters may contribute only 35 % of the deep radiocarbon budget, even if their volumetric contribution is 75 %. In our estimation, the most promising approaches involve using CFC-11 to constrain the amount of deep water formed in the Southern Ocean. Finally, we highlight the broad utility of PO * 4 as a tracer of deep water masses, including descending plumes of Antarctic Bottom Water and large-scale patterns of deep ocean mixing, and as a tracer of the efficiency of the biological pump.
1 Remembering Ernst (W.B.) In 1987, Klaus Hasselmann was invited to Lamont-Doherty to present three lectures on climate. The first two dealt with what he referred to as PIPS and POPS, but they did not particularly pique my interest. However, the third one hit the mark.
In it Klaus laid out the distribution of properties generated by Ernst Maier-Reimer's ocean circulation model (Maier-Reimer and Hasselmann, 1987). I was particularly interested in its ability to reproduce the distribution of natural radiocarbon in the ocean. But the plots he showed were at first glance incomprehensible. It turned out that rather than presenting differences from the 14 C to C ratio in atmospheric CO 2 , they were referenced to that in mean ocean water. After the lecture, I offered to come to Hamburg to help Maier-Reimer switch to a mode of presentation understandable to those conversant with the 14 C measurements. And so it was that I spent three weeks with Ernst probing not only the 14 C to C distribution produced by his model but also that of O 2 and SiO 2 . For me it was a fantastic learning experience. Not only did Ernst have an amazing mind but he had a knack for teaching by tweaking his model. Thus began a lasting collaboration and friendship.

4
This led to an interest in determining the contributions of NADW and AABW to the ventilation of the deep Pacific and Indian oceans. As the ratio of O 2 utilization to PO 4 release during respiration appears to be nearly constant throughout the ocean's interior (Takahashi et al., 1985;Anderson and Sarmiento, 1994), Broecker and colleagues (Broecker et al., 1985a(Broecker et al., , 1998 proposed a conservative property PO * 4 : As only differences between PO * 4 values are of importance, the choice of the constant 1.95 is arbitrary. Hence zero would have been more convenient. Other choices for the O 2 consumption to PO 4 remineralization ratio are also possible (Hupe and Karstensen, 2000) but have little impact on our  global-scale calculations, so we stick with the formulation of Broecker et al. (1998) above.
The attraction of PO * 4 as a water mass tracer is that although the deep waters formed in the northern Atlantic range widely in temperature, all the contributors have PO * 4 values close to 0.7 (Fig. 1). Further, the deep waters (i.e., > 2000 m) in the deep Pacific and Indian oceans have PO * 4 values close to 1.4. Hence, if the PO * 4 for deep waters formed in the Southern Ocean were known, the relative amounts of deep water produced in the two key source regions could be established.
Based on PO * 4 , Broecker et al. (1998) concluded that the deep Pacific and Indian oceans received about half of their water from the northern Atlantic and half from the Southern Ocean. However, Johnson (2008), Gebbie and Huybers (2010), DeVries and Primeau (2011), and Khatiwala et al. (2012), using more complex inversions of multiple tracers and model-data synthesis, concluded that only about onequarter of this water came from the northern Atlantic.
Here we attempt to resolve this discrepancy by reexamining the PO * 4 -based approach. We show that much of the mismatch may be resolved by consideration of what "counts" as southern-sourced water. Crucial to this discussion is the extent to which deep waters acquire their tracer signatures by ventilation in the surface ocean or by entrainment during descent. Our discussion is focussed on the volumetric contribution of northern and southern water masses to the deep Indo-Pacific, rather than water fluxes. Flux information -for tracers or for parcels of water -must be informed by estimates of residence time or formation rates, and we briefly discuss the potential of radiocarbon and chlorofluorocarbons (CFCs) to provide such information.  (Johnson, 2008;Gebbie and Huybers, 2010;Khatiwala et al., 2012) suggest that these depths are predominantly a two-component mixture of deep North Atlantic and Southern Ocean waters. To help identify recently ventilated dense waters we also examined CFC-11 and neutral density. The mean PO * 4 value for deep (> 1500 m) recently ventilated (CFC-11 > 0.5 pmol kg −1 ) waters in the North Atlantic ( value is achieved if water upwelling in the Southern Ocean is cooled to the freezing point, has about 65 % of its O 2 deficiency replenished, and loses little of its PO 4 to sinking organics (see Table 1).
However, while PO * 4 values of 1.95 characterize wellventilated Antarctic shelf waters, these entrain up to 3 times their volume in circumpolar deep water as they cascade down the continental slope (Carmack and Foster, 1975;Orsi et al., 1999); indeed PO * 4 beautifully highlights this process (Fig. 4). As a result, by the time Antarctic bottom water enters the Antarctic Circumpolar Current (ACC) it has much lower PO * 4 : Weddell Sea bottom waters have PO * 4 of ∼ 1.8, and deep Ross Sea waters ∼ 1.6 (Figs. 3 and 4). This basinal difference may be attributed to less input of NADWinfluenced circumpolar deep water and higher local ventilation rates in the Weddell Sea, elevating PO * 4 in this more enclosed basin. The average circumpolar PO * 4 for recently ventilated (CFC-11 > 0.5 pmol kg −1 ) waters that have made   Figure 3. PO * 4 sections extending out from the Antarctic continent for the Weddell and Ross Seas. As can be seen, water with a value close to 1.95 is descending in a narrow margin-hugging plume.
Repeating the PO * 4 mass balance calculation of Broecker et al. (1998) with the Southern Ocean bottom water value of 1.64 suggests that the deep Indo-Pacific is filled by 75 % southern-sourced water and 25 % NADW, with an uncertainty of ±9 % (1 SD). This is within error of the values obtained by Johnson (2008), Gebbie and Huybers (2010), and Khatiwala et al. (2012). However if we use the wellventilated shelf water value of 1.95, the north-south balance is closer to 50-50 (Broecker et al., 1998). This illustrates that while the volume of what are typically considered southern deep waters in the Indo-Pacific may substantially outweigh that of NADW, much of this water is entrained in the subsurface and does not reflect full Southern Ocean ventilation. Note that while sinking waters in the North Atlantic also entrain surrounding waters on descent, these are of recent northern origin. Entrainment in the North Atlantic thus does not substantially influence the PO * 4 signature of NADW nor the inference that this water is fully ventilated in the North Atlantic.
Differences in the extent to which the Southern Ocean end member is locally ventilated may thus explain much of the difference between the north-south balance obtained by Broecker et al. (1998) versus Johnson (2008, Gebbie and Huybers (2010), and Khatiwala et al. (2012). Johnson (2008) uses bottom water end member values for AABW, so it is un-  (Warren, 1981). This is also picked out by selected depth profiles along these sections in (a), with the black dots showing a profile further out from the shelf edge. Entrainment of low-PO * 4 waters in the subsurface reduces southern deep water PO * 4 , from 1.95 on the shelf to ∼ 1.65 at depth. This can also be seen in the histograms in (c) (encompassing larger areas than those shown in the maps and sections), which show two distinct PO * 4 populations in the top 1000 m, which mix to give the more homogenous values at depth. Note that Weddell Sea waters have higher PO * 4 than Ross Sea waters, likely due to less influence of low-PO * 4 NADW and higher local deep water formation rates, elevating PO * 4 throughout this more enclosed basin. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015), with sections contoured using isopycnic gridding. surprising that our estimates using a Southern Ocean bottom water value are similar to his. Gebbie and Huybers (2010) and Khatiwala et al. (2012) use surface mixed layer conditions south of the ACC (Orsi et al., 1995), taken from gridded climatologies (WOA: Conkright et al., 1994;WOCE: Gouretski and Koltermann, 2004). As discussed by Gebbie and Huybers (2010), gridded data struggle to capture shelf features and dense overflow waters, and may thus miss some of the end member values characteristic of the ventilated Southern Ocean interior (Warren, 1981). More crucially, high adiabatic upwelling rates (Toggweiler and Samuels, 1995;Marshall and Speer, 2012) and deep mixed layers (Gordon and Huber 1990;Dong et al., 2008) may also lead to inclusion of upwelled northern waters in these Southern end members, despite little property modification in the Southern Ocean surface. These issues may explain why the southern proportions of Gebbie and Huybers (2010) and Khatiwala et al. (2012) are larger than those using the ventilated PO * 4 end member (as in Broecker et al., 1998) and lie close to our estimates using bottom water PO * 4 values. Note that we do not wish to imply that Antarctic shelf water is necessarily the most appropriate or only component to characterize ventilated Southern Ocean waters. As pointed out by Johnson (2008), shelf waters may have too unique a set of tracer properties to usefully capture the range of Southern Ocean water mass characteristics mixed into the deep Indo-Pacific. Furthermore while a 50-50 mixture between NADW and Antarctic shelf waters can produce the PO * 4 signature of deep Indo-Pacific waters, it gives too high a salinity, so more southern-sourced water is required. Processes besides shelf water formation may also help ventilate deep southern-sourced waters, such as deep winter mixing in the open Southern Ocean (Gordon and Huber, 1990;Dong et al., 2008) (Schlitzer, 2015), with sections contoured using isopycnic gridding.
The key issue that we aim to bring awareness to is discussion of what "counts" as ventilated Southern water. Implicit in the Gebbie and Huybers (2010) and Khatiwala et al. (2012) studies is that any waters reaching the Southern Ocean mixed layer may be considered Southern Ocean waters. However, these waters may experience little equilibration with Antarctic surface conditions, including cooling, gas exchange, and nutrient use, depending on their transit time through the Southern Ocean surface and the relaxation time of the tracer of interest. Therefore while they may count in the inventory of water volume entering the deep Indo-Pacific from the Southern Ocean (Talley, 2013;Marshall and Speer, 2012;Lumpkin and Speer, 2007), they may only partially reflect the exchanges of heat and CO 2 key to the Southern Ocean's role in climate (Stocker and Johnsen, 2003;Marinov et al., 2006;Barker et al., 2009;Sigman et al., 2010;Ferrari et al., 2014).

Ventilation and the radiocarbon budget
The difference between Southern Ocean water mass volume and tracer ventilation is particularly pronounced in the deep radiocarbon budget. Of the 220 mol yr −1 of 14 C undergoing radiodecay in the deep sea, about 20 mol yr −1 is resupplied by particle rain. As NADW supplies about 130 mol 14 C yr −1 , this leaves about 70 mol 14 C yr −1 to be supplied from the Southern Ocean (see Table 2). Ventilation of radiocarbon is thus dominated by the North Atlantic, even if the Southern Ocean contributes greater volume. This is due to 14 C's long equilibration time and the limited exchange time between Southern Ocean surface waters and the atmosphere. Waters upwelled into the Southern Ocean surface thus do not reach equilibrium for 14 C and radiocarbon gradients between surface and deep waters are very small (Broecker et al., 1985b). This, along with the presence of 14 C produced by H-bomb testing, also introduces large uncertainty into any attempt to use radiocarbon to quantify the contribution of Southern Ocean waters to the deep Indo-Pacific. The importance of

Constraints based on CFCs
Further insights into Southern Ocean ventilation may be obtained using CFC data, which also offer the potential to constrain flux information. As with 14 C, the degree of surface water saturation (Schlosser et al., 1991) must be carefully considered if the input flux of CFC tracer is to be converted to a ventilation flux for southern ocean water volume (England, 1995). However CFCs have the advantages over 14 C of a much larger surface to deep gradient and faster and less complicated equilibration. CFC-based estimates of the flux of ventilated Southern Ocean water give values of ∼ 15 Sv (Orsi et al., 2002;Schlitzer, 2007). This is similar to values for net production of NADW (Broecker et al., 1998;Ganachaud and Wunsch, 2000;Smethie and Fine, 2001), so it appears to support roughly equal ventilation of the deep ocean by the northern Atlantic and the Southern Ocean (Broecker et al., 1998;Peacock et al., 2000;Orsi et al., 2001). However this does not rule out a much higher water flux from the south (Sloyan and Rintoul, 2001;Lumpkin and Speer, 2007;Talley, 2013) -just not full equilibrium with Southern Ocean surface conditions. Furthermore some of the ∼ 16 Sv of NADW may be lost to mixing or entrainment into Antarctic Intermediate or Bottom Waters (Primeau and Holzer, 2006), so the ventilation flux into the deep Indo-Pacific is likely still weighted towards the South. We also note that if diffusion down isopycnals in the open Southern Ocean is an important contributor to regional ventilation (Abernathy and Ferreira, 2015), this may not be as easily picked up as the CFC signal in shelf waters (Fig. 5). The reason is that low CFC-11 concentrations in a large volume may match high CFC-11 concentrations in a small volume. Finally, even if the fluxes of northern and southern-sourced waters into the deep Indo-Pacific are similar, it is possible for southern-sourced waters to form a larger volumetric contribution if they have a longer residence time (Johnson, 2008), which is quite plausible given their injection onto deeper density surfaces.

PO * 4 and other tracers of the global overturning circulation
The large dynamic range of PO * 4 in the deep ocean makes it an effective tracer of the large scale circulation and mixing processes that make up the global overturning circulation. Its utility is highlighted by the PO * 4 sections, surfaces, and tracer-tracer plots in Figs. 6-8, supplementary Figs. A1-A6 in Appendix A, and the SOCCOM data product (Verdy and Mazloff, 2017). As can be seen, low-PO * 4 water from the North Atlantic mixes with high-PO * 4 water formed in the Southern Ocean. This mixing occurs along shared isopycnals in the ACC (Fig. 7;Abernathy and Ferreira, 2015), over rough seafloor topography (Naveira-Garabato et al., 2004;Roemmich et al., 2009), and in the deep surface mixed layer of the Southern Ocean (Gordon and Huber 1990;Dong et al., 2008). These mixing patterns are also well illustrated on cross plots of PO * 4 with salinity and potential temperature (Figs. 8 and A1-A3). NADW is identifiable as a salinity and PO * 4 maximum sandwiched between fresher and higher-PO * 4 southern waters above and below. Mixing between these northern and southern waters is well illustrated by the linear trends in the PO * 4 -salinity plot (Fig. 8). By the time circumpolar deep waters reach the Drake Passage, they have been somewhat homogenized, though a PO * 4 minimum at middepths remains, tracing the persistent influence of North Atlantic waters (Fig. 6).
Other features of interest that are well highlighted by PO * 4 include the input of very low-PO * 4 deep water from the Mediterranean Sea into mid-depths of the North Atlantic (Figs. 7 and A4); the penetration of relatively high-PO * 4 water with a strong southern influence into the deep NE Atlantic (Figs. 7, A1, A6); and the formation of mid-depth circumpo-J. W. B. Rae and W. Broecker: PO * 4 and deep water formation 3785 Figure 6. PO * 4 sections for the western Atlantic and for a series of quadrants of the Southern Ocean. Low-PO * 4 waters entering the Southern Ocean from the Atlantic and the high-PO * 4 waters generated in the Southern Ocean are blended in the Antarctic Circumpolar Current, forming circumpolar deep water. However a PO * 4 high at the seafloor and low at ∼ 2000 m continue to trace the influence of AABW and NADW respectively. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015), with sections contoured using isopycnic gridding.   . A global hydrographic section for potential temperature, salinity, PO * 4 , and silicate. Cross plots show all the data in this section with neutral density greater than 27.2 kg m −3 ; the colours of the dots refer to the scale shown to the right of the cross plots. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015), with sections contoured using isopycnic gridding. J. W. B. Rae and W. Broecker: PO * 4 and deep water formation 3789 Figure A2. Indian Ocean hydrographic section for potential temperature, salinity, PO * 4 , and silicate. Cross plots show all the data in this section with neutral density greater than 27.2 kg m −3 ; the colours of the dots refer to the scale shown to the right of the cross plots. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015), with sections contoured using isopycnic gridding. J. W. B. Rae and W. Broecker: PO * 4 and deep water formation Figure A3. Pacific hydrographic section for potential temperature, salinity, PO * 4 , and silicate. Cross plots show all the data in this section with neutral density greater than 27.2 kg m −3 ; the colours of the dots refer to the scale shown to the right of the cross plots. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015), with sections contoured using isopycnic gridding. J. W. B. Rae and W. Broecker: PO * 4 and deep water formation 3791 Figure A4. Potential temperature, salinity, PO * 4 , and silicate on the 27.6 isopycnal horizon. The depth of this horizon is shown in Fig. 8 and averages ∼ 1000 m in the basins and is in the mixed layer in the Southern Ocean. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015). Figure A5. Potential temperature, salinity, PO * 4 , and silicate on the 28.0 isopycnal horizon. The depth of this horizon is shown in Fig. 8 and averages ∼ 2500 m in the basins and ∼ 250 m in the mixed layer in the Southern Ocean. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015). J. W. B. Rae and W. Broecker: PO * 4 and deep water formation Figure A6. Potential temperature, salinity, PO * 4 , and silicate on the 28.3 isopycnal horizon. The depth of this horizon is shown in Fig. 8 and averages ∼ 4000 m in the basins and ∼ 400 m in the mixed layer in the Southern Ocean. Data are from GLODAPv2 (Key et al., 2015;Olsen et al., 2016) with profiles, maps, and sections plotted in ODV (Schlitzer, 2015).