Enhanced phosphorus recycling during past oceanic anoxia amplified by low rates of apatite authigenesis

Enhanced recycling of phosphorus as ocean deoxygenation expanded under past greenhouse climates contributed to widespread organic carbon burial and drawdown of atmospheric CO2. Redox-dependent phosphorus recycling was more efficient in such ancient anoxic marine environments, compared to modern anoxic settings, for reasons that remain unclear. Here, we show that low rates of apatite authigenesis in organic-rich sediments can explain the amplified phosphorus recycling in ancient settings as reflected in highly elevated ratios of organic carbon to total phosphorus. We argue that the low rates may be partly the result of the reduced saturation state of sediment porewaters with respect to apatite linked to ocean warming and acidification and/or a decreased availability of calcium carbonate, which acts as a template for apatite formation. Future changes in temperature and ocean biogeochemistry, induced by elevated atmospheric CO2, may similarly increase phosphorus availability and accelerate ocean deoxygenation and organic carbon burial.

therefore, also reflects less extensive euxinia during the former event.Specifically, for the PETM the sulfur isotope excursion is interpreted as an increase in the volume of Oxygen Minimum Zones (O2 < 20 μM) by a factor 10 to 20.For the two OAEs, the respective excursions indicate a seafloor area extent of euxinia similar to the range predicted by molybdenum isotopes.Uranium isotope records also allow for a comparison between OAE2 and the PETM, in terms of seafloor anoxia extent.For the PETM, uranium isotopes constrain the maximum extent of anoxia at 2% of the seafloor (61).While (87) estimate a similar extent for OAE2 from their uranium isotope record, (59) use new uranium isotope records to model a maximum seafloor anoxia extent of 15% for the same event.The deposition of sapropels was restricted mostly to the eastern Mediterranean and their extent can therefore not be compared easily to large-scale events such as the OAEs and the PETM.Redox conditions can, however, be compared between the different sapropels.Specifically, uranium and molybdenum isotope values for the S1 and S5 sapropels indicate that conditions during S5 were more strongly reducing than during S1 (36).During peak sapropel conditions, anoxic seafloor conditions during S5 occurred above 1000m whereas for S1 the remained well below this (31).Specifically for the S5, the chemocline likely occurred around 150-300m with proxy records indicating euxinia in the photic zone (31).For the S1, a well-document oxygenation event splits the sapropel layer in two parts (S1a and S1b) (31 and references therein).
The redox conditions for the sites discussed in this study (chosen for their availability of CORG/PTOT and other redox data) vary (see Table S11, also for references).For the T-OAE, proxy records indicate at least intermittent euxinia at all four sites.Even intermittent euxinia, taking into consideration the long duration of the two events, could mean persistence of euxinia over thousands to tens of thousands of years.During OAE2, there is evidence of photic zone euxinia for the majority of our sites.All sites, with the exception of ODP Site 1276, Wunstorf and Bass River, experienced water column euxinia and only Wunstorf and Bass River were likely sulfide free even in the sediments.At Tarfaya and Wunstorf, the deposition of black shales and the associated redox conditions varied on orbital timescales (tens of thousands of years).
Molybdenum values at four sites, DSDP Sites 367 and 530, and ODP Sites 641 and 1260, indicate permanent euxinia.We note that for the T-OAE and OAE2 molybdenum values are likely affected by basinal restriction (37,72).As a result, even when Mo values are below the 100 ppm threshold, it's possible that they did in fact experience persistently euxinic conditions.In contrast to the T-OAE and OAE2, only three PETM sites show indications of being intermittently to permanently euxinic (the Arctic, Kheu River and Guru Fatima), while photic zone euxinia occurred (intermittently or for longer periods of time) at two sites (the Arctic and IB10).All other sites had at most sulfidic porewaters.At our single i-282c site conditions were permanently euxinic with photic zone euxinia occurring as well.One of the S5 sites (PS25PC) was intermittently to permanently euxinic, while KC19C was likely at least anoxic.For S1, KC19C may have been anoxic and at MS21PC porewaters, and potentially bottom waters, were sulfidic.

PHREEQC SENSITIVITY RANGES
We used measured porewater concentrations for the modern sediments and estimated values for the ancient sediments as input for the PHREEQC calculations, as detailed below.Obviously, porewater and seawater concentrations and environmental characteristics for ancient marine systems cannot be measured directly and hence must be derived from, e.g., proxy and modelling studies.Here, we use a range of values for [PO4], pH, temperature and key solute concentrations.Where possible, these ranges are based on estimates for seawater and porewater values from earlier studies.
For the modern Arabian Sea and Black Sea, porewater concentrations were obtained through sampling and analyses in 2009 (32) and 2015 (11), respectively.For OAE2, the PETM, and the i-282c and S5 sapropels, we estimated minimum and maximum porewater concentrations of major ions from model and proxy data for bottom waters (Table S4).Note that we did not include the T-OAE because of a lack of sufficient data on seawater composition.For [PO 4 ], our range (1 μmol l -1 and 11 μmol l -1 ) captures the values used by (88) for OAE2 simulations, the low bottom water [PO4] of the modern SE Mediterranean (~2 μmol l -1 ) (51), the maximum porewater concentrations estimated for sapropel S1 based on diagenetic modelling (6.5 μmol l -1 ) (89) and the average value for modern porewaters in sediments of the euxinic basin of the Black Sea (~10 μmol l -1 ), where CFA authigenesis occurs.For OAE2, we also tested the effect of the much higher [PO4] concentrations that are found in porewaters of the Arabian Sea (80 μmol l -1 ).Our range of values is thus wide enough to account for both low [PO4] in porewaters in deep sea sediments and increased porewater [PO4] following deoxygenation.
The maximum pH values used for OAE2 (7.6) and the PETM (7.7) correspond to the estimated, average ocean pH during their respective time periods (23), while for sapropels the value corresponds to the water column pH of the modern Eastern Mediterranean (8.2) (55).The minimum pH value used for all ancient sediments is 6.9 and accounts for the generally lower pH values in porewaters (56).These ranges cover the pH values for modern marine surface sediments (top 10 cm) (23).
The chosen temperature range for OAE2 is based on proxy and model reconstructions of temperature during the event (48,49,90,91).The range of bottom water temperatures for the PETM is similar to OAE2 (92,93).Therefore we use a minimum temperature of 14˚C and a maximum temperature of 25˚C for both events.Sapropel i-282c was deposited during a period of warming (54) and the deposition of sapropel S5 corresponds to a solar insolation maximum (94).We, therefore, assign a maximum temperature of 17˚C to our sapropel calculations, which is a few degrees higher than the modern bottom value of 14˚C for the Eastern Mediterranean south of Crete (55).The full list of values and their references can be found in Table S4.
We also tested the effect of changes in alkalinity and [Ca 2+ ], [Mg 2+ ] and [SO4 2-] on the saturation index of CFA using the corresponding average ocean value for each time period (23).Overall the effects of changes in these parameters were small to negligible when compared to those of temperature, pH and [PO4].
For OAE2, an increase in [Ca 2+ ] from 17 mmol Ca kg -1 to 29 mmol Ca kg -1 , the maximum range tested here, had the largest effect of the three ion concentrations, yet we only observed an increase in SI of ~2 units.Furthermore, as ocean [Ca 2+ ] during the ancient events (>10 mmol Ca l -1 ) was higher than that in the modern Arabian Sea (~10 mmol Ca l -1 ) and Black Sea (~7 mmol Ca l -1 ), changes in [Ca 2+ ] cannot have contributed to the reduction in CFA authigenesis during past deoxygenation events.
Within the ranges tested, the effect of temperature on SI was an increase of ~13 units for OAE2 and the PETM and ~3 units for the sapropels, or an increase of roughly 1 unit per 1˚C.Similarly, the effect of a change in pH was 4.7 for OAE2, 5.3 for the PETM and ~8 for the sapropels.Within the 1 -11 μmol l -1 range for [PO4], the change in SI was equal to 5 for all ancient sediments.However, as we note in the main text, the effect of changes in [PO4] was largest at low concentrations, whereas the effect of a change in pH and temperature remained constant.This is why, even with much higher [PO4] (80 μmol l -1 ), the change in SI is only 4 units.For the sapropels, due to the larger range for pH and the smaller temperature range, the effect of [PO4] is relatively more important.However, for OAE2 and the PETM the combined effect of temperature and pH is more significant, producing lower SI values irrespective of the [PO4].

REACTION FORMULA CODE
Primary redox reactions Secondary and other reactions Tabel S6.Reaction equations implemented in the reactive transport model.

EQUATION CODE
Primary redox reaction equations

E6
Secondary redox and other reaction equations Table S7.Key environmental parameters, elemental ratios and boundary conditions used in the reactive transport model relevant to the scenario.All other settings were as in (32).

DESCRIPTION SYMBOL VALUE/EXPRESSION UNIT SOURCE
Porosity at the surface Φ0 0.
i-282c (ODP Site 969)(38) are shown in Fig.S1.Supplementary FigureS1.Calcium carbonate content (CaCO3; solid line) and CORG/PTOT values (dashed line) for sapropel i-282c sediments from ODP Site 969(38).The grey area indicates the position of the sapropel.TableS1.Complete list of sources for the data shown in: Fig.1(Corg/Ptot), Fig.2(Corg/Ptot, Fe/Al and Mo) and Supp.Fig.3 (CaCO3).Reference numbers correspond to the reference list in the main text.Note that for previously unpublished data (indicated as This study) we provide the reference number to the original study for which these were measured between double brackets.

Table S4 .
Chemical species included in the reactive transport model.

Table S8 .
Reaction parameters used in the model.
(32)2); b:(35); c: based on range for modern sediments, Fig.3; d: based on sedimentation rate for site 1B in OMZ in(32); e: no activity of benthic fauna and no oxygen in fully anoxic setting.

Table S9 .
Input values for the PHREEQC Saturation Index (SI) calculations as shown in Fig.3A.For OAE2, the PETM and the sapropels, a number of different combinations were tested, with values and SI results that fall between the ones presented in this table.For the Black Sea and Arabian Sea, calculations were performed for four depths between 0.25 cm and 32.5 cm.The average SI value for each location is shown here.Numbers between brackets refer to references as given below the table.Full references are given in the Supplementary Information.

Table S10 .
Calcium carbonate contents (CaCO3: average, maximum, minimum) for all sites presented in Fig.2.For the T-OAE, OAE2, PETM and for sapropels, the pre-event/sapropel values are also given.The values presented for the Arabian Sea and Black Sea cover the sampled depths at each station.References are listed in TableS1.

Table S11 .
(34)ilation of available redox condition indicators for the sites used in this study.The indicators presented here are used specifically to distinguish between oxic, hypoxic, anoxic and euxinic conditions, using the interpretations from the studies in which each indicator was published.The redox interpretation for molybdenum concentrations is based on(34).Note that molybdenum concentration values for the T-OAE and OAE2 may be affected by basinal restriction.