Introduction

Ocean circulation and ocean ventilation are crucial drivers of Earth’s climate system. Ocean ventilation is the process by which surface waters, recently in contact with the atmosphere, are injected into the ocean interior and transported away from the source (aging of water masses). Solubility, biological, and alkalinity pumps are the main mechanisms that foster the storage of roughly 50 times more carbon in the deep-sea compared to the atmosphere1,2, making the deep-sea the largest active Earth surface carbon reservoir. Accordingly, Quaternary glacial and interglacial variations in atmospheric CO2 concentrations have been attributed to changes in the sink and source properties of the Earth surface carbon cycle, particularly to the marine carbon cycle3,4,5. Although there appears to be some heterogeneity in the different sectors of the Southern Ocean6,7,8, there is growing evidence for the presence of an isolated carbon reservoir during the last glacial period that may have accumulated re-mineralized (respired) organic carbon, 8,9,10,11,12,13,14. In addition, it has been suggested that the deep ocean possibly absorbed 730–980 Pg of dissolved inorganic carbon (DIC) during the Last Glacial Maximum (LGM, 23–19 ka) of which one third may be accounted for by a transfer from thermocline and intermediate waters15. This demonstrates that the deep-sea carbon inventory is heterogeneous. It is well admitted that the deep-sea carbon pool did age significantly during the LGM, but there is a profound uncertainty regarding the exchange between surface and deep ocean, i.e. the role of intermediate and thermocline waters.

Reconstructions of thermocline and intermediate water ventilation are still sparse and often exhibit conflicting results14,15,16,17,18,19,20,21,22, leaving their importance in particular enigmatic. Inconsistencies in glacial and deglacial ventilation records may at least partly be the result of using foraminiferal radiocarbon dates relative to the atmosphere. Cold-water corals (CWC) have been shown to serve as a robust archive for several geochemical proxies23,24,25,26,27,28,29,30 in particular for the thermocline and deeper waters. Their aragonite skeletons contain relatively high concentrations of uranium, allowing accurate age determination through 230Th/U dating26. Combined 230Th/U and 14C measurements enable us to determine past ocean 14C/12C ratios and thus are a proxy for accurate and precise ∆14C and corresponding ventilation ages29,30. Moreover, CWC aggradations often occur near the boundaries of thermocline and intermediate waters, which are sensitive to large-scale oceanographic perturbations31.

Here we present the first ventilation ages derived from paired 230Th/U and 14C measurement on CWCs retrieved from thermocline waters (here: the permanent thermocline is the transitional zone where surface and deep waters mix) at 450 m water depth off the Maldives in the equatorial Indian Ocean32 (Fig. 1). These measurements allow for the determination of ∆14C and Benthic-Atmospheric (Batm) 14C ventilation ages for each sample (based on IntCal2033). Finally, we compare our results with glacial radiocarbon simulations applying an ocean general circulation model including ∆14C32. The Indian Ocean is only ventilated from the south and thus a cul de sac, making it an important compartment for thermohaline circulation and especially for reconstructing past ∆14C of thermocline waters originating in the Southern Oceans (Fig. 2). Our new coral data provides answers to questions regarding how variable equatorial Indian Ocean thermocline ventilations ages have been and whether they have contributed to the drawdown of carbon during the LGM.

Figure 1
figure 1

World map showing the different cold-water coral locations that have been analysed by paired 230Th/U and 14C measurements and are presented here: the equatorial Atlantic10 (750–1492 m water depth), off Tasmania35 (1430–1950 m), off W-Australia23, (675–1788 m), Drake Passage10 (750–2100 m) as well as off the Maldives (this study, 450 m). Also shown are the source regions of the Subtropical Mode Water (red) and Subantarctic Mode Water (blue) according to ref63 as well as the Suptropical Front (STF) and Subantartic Front (SAF).The map was generated with the webODV Explore64.

Full size image
Figure 2
figure 2

Modern distribution of seawater ∆14C in the Indian Ocean. Position of the chosen section is shown in the small map (left). Also shown is on the same section the salinity distribution. Data was taken from the global dataset GLODAP v2. 65,66 and plotted with webODV Explore64. The red stars indicate the sampling site of this study.

Full size image

Coupled 230Th/U and 14C of cold-water corals from the Maldives

In accordance to previous results32, our 230Th/U age determinations demonstrate that the analysed CWCs fall into a very narrow time interval close to the LGM with a minimum age of 20.413 ± 0.059 ka and maximum age of 22.096 ± 0.055 ka (Table 1s).The corresponding radiocarbon ages (calibrated against IntCal2033) are systematically older and reveal ages from 22.071 ka to 23.503 ka (Fig. 2, Table 2s). Our dataset reveals two prominent features. Firstly, Indian Ocean thermocline waters off the Maldives in 450 m water depth appear to be extremely variable with a range in the calculated ∆14C between + 109‰ and + 392‰ within a time span of less than 1.5 ka. Secondly, they are depleted compared to the IntCal2033 atmospheric 14C curve, and most of the (surface) Marine2034 curve at their corresponding calendar ages (Fig. 3). This 14C depletion of thermocline water is shown in ∆∆14C (i.e. the ∆14C difference between atmosphere and corals) values as low as − 250‰ to − 345‰ , corresponding to Batm ages of up to 2100 years. The observed variability in Batm ages cannot be related to species, but tend to cluster with higher Batm ages at the slightly deeper site (Malé Vaadhoo channel), although both are only < 100 km apart from each other and are both located on the eastern side of the Maldives.

Figure 3
figure 3

(A) ∆14C records from the equatorial Atlantic10, Tasmania35, and southwest Australia23 and from the Maldives (this study, from the shallowest water depth of 450 m). For comparison, the ∆14C record are plotted against the calibration curves IntCal2033 (black) and Marine2034 (grey), (B) Calculated ∆∆14C (∆14C against atmospheric ∆14C at the respective interval) plotted together with the above-mentioned records. Note we only plot ∆14C and ∆∆14C reconstructions that are based on paired 230Th/U and 14C analyses. Note we have taken out one data point at ~ 17.5 by ref23 that exhibits an extremely large error in ∆14C and corresponding ∆∆14C.

Full size image

Moreover, our ∆∆14C and Batm values for the Indian Ocean thermocline at the LGM are similar to those observed off Tasmania in far deeper water depths of 1430–1950 m, off SW-Australia in water depth as deep as 1788 m and from the Drake Passage (750–2100 m water depth10,23,35,36, Figs. 1, 3 and 4). Note that presently these water masses have moderately lower (− 25‰ to − 50‰ ) seawater ∆14C values as compared to the pre-bomb thermocline waters of the Indian Ocean37,38. Furthermore, the thermocline ∆∆14C values and Batm ages off the Maldives during the LGM are more depleted in radiocarbon as intermediate to deep-water masses in the equatorial Atlantic10, and are more depleted compared to modern and Holocene reservoir ages in source region of the SAMW 39,40 (Figs. 3 and 4).

Figure 4
figure 4

(A) Benthic–Atmosphere ages (ventilation ages) records from ref10,23,35 and this study at shallowest water depth of 450 m based on the calculated ∆∆14 values, calibrated against IntCal2033. Note we only plot Batm age reconstructions that are based on paired 230Th/U and 14C analyses.

Full size image

Comparison with model results

For further analysis, we simulated the temporal evolution of radiocarbon in the equatorial thermocline. Our coral based ∆∆14C (supplementary material) values and Batm ages broadly agree with the radiocarbon simulations but some discrepancies are visible (Fig. 5 and 2s). In particular, the temporal variability of simulated Batm ages is considerably smaller than reconstructed. Simulated Batm ages vary roughly from 1000 to 1500 years in the interval covered by the corals, whereas the corals exhibit a Batm range from 500 to 2300 years. Correspondingly, within an interval of less than 1.5 ka, the coral based Batm ages are outside the uncertainty bounds spanned by the various model scenarios (Fig. 5, ∆∆14C in the supplementary material).

Figure 5
figure 5

Comparison of radiocarbon model simulations and the reconstructed Batm ages. Model simulations are depicted in blue for PD (present control), yellow for GS (glacial ocean) and green for CS (glacial stadial). For details, please see text. Reconstructed ∆∆14C values are shown the supplementary material.

Full size image

This indicates that the simulations underestimate the past radiocarbon variability of the Indian Ocean thermocline. While previous 14C simulations for the LGM were roughly consistent with benthic 14C values reconstructed on other locations41, our new data from the Indian Ocean highlights that the 14C history of glacial thermocline waters is complex. Thermocline waters are at the transient zone between the surface mixed layer and the deep-ocean. Especially in the glacial ocean, where deeper waters stored additional radiocarbon depleted carbon8,9,11, the equatorial thermocline of the Indian Ocean tends to reflects both atmospheric ∆14C and deep ocean ∆14C. However, even if radiocarbon depleted but carbon rich deep-water reservoirs are a pervasive feature of the glacial ocean, high ventilation ages in near surface waters are a difficult phenomenon to explain. In the following, we consider three hypotheses that could explain the variability and depletion of 14C reconstructed for the glacial thermocline of the Indian Ocean: (1) in-situ aging, (2) advection of 14C-depleted mid-depth water masses, as well as (3) local upward mixing of 14C-depleted carbon.

In situ aging

Lowest ventilation ages recorded in our dataset plot near or in-between the Intcal2033 and Marine2034 ∆14C curves as expected for thermocline water masses that have been in contact with the atmosphere. However, the observed variability in ventilations ages suggest a strong but variable aging of thermocline waters. Can the observed radiocarbon decline be explained by an in-situ aging from an isolated thermocline water? We discount this hypothesis for the following reasons, (a) our sites here are not horizontally isolated from other ocean basins, (b) in-situ aging is at odds with the amplitude and rapidity of our reconstructed ∆∆14C variations (about 300‰ within 1500 years).

Advection of intermediate and mode waters

The decadal to centennial scale variability seen in our 14C record could be explained by the advection of southern sourced mode waters16,17. Here, the principal mechanism is the upwelling of carbon and nutrient-rich water in the Southern Ocean, which is subsequently transported to the equatorial thermocline by the Antarctic Intermediate Water (AAIW) and the Subantarctic Mode Water (SAMW)16,17. In the Equatorial Pacific, the advection of such Southern Ocean radiocarbon depleted waters was synchronous with deep-water ventilation changes22. However, even though this mechanism has been proposed for periods of abrupt climatic perturbations such as the Younger Dryas and Heinrich Stadials I and II, reconstructed ventilation ages of the intermediate northern Indian Ocean do not exhibit any larger excursions during the LGM17. Further evidence comes from a neodymium isotope based reconstructions showing, that advances of AAIW in the equatorial Indian Ocean are restricted to the deglaciation and did not occur during the LGM42.

It has been suggested that increased glacial reservoir ages could be related to decreased air-sea equilibration during the LGM43. However, the amplitude of our reconstructed ventilation changes rather supports the hypothesis of altered glacial deep-sea overturning and increased CO2 storage, as recently suggested by a comprehensive compilation of glacial deep-sea 14C records11.

Nevertheless, with the present dataset we cannot rule out that radiocarbon depleted mid-depth waters, either SAMW or AAIW, may have partly contributed to the observed variability in the thermocline ventilations ages.

Abyssal upward mixing of 14C depleted carbon

Our reconstructed variable and increased ventilation ages of thermocline waters in the equatorial Indian Ocean during the LGM can be attributed to upward mixing of deep waters. As the Indian Ocean is solely ventilated from the south44, modern Indian Deep Water (IDW) is formed from abyssal waters such as Antarctic Bottom Water via diapycnal mixing in the interior37,38,39,40,41,42,43,44,45., thereby increasing the volume of southern sourced water masses at shallower water depths (Figs. 2 and 6). Thus, abyssal upwelling controls the distribution pattern of DIC and 14C concentrations, revealing a gradual aging that ends up in the upper deep-water of the northern Indian Ocean45,46,47,48,49. Consequently, modern IDW is considered as a key supplier of carbon for the Southern Ocean upwelling37. During the last glacial period and in particular during the LGM, southern sourced waters expanded into deep and abyssal depth of the Indian Ocean, displaced the ambient Atlantic source water mass and thereby significantly increased the carbon storage capacity of the deep39,50,51,52. A replacement by a southern sourced deep-water mass could therefore be accompanied by poor ventilation and in turn by a lack of oxygen replenishment. As a water mass remains isolated from the atmosphere, 14C decays while oxygen is consumed due to oxidation of organic matter. Thus, we would expect water mass aging to be accompanied by decreasing oxygen concentrations. Indeed, there is evidence for anoxic bottom waters during the LGM in the deep Indian Ocean49,52. Accordingly, poorly ventilated deep-water masses and anoxic conditions point towards an extremely radiocarbon depleted deep-water in the abyssal and deep Indian Ocean, which may have extended into the thermocline leading to temporally very variable ventilation ages.

Figure 6
figure 6

Simplified ocean circulation sketch of the modern (a) and LGM (b) Indian Ocean. Modern circulation pattern is based on ref44,46 (top panel). The modern circulation in the Indian Ocean is characterised by inflowing Antarctic Bottom Water (AABW) at depth, and inflowing shallower North Atlantic Deep-Water (NADW, green). Towards the Indian continent, AABW diffuses upward and flows back southward as the Indian Deep Water (IDW) above the NADW. Above the IDW, southern sourced mid-depth waters such as Antarctic Intermediate Water and Sub Antarctic Mode Water penetrates into the lower latitudes. In thermocline waters (TW), surface and deep waters mix. During the LGM, our ventilation ages suggest that the AABW and the IDW episodically diffused upward and thereby potentially reduced advances of southern sources mid-depth waters and expanded into TW, although with high temporal variability. Also shown are near surface waters originating in the Northern Indian Ocean such as the Indian Equatorial Water (IEW) and the Red Sea–Persian Gulf Intermediate Water (RSPGIW)67.

Full size image

During the LGM, radiocarbon depleted but carbon rich waters have been identified in the Southern Ocean such as the Drake Passage, in the Indian Sector of the Southern Ocean as well as off Tasmania, but with Batm ages lower than < 3000 years10,35,36. Mid-depth waters tend to shoal on the pathway into the tropics44. Thus, a 14C Southern Ocean signal would be diluted with 14C enriched low-latitude surface waters during the pathway into the Indian Ocean, making it difficult to generate Batm ages of up to ~ 2100 years in the equatorial thermocline. This would in turn imply that a substantially older deep-water mass is required to explain the observed 14C depletion in thermocline waters. Very high ventilation ages near the LGM (> 4000 years) have been identified in the (SW) Pacific Ocean8,14,39, but also in the northern deep and abyssal Indian Ocean by using fossil foraminiferal ∆14C ages51,53. These extremely old deep- and abyssal water masses may thus be the most likely potential radiocarbon depleted source to cause, by upward mixing with the overlying water mass, the accumulation of 14C-depleted DIC in the equatorial thermocline of the Indian Ocean (Fig. 6).

Moreover, these radiocarbon depleted thermocline waters at the LGM may have also contributed to the deglacial release (Younger Dryas and Heinrich Stadial 1) of 14C depleted intermediate water masses in the Arabian Sea17, implying strong local differences of carbonate system characteristics.

Taken together, our new equatorial thermocline Indian Ocean 14C data points towards extensive, but variable mixing of the Indian Ocean equatorial thermocline with extremely 14C-depleted abyssal waters. Our study therefore shows that the deep Indian Ocean carbon reservoir, although temporally restricted, expanded to thermocline waters and thus contributed to the drawdown of atmospheric CO2 at the end of the last glacial period. The dynamic nature of this oceanographic phenomenon suggest that this extended carbon pool is regionally variable. Accordingly, future studies should intensively try to identify regional differences and depth constraints of carbon pool extension especially in the Indian Ocean.

Online methods

Cold-water coral samples

This study analysed scleractinian cold-water corals retrieved during research cruise SO236 to the Maldives Archpielago. In particular, coral samples were collected by a video-guided grab and a box corer in the Vaadhoo Channel (SO236-007, 04°09.07ʹN, 73°29.28ʹE, 443 m water depth) and the Kardiva Channel (SO236-017-TVG, 04°51.26N, 73°28.05ʹE, 453–457 m water depth. Initial radiocarbon datings32 revealed calibrated ages near the LGM between 22.54 and 21.4 ka. Thus, this sample set provides the unique opportunity to study ventilation ages at thermocline depth of the equatorial Indian Ocean during the LGM. Well-preserved coral skeletons (Desmophyllum pertusum, Enallopsammia rostrata and Madrepora oculata) were cleaned mechanically in order to remove potential containments (e.g., ferro-manganese coatings, borings, epibionts). Samples have been screened for their mineralogy with a PANalytical X’Pert PRO diffractometer, equipped with a copper X-ray tube revealing that all samples remained in their initial aragonitic mineralogy.

230Th/U ages determinations

Samples were chemically cleaned in a weak acid leach28. The 230Th/U measurements were carried out at the Institute of Environmental Physics at Heidelberg University (IUP, Germany) on a multi-collector inductively coupled plasma mass spectrometer (ThermoFisher, Neptune Plus)28. The reference material HU-1 was measured for the reproducibility assessment of the mass-spectrometry measurements54. Note, we assume HU-1 to be in secular equilibrium, which contrasts with observations by ref54 and causes a 1.5‰ difference in the absolute value of δ234U. For age determination this difference has no consequence, as we use the half-lives of ref54 for age determination, hence we presume a different isotopic composition for our batch of HU-1 if compared to the data published by ref54. In total, 13 samples were analysed revealing all only minor residual contaminations (232Th < 4 ppb). Nevertheless, an initial 230Th correction was applied prior to age calculations using a 230Th/232Th activity ratio for the upper thermocline waters of 8 ± 418. Age determinations and uncertainty assessment were carried out using iterative solution of the decay equations and error propagation using Monte Carlo simulations26. The initial 234U/238U activity ratios of all measured corals are, when transferred into δ234U notation (i.e., ‰ deviation from secular equilibrium), within uncertainty in a narrow band of ± 10‰ compared to the value of modern seawater (145.0 ± 1.5‰55), suggesting a closed system behaviour for the exchange of U between the skeletons and seawater.

Radiocarbon measurements

The extraction of CO2 from the CWC samples was carried out at the IUP, Heidelberg University, Germany, following the method described in56. The final iron–graphite compound was measured on an accelerator mass spectrometer (AMS, MICADAS) at the Curt-Engelhorn-Center Archaeometry (CEZA), Mannheim, Germany57,58. Calculation of ∆14C, ∆∆14C and Benthic-Atmosphere (Batm) ages 10,29,30 is based on IntCal2033.

Modelling

The radiocarbon measurements were compared with ∆14C values simulated using an enhanced version of the Hamburg Large Scale Geostrophic ocean general circulation model59; for the enhancements and implementation of ∆14C see refs.60,61 and further references therein. The model has an effective horizontal resolution of 3.5° and 22 layers in the vertical. It considers recent (PD), cold stadial (CS) and glacial (GS) climatic background conditions which result in upper and lower ocean ventilation intensities. The simulations were carried out with transient values of atmospheric ∆14C34 and pCO262 evaluated nearest to the coral sites.