doi:.38/nature777 Carbonate dissolution in TNO57-21 There are several possible ways to reconstruct the history of carbonate dissolution in marine sediments although each has potential pitfalls, such as varying primary environmental factors e.g. 1, 2. The records of %CaCO 3 3, %>63 (coarse fraction) and foraminiferal fragmentation from TNO57-21 (Fig. S3) suggest an interval of dissolution from around 17 to 14.5kyr ago. The change from corrosive to less-corrosive bottom waters at ~14.5 kyr ago is in line with a switch from southernsourced (low [CO = 3 ]) to northern-sourced (higher [CO = 3 ]) deep waters over this site at this time, as described by Piotrowski et al. 4 ). %CaCO 3 and % coarse fraction are sensitive to primary changes in the dominating organisms (i.e. the relative contributions from e.g. diatoms, coccolithophores and foraminifera), which evidently varied during the deglacial period. Although different species of foraminifera may also fragment to differing degrees e.g. 5, 6 we believe that the assemblage is varied enough throughout the entire record so as not to affect the overall tendency to fragment. We therefore interpret the fragmentation record to reflect changes in dissolution at this site (and in particular foraminiferal dissolution). The record of foraminiferal fragmentation in TNO57-21 reveals that dissolution is significant although it should be noted that the definition of fragmentation used here is given by: fragments % fragments = x fragments + whole shells (S1) Thus a value of fragmentation of 7% does not mean that 7% of the original shells have fragmented since any shell will break into more than one fragment. This issue was addressed by Le and Shackleton 7, who suggested an alternative definition of fragmentation: fragments/ x % fragments = x fragments/ x + whole shells (S2) 1
doi:.38/nature777 Where x represents the typical number of fragments per shell (in their work this was assigned a value of 8). Thus a value of fragmentation of 7% by use of equation S1 would translate to a value of ~23% by use of S2 if x = 8 (implying that about 23% of the original shells had fragmented). Since the core is from a drift site it has very high sedimentation rates (>cm/kyr) and we can be confident that our planktonic 14 C ages are not influenced by fragmentation 8. The interval of most intense dissolution occurs within the warm interval defined as HS1 (see main text). Although dissolution and fragmentation can affect planktonic faunal assemblages 5, 6 we do not believe that the main features in our faunal records are an artefact of dissolution (the same argument could be used to suggest that fragmentation is not primarily driven by faunal changes). For example, the decrease in polar species during HS1 (Fig. S4A) is not aligned with the change in fragmentation. Furthermore, since N. pachyderma (s) (the dominant polar species in TNO57-21) has a robust shell 5, fragmentation would tend to increase its relative abundance in the assemblage. Warmer species on the other hand tend to have more fragile shells and might be expected to fragment earlier through dissolution. Hence we might expect a decrease in these species during HS1 if dissolution were a dominating control. G. ruber in particular has a very fragile shell 5 but reaches its maximum abundance in TNO57-21 during the peak in fragmentation (Fig. S4B). This may in fact lead us to suggest that the abundance of warm species during HS1 was originally higher than we observe. Although benthic foraminiferal shells tend to be more resistant to dissolution and fragmentation than planktonics e.g. 1 we do not believe that the peaks in benthic forams (within HS1 and HS2) are an artefact of dissolution. Our reasoning is twofold: the peaks in benthic abundance do not coincide with peaks in fragmentation, for example the rapid increase in benthic shells during HS1 occurs while fragmentation is relatively constant and more than years before the large increase in fragmentation at ~17 kyr ago). Secondly, E. exigua is a rather small and thin-shelled foraminifera. Since the benthic peaks coincide with an increase in relative abundance of E. exigua we believe these peaks are not driven by dissolution but instead reflect changes in surface water productivity. The unexpectedly low values of foraminiferal Mg/Ca during HS1 (Fig. S4C) are precisely aligned with the interval of increased fragmentation within HS1. This is likely to be a result of partial dissolution e.g. 9 (Fig. S5). We believe that it is more preferable to adjust the record of Mg/Ca for the effects of dissolution rather than simply discard the interval of intense fragmentation during HS1. In order to adjust Mg/Ca records for the effects of dissolution it is necessary to quantify the extent of dissolution within each sample (since we know dissolution can vary through time). Various approaches have been described 9-12. Those involving water depth 11 (which is essentially invariant 2
doi:.38/nature777 through time) are not ideal as we know dissolution does vary temporally. Normalising to a constant foraminiferal shell-weight may be complicated by variability in initial shell weight through time 13. Mekik et al., 12 have suggested using a fragmentation index based on a single species (G. menardii) to correct for the effects of dissolution on Mg/Ca. Since we do not have a species that is both easy to recognise once fragmented and present throughout the whole record, we are prevented from using a single species approach. We therefore use our record of fragmentation to assess the effects of dissolution on Mg/Ca. We used an empirical calibration of Mg/Ca loss (unpublished data from H. Johnstone et al. manuscript in prep) versus fragmentation 6 in surface sediments from the Ontong Java Plateau (OJP), West Equatorial Pacific (Fig. S6), to adjust our down-core Mg/Ca ratios in TNO57-21 (Fig. S7) (we use a consistent definition of fragmentation throughout). We do not have a suitable depth transect of core-top samples in the region of TNO57-21 to produce a more ideal calibration. Instead we test the sensitivity of the adjustment to various linear response curves of Mg/Ca loss versus fragmentation (Fig. S7B). The adjustment procedure causes a recovery of warmer temperatures (increased Mg/Ca) during the interval of most intense dissolution although all Mg/Ca ratios are increased as a consequence of the linear calibration employed. Given the uncertainty in our empirical calibration, we have used the best linear fit through the core-top data ( %Mg/Ca loss = - 2.75 +.5 * %fragmentation) for the adjustment shown in the main text (Fig. 2). Even with the adjustment, there is still considerable scatter of Mg/Ca within HS1. We believe that this scatter is most likely to be an artefact of dissolution since it is somewhat reduced by the adjustment. It is possible, given the strong influence of dissolution during this interval, that the method of adjustment is imperfect. Importantly, the adjustment procedure does not cause significant changes to the overall structure of the Mg/Ca record (except during the significant changes in fragmentation which occur either side of the dissolution peak). In combination with the recovery of warmer temperatures during HS1 this gives us confidence that our general approach is fairly robust. In any case our general conclusions are not critically dependent on the adjustment procedure. Diatom counts We have performed diatom counts on TNO57-21 (Fig. S8D). Apart from a general decrease during the termination, also reflected in the decreased IRD/iceberg transport rate to TNO57-21 14 (Supp. 3
doi:.38/nature777 Fig. S8B), we see an increase in the abundance of Hyalochaete Chaetoceros ssp. resting spores (CRS) during HS2. In the modern Southern Ocean/South Atlantic CRS are typically only found in sediments close to the Antarctic Peninsula, the southern margin of the Falklands Plateau and the Argentine Basin, reflecting the neritic habitat of this sub-genus 15, 16. A tongue of increased abundance east of these locations reflects lateral transport by the ACC but the sub-genus is quite rare in the open ocean east of ~ºW. Two records from the south of TNO57-21 reveal a pulse in the abundance of CRS coincident with Antarctic warming during early TI (i.e. HS1) 17. This was ascribed to the presence of increased meltwater as a result of sea-ice retreat at this time 17 but we would suggest that increased transport along the ACC could give rise to a similar observation. We interpret our observation of increased CRS during HS2 to reflect enhanced transport along a vigorous ACC. The absence of a peak in CRS abundance during HS1 probably reflects the overall deglacial warming and the replacement of diatoms with carbonate producers at the site of TNO57-21 (Supp. Fig. S8E). References 1. Peterson, L. C. & Prell, W. L. Carbonate Dissolution in Recent Sediments of the Eastern Equatorial Indian-Ocean - Preservation Patterns and Carbonate Loss above the Lysocline. Marine Geology 64, 259-29 (1985). 2. Barker, S. in Encyclopedia of Quaternary Science (ed. Elias, S. A.) 1711 (Elsevier, 26). 3. Sachs, J. P. & Anderson, R. F. Increased productivity in the subantarctic ocean during Heinrich events. Nature 434, 1118-1121 (25). 4. Piotrowski, A. M., Goldstein, S. L., Hemming, S. R. & Fairbanks, R. G. Temporal relationships of carbon cycling and ocean circulation at glacial boundaries. Science 7, 1933-1938 (25). 5. Berger, W. H. Planktonic foraminifera: selective solution and the lysocline. Marine Geology 8, 111-138 (197). 6. Le, J. N. & Thunell, R. C. Modelling planktic foraminiferal assemblage changes and application to sea surface temperature estimation in the western equatorial Pacific Ocean. Marine Micropaleontology 28, 211-229 (1996). 7. Le, J. & Shackleton, N. J. Carbonate dissolution fluctuations in the western Equatorial Pacific during the late Quaternary. Paleoceanography 7, 21-42 (1992). 8. Barker, S., Broecker, W., Clark, E. & Hajdas, I. Radiocarbon age offsets of foraminifera resulting from differential dissolution and fragmentation within the sedimentary bioturbated zone. Paleoceanography 22 (27). 9. Barker, S., Cacho, I., Benway, H. & Tachikawa, K. Planktonic foraminiferal Mg/Ca as a proxy for past oceanic temperatures: a methodological overview and data compilation for the Last Glacial Maximum. Quaternary Science Reviews 24, 821-834 (25).. Rosenthal, Y. & Lohmann, G. P. Accurate estimation of sea surface temperatures using dissolution-corrected calibrations for Mg/Ca paleothermometry. Paleoceanography 17, doi:.29/21pa749 (22). 4
doi:.38/nature777 11. Dekens, P. S., Lea, D. W., Pak, D. K. & Spero, H. J. Core top calibration of Mg/Ca in tropical foraminifera: Refining paleotemperature estimation. Geochemistry Geophysics Geosystems 3, U1-U29 (22). 12. Mekik, F. & Francois, R. Tracing deep-sea calcite dissolution: Agreement between the Globorotalia menardii fragmentation index and elemental ratios (Mg/Ca and Mg/Sr) in planktonic foraminifers. Paleoceanography 21 (26). 13. Barker, S. & Elderfield, H. Foraminiferal calcification response to glacial-interglacial changes in atmospheric CO 2. Science 297, 833-836 (22). 14. Kanfoush, S. L. et al. Millennial-scale instability of the antarctic ice sheet during the last glaciation. Science 288, 1815-1818 (2). 15. Crosta, X., Pichon, J. J. & Labracherie, M. Distribution of Chaetoceros resting spores in modern peri-antarctic sediments. Marine Micropaleontology 29, 283-299 (1997). 16. Zielinski, U. & Gersonde, R. Diatom distribution in Southern Ocean surface sediments (Atlantic sector): Implications for paleoenvironmental reconstructions. Palaeogeography Palaeoclimatology Palaeoecology 129, 213-25 (1997). 17. Bianchi, C. & Gersonde, R. Climate evolution at the last deglaciation: the role of the Southern Ocean. Earth and Planetary Science Letters 228, 47-424 (24). 18. Sachs, J. P. & Anderson, R. F. Fidelity of alkenone paleotemperatures in southern Cape Basin sediment drifts. Paleoceanography 18, art. no.-82 (23). 19. CLIMAP. in Map and Chart Series MC-36 1-18 (Geological Society of America, 1981). 2. Barker, S., Greaves, M. & Elderfield, H. A study of cleaning procedures used for foraminiferal Mg/Ca paleothermometry. Geochemistry Geophysics Geosystems 4, doi.29/23gc559 (23). 21. Barker, S. Planktonic foraminiferal proxies for temperature and pco 2, PhD thesis thesis, 136 pp, University of Cambridge (22). 22. Stuiver, M. & Grootes, P. M. GISP2 oxygen isotope ratios. Quaternary Research 53, 277-283 (2). 23. EPICA. One-to-one coupling of glacial climate variability in Greenland and Antarctica. Nature 444, 195-198 (26). 24. Spahni, R. et al. Atmospheric methane and nitrous oxide of the late Pleistocene from Antarctic ice cores. Science 3, 1317-1321 (25). 25. Parrenin, F. et al. The EDC3 chronology for the EPICA dome C ice core. Climate of the Past 3, 485-497 (27). 26. Loulergue, L. et al. New constraints on the gas age-ice age difference along the EPICA ice cores, -5 kyr. Climate of the Past 3, 527-54 (27). 27. Andersen, K. K. et al. A 6 year Greenland stratigraphic ice core chronology. Clim. Past Discuss. 3, 1235-126 (27). 28. Monnin, E. et al. Atmospheric CO2 concentrations over the last glacial termination. Science 291, 112-114 (21). 29. Stuiver, M. & Reimer, P. J. Extended 14 C data-base and revised Calib 3. 14 C age calibration program. Radiocarbon 35, 215-2 (1993).. Hughen, K. A. et al. Marine4 marine radiocarbon age calibration, -26 cal kyr BP. Radiocarbon 46, 59-86 (24). 31. Fairbanks, R. G. et al. Radiocarbon calibration curve spanning to 5, years BP based on paired Th-2/U-234/U-238 and C-14 dates on pristine corals. Quaternary Science Reviews 24, 1781-1796 (25). 32. Charles, C. D., Lynch-Stieglitz, J., Ninnemann, U. S. & Fairbanks, R. G. Climate connections between the hemispheres revealed by deep sea sediment core ice core correlations. Earth and Planetary Science Letters 142, 19-27 (1996). 5
doi:.38/nature777 Table S1. 14 C dates and age model for TNO57-21. Depth (cm) Depth (cm) Species 14C age error Calib. (a) Cal. Age Lower Upper Comment (b) TNO57-21 RC11-83 (yr) (yr) (yr) (yr) (yr) -5 G. bull 9,77 45 1,439,386,56 OS-64528 1-115 G. bull, 5 1 11,128 11,91 11,177 OS-64527 12-125 G. bull,9 45 1 12,73 11,96 12,114 OS-64533 1-135 G. bull 12,2 75 1 13,439 13,335 13,59 OS-64534 14-145 G. bull 12,7 55 1 13,946 13,869 14,15 OS-64535 15-155 G. bull 12,95 55 1 14,296 14,115 14,432 OS-64536 16-165 G. bull 13,15 7 1 14,712 14,574 14,943 OS-64523 18-185 G. bull 14,75 65 1 16,869 16,668 17,79 OS-64632 19-195 G. bull 14,85 6 1 17,14 16,798 17,229 OS-62388 2-25 G. bull 15,65 8 1 18,343 18,135 18,583 OS-64633 2-215 G. bull 16,5 65 1 18,78 18,722 18,84 OS-64634 22-225 G. bull 16,7 6 1 19,277 19,186 19,388 OS-64635 2-235 G. bull 16,9 8 1 19,451 19,371 19,522 OS-64636 24-245 G. bull 17,4 7 1 19,948 19,86 2,2 OS-64524 25-255 G. bull 18, 85 1 2,533 2,376 2,669 OS-64637 26-265 G. bull 18,2 85 1 2,763 2,556 2,916 OS-64522 27-275 G. bull 18,4 8 1 21,15 2,816 21,26 OS-64499 29-295 G. bull 19, 9 1 22,267 22,18 22,353 OS-64521-5 G. bull 19,6 9 1 22,496 22,397 22,585 OS-64519 7 189 G. bull 9,99 5 1,646,57,7 C96 128 221 G. bull 11,9 5 1 13,25 13,158 13,254 C96 135 231 G. bull 12,12 5 1 13,353 13,299 13,4 C96 147 248 G. bull 12,9 5 1 14,19 14,6 14,275 C96 156 262 G. bull 13,2 5 1 14,888 14,747 15,19 C96 18 298 G. bull 14,3 5 1 16,324 16,116 16,55 C96 25 335 G. bull 15,62 5 1 18,32 18,133 18,545 C96 256 426 Mixed 17,7 1 2,22 2,91 2,344 C96 285 477 Mixed 19,6 1 22,498 22,39 22,593 C96 334 546 Mixed 21, 1 24,852 24,575 25,29 C96 44 644 Mixed 24, 2 28,172 27,991 28,353 C96 484 756 Mixed 28,4 2 2 33,155 32,893 33,417 C96 55 844 Mixed 32, 2 2 36,77 36,522 37,18 C96 616 943 Mixed 35,8 2 4,534 4,25 4,863 C96 Notes: (a) Calibration was performed using either (1) Calib 29 v.5..1 with the Marine4 calibration curve and R = 2 yr or (2) Fairbanks et al. 31 with a reservoir age of 6 yr. (b) NOSAMS accession number (OS), C96 = Charles et al. 32. Calendar age range is 1. 14 C errors for the Charles et al. 32 data are estimated. Shaded values were not used in the age model for TNO57-21. 6
Figure S4. (a) Polar species in TNO57-21). (b) total warm species (upper purple curve) and G. doi:.38/nature777 Depth RC11-83 (m) 2 4 6 8 7 a TNO57-21 6 RC11-83 7 6 b TNO57-21 RC11-83 %CaCO 3 5 4 %CaCO 3 5 4 2 2 2 cm 1 2 3 4 5 6 Depth TNO57-21 (m) 1 2 3 4 5 6 Depth (m, TNO57-21 scale) Calendar age (kyr) 4 35 25 2 15 + TNO57-21 (MarineCal4) RC11-83 (MarineCal4) RC11-83 (Fairbanks5) TNO57-21 Age Model Calendar age (kyr) 25 2 15 + TNO57-21 (MarineCal4) RC11-83 (MarineCal4) RC11-83 (Fairbanks5) TNO57-21 Age Model 5 c Supplementary 1 2figure 3annotations 4 5 6 Depth TNO57-21 (m) 1 2 3 4 Depth TNO57-21 (m) Figure S1. Age model derivation. (a) Records of %CaCO 3 from TNO57-21 18 and RC11-83 19 contain significant structure for tuning. (b) Both records on a common depth scale. Tuning points are represented by vertical arrows. (c) Calibrated 14 C ages versus depth in TNO57-21. (D) Age model used in this study. Vertical error bars in parts (c) and (d) represent the 1 uncertainty in the calibrated calendar age (see Table S1). Horizontal error bars represent the sampling interval for TNO57-21 (5cm) and the correlation error for RC11-83 (±cm) respectively. d Figure S2. Quality control for Mg/Ca measurements. Fe/Mg can be used to monitor the potential influence of silicate contamination in foraminiferal Mg/Ca ratios 2. Fe/Mg >.1 suggests contamination may be significant. Here we reject a single sample which has a particularly high Mg/Ca ratio and Fe/Mg >.4. Two other samples with Fe/Mg >.1 do not appear to have anomalously elevated Mg/Ca ratios. Figure S3. Dissolution in TNO57-21. Records of %CaCO 3 3, coarse fraction (%>63 m) and foraminiferal fragmentation reveal an interval of more intense dissolution between ~17 to ~14.5 kyr ago. 7
doi:.38/nature777 2.5.5 Mg/Ca (mmol/mol) 2 1.5 1 Mg/Ca Fe/Mg.4.3.2.1 Fe/Mg (mol/mol).5 1 1.5 2 2.5 3 3.5 Core depth (m) Figure S2. Quality control for Mg/Ca measurements. Fe/Mg can be used to monitor the potential influence of silicate contamination in foraminiferal Mg/Ca ratios 2. Fe/Mg >.1 suggests contamination may be significant. Here we reject a single sample which has a particularly high Mg/Ca ratio and Fe/Mg >.4. Two other samples with Fe/Mg >.1 do not appear to have anomalously elevated Mg/Ca ratios. 8
doi:.38/nature777 % CaCO 3 % fragments 12 14 16 18 2 22 24 26 7 6 5 4 2 Supplementary figure annotations Figure S1. Age model derivation. (a) Records of %CaCO 3 from TNO57-21 18 and RC11-83 19 contain significant structure for tuning. (b) Both records on a common depth scale. 4Tuning points are represented by vertical arrows. (c) Calibrated 14 C ages versus depth in TNO57-21. (D) Age 2 model used 2 in this study. Vertical error bars in parts (c) and (d) represent the 1 uncertainty in the calibrated calendar age (see Table S1). Horizontal error bars represent the sampling interval for TNO57-21 (5cm) and the correlation error for RC11-83 (±cm) respectively. 4 Figure S2. Quality control for Mg/Ca measurements. Fe/Mg can be used to monitor the potential 5 influence of silicate contamination in foraminiferal Mg/Ca ratios 2. Fe/Mg >.1 suggests contamination 6 may be significant. Here we reject a single sample which has a particularly high Mg/Ca ratio and Fe/Mg >.4. Two other samples with Fe/Mg >.1 do not appear to have anomalously elevated Mg/Ca ratios. 7 12 14 16 18 2 22 24 26 Age (kyr) Figure S3. Dissolution in TNO57-21. Records of %CaCO 3 3, coarse fraction (%>63 m) and foraminiferal fragmentation reveal an interval of more intense dissolution between ~17 to ~14.5 kyr ago. 12 8 6 % > 63 µm Figure S4. (a) Polar species in TNO57-21). (b) total warm species (upper purple curve) and G. ruber (lower red curve). (c) Mg/Ca ratios of G. bulloides. (d) Fragmentation. Figure S5. Dissolution can cause a significant decrease in Mg/Ca of G. bulloides. (a) A core-top transect of Mg/Ca ratios in G. bulloides and G. inflata from the North Atlantic reveals significant scatter 21. Particularly low values are found in deep water locations (b), indicating the effect of dissolution on Mg/Ca. (c) Removal of the latitudinal trend (a function of temperature) from the core-top data reveals a trend of decreasing Mg/Ca with increasing water depth (and therefore dissolution) for deeper samples. In this case dissolution can cause a ~4% decrease in Mg/Ca of G. bulloides. Hatched region is approximate depth of saturation horizon in this region. (d) The scatter seen in part (a) seems common to both species and probably reflects differing states of undersaturation within sedimentary pore-waters. (e) A laboratory leaching experiment demonstrates the effect of partial dissolution on Mg/Ca in G. bulloides 21. In this experiment 4 individual tests of G. bulloides were selected and cleaned. Before dissolution, each sample was split into two. The first half was dissolved and analysed, the second was subjected to multiple dilute acid leaches in order to induce partial dissolution. Treatment was continued until roughly half the remaining sample had 9
doi:.38/nature777 a % warm species 7 6 5 4 3 2 1 b 2 4 5 % polar species c Supplementary figure annotations Figure S1. Age model derivation. (a) Records of %CaCO 3 from TNO57-21 18 and RC11-83 19 contain significant structure for tuning. (b) Both records on a common depth scale. Tuning points are represented by vertical arrows. (c) Calibrated 14 C ages versus depth in TNO57-21. (D) Age model used in this study. Vertical error bars in parts (c) and (d) represent the 1 uncertainty in the calibrated calendar age (see Table S1). Horizontal error bars represent the sampling 1 interval for TNO57-21 (5cm) and the correlation error for RC11-83 (±cm) respectively. 2 d Figure S2. Quality control for Mg/Ca measurements. Fe/Mg can be used to monitor the potential influence of silicate contamination in foraminiferal Mg/Ca ratios 2. Fe/Mg >.1 suggests % fragments contamination 4 may be significant. Here we reject a single sample which has a particularly high Mg/Ca ratio and Fe/Mg >.4. Two other samples with Fe/Mg >.1 do not appear to have 5 anomalously elevated Mg/Ca ratios. 6 Figure S3. Dissolution in TNO57-21. Records of %CaCO 3 3, coarse fraction (%>63 m) and foraminiferal 7 fragmentation reveal an interval of more intense dissolution between ~17 to ~14.5 kyr 12 14 16 18 2 22 24 26 ago. Age (kyr) 2 Mg/Ca (mmol/mol) Figure S4. (a) Polar species in TNO57-21). (b) total warm species (upper purple curve) and G. ruber (lower red curve). (c) Mg/Ca ratios of G. bulloides. (d) Fragmentation. Figure S5. Dissolution can cause a significant decrease in Mg/Ca of G. bulloides. (a) A core-top transect of Mg/Ca ratios in G. bulloides and G. inflata from the North Atlantic reveals significant scatter 21. Particularly low values are found in deep water locations (b), indicating the effect of dissolution on Mg/Ca. (c) Removal of the latitudinal trend (a function of temperature) from the core-top data reveals a trend of decreasing Mg/Ca with increasing water depth (and therefore dissolution) for deeper samples. In this case dissolution can cause a ~4% decrease in Mg/Ca of G. bulloides. Hatched region is approximate depth of saturation horizon in this region. (d) The scatter seen in part (a) seems common to both species and probably reflects differing states of undersaturation within sedimentary pore-waters. (e) A laboratory leaching experiment demonstrates the effect of partial dissolution on Mg/Ca in G. bulloides 21. In this experiment 4 individual tests of
doi:.38/nature777 5 4 G. inflata G. bulloides Mg/Ca (mmol/mol) 3 2 Water depth (m) 2 4 1 6 a 55 5 45 4 Latitude (ºN) 35 5 6 6 b 55 5 45 4 Latitude (ºN) 35 2 25 G. bulloides G. inflata - -2 Water depth (m) 35 4 45 5 55 5 4 2 '% Mg/Ca loss' c '%Mg/Ca loss' bulloides - 2 4 5 5 4 2 - '%Mg/Ca loss' inflata d -2 - Mg/Ca G. bulloides (leached) 3. 2.6 2.2 e 1.8 1.8 2.2 2.6 3. Mg/Ca G. bulloides (un-leached) Figure S5. Dissolution can cause a significant decrease in Mg/Ca of G. bulloides. (a) A core-top transect of Mg/Ca ratios in G. bulloides and G. inflata from the North Atlantic reveals significant scatter 21. Particularly low values are found in deep water locations (b), indicating the effect of dissolution on Mg/Ca. (c) Removal of the latitudinal trend (a function of temperature) from the 11
doi:.38/nature777 core-top data reveals a trend of decreasing Mg/Ca with increasing water depth (and therefore dissolution) for deeper samples. In this case dissolution can cause a ~4% decrease in Mg/Ca of G. bulloides. Hatched region is approximate depth of saturation horizon in this region. (d) The scatter seen in part (a) seems common to both species and probably reflects differing states of undersaturation within sedimentary pore-waters. (e) A laboratory leaching experiment demonstrates the effect of partial dissolution on Mg/Ca in G. bulloides 21. In this experiment 4 individual tests of G. bulloides were selected and cleaned. Before dissolution, each sample was split into two. The first half was dissolved and analysed, the second was subjected to multiple dilute acid leaches in order to induce partial dissolution. Treatment was continued until roughly half the remaining sample had been dissolved. The residual sample was then fully dissolved and analysed. Even though leaching in this way may be expected to be less effective at lowering Mg/Ca due to the very high rate of dissolution, a clear decrease is observed in nearly every case (dashed line represents a % decrease with respect to the un-leached samples). 12
doi:.38/nature777 Water depth (m) 15 2 25 35 4 45 6 5 N. dutertrei P. obliquiloculata 4 2 '% Mg/Ca loss' a % fragments 2 4 6 8 6 5 P. obliquiloculata N. dutertrei (-2.75+.4x) (-2.75+.5x) (-2.75+.6x) 4 2 Mg/Ca % loss b y = -4.72 +.5283x R=.9675 y = -1.4586 +.49595x R=.95289 Figure S6. Calibration of Mg/Ca decrease versus fragmentation for two species of planktonic foraminifera from the Ontong Java Plateau (West Equatorial Pacific). (a) Mg/Ca in N. dutertrei and P. obliquiloculata versus water depth. Mg/Ca data are plotted as % Mg/Ca loss with respect to the shallowest sample in each case (i.e. a minimum). Hatched region is approximate depth of saturation horizon in this region. Mg/Ca data are from Heather Johnstone et al. (manuscript in preparation) (b) Mg/Ca loss versus fragmentation from the OJP. Both species reveal a linear trend. Coloured lines represent the sensitivity tests in Fig. S7. Fragmentation data are from Le and Thunell 6. 13
doi:.38/nature777 6 5 '% Mg/Ca loss' 4 2 (-2.75+.4x) (-2.75+.5x) (-2.75+.6x) (-2.75+.8x) Barker et al., Figure S7 a b 16 Mg/Ca (mmol/mol) 2 3prm 3prm (-2.75+.4x) 8 3prm (-2.75+.5x) 3prm (-2.75+.6x) 1 3prm (-2.75+.8x) 6 been dissolved. The residual sample was then fully dissolved and analysed. Even though leaching in this way may be expected to be less effective at lowering Mg/Ca due to the very 14 high rate of 2 dissolution, a clear decrease is observed in nearly every case (dashed line represents a % 13 decrease with respect to the un-leached samples). 12 Mg/Ca (mmol/mol) 11 Figure S6. Calibration of Mg/Ca decrease versus fragmentation for two species of planktonic foraminifera from the Ontong Java Plateau (West Equatorial Pacific). (a) Mg/Ca in N. dutertrei and 3prm 9 P. obliquiloculata versus water depth. Mg/Ca data are plotted as % Mg/Ca loss with respect to the 3prm (-1.5+.5x) 8 shallowest sample in each 3prm case (-2.75+.5x) (i.e. a minimum). Hatched region is approximate depth of saturation 1 3prm (-4+.5x) 7 horizon in this region. Mg/Ca data are from Heather Johnstone et al. (manuscript in preparation) (b) 6 Mg/Ca loss versus fragmentation from the OJP. Both species reveal a linear trend. Coloured lines 12 14 16 18 2 22 24 26 represent the sensitivity tests in Fig. S7. Fragmentation Age (kyr) data are from Le and Thunell 6. c 14 12 Temperature (ºC) Temperature (ºC) Figure S7. Adjustment of measured Mg/Ca ratios in TNO57-21 using an empirical calibration of Mg/Ca loss versus fragmentation (Fig. S6). (a) Calculated Mg/Ca loss using our fragmentation counts from TNO57-21. Differently coloured curves represent various sensitivities of the calibration. (b) Measured (lower blue curve) and adjusted Mg/Ca ratios for various sensitivities of Mg/Ca loss versus fragmentation. (c) Same as (b) but for different intersects of the linear calibration shown in Fig. S6. Figure S8. Increased abundance of Chaetoceros ssp. resting spores during HS2 is interpreted as an increase in the transporting efficiency (i.e. vigour) of the ACC. (a) Greenland temperature 22 ; (b) IRD from TNO57-21 14 ; (c) Polar foraminiferal species; (d) Diatom counts; (e) %CaCO 3 from TNO57-21 18 ; (f) benthic foraminifera fauna; (g) Antarctic temperature 23. 14
doi:.38/nature777 δ 18 O ( ) -34-36 -38-4 Age (kyr) 15 2 25 HS1 HS2 a % polar species been dissolved. The residual sample was then fully dissolved and analysed. Even 5 though leaching in 7 this way may be expected to be less effective at lowering Mg/Ca due to the very high rate of 6 dissolution, a clear decrease is observed in nearly every case (dashed line represents a % decrease with respect to the un-leached 5 samples). % CaCO 3 Figure S6. Calibration 7 of Mg/Ca decrease versus fragmentation for two species of planktonic 6 foraminifera from the Ontong Java Plateau (West Equatorial Pacific). (a) Mg/Ca in N. dutertrei and 2 5 P. obliquiloculata versus water depth. Mg/Ca data are plotted as % Mg/Ca loss with respect to the 4 shallowest sample in each case (i.e. a minimum). Hatched region is approximate depth of saturation horizon in this region. Mg/Ca data are from Heather Johnstone et al. (manuscript in preparation) (b) 2 Mg/Ca loss versus fragmentation f from the OJP. Both species reveal a linear trend. Coloured lines represent the sensitivity tests in Fig. S7. Fragmentation data are from Le and Thunell 6. -44 Figure S7. Adjustment -46 of measured Mg/Ca ratios in TNO57-21 using an empirical calibration of δ 18 O ( ) -42 2 4 5 4 b c d e g Total diatoms Chaetoceros spp. Mg/Ca loss versus fragmentation (Fig. S6). (a) Calculated Mg/Ca loss using our fragmentation -48 counts from TNO57-21. Differently coloured curves represent various sensitivities of the -5 calibration. (b) Measured (lower blue curve) and adjusted Mg/Ca ratios for various sensitivities of -52 Mg/Ca loss versus fragmentation. (c) Same as (b) but for different intersects of the linear calibration 15 2 25 shown in Fig. S6. Age (kyr) 4 8 12 25 2 15 IRD (#/g) 6 valves per gram benthic shells per gram Figure S8. Increased abundance of Chaetoceros ssp. resting spores during HS2 is interpreted as an increase in the transporting efficiency (i.e. vigour) of the ACC. (a) Greenland temperature 22 ; (b) IRD from TNO57-21 14 ; (c) Polar foraminiferal species; (d) Diatom counts; (e) %CaCO 3 from TNO57-21 18 ; (f) benthic foraminifera fauna; (g) Antarctic temperature 23. Figure S9. Tuning the gas records between Dome C and EDML. (a) The methane records from Dome C 24 (EDC3 age model 25, 26 ) and EDML 23 (GICC5 age model 27 ) are misaligned during the deglacial period; (b) Shifting the EDC3 age model by 5 years brings the records into line and allows us to place the CO 2 record from Dome C 28 on the same timescale as the EDML temperature record (see main text Figure 3). 15
doi:.38/nature777 Figure S6. Calibration of Mg/Ca decrease versus fragmentation for two SUPPLEMENTARY species of planktonic INFORMATION foraminifera from the Ontong Java Plateau (West Equatorial Pacific). (a) Mg/Ca in N. dutertrei and P. obliquiloculata versus water depth. Mg/Ca data are plotted as % Mg/Ca loss with respect to the shallowest sample in each case (i.e. a minimum). Hatched region is approximate depth of saturation horizon in this region. Mg/Ca data are from Heather Johnstone et al. (manuscript in preparation) (b) Mg/Ca loss versus fragmentation from the OJP. Both species reveal a linear trend. Coloured lines EDC3 age (kyr) EDC3 age (kyr) represent the sensitivity tests in Fig. S7. Fragmentation data are from Le and Thunell 6. 15 2 25 15 2 25 8 8 CH4 (ppbv) 7 6 5 Figure S7. Adjustment of measured Mg/Ca ratios in TNO57-21 using an empirical calibration of Mg/Ca loss versus fragmentation (Fig. S6). (a) Calculated Mg/Ca loss using our fragmentation counts from TNO57-21. Differently coloured curves represent various sensitivities of the EDML EDC CH4 (ppbv) 7 6 5 EDML EDC calibration. (b) Measured (lower blue curve) and adjusted Mg/Ca ratios for various sensitivities of Mg/Ca loss versus fragmentation. (c) Same as (b) but for different intersects of the linear calibration shown in Fig. S6. 4 Figure S8. Increased abundance of Chaetoceros ssp. resting spores during HS2 is interpreted as an a increase in the transporting efficiency (i.e. vigour) of the ACC. (a) Greenland temperature 22 ; (b) 15 2 25 GICC5 age (kyr) 4 b 15 2 25 GICC5 age (kyr) IRD from TNO57-21 14 ; (c) Polar foraminiferal species; (d) Diatom counts; (e) %CaCO 3 from TNO57-21 18 ; (f) benthic foraminifera fauna; (g) Antarctic temperature 23. Figure S9. Tuning the gas records between Dome C and EDML. (a) The methane records from Dome C 24 (EDC3 age model 25, 26 ) and EDML 23 (GICC5 age model 27 ) are misaligned during the deglacial period; (b) Shifting the EDC3 age model by 5 years brings the records into line and allows us to place the CO 2 record from Dome C 28 on the same timescale as the EDML temperature record (see main text Figure 3). 16