Supporting Online Material for

Transcription

1 Supporting Online Material for Glacial Silicic Acid Concentrations in the Southern Ocean Michael J. Ellwood,* Martin Wille, William Maher *To whom correspondence should be addressed. This PDF file includes: Materials and Methods SOM Text Figs. S1 to S7 Tables S1 to S4 References Published 21 October 2010 on Science Express DOI: /science

2 Supporting Online Material to: Glacial silicic acid concentrations in the Southern Ocean Michael J. Ellwood, Martin Wille, and William Maher Methods Spicule isolation and digestion Calcium carbonate present in sediment core samples was eliminated by titrating with hydrochloric acid followed by digestion with hydrogen peroxide. Samples were boiled for 5 minutes in sodium hexametaphosphate (1%) solution, diluted, and the process repeated. Sediment samples were sieved to 150 m and spicules were picked from the >150 m fraction under a binocular microscope. Before analysis, spicules were chemically cleaned by rinsing with deionised water followed by digestion for one hour in a hot (90 C) solution of 0.1% hydroxylamine hydrochloride in 1% acetic acid, followed for a second hour in a solution of 0.1% sodium fluoride in 1% acetic acid. The final cleaning step involved heating in a strong acid solution (50% HNO 3 :HCl, 1:1). Sponge spicules were dissolved in 0.5 ml of 2 M sodium hydroxide solution. Silicon and germanium determination Silicon concentrations were determined colourmetrically while germanium concentrations were determined by isotope dilution (S1) using an automated hydride generations system (S2). Germanium:silicon (Ge:Si) reproducibility for sponge samples was good at better than 12%. 1

3 Silicon isotope determination One ml graduated Pasteur pipettes filled with 0.5 ml of cation exchange resin (Dowex AG-X8, mesh) were used to remove sodium from the sample digest. After cleaning the resin with HCl and deionised water (S3), each column was loaded with 0.75 ml of sample solution. Silicon was eluted from the column using 1.5 ml of deionised water and acidified with 0.25 ml of 2 M HNO 3. Silicon isotopic ratios were determined by multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS) (Finnigan Neptune, Germany) operated in mediumresolution (M/ΔM ~2000) and dry plasma mode. The sample was introduced into the MC-ICP-MS via an ESI-Apex desolvation nebulizer fitted with a Teflon inlet system and a demountable torch fitted with an alumina injector to minimize the silicon background. Data acquisition and reduction were carried out using a standard-samplestandard bracketing technique (S4). Prior to each sample run, blank measurements are made to ensure that the combined blank and background contribute less than 1% of the total sample signal. The reproducibility of the δ 30 Si results (δ 30 Si = [( 30 Si/ 28 Si sample / 30 Si/ 28 Si std -1) x1000]) normalized to the NBS28 standard were ± 0.23 (2 SE, n = 54) and is based on regression error from the plot of Si versus Si (Fig. S1). Measurements of an inter-laboratory diatomite standard prepared and measured on different days produced average values for Si of 0.64 ± 0.08 and Si of 1.29 ± 0.14 (2σ S.D), which is in good agreement with values obtained by other laboratories (S5). 2

4 Calculation of changes in seawater silicic acid concentration based on the sponge spicule silicon isotope record To calculate the change in silicic acid concentration during this period, we assumed a Southern Ocean silicon isotope seawater composition of 1.1 and combined this with the 30 Si sponge versus silicic acid concentration relationships obtained by Wille et al. (S6) and Hendry et al. (S7). We assumed a seawater silicon isotope value of 1.1 because the exact seawater value for each site is unknown. Note that our Holocene 30 Si values for E33-22 and ODP177 site 1089 are about 0.5 to 1 more positive (heavier) than contemporary sponge values, and probably results from regional variation is in the silicon isotope composition of seawater. We have also assumed that the silicon isotope composition of the deep ocean has not varied significantly over the past 150,000 years, as suggested by modeling (S8)(Table S3). Ten-box model setup (PANDORA) Model setup and justification To model silicic acid, silicon isotope and germanium cycling in the ocean, we set up a ten-box model based on the PANDORA model architecture of Broecker and Peng (S9) (Fig. S3). We then undertook a number of perturbation experiments with the specific aim of determining the influence changes in the uptake stoichiometry of silicic acid relative to that of phosphate and nitrate have on silicon inventory, germanium inventory and the silicon isotope composition of the ocean. The silicic acid leakage hypothesis purports the idea that Southern Ocean surface waters were fertilized with iron from atmospheric dust during glacial times, which result in a reduction in silicic acid utilization by diatoms relative to that of phosphate and nitrogen (S10-12). This reduction allowed silicic acid to leak out of the Southern 3

5 Ocean via mode and intermediate water masses to low latitudes where it subsequently fueled an increase in diatom production: currently diatom production in low latitude surface waters, such as in the Eastern Equatorial Pacific and equatorial Atlantic, are limited by the availability silicic acid and iron (S11, 13-14). This increase in low latitude production diatom production, and associated organic carbon production, is thought to be at the expense of coccolithophore production thereby diminishing calcium carbonate production and hence the carbonate pump (S10-11). An increase in the organic carbon to calcium carbonate export ratio leads to increased carbon dioxide uptake by the ocean and hence a lower atmospheric carbon dioxide concentration (S11). Silicon and germanium inputs The inputs and outputs for silicon and germanium in to and out of the model are the same those outlined by Sutton et al. (S15) and are as follows: the flux of silicon entering the ocean was set to 7.05 x10 12 moles y -1 and was composed of four sources: 1) riverine (5.6 x10 12 moles y -1 ), 2) hydrothermal (0.55 x10 12 moles y -1 ), 3) submarine weathering (0.4 x10 12 moles y -1 ) and 4) aeolian (0.5 x10 12 moles y -1 ) (S16-19). The model was then initiated and the remineralisation efficiency of biogenic opal adjusted so that the steady-state amount of silicic acid residing within the ocean was equivalent to a global average concentration of 70 µmol L -1 (S19). For germanium, we set the input flux to 9.5 x10 6 moles y -1 and was composed of four sources: 1) riverine (3.0 x10 6 moles y -1 ), 2) hydrothermal (6.0 x10 6 moles y -1 ), 3) submarine weathering (0.2 x10 6 moles y -1 ) and 4) aeolian (0.3 x10 6 moles y -1 ) (S17). We assumed that the loss of germanium from the ocean was via biogenic particles and non-biotic processes (S15, 17-18, 20). As noted by Hammond et al. (S17), the removal of germanium by these 4

6 non-biotic processes is a case of less efficient germanium transfer, relative to silicon, from oceanic sediments into the overlying water column post opal dissolution. To account for these non-biotic process, we adjusted the remineralization efficiency of germanium associated with biogenic opal dissolution so that at steady-state the overall oceanic Ge:Si ratio equaled 0.76 x10-6 (Table S2)(S15). To mimic Ge:Si fractionation in sponges, we assumed a Michaelis-Menten uptake process whereby the uptake of germanium is via the silicic acid uptake system, albeit at slightly slower rate than that of silicic acid (S21). Silicon isotope cycling The ten-box model was also modified to accommodate silicon isotope fractionation (S22). During opal production we assumed that silicon isotope fractionation equaled (S23). For the four major silicon inputs, we assumed the following silicon isotope compositions: 1) riverine ( ) (S24), 2) hydrothermal ( ) (S25), 3) submarine weathering/basalt ( ) and 4) aeolian ( ) (S25), and a reference (NBS28) silicon isotope value of To mimic silicon isotope fractionation in sponges, we used the 30 Si sponge versus silicic acid concentration relationship obtained by Wille et al. (S6). To simulate changes in silicic acid and germanium utilization and their inventories during glacial times, we adjusted three parameters in timed model runs: 1) the utilization of silicon (and germanium) in the surface ocean, 2) an increase in riverine silicon inputs and 3) an increase in germanium loss from the ocean via the nonopaline sink. 5

7 Model results A steady-state run the contemporary ocean Before undertaking perturbation experiments with the ten-box model, we ran the model to steady-state and compared model outputs to observations and other model outputs employing a similar architecture (Table S2). On the whole, the model captured the distribution of both silicon and germanium in the ocean and replicated Ge:Si fractionation in surface and deep waters (S15). The model also did a good job of replicating the silicon isotope composition of each oceanic basin, with our results being comparable to those of Reynolds (S22)(Table S2). Silicon isotope and Ge:Si perturbations experiments In the first timed model run, the model was run to steady-state in its perturbed state such that it either resembled the: a) utilization of phosphate in the surface ocean, or b) the lower global Ge:Si ratio observed for the glacial ocean (Table S3). To mimic phosphate utilization in surface waters, as suggested by the silicic acid leakage hypothesis (S12), we adjusted the residence time ( ) (this represents the rate of nutrient utilization and hence productivity in surface waters (S9)) of silicic acid in: 1) the surface Antarctic box, and 2) all surface ocean boxes. Adjusting the residence time of silicic acid, and hence diatom utilization of silicic acid, in the surface Antarctic box resulted in a global ocean silicic acid inventory increase of 6% (6 Pmoles) (Table S3). Likewise, when the residence time of silicic 6

8 acid in all the surface ocean boxes was increased to match that of phosphate, the result was a global ocean silicic acid inventory increase of 19% (18 Pmoles). Adjusting the residence time of silicic acid in the surface ocean also resulted in a redistribution of opal export and hence opal sedimentation (Fig. S5). In the first experiment, when the residence time of silicic acid in the surface Antarctic box was doubled from 50 to 100 years, the amount of silicon lost from the ocean via opal export from the surface Antarctic box decreased by 41% (0.5 Pmoles)(Table S3). Conversely, the amount silicon lost from the ocean via opal export from the intermediate and low latitude boxes increased (Fig. S5). In the second experiment when the residence time of silicic acid in all the surface boxes was increased, the amount of silicon lost from the ocean via opal export decreased in all boxes except for the surface Pacific box and the surface Atlantic box (Fig. S5, Table S3). In these boxes the amount silicon lost from the ocean via opal export increased by 46% (0.8 Pmoles) and 111% (0.6 Pmoles), respectively. In both experiments changes silicic acid utilization led to an increase in the global ocean silicic acid inventory, a reduction in the opal export and sedimentation flux for Antarctic region, and an increase in opal export and sedimentation flux for intermediate and low latitude regions, i.e. the surface Atlantic and Pacific (Fig S5). These changes in the opal export and sedimentation fluxes are consistent with the silicic acid leakage hypothesis and with the Southern Ocean and Atlantic sediment opal records for the last glacial period. Overall, opal accumulation for the Southern Ocean was reduced during last glacial period while there was an increase in opal 7

9 accumulation for the equatorial Atlantic (S26-27). For the equatorial Pacific, opal accumulation is lower during last glacial period (S27-28), however, elevation of chronic iron limitation along with increased silicic acid supply likely caused by an overall increase in diatom production, but with a lower Si:C export ratio and lower sediment opal accumulation (S29). Although there was a redistribution of silicic acid between surface boxes and major changes in opal sedimentation, only a small change (up to 0.03 ) in the overall oceanic silicon isotope composition was observed in both perturbation experiments (Fig. S6, Table S3). However, larger regional changes in the silicon isotope composition of seawater were observed for the experiment where was varied in all the surface ocean boxes (Fig. S5). For the surface Antarctic, the surface Pacific, the surface Atlantic, the intermediate Indo-Pacific and the intermediate Atlantic, the silicon isotope composition of seawater in these boxes was lower (Table S3). Correspondingly, the silicon isotope composition of opal produced in, exported and deposited on the seafloor from these boxes is also lower (Fig. S5). Thus, the decline in the silicon isotope composition of diatom opal produced in these boxes is consistent with paleo diatom opal records from the Atlantic and Indian sectors of the Southern Ocean and the Eastern Equatorial Pacific (see Fig. 4, main text) (S12, 29-32). The rapid change in the silicon isotope composition produced in model runs compared to that of the sediment record reflects the instantaneous change in silicic acid utilization within the model whereas in the real ocean this change it likely to have occurred over a few thousand years as the iron concentration of the surface ocean changed, thus altering the silicic acid requirement of diatoms. 8

10 We also examined the effect of reducing opal accumulation in Antarctic sediments on the silicon isotope composition of diatoms. This reduction was simulated by adjusting the amount of opal dissolution within the deep Antarctic box such that a 50% reduction in opal burial was achieved (Table S3). This 50% reduction in the opal accumulation in the Antarctic led to an increase in opal burial outside the Antarctic, but only a small change in the silicon isotope composition (<0.1 ) of the redistributed opal (Fig. S5 and S6), which was much smaller than the observed glacial-interglacial change (see Fig. 3 and 4, main text)(s12, 29-32). Ge:Si experiments To mimic the low global Ge:Si ratio observed for the glacial ocean, we either increased the riverine input of silicon (2.65 Tmol y -1 ) or increased the amount of germanium lost from the ocean via the non-opal sink by adjusting the germanium remineralisation efficiency until and the Ge:Si ratio equal 0.55 x10-6 (Fig. S7) (S33-34). Note that an increase in the riverine silicon input resulted in an increase in the global silicic acid inventory of 37 % (Table S4) but only a small change (0.09 ) in the silicon isotope composition of the ocean (Fig. S6, Table S3). Silicon isotope excursion during MIS 5e The silicon isotope excursion at the onset of MIS 5e for core E33-22 (Pacific sector) coincides with the penultimate deglacial transition where changes in the silicon isotope and Ge:Si composition of diatoms also occur (Fig S4). There is, however, a slight increase in sponge Ge:Si record across this transition suggesting that this isotope excursion is real. Although the exact mechanism(s) associated with this 9

11 silicon isotope excursion are not fully understood, it reflects a change in deep water ventilation which led to an increase in silicon-rich water bathing this core site. 10

12 Si ( ) E33-22 ODP177 site Si ( ) Fig. S1. Mass-dependent fraction (MDF) of Si versus Si plot for sponge spicules analyzed in this study. Error bars are 2 standard deviation, ± 0.23 for S and ±0.12 for Si. Dotted line is the MDF line represented by Si = ( Si), r 2 =

13 Th-normalized opal flux (g cm -2 ka -1 ) Si sponge (inverted) Age (ky) Fig. S2. Silicon isotope and thorium normalized opal flux for the South Atlantic sector of the Southern Ocean. Silicon isotope results are for ODP177 site 1089 (diamonds) and 230 Th-normalized opal flux for core TN057-13PC (circles) (S39). Note that the silicon isotope scale has been reversed. 12

14 Fig. S3. Ten-box model (PANDORA) which operates something like the real system (S9). Blue lines indicate water flow and values are in Sverdrups. Red lines indicate opal export from the surface ocean to deeper boxes and sediments. Green lines indicate the four major oceanic germanium and silicon inputs (riverine, hydrothermal, eolian and submarine weathering). Box numbers (bolt italics) for each oceanic region are as follows: 1 North Atlantic, 2 Surface Atlantic, 3 Intermediate Atlantic, 4 Surface Antarctic, 5 Intermediate Indo-Pacific, 6 Surface Pacific, 7 North Pacific, 8 Deep Indo-Pacific, 9 Deep Antarctic, and 10 Deep Atlantic. 13

15 18 O planktonic ( ) Ge:Si diatom (x10-6 ) Ge:Si sponge (x10-6 ) Age (ky) MIS A B C D E Atlantic sector 0 F Pacific sector Age (ky) C 30 benthic ( ) Si diatom ( ) 30 Si sponge ( ) Fig. S4. Comparison of isotope and trace element records for the Southern Ocean covering the period 100 to 150ka. (A) Stable 18 O isotope records for planktonic foraminifera (blue line, Pacific sector; purple line, Atlantic sector)(s35-36). (B) Stable 13 C isotope records for benthic foraminifera (S35-36) (purple line, Atlantic sector). 14

16 (C) Diatom germanium:silicon records for cores in the Atlantic (RC13-259; 53 53'S; 04 56'W, purple line) (S33) and Pacific (E17-9; 63 05'S 'W, blue line) (S33) sector of the Southern Ocean. D) Diatom silicon isotope record for core RC (S12). (E) Sponge germanium:silicon records for core E33-22 (circles, Pacific sector) and ODP177 drill site 1089 (diamonds, Atlantic sector). (F) Sponge silicon isotope record core E33-22 (circles, Pacific sector) and ODP177 drill site 1089 (diamonds, Atlantic sector). 15

17 Fig. S5. Changes in the opal burial flux and the silicon isotope composition of buried opal relative to the modern ocean. (A) Percentage change in the opal burial flux of opal (glacial modern) associated with each surface water box from where the opal was exported (Table S3). (B) Change in the silicon isotope composition of buried opal (glacial modern) associated with each surface water box from where the opal was exported (Table S3). Abbreviations are as follows: Surf Atl = Surface Atlantic, 16

18 Intermed Atl = Intermediate Atlantic, Surf Ant = Surface Antarctic, Intermed Indo- Pac = Intermediate Indo-Pacific, Surf Pac = Surface Pacific, and N Pac = North Pacific. 17

19 Fig. S6. Change in the silicon isotope composition of seawater (glacial modern) where various parameters have been altered (Table S3). Abbreviations are as follows: N Atl = North Atlantic, Surf Atl = Surface Atlantic, Intermed Atl = Intermediate Atlantic, Surf Ant = Surface Antarctic, Intermed Indo-Pac = Intermediate Indo- Pacific, Surf Pac = Surface Pacific, N Pac = North Pacific, Deep Indo-Pac = Deep Indo-Pacific, Deep Ant = Deep Antarctic, Deep Atl = Deep Atlantic and Oceanic ave = Oceanic average. 18

20 Age (ky) Ge:Si diatom (x10-6 ) Ge:Si sponge (x10-6 ) A B Age (ky) Fig. S7. Ge:Si records from diatoms and sponges along with ten-box model results. A) Diatom Ge:Si records for cores located in the Atlantic sector of the Southern Ocean (RC13-259)(S33) and the Pacific sector of the Southern Ocean (E17-9)(S33) along with model results for a change in silicic acid utilization (red curve), a change the germanium inventory of the global ocean (blue curve), and a change the silicon inventory of the global ocean (purple curve). Note that in the timed model runs, we adjusted the silicon inputs into or germanium lost from the ocean such that the glacial diatom Ge:Si ratio was 0.55 x10-6. B) Sponge Ge:Si records for ODP177 site 1089 and E33-22 along with model results for a change in silicic acid utilization (red curve), a change the germanium inventory of the global ocean (blue curve), and a change the silicon inventory of the global ocean (purple curve). Note we have adjusted model Ge:Si results such that the glacial Ge:Si ratio was 0.14 x

21 Table S1. Silicon isotope and Ge:Si data for sponges spicules isolated from cores ODP177 site 1089 and E ODP177 site 1089 Hole Core Section Top (cm) Interval Bottom (cm) Depth (mbsf) Age $ (ka) 30 Si 29 Si Ge:Si (x10-6 ) A A nd nd a A A a a A A A A A A a a A a a A A A a A a a D nd A nd A nd A nd B nd B nd D nd D nd D nd B nd B nd E33-22 Depth (cm) Age # (ka) 30 Si 29 Si Ge:Si (x10-6 ) a a a a nd nd

22 nd nd (0.522) (0.890) nd nd nd nd b nd nd nd nd nd nd (0.652) nd nd (0.388) nd nd nd nd nd nd nd nd nd nd nd nd nd nd a a a $ Age model taken from Hodell et al. (S35). # Age model taken from Nimmemann and Charles (S36). a Average of duplicate or triplicate analysis. b A questionable number as it falls off the mass-dependent fraction line for Si versus Si. nd = not determined Sample in brackets appear to be contaminated for germanium 21

23 Table S2. Pandora model results for silicon and germanium and silicon isotopes. Pandora * Renyolds Pandora # Oceanic region Box Si (µmol L -1 ) 30 Si Si (µmol L -1 ) 30 Si Ge:Si (x 10-6 ) model Observed & North Atlantic nd Surf. Atlantic Intermed. Atlantic ~1 Surf. Antarctic Intermed. Indo Pacific Surf. Pacific ~2 North Pacific Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average * Results from the ten-box Pandora model with silicon and germanium input and output (export) fluxes. Results from Reynolds (S22) using the Pandora model without silicon input and output fluxes. Note that Reynolds (S22) used a high oceanic silicon inventory, hence the higher oceanic silicon concentration average. # Based on data from Froliech et al. (S37) and Sutton et al. (S15). nd = no data 22

24 Table S3. Model output results for changes in silicic utilization and riverine silicon supply to the global ocean. Contemporary Ocean Standard run steady state conditions Box - PO 4 PO 4 (µmol L -1 ) * - Si Si (µmol L -1 ) 30 Si Sedimentation flux (mol y -1 ) 30 Si North Atlantic Surf. Atlantic x Intermed. Atlantic x Surf. Antarctic x Intermed. Indo-Pacific x Surf. Pacific x North Pacific x Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average x10 12 Last Glacial Maximum Reduced Si utilization ( ) in the surface Antarctic box only Box - PO 4 PO 4 (µmol L -1 ) * - Si Si (µmol L -1 ) 30 Si Si export flux (mol y -1 ) 30 Si North Atlantic Surf. Atlantic x Intermed. Atlantic x Surf. Antarctic x Intermed. Indo-Pacific x Surf. Pacific x North Pacific x Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average x10 12 Reduced Si utilization ( ) in all boxes Box - PO 4 PO 4 (µmol L -1 ) * - Si Si (µmol L -1 ) 30 Si Si export flux (mol y -1 ) 30 Si North Atlantic Surf. Atlantic x Intermed. Atlantic x Surf. Antarctic x Intermed. Indo-Pacific x Surf. Pacific x North Pacific x Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average x10 12 Increased riverine input of Si 23

25 Box - PO 4 PO 4 (µmol L -1 ) * - Si Si (µmol L -1 ) 30 Si Si export flux (mol y -1 ) 30 Si North Atlantic Surf. Atlantic x Intermed. Atlantic x Surf. Antarctic x Intermed. Indo-Pacific x Surf. Pacific x North Pacific x Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average x10 12 Increased riverine input of Si and changed Si utilization ( ) Box - PO 4 PO 4 (µmol L -1 ) * - Si Si (µmol L -1 ) 30 Si Si export flux (mol y -1 ) 30 Si North Atlantic Surf. Atlantic E Intermed. Atlantic x Surf. Antarctic x Intermed. Indo-Pacific x Surf. Pacific x North Pacific x Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average x % reduction in opal burial in the Antarctic Box - PO 4 PO 4 (µmol L -1 ) * - Si Si (µmol L -1 ) 30 Si Si export flux (mol y -1 ) 30 Si North Atlantic Surf. Atlantic x Intermed. Atlantic x Surf. Antarctic x Intermed. Indo-Pacific x Surf. Pacific x North Pacific x Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average x10 12 * Based on a riverine input flux of 10 x10 10 mol y -1 and an average oceanic inventory of 2.92 x10 15 moles of PO 4 (S38). Opal burial from the deep Antarctic box was reduced by 50% by adjusting opal remineralisation within the deep Antarctic box. 24

26 Table S4. Model output results for changes in silicic utilization, riverine silicon input and germanium loss from the global ocean. Contemporary ocean Standard run steady state conditions PO 4 Si Ge:Si Box - PO 4 (µmol L -1 ) * - Si (µmol L -1 ) (x10-6 ) North Atlantic Surf. Atlantic Intermed. Atlantic Surf. Antarctic Intermed. Indo-Pacific Surf. Pacific North Pacific Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average Last glacial maximum Reduced Si utilization ( ) in all boxes PO 4 Si Ge:Si Box - PO 4 (µmol L -1 ) - Si (µmol L -1 ) (x10-6 ) North Atlantic Surf. Atlantic Intermed. Atlantic Surf. Antarctic Intermed. Indo-Pacific Surf. Pacific North Pacific Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average Increased germanium loss from the ocean PO 4 Si Ge:Si Box - PO 4 (µmol L -1 ) - Si (µmol L -1 ) (x10-6 ) North Atlantic Surf. Atlantic Intermed. Atlantic Surf. Antarctic Intermed. Indo-Pacific Surf. Pacific North Pacific Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average Increased riverine input of Si 25

27 Box - PO 4 PO 4 (µmol L -1 ) - Si Si (µmol L -1 ) Ge:Si (x10-6 ) North Atlantic Surf. Atlantic Intermed. Atlantic Surf. Antarctic Intermed. Indo-Pacific Surf. Pacific North Pacific Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average Increased riverine input of Si and changed Si utilization ( ) Box - PO 4 PO 4 (µmol L -1 ) - Si Si (µmol L -1 ) Ge:Si (x10-6 ) North Atlantic Surf. Atlantic Intermed. Atlantic Surf. Antarctic Intermed. Indo-Pacific Surf. Pacific North Pacific Deep Indo-Pacific Deep Antarctic Deep Atlantic Oceanic average * Based on a riverine input flux of 10 x10 10 mol y -1 and an average oceanic inventory of 2.92 x10 15 moles of PO 4 (S38). 26

28 References and Notes S1. R. A. Mortlock, P. N. Froelich, Determination of germanium by isotope dilution hydride generation inductively coupled plasma mass spectrometry. Anal. Chim. Acta 332, 277 (1996). S2. M. J. Ellwood, W. A. Maher, An automated hydride generation-cryogenic trapping-icp-ms system for measuring inorganic and methylated Ge, Sb and As species in marine and fresh waters. J. Anal. At. Spectrom. 17, 197 (2002). S3. S. van den Boorn, P. Z. Vroon, C. C. van Belle, B. van der Wagt, J. Schwieters, M. J. van Bergen, Determination of silicon isotope ratios in silicate materials by high-resolution MC-ICP-MS using a sodium hydroxide sample digestion method. J. Anal. At. Spectrom. 21, 734 (2006). S4. F. Albarede, P. Telouk, J. Blichert-Toft, M. Boyet, A. Agranier, B. Nelson, Precise and accurate isotopic measurements using multiple-collector ICPMS. Geochim. Cosmochim. Acta 68, 2725 (2004). S5. B. C. Reynolds, J. Aggarwal, L. André, D. Baxter, C. Beucher, M. A. Brzezinski et al., An inter-laboratory comparison of Si isotope reference materials. J. Anal. At. Spectrom. 22, 561 (2007). S6. M. Wille, J. Sutton, M. J. Ellwood, M. Sambridge, W. Maher, S. Eggins et al., Silicon isotopic fractionation in marine sponges: A new model for understanding silicon isotopic variations in sponges. Earth Planet. Sci. Lett. 292, 281 (2010). S7. K. R. Hendry, R. B. Georg, R. E. M. Rickaby, L. F. Robinson, A. N. Halliday, Deep ocean nutrients during the Last Glacial Maximum deduced from sponge silicon isotopic compositions. Earth Planet. Sci. Lett. 292, 290 (2010). 27

29 S8. C. L. De La Rocha, M. J. Bickle, Sensitivity of silicon isotopes to wholeocean changes in the silica cycle. Mar. Geol. 217, 267 (2005). S9. W. S. Broecker, T.-H. Peng, The Role of CaCO3 Compensation in the Glacial to Interglacial Atmospheric CO2 Change. Global Biogeochem. Cy. 1, 15 (1987). S10. Y. Nozaki, Y. Yamamoto, Radium 228 based nitrate fluxes in the eastern Indian Ocean and the South China Sea and a silicon-induced "alkalinity pump" hypothesis. Global Biogeochem. Cy. 15, 555 (2001). S11. K. Matsumoto, J. L. Sarmiento, M. A. Brzezinski, Silicic acid leakage from the Southern Ocean: A possible explanation for glacial atmospheric pco 2. Global Biogeochem. Cy. 16, (2002). S12. M. A. Brzezinski, C. J. Pride, V. M. Franck, D. M. Sigman, J. L. Sarmiento, K. Matsumoto et al., A switch from Si(OH) 4 to NO - 3 depletion in the glacial Southern Ocean. Geophys. Res. Lett. 29, doi: /2001gl (2002). S13. K. H. Coale, S. E. Fitzwater, R. M. Gordon, K. S. Johnson, R. T. Barber, Control of community growth and export production by upwelled iron in the equatorial Pacific Ocean. Nature 379, 621 (1996). S14. J. H. Martin, K. H. Coale, K. S. Johnson, S. E. Fitzwater, R. M. Gordon, S. J. Tanner et al., Testing the iron hypothesis in ecosystems of the Equatorial Pacific Ocean. Nature 371, 123 (1994). S15. J. Sutton, M. J. Ellwood, W. Maher, P. L. Croot, Oceanic distribution of inorganic germanium relative to silicon: Germanium discrimiation by diatoms. Global Biogeochem. Cy. 24, doi: /2009gb (2010). 28

30 S16. D. J. DeMaster, The accumulation and cycling of biogenic silica in the Southern Ocean: revisiting the marine silica budget. Deep-Sea Res. PT II 49, 3155 (2002). S17. D. E. Hammond, J. McManus, W. M. Berelson, Oceanic germanium/silicon ratios: Evaluation of the potential overprint of temperature on weathering signals. Paleoceanography 19, doi: /2003pa (2004). S18. S. L. King, P. N. Froelich, R. A. Jahnke, Early diagenesis of germanium in sediments of the Antarctic South Atlantic: In search of the missing Ge sink. Geochim. Cosmochim. Acta 64, 1375 (2000). S19. P. Treguer, D. M. Nelson, A. J. Vanbennekom, D. J. Demaster, A. Leynaert, B. Queguiner, The Silica Balance in the World Ocean - a Reestimate. Science 268, 375 (1995). S20. D. E. Hammond, J. McManus, W. M. Berelson, C. Meredith, G. P. Klinkhammer, K. H. Coale, Diagenetic fractionation of Ge and Si in reducing sediments: The missing Ge sink and a possible mechanism to cause glacial/interglacial variations in oceanic Ge/Si. Geochim. Cosmochim. Acta 64, 2453 (2000). S21. M. J. Ellwood, M. Kelly, W. A. Maher, P. De Deckker, Germanium incorporation into sponge spicules: Development of a proxy for reconstructing inorganic germanium and silicon concentrations in seawater. Earth Planet. Sci. Lett. 243, 749 (2006). S22. B. C. Reynolds, Modeling the modern marine delta30si distribution. Global Biogeochem. Cycles 23, (2009). 29

31 S23. C. L. De La Rocha, M. A. Brzezinski, M. J. DeNiro, Fractionation of silicon isotopes by marine diatoms during biogenic silica formation. Geochim. Cosmochim. Acta 61, 5051 (1997). S24. R. B. Georg, A. J. West, A. R. Basu, A. N. Halliday, Silicon fluxes and isotope composition of direct groundwater discharge into the Bay of Bengal and the effect on the global ocean silicon isotope budget. Earth Planet. Sci. Lett. 283, 67 (2009). S25. C. L. De La Rocha, M. A. Brzezinski, M. J. DeNiro, A first look at the distribution of the stable isotopes of silicon in natural waters. Geochim. Cosmochim. Acta 64, 2467 (2000). S26. L. I. Bradtmiller, R. F. Anderson, M. Q. Fleisher, L. H. Burckle, Opal burial in the equatorial Atlantic Ocean over the last 30 ka: Implications for glacialinterglacial changes in the ocean silicon cycle. Paleoceanography 22, PA4216 (2007). S27. L. I. Bradtmiller, R. F. Anderson, M. Q. Fleisher, L. H. Burckle, Diatom productivity in the equatorial Pacific Ocean from the last glacial period to the present: A test of the silicic acid leakage hypothesis. Paleoceanography 21, PA4201 (2006). S28. S. S. Kienast, M. Kienast, S. Jaccard, S. E. Calvert, R. François, Testing the silica leakage hypothesis with sedimentary opal records from the eastern equatorial Pacific over the last 150 kyrs. Geophys. Res. Lett. 33, doi: /2006gl (2006). S29. L. E. Pichevin, B. C. Reynolds, R. S. Ganeshram, I. Cacho, L. Pena, K. Keefe et al., Enhanced carbon pump inferred from relaxation of nutrient limitation in the glacial ocean. Nature 459, 1114 (2009). 30

32 S30. C. L. De La Rocha, M. A. Brzezinski, M. J. DeNiro, A. Shemesh, Siliconisotope composition of diatoms as an indicator of past oceanic change. Nature 395, 680 (1998). S31. C. P. Beucher, M. A. Brzezinski, X. Crosta, Silicic acid dynamics in the glacial sub-antarctic: Implications for the silicic acid leakage hypothesi. Global Biogeochem. Cy. 21, doi: /2006gb (2007). S32. X. Crosta, C. Beucher, K. Pahnke, M. A. Brzezinski, Silicic acid leakage from the Southern Ocean: Opposing effects of nutrient uptake and oceanic circulation. Geophys. Res. Lett. 34, L13601 (2007). S33. R. A. Mortlock, C. D. Charles, P. N. Froelich, M. A. Zibello, J. Saltzman, J. D. Hays et al., Evidence for lower productivity in the Antarctic Ocean during the last glaciation. Nature 351, 220 (1991). S34. P. N. Froelich, V. R. Vlanc, R. A. Mortlock, S. N. Chillaud, W. Dustan, A. Udomkit et al., River fluxes of dissolved silica to the ocean were higher during glacials: Ge/Si in diatoms, rivers and oceans. Paleoceanography 7, 739 (1992). S35. D. A. Hodell, C. D. Charles, J. H. Curtis, P. G. Mortyn, U. S. Ninnemann, K. A. Venz, in Proceedings of the Ocean Drilling Program, Scientific Results Volume, R. Gersonde, D. A. Hodell, P. Blum, Eds. (2002). S36. U. S. Ninnemann, C. D. Charles, Regional differences in Quaternary Subantarctic nutrient cycling: Link to intermediate and deep water ventilation. Paleoceanography 12, 560 (1997). S37. P. N. Froelich, R. A. Mortlock, A. Shemesh, Inorganic germanium and silica in the Indian Ocean: Biological fractionation during (Ge/Si)opal formation. Global Biogeochem. Cy. 3, 79 (1989). 31

33 S38. G. M. Filippelli, The Global Phosphorus Cycle. Rev. Mineral. Geochem. 48, 391 (2002). S39. R. F. Anderson, S. Ali, L. I. Bradtmiller, S. H. H. Nielsen, M. Q. Fleisher, B. E. Anderson et al., Wind-Driven Upwelling in the Southern Ocean and the Deglacial Rise in Atmospheric CO2. Science 323, 1443 (2009). 32