Supplementary Information: 1) Separation of diatom opal from sediments Diatom opal was separated from sediments using a protocol adapted from Morley et al., 2004. Approximately 10 g of wet sediment was subsampled into a 50 ml centrifuge tube. Organic material was removed by sonicating and heating the sediment in 30 % H2O2 (Aristar) for at least two hours. The suspension was then centrifuged, the supernatant removed and rinsed with distilled water. Approximately 20 ml of 5 % HCl was added and left overnight to remove carbonate. The suspension was then centrifuged, the supernatant removed and rinsed with distilled water. Approximately 15 ml sodium polytungstate (density 2.2 gcm-3) was added to the sediment, sonicated and centrifuged. The light fraction was collected in a 15 ml centrifuge tube. The separation process was repeated four times with the heavy fraction to maximise yield and repeated with the light fraction until visual inspection showed it to be clean (i.e. no heavy material remained). The light fraction was filtered using a 10 micron membrane, rinsed at least six times with distilled water and allowed to dry in a clean oven [Morley, et al., 2004]. The residue was chemically cleaned using the protocol of Shemesh et al., 1988 and Ellwood & Hunter, 1999. The working reagents for the cleaning process were: 1) 1% hydroxylamine chloride in acetic acid; 2) 0.1% w/w NaF solution; 3) 50% v/v strong-acid solution of ultrapure HNO3 and HCl. The dried sample was added to 50 ml centrifuge tube. 5 ml of hydroxylamine chloride-acetic acid solution was added with 20 ml 18O Milli-Q water, swirled and heated for one hour in a boiling water bath. 20 ml of chilled Milli-Q was added, the mixture swirled immediately and sonicated for 1 minute before centrifugation for 3 minutes. The supernatant was discarded. 45 ml of Milli-Q was added and the sonication/centrifugation procedure repeated. 5 ml of the etching NaF solution was then added, and heated in a boiling water bath for 1 hour. 40 ml of cool Milli-Q was added, swirled, sonicated, centrifuged and the supernatant removed as above. The residue was rinsed again in Milli-Q water. 5 ml of the strong acid solution was added and heated in a sub boiling water bath for two hours and the rinsing procedure repeated as above. The samples were rinsed 5 times in Milli-Q [Ellwood and Hunter, 1999; Shemesh, et al., 1988]. Chemical cleaning of diatoms separated from clays carried out in a Class 100 clean laboratory. All plasticware was acid-cleaned in 10% HNO3 for at least 24 hours. After chemical cleaning, less than 2 mg of sample was extracted and added to 0.2 ml 0.25 N HF (FEP distilled), sonicated, centrifuged and the supernatant removed. The sample is then rinsed again in Milli-Q water, sonicated, centrifuged and the supernatant removed. The sample is then totally digested in 1 ml dilute HF at room temperature before analysis. 2) Q-ICP-MS Analysis Analysis was carried out using a ratio method adapted from Harding et al., (2006) on the Perkin Elmer Elan 6100DRC Q-ICP-MS at the University of Oxford (Table A). RF Power = 1150 W Nebulizer = Microflow PFA-100 100 µl min-1 Spray chamber = Quartz cyclonic Argon plasma gas flow rate = 15 l min-1 Argon auxiliary gas flow rate = 1.0 l min-1 Argon nebulizer gas flow rate = 0.9-0.99 l min-1 Sample cone = Platinum, aperture 1.1 mm Skimmer cone = Platinum, aperture 0.9 mm Detector = SimulScan detection system Peristaltic pump speed (spray chamber drain tube) = 8rpm CeO/Ce and Ba++/Ba ratios = <3 % Table A: Operating conditions for the Q-ICP-MS When this ratio method is used for studying metal/Ca ratios in foraminiferal calcite, the pulse detector can be used to measure trace elements and Ca simultaneously by measuring a minor isotope representing < 1 % of natural Ca to prevent saturation of the detector [Harding, et al., 2006]. However, there is no suitable isotope of Si for this purpose. Silicon-29 is the most suitable for signal-to-blank ratios but represents 4.7 % of natural silicon. Instead, a two step process is used, taking advantage of the relatively high concentrations of Al in diatom opal from coastal sediments: 1) the (Al/Si) ratio is measured in a diluted aliquot of sample (10-20 ppm Si) and 2) the (Metal/Al) ratio is measured in a separate aliquot of the same sample (100-200 ppm Si). The (Metal/Si) ratio is then calculated by multiplication. Long term machine drift is reduced by bracketing each sample with an external, matrix-matched standard solution. Correcting for this drift results in improved accuracy and precision (Table B, see 2007pa001576-ts01.txt). Similarly, the short-term drift is reduced by bracketing each trace metal with a Si measurement. The Q-ICP-MS is specifically tuned to the elements of interest and the Si signal is attenuated by increasing the mass resolution at the 29Si peak. The isotopes measured were checked for isobaric and polyatomic interferences. If 1) the calculated error is large due to machine instability (RSDs in excess of 10-15 %), or 2) the calculated (Zn/Si)opal is greatly elevated, the measurement is repeated. Samples with persistently elevated (Zn/Si)opal were rejected if sequential dissolution indicated surface contamination. A machine blank of 1% HNO3 is run per 5 sample-standard pairs. Laboratory blanks indicate no significant contamination of Si and small amounts of Al and Zn which are at least an order of magnitude lower than the opal signal (Table C). Blank corrections, (Metal/Si) ratios and errors are calculated offline. Si Machine blank intensity = 15 000-20 000 Al Machine blank intensity = 100-300 Zn Machine blank intensity = 10-25 Si Lab blank intensity = 21 800 Al Lab blank intensity (20 ppm Si) = 640 Al Lab blank intensity (200 ppm Si) = 1400 Zn Lab blank intensity = 60 Si Sample intensity = 1 000 000 Al Sample intensity (20 ppm Si) = 150 000- 200 000 Al Sample intensity (200 ppm Si) = 1 500 000-2 000 000 Zn Sample intensity = 400-1000 Table C: Typical blank and sample intensities for box core opal analysis. Intensities are given in counts per second. Solutions and standards were prepared using Milli-Q water and sub boiling quartz distilled 16 N HNO3. Machine blanks of other 1% FEP distilled acids indicate the contribution of quartz distillation to the Si blank is negligible. All FEP bottles were refluxed with ~2 N HNO3 at ~70 °C for at least 2 days and rinsed 6 times with Milli-Q water. All other plasticware was soaked in ~2 N HNO3 for at least 48 hours and rinsed 6 times with Milli-Q water. The multielement stock was prepared gravimetrically by spiking 500 ml of 1000 ppm Si standard (Greyhound Chromatography and Allied Chemicals, Merseyside, UK) with concentrations of trace elements similar to expected opal concentrations. Accurate concentrations of trace elements were measured using standard addition (Si = 988 ppm; Al = 1.18 ppm; Zn = 2.6 ppb). The following arguments demonstrate the method is reliable: 1) SEM images show the opal to be clean of clay particles (see Manuscript; Figure 2); 2) Sequential dissolution experiments show (Al/Si)opal (Figure I; 2007pa001576-fs01.jpg) and (Zn/Si)opal ratios plateau to a constant value (see Manuscript; Figure 3); 3) Poor correlation between Al and Zn intensities suggest the Zn is not sourced from clays or surface adsorption (see Manuscript; Figure 7), in contrast to other metals (Figure II; 2007pa001576-fs02.jpg); 4) Yield experiments demonstrate no Si is lost through the addition of dilute HF (Figure III; 2007pa001576-fs03.jpg); 5) Repeat measurements show a high level of machine and sample reproducibility both within-analysis and long-term (Figure IV, V; 2007pa001576-fs04.jpg, 2007pa001576-fs05.jpg). 3) Lead-210 Analysis The 210Pb and 214Pb activities were measured using a Caberra well-type ultra-low background HPGe gamma ray spectrometer, spectra analysed using a Genie 2000 system and accumulated using a 16K channel integrated multichannel analyser (detection limits are ~15 Bq kg-1). The activity was calibrated using a bentonite clay spiked with a gamma-emitting radionuclide standard QCYK8163 and checked against the reference material IAEA135.