Methods are described in a forthcoming manuscript by Cohen et al. Dissolved nitrate+nitrite, nitrite and silicate were measured on an Alpkem Rapid Flow Analyzer, and ammonium and phosphate were measured on a Technicon AutoAnalyzer II at Oregon State University.

For the dissolved metal analysis, filtered seawater was acidified to pH 1.8 using hydrochloric acid, and stored for 6 months at room temperature in the dark. Seawater preconcentration was performed using the seaFAST automated preconcentration system followed by quantification via inductively coupled plasma mass spectrometry. Reagents consisted of a 4M ammonium acetate pH 6.0 buffer prepared using high purity ammonium hydroxide and acetic acid, a 1% nitric acid rinse solution, 10% nitric acid elution acid, and a second internal standard 10% nitric elution acid solution containing 10 ppb indium (115In). Polypropylene conical tubes used with the autosampler were HCl acid-soaked and pH 2-rinsed prior to use. Process blanks consisted of Milli-Q HCl-acidified to pH 2 , and were run alongside samples. A stable isotope cocktail, which consisted of 57Fe, 61Ni, 65Cu, 67Zn and 111Cd, was spiked (50 µL) into each 15 mL sample to account for recovery and matrix effects. Following offline seaFAST preconcentration to 500 μL, the samples were analyzed using an iCAP Q inductively coupled plasma-mass spectrometer (ICP-MS). A six-point external standard curve was used with multi-element and indium (In) standards, diluted to range from 1-10 ppb in 5% nitric acid. Dissolved metal concentrations were determined using isotope dilution. Nitric acid (5%) injection blanks were subtracted from sample metal cps values except for Zn, in which injection blanks were generally higher than process blanks. Accuracy was determined using the 2009 Geotraces surface coastal (GSC) seawater intercalibration standard (n=5): dFe = 1.65 ± 0.19 nM [GSC = 1.56 ± 0.12 nM], dZn = 1.65 ± 0.17 nM [GSC = 1.45 ± 0.10 nM], dCu = 1.38 ± 0.16 nM [GSC = 1.12 ± 0.15 nM], dCd = 0.37 ± 0.03 nM [GSC = 0.37 ± 0.02 nM], dNi = 4.28 ± 0.17 nM [GSC = 4.5 ± 0.21 nM], and dMn = 2.42 ± 0.29 nM [GSC = 2.23 ± 0.08 nM]. The limit of detection was determined by calculating 3x the standard deviation of process blanks dataset-wide: dFe = 0.23 nM (n=14), dZn = 0.35 nM (n=16), dCu = 0.03 nM (n=18), dCd = 0.0103 nM (n=16), dNi = 0.07 nM (n=18), dMn = 0.03 nM (n=18). Blanks that were overtly contaminated with Fe (4 of 18), Zn (2 of 18) or Cd (2 of 18) were not included in the LOD estimation. In the case of Zn, high deep water (>1,000 m) concentrations altered the spiked:stable isotope ratio, and accurate concentrations were not able to be obtained.

For the particulate metal analysis, whole 142 mm filters were extracted using an acid leachable digestion and quantified with ICP-MS. Process blanks were prepared in the laboratory and consisted of acid-cleaned Supor filters that were soaked for 1 week in oligotrophic seawater to condition with salt and remove residual metals introduced from high purity cleaning acids. Plastic forceps were used to place thawed filters into acid-cleaned polypropylene tubes containing 8mL of 5% nitric acid containing 1 ppb In. Filters were digested at 140°C for 3-4 hours, until less than ~4 mL was remaining. Filters were removed and the remainder of the solution was allowed to evaporate in the heat block until completely dry. Two mL of fresh 5% nitric acid was added to the precipitates. Tubes were vortexed, and 900 μL replicates of each sample were pipetted into an 81-well acid-rinsed plate alongside SPEX + In diluted standards. Metal quantification was performed using the iCAP Q ICP-MS. Particulate metal concentrations were calculated following Cox et al. 2014.