Our sampling methodology was in alignment with the GEOTRACES cookbook. Seawater samples were collected by onboard supertechs with the GEOTRACES Trace Element Carousel sampling system (GTC). In short, this includes a trace metal clean rosette bearing 24 12-liter (L) Go-Flo bottles (General Oceanics), a Vectran cable, winch and A-frame system, and a clean sampling van (as in Cutter and Bruland 2012). Surface samples were collected using a trace metal clean towfish upon station arrival. Cobalt (Co) samples were filtered in the clean van using a 0.2 micromolar (µM) Acropak-500 capsule (Pall) into 60 milliliters (mL) LDPE bottles (Nalgene) that had been prepared with the following process: 1 week soak in Citranox detergent, Milli-Q water (Millipore) rinse, 2 week soak in 10% HCl (JT Baker, Reagent grade, in Milli-Q), <0.1% HCl rinse. Sample bottles were filled completely, leaving no headspace. Samples to be analyzed for labile cobalt (lCo) were stored in doubled plastic bags at 4 degrees Celsius (°C) until analysis. Total cobalt (dCo) samples were packed in groups of up to six bottles. Each group of bottles was put in an open plastic bag, which was then placed and sealed inside a heat-sealable bag containing one oxygen-absorbing satchel (Mitsubishi Gas Chemical RP-3K) for each bottle, then stored at 4°C.

Dissolved cobalt analysis was performed in a trace metal clean lab at Woods Hole Oceanographic Institution within 10 months of cruise completion using competitive ligand exchange cathodic stripping voltammetry (CLE-CSV) with a hanging mercury drop electrode (HMDE). This method was developed by Saito and Moffett (2001); also see Saito et al., 2010 and Hawco et al., 2016, for method modifications. For dCo measurements only, samples were irradiated in acid-rinsed quartz tubes for 1 hour using a Metrohm 705 UV Digestor to degrade organic Co ligands prior to analysis. For both dCo and lCo measurements, 11 mL of sample was aliquoted into acid-washed 15 mL polypropylene vials. 38.8 microliters (µL) of 0.1 M dimethylglyoxime (DMG, Sigma Aldrich) in Optima-grade methanol (Fisher Scientific) was added. To eliminate dCo contamination from DMG, DMG was first dissolved in 10^-3 M EDTA, then recrystallized over ice and dried before being redissolved in methanol. For lCo samples, DMG was allowed to equilibrate with the sample seawater for at least 8 hours before proceeding. DMG binds free or weakly bound Co as Co(DMG)2, which was the actual measured species in this method. Following DMG addition, 130 µL of 0.5 M N-(2-hydroxyethyl)piperazine-N-(3-propanesulfonic acid) (EPPS, Sigma Aldrich) was added. EPPS was passed through a Chelex 100 resin column prepared through repeated rinses with Optima HCl and NH4 as described by Price et al. (1989). 8.5 mL of sample solution was then transferred using an autosampler (Metrohm 858 Sample Processor) to the Teflon analysis cup of a Metrohm 663 VA Stand fitted with an HMDE. The VA stand interacted with a Metrohm µAutolabIII through an IME663 interface. All instruments were managed using the Nova 2.1.6 software (Metrohm). 1.5 mL of 1.5 M NaNO2 (Merck) solution, also prepared with rinsed Chelex 100 resin (Price et al., 1989), was also added using the autosampler, then this mixture was purged with HEPA-filtered Ultra-high Purity N₂ (Airgas) for 180 seconds. Each sample underwent multiple scans sequences, described as follows. The sample mixture was purged with N2 for 20 seconds. A new Hg drop surface was produced by the working electrode. The sample mixture was then stirred while a voltage of -0.6 volts (V) was applied through the Hg drop for 90 seconds to promote deposition of the Co(DMG)2 complex. Following a 15 second equilibration period, a fast linear sweep from -0.6 to -1.4 V at 5 votls per second (V s-1) reduced Co(II) in the Co(DMG)2 complex to Co(0), which resulted in a reduction peak centered on -1.15 V. Triplicate scans were performed for each sample to determine method precision. 4 subsequent additions of 25 picomolar (pM) CoCl2 (Fisher Scientific) were then added to the sample by the autosampler and each followed by a scan in order to create a standard curve that was specific to each sample.

Analytical blanks were prepared by UV-irradiating filtered seawater for 1 hour, then passing irradiated seawater through a column of cleanly prepared Chelex 100 resin (Biorad) (Price et al., 1989). Seawater was then irradiated again to eliminate any organic material that may have been introduced by the chelexing process. Blanks were then run in the same manner as dCo samples. Multiple blanks were run for each unique combination of reagent batches (DMG, EPPS and NaNO2).

All metadata was copied from the finalized cruise event log.

Other Information:

Limit of detection of Cobalt: 2.7 pM

Intercalibration: Several GEOTRACES community intercalibration standards (SAFe S, D1 and D2) were run intermittently throughout the sample processing period. As intercalibration standards are acidified for storage, samples were adjusted to pH 7.5-8.5 with dropwise additions of a negligible volume of ammonia hydroxide (NH4OH) and UV-irradiated immediately before analysis. They were then measured as dCo samples. All measured standards were within one standard deviation of the consensus values.

SAFe D1 - measured = 42.3 ± 1.4 pM (n=3); consensus = 44.3 ± 4.6 pM (converted from pmol/kg)

SAFe D2 - measured = 45.3 ± 1.2 pM (n=2); consensus = 44.6 ± 2.8 pM (converted from pmol/kg)

SAFe S1 - measured = 4.7 ± 2.8 pM (n=4); consensus = 4.7 ± 1.2 pM (converted from pmol/kg)

Occupation of GEOTRACES crossover station: GP17-OCE Station 1, located at 19° S, 152°W, was a crossover station with the 2018 GP15 expedition. Analysis for both cruises was performed in the Saito lab using the method described for this dataset. The total dCo depth profiles from both occupations are well-matched. A mean offset of 4 pM is present deeper in the water column (≥2200 meters (m)), which is notable but not statistically significant (two tailed t-test, t =2.23, P = 0.051). The average deep water (≥ 2200 m) dCo value is 26.7 ± 3.4 pM (n = 7) for the GP17-OCE measurements and 22.7 ± 2.5 pM (n = 5) for GP15. In both profiles, dCo is undetectable at the surface. In the GP15 profile, the dCo maximum is located at 500 m, while the GP17-OCE measurements reach their maximum at 625 m. dCo remains elevated from the depth of the dCo maximum to around 1100 m in both profiles. The dCo values measured in this mesopelagic maximum are higher for the GP17-OCE expedition than the GP15 expedition. Variability in the upper water column is expected due to interannual and seasonal changes in mixing and production, the former of which is evident from the deviations present in the salinity profile in the upper 400 m between the two cruises. Changes in production and remineralization efficiency could also contribute to the comparatively increased mid-depth dCo maximum we observe in the GP17-OCE data. Regardless, the major vertical features observed on GP15 are robustly replicated from the GP17-OCE re-occupation, and dCo concentrations are comparable throughout the water column. The results of this crossover station comparison provide good confidence for the inter-cruise consistency of our method.

Data consistency: Samples were analyzed with technical triplicates performed through repeated 'no-addition' scans by the electrode system. Consistency of the instrument was monitored by regular measurement of house standards made from 2 batches of UV-irradiated Equatorial Pacific seawater collected on the KN195-05 expedition (dCo = 18.02 ± 3.9 pM, n = 18 and dCo = 12.46 ± 2.0 pM, n = 23).