Sampling and analytical procedures:
Sample bottles were LDPE and were cleaned in accordance with the GEOTRACES cookbook (http://www.geotraces.org/images/stories/documents/intercalibration/Cookbook.pdf). Sample bottles were stored in dilute (~0.1 N) trace metal grade nitric acid, and were emptied before shipping. Fish surface samples were taken using an all-plastic "towed fish" pumping system as described in Bruland et al. (2005) at approximately 3 m depth. Samples were filtered through an Acropak-200 Supor capsule filter (0.2 µm pore size made of polyethersulfone), and sample bottles were rinsed three times before filling. Samples were acidified at sea with the equivalent of 2 mL optima 12M HCl per liter of seawater (resulting in pH 1.7-1.8). Samples were sitting acidified for several months before analysis back in the laboratory.

Samples were analyzed with an adaptation of Biller and Bruland (2012) as described in Parker et al (2016). For this method, 24 mL of sample was pipetted into 30 mL teflon jars, and spiked with In to 5 nM (natural concentrations of In are sub pM, so this spike overwhelms the natural concentrations). This In spike was developed by Middag et al. (2015) and we used it as a check of consistent recovery across different resin columns and sampling days.

Samples were pre-concentrated on resin columns of Nobias-chelate PA1 resin (IDA and EDTA functional groups). We used 8 columns at a time, and therefore pre-concentrated 8 samples at a time. Columns were stored in dilute (~0.1N) nitric acid, and every day before samples were analyzed, the columns were cleaned with alternating pH 6 seawater, 1N nitric acid eluent, and pH 6 rinse/conditioning solution (0.05M NH4Ac).

Just before pre-concentration on resin columns, 8 samples were buffered to pH 5.7-6.2 with an NH4Ac buffer, inverted and weighed. For each set of samples, the columns were conditioned with the rinse/conditioning solution (to get the columns the correct pH before the sample arrives), then loaded with the sample (~40 minutes), then rinsed with the rinse/conditioning solution (to remove the salts) and then eluted with ~1 mL of 1M quartz-distilled nitric acid with 10 ppb Rh internal standard. The eluent vials were weighed before and after elution, and the (now empty) sample jars are weighed as well. This method results in a concentration factor of about 24.

Column blanks were determined for each column by following the same procedure for samples, but loading a low-metal seawater for just 1 min rather than the 40 min a sample usually takes to load. Reagent blanks were determined to account for the reagents that were added directly to the sample. They were measure by analyzing low-metal seawater twice normally (rx1) and twice with double the reagents (rx2; reagent doubled include the acid used to originally acidify the sample, the buffer used to get the sample to pH 5.7-6.2, and the In spike). In this way, the rx2 sample was the correct pH for analysis, but contained twice the amount of blank from the reagents. The difference between the rx2 and rx1 was used as the reagent blank.

Standards were made by spiking low-metal seawater and analyzing the standards in the same manner as the samples. The resulting standard addition curve was compared with a standard curve made in clean eluent. This allowed us to calculate and monitor the percent recovery of each metal over time. We also had a Mo standard curve that allowed us to correct for the MoO interference on Cd.

Detection Limits and Accuracy:
Detection limits were calculated using 3 times the standard deviation of the column blanks. This was done across three analytical runs that focused on this dataset; the precision reported in this dataset are the average detection limit across those three analytical runs. See the Supplemental File "Detection_limits.csv".

For Mn, Ni, Cd, and Pb, accuracy is assessed by measuring reference samples and comparing with consensus values. For Sc, Y, Ce, and La, for which there are no consensus values, accuracy is assessed by measuring reference samples and comparing with values measured previously by the PI and by scrutinizing the blanks and standards for any anomalies. Though there has yet to be an official intercalibration of these metals, previously we have measured Sc, Y, and La depth profiles that are consistent with other published datasets. Official intercalibration of these datasets with other PMT analysts for as many metals as possible are to come.

Known Issues/Problems:
The northern half of the transect was analyzed without UV irradiating the samples, and the southern half of the transect was analyzed with UV irradiating the samples (2 hours in teflon jars with quartz lids). This is not expected to make a difference for any of the metals reported here and was only done to better analyze Cu and Co, for which it is necessary with this method (Biller and Bruland, 2012).