Sampling Methods at Sea:
The GA13 team followed cookbook methods for sampling from the conventional Niskin rosette. To sample for ²³⁰Th, ²³²Th, and ²³¹Pa, samples were taken from the stainless steel rosette and filtered directly from the Niskin bottle through a 0.45 micron AcroPak 500 filter. Tygon tubing was used to attach the filter to the opening of the Niskin bottle. The samples were collected into 4-liter (L) cubitainers. These cubitainers were then double bagged into plastic bags. These samples will then be shipped back to the lab, and stored until analysis.

Samples were not acidified at sea, but once returned to the lab in Mississippi, the samples were acidified to a pH < 2 with Optima HCl and left to equilibrate at least 3 months before analysis.

Analytical Methods:
In the on-shore laboratory, seawater samples were weighed to determine sample size, taking into account the weight of the cubitainer and of the added acid. Then, weighed aliquots of the artificial isotope yield monitors Th-229 (10 picograms (pg)) and Pa-233 (~0.8 pg) and 10 milligrams (mg) dissolved Fe were added to each sample. After allowing 1 day for spike equilibration, the pH of each sample was raised to 8-9 by adding ~10-14 milliliters (mL) of concentrated NH4OH (Fisher Scientific OPTIMA grade) which caused iron (oxy)hydroxide precipitates to form. Each sample cubitainer was fitted with a nozzle cap, inverted, and the Fe precipitate was allowed to settle for 2 days. After 2 days, the nozzle caps were opened and the pH~8-9 water was slowly drained, leaving only the iron oxyhydroxide precipitate and 250-500 mL of water. The Fe precipitate was transferred to centrifuge tubes for centrifugation and rinsing with Milli-Q H2O (>18 MΩ) to remove the major seawater ions. The precipitate was then dissolved in 8M HNO3 (Fisher Scientific OPTIMA grade) and transferred to a Teflon beaker for acid digestions. First the nitric sample solution was dried to near dry at 180-200°C. The sample was then taken up in 1-2 mL 8 M HNO3, the beakers capped and the samples refluxed at 180°C for at least 3 hours. The sample was then cooled, uncapped, retaining all sample drops in the beaker, heated again to 180°C for an HF (Optima) addition of 1 mL. This solution was dried at 180°C to a white precipitate that is dissolvable in optima HCl. After total dissolution of the sample, another precipitation of iron (oxy)hydroxide followed and the precipitate was washed with Milli-Q H2O, centrifuged, and dissolved in 8M HCl (Fisher Scientific OPTIMA grade) for a series of anion-exchange chromatography using 6 mL polypropylene columns each containing a 1 mL bed of Bio-rad resin (AG1-X8, 100-200 mesh size) and a 45 μm porous polyethylene frit (Anderson et al., 2012). The final column elutions were dried down at 180-200°C in the presence of 2 drops of concentrated HNO3 (Fisher Scientific OPTIMA grade) and taken up in 1.0 mL of 0.32 M HNO3 (Fisher Scientific OPTIMA grade) for mass spectrometric analysis. Digestions and columns were done in a standard fume hood, but whenever samples were sealed (i.e., no acid fumes) they were handled in a benchtop HEPA-filtered laminar flow hood.

Concentrations of Th-232, Th-230, and Pa-231 were calculated by isotope dilution, relative to the calibrated tracers Th-229 and Pa-233 added at the beginning of sample processing. Analyses were carried out on a Thermo-Finnigan ELEMENT XR Single Collector Magnetic Sector ICP-MS. This model lacks the high-performance Interface pump (Jet Pump Aridus I™), but we did utilize the specially designed sample (Jet) and skimmer (X) cones which increased sensitivity. All measurements were made in low resolution mode (∆m/M≈300), peak jumping in Escan mode across the central 5% of the flat-topped peaks. Measurements were made on a MasCom™ SEM; Th-229, Th-230, Pa-231, and Pa-233 were measured in Counting mode, while the Th-232 signals were large enough that they were measured in Analog mode. Two solutions of SRM129, a natural U standard, were run multiple times throughout each run. One solution was in a concentration range where U-238 and U-235 were both measured in Counting mode, allowing us to determine the mass bias/amu (typical values varied from -0.5%/amu to 0.2%/amu). In the other, more concentrated solution, U-238 was measured in Analog mode and U-235 was measured in Counting mode, yielding a measurement of the Analog/Counting Correction Factor (typical values varied from 0.9 to 1.1). These corrections assume that the mass bias and Analog/Counting Correction Factor measured on U isotopes can be applied to Th and Pa isotope measurements. Each sample measurement was bracketed by measurement of an aliquot of the run solution (0.32 M HNO3), which was used to correct for the instrumental background count rates. Tailing of Th-232 into the minor Th and Pa isotopes was monitored by counting at the half-masses surrounding Th-230 and Pa-231. Tailing corrections were typically small (<0.5% and often negligible).

Water samples were analyzed in batches of 12 to 20 (5 batches total). Procedural blanks were determined by processing 4-5 L of Milli-Q H2O in an acid-cleaned cubitainer acidified to pH ~2 with 6 M HCl (Fisher Scientific OPTIMA grade) as a sample in each batch (n = 10 total procedural blanks). In addition to the procedural blanks, with every batch an aliquot of one of two intercalibrated working standard solutions of Th-232, Th-230, and Pa-231, SW STD 2010-1 referred to by Anderson et al. (2012) and SW STD 2015-1 which has ~6 times lower Th-232 activity, were added to acidified MQ-H2O and treated like a sample. Samples were corrected using the procedural blank analyzed during each batch. Procedural blank, limit of detection and the results of the reference material solutions are reported in the Supplemental File "987617_v1_ga13_dissolved_th_pa_LODs.pdf". The limit of detection (LOD) is the smallest quantity of each isotope in samples that can reliably be detected or that can be statistically distinguished from a procedural blank. The LOD was considered to be 2 standard deviations above the average of the procedural blanks and we have scaled the limit of detection into the equivalent concentration in a 5 liter sample. Our results for SWS2010-1 are within the consensus range from the intercalibration exercise (Anderson et al., 2012). Consensus values for SWS2015-1 have not been yet been coordinated but they agree with the reports of the LDEO lab. A total of 134 samples were collected. Human error during the beginning of the analysis period caused a certain number of batches/elemental fractions to be lost during column chromatography (in the case of Pa-231) or clearly contaminated (in the case of Th-232/Th-230) and this resulted in 34 Pa-231 samples, 12 Th-232 samples and 2 Th-230 samples to be reported as missing data (flag = 9).