Sampling and analytical procedures:
The sampling methodology followed the GEOTRACES cookbook guidelines. Large and small particle particulates were collected using dual-filter head in situ McLane pumping systems (also see Lam datasets from HLY1502 for additional pumping system details). Generally, shallow total Th-234 samples (<1000 m) were collected from Niskins on the ODF Rosette and deep total Th-234 samples (>1000 m) were collected from Niskin bottles hung above the in situ pumps. However, on occasions where pump associated Niskins did not close prior to recovery, the ODF Rosette was used to obtain deep water for total Th-234. The following sampling methods used on HLY1502 for total and particulate Th-234 followed closely those detailed in Black et al., 2018. Where cruise-specific differences exist, they have been noted here.

Total Thorium-234 (Bottle, Pump, Sub-ice Pump, UW) Sampling:
Total Th-234 samples were taken at 31 of the 66 stations occupied on the HLY1502 campaign. Typically, these samples came from single Niskin bottles attached to a wire (i.e. above the McLane pumps) or Niskins in the standard ODF rosette configuration These totals are listed under ‘Th_234_T_CONC_BOTTLE. At shallow shelf and near-slope stations (2, 3, 6, 10, 60, 61, 66), total Th-234 samples were taken at only 4 to 12 discrete depths, typically using the ODF rosette. The Pacific endmember Station 1 similarly had a single ODF cast for total Th-234 with 12 depths. A single sample was also taken at this location from the ship’s underway system (0.5 m). At most of the deeper basin stations (14, 19, 26, 30, 32/34, 38, 43, 46, 48, 52, 56, and 57), the full water column was sampled with either 2 casts (1 ODF and 1 Niskin above pumps, 20 depths total) or 3 casts (1 ODF and 2 Niskin above pumps, 28 depths total).

For better resolution within the marginal ice zone and areas with permanent ice cover, total Th-234 samples were taken from the ODF rosette and 'Th_234_T_CONC_PUMP' or 'Th_234_T_CONC_SUBICE_PUMP' samples were taken using the Be-7 submersible pumping system. In the marginal ice zone on the northern transect (stations 7, 8, 9, 12, 17), additional samples at 1-2 depths were taken using the submersible pumping system (See the Be-7 report by Kadko group for more details). On the southbound transect, marginal ice zone samples came from 4 depths at stations 51, 53, 54 using the ODF rosette. For under-ice samples, a hole was made using a gas-powered ice auger and the submersible pump was lowered into the hole. Samples were taken at three discrete depths at ice stations 31 and 33. Note that the beryllium pump samples were not given GEOTRACES numbers and this field in the datasheet has 'NaN'.

At each discrete depth, ~4 L of water was taken from the corresponding Niskin bottle or pump tubing after rinsing the sample bottle three times. The 4L sample bottles were mass-volume calibrated prior to the cruise and were filled to the marked calibration line.

Particulate Thorium-234 and Thorium-228 (SPT and LPT) Sampling:
Size-fractionated particulate Th-234 and Th-228 samples were taken at 20 of the 66 stations occupied using high-volume McLane pumps. The filter heads each contained a 51 μm pore size pre-filter followed by either a Supor filter or a pre-combusted and acid-leached QMA filter with a nominal pore size of 1 μm. Filter heads were pumped down and removed from the filter heads in the designated trace metal clean ‘bubble’ space by the Lam group (see particulate trace metal dataset information from Lam group). The filters were placed in plastic 142 mm petri dishes and brought to the short-lived radionuclide van (Café Thorium) for processing. The material on the 51 μm pre-filter from the Supor filter head was rinsed onto silver (Ag) filters using 0.1 μm filtered seawater and dried. The 142 mm QMA filter was oven dried and subsampled with a 25 mm punch for Th-234 analysis. The remainder of the 142 mm filter was sealed with tape and stored for Th-228 counting months later. The average sample volume through the 51 μm pre-filter was 402 L and for the area of the QMA subsample was 34 L. The average sample volume through the area of the entire 142 mm QMA was 871 L. These volume averages only include samples flagged as (2) or (3), and not (4) or (9). See data flags section for further information.

Total Thorium-234 General Analytical Procedures:
Th-234 was determined by the widely-adopted 4 L method (Buesseler et al., 2001), which has been utilized previously for other GEOTRACES efforts (e.g. Owens et al., 2015 and Black et al., 2018). An exact 1 mL aliquot of Th-230 (50.39 dpm per g) was used as the yield monitor and added during initial acidification of the samples. QMAs (25 mm) were used to collect the precipitate from the 4L process and immediately dried. Once dried, they were mounted onto plastic 25 mm discs, covered with a mylar layer and 2 layers of aluminum foil, and immediately beta counted at sea. The filters were counted again 5 to 6 months later to quantify the background radioactivity due to the beta decay of long-lived natural radionuclides that are also precipitated. The mean value of the at-sea counts (decay-corrected to the time of collection) minus the background value for each filter is reported as the Th-234 activity (mBq per kg). Activities for Th-234 are generally reported in dpm per L, but have been converted here using a standard density of 1.025 kg per L and 1 dpm = 16.667 mBq. Data are decay corrected to the mid-point time between when bottles 1 and 12 were fired for shallow ODF casts and when the messenger was dropped for deep pumping casts.

To determine Th-234 activity deficits, U-238 (its parent isotope) activities were calculated using a standard uranium-salinity relationship (Owens et al., 2011). Salinities measured on this campaign ranged from 24.4 to 35.1 and calculated U-238 activities from 1.60 dpm per L to 2.44 dpm per L. While this salinity range is rather large compared to those found on previous GEOTRACES campaigns in the Pacific and Atlantic, Not et al. (2012) showed that the relationship held for sea ice, sea ice brine, and subsurface water samples from the Arctic ranging in salinity from ~0 to 135. The efficiency of the beta detectors was determined by minimizing the Th-234 deviation from U-238 for samples collected from regions of the water column where Th-234 and U-238 are expected to be at equilibrium. These included depths below 1000 m and above 400 m off the seafloor that were not near the coastal shelf. For these sample depths (n= 42) the mean derived U-238 activity and standard deviation (s.d.) were 2.431 ± 0.002 dpm per L, a value well within observed natural ranges (Owens et al., 2011).

The reported Th-234 activities were corrected for the chemical recovery efficiency of the 234Th-Mn precipitate method. To determine the percent recovery of the added Th-230 tracer, the method detailed in Pike et al. (2005) was followed without the initial ion exchange column chemistry steps. Filters were leached in a nitric acid-hydrogen peroxide solution and 2 g of a Th-229 yield monitor (activity of either 68.87 dpm per g or 76.27 dpm per g) was added. Samples were then sonicated for 20 min, allowed to stand covered overnight, diluted, and prepared for analysis by ICP-MS. The mean chemical recovery for all reported values was 88.7% and the median recovery was 92%.

Particulate Thorium-234 and Thorium-228 General Analytical Procedures:
Once the silver filters and the 25 mm QMA subsamples were dried, they were mounted onto plastic 25 mm discs, covered with a mylar layer and 2 layers of aluminum foil, and immediately beta counted at sea. They were counted again 5 to 6 months later at the Buesseler beta counting facility at Woods Hole Oceanographic Institution. All data were decay corrected back to the mid-pumping times.

The basic analytical methodology for small particle Th-228 (142 mm QMA filters) has been detailed in Maiti et al. (2014). This method was adapted to measure Th-228 on large particle Ag filters (25 mm) by Dr. Black and the members of the Buesseler and Charette labs at Woods Hole Oceanographic Institution. Details of the measurement chamber construction and the calibration of the Radium Delayed Coincidence Counters can be found in the appendices of Black (2017). The RaDeCC is an alpha scintillation counter that distinguishes decay events of short-lived radium daughter products based on their contrasting half-lives. This system was pioneered by Giffin et al. (1963) and adapted for radium measurements by Moore and Arnold (1996). The RaDeCC method was chosen here because it is well suited for sequential measurements that involve Th-228 as an intermediate (e.g. Th-234, Th-228, particulate carbon), as there is no sample loss or chemical interaction. Large particle samples (Ag filters) were used for Th-234 beta counting at-sea and in the lab 5-6 months later. They were then demounted, weighed, and placed into a 25-mm chamber for use with the RaDeCC systems. The 142 mm QMA filters were removed from the petri storage dishes and individually counted in larger chambers made for the RaDeCC systems. All particulate samples were counted for an average of 23 hours. After measurement of the 25 mm Ag filters, masses were recorded again to ensure than any mass loss could be monitored (although no significant mass loss was found).

Blanks, uncertainties, internal consistency, and detection limits:
Particulate and total Thorium-234:
Thirty-three blank particulate samples (dipped blanks) were collected for each particle size using extra filter heads deployed with the McLane pumps, but without a connection to the pumping systems. On ship, blank QMA filters averaged, in counts per minute, 0.33 cpm ± 0.05 (s.d.) and after 5 to 6 months the background count average was 0.29 cpm ± 0.04. The on-ship cpm values were within 1 s.d. of the final cpm values, typical ‘empty’ detector cpm (see below), and cpm for non-dipped blank filters (i.e. unused QMA filters for total and small particle analysis and unused Ag filters for large particle analysis)

There was a minute difference between the Ag filter blanks when first measured on-ship (0.30 cpm ± 0.04) and after 5 to 6 months (0.25 cpm ± 0.03), however, no correction was made to the data. These blank averages fall about the average for empty detectors (i.e. the detectors are run with no samples inside for a period of 24-48 hours). The average and s.d. of the empty detectors just prior to the running of any samples on this campaign were 0.28 cpm ± 0.04, which is indistinguishable from the initial blank filter measurement average. Furthermore, there has been no evidence of significant addition (e.g. sorption) of Th-234 to the blank Ag filters on previous GEOTRACES campaigns and a ‘blank’ value of 0.05 cpm is only 1% of the average (3.7 cpm) and 2% of the median (2.1 cpm) for the sample Ag filters. Only 10 Ag sample filters had cpm less than 0.5 (potential blank adjustment = 10%). Our uncertainties, which are discussed more in the next section, are set at a minimum of 5% even when propagated counting uncertainties are lower. While we don’t think the data supports a significant Ag filter blank, a slight increase in the cpm would be within our assumed uncertainties for almost all of the samples.

Limits of detection are not reported because they are not applicable to the Th-234 beta counting method and for total Th-234, specifically, there is never an instance where the 4L volume results in a shipboard sample activity that is anywhere close to the limits of the detectors or a ‘blank’ or unused QMA value. A 'non-detect' for Th-234 or a case where there is no Th-234 present (initially or after 6 months of decay) will still result in a measurable amount of background radioactivity due to the beta decay of long lived natural radionuclides that are also collected on the pump filters. These background values are utilized and therefore, they are not reported as a non-detection of Th-234. The net cpm for total and particulate Th-234 samples here was always higher than 0.05 cpm and in almost all cases was well above this value. Only 4 particulate samples had net cpm between 0.06 cpm and 0.2 cpm. See the previous data flag explanation for our sample volume limits (i.e. the pump volume below which the data are likely unreliable and unrepresentative).

Five 'low-level' uranium standards, with activities close to those measured for total and particulate samples, and five 'high-level' standards ranging from 238 dpm to 365 dpm were run on the RISØ detectors to confirm correct operation and to determine detector to detector variability. These uranium standards have been used for all GEOTRACES cruises performed by the Buesseler lab. These standards were run at the beginning and end of this cruise, as well as periodically during the cruise when sample demands were lower. Analysis of the lower activity uranium standard data suggested that a minimum 5% detector uncertainty should be used. Since the counting uncertainty for total Th-234 samples was always below 5% (square root of the number of counts), the uncertainty on each total Th-234 measurement was set at 5%. Some of the particulate Th-234 samples with relatively low activities had counting uncertainties above 5% and in these cases the counting uncertainty was used as the final measurement uncertainty.

Counting uncertainty is generally the largest source of uncertainty so whenever possible samples were counted until errors were below 5%. For the low-volume (i.e. 4L) total Th-234 samples, all filters were beta counted twice for a minimum of 12 h at sea. As long as the calculated gross counts per minute from these 2 measurements were within 10%, they were averaged for the at-sea Th-234 value. Instances where the replicates were different by more than 10% were individually evaluated (i.e. the raw counting data) and re-counted as needed. The few instances where the 10% difference was noticed occurred where the activities were lowest and the Th-234 deficits relative to U-238 activities the largest (i.e. over the Arctic shelves). Depths were sometimes occupied twice on different casts at the same station, such as with Station 30 (see crossover evaluation). The 225 m depth was sampled in this way and the independent replicates (i.e. two separate bottles at two separate times) are well within uncertainties (see blue dots in 3-paneled figure in Question 7). Good agreement and consistency between overlapping depths on subsequent casts at a given station were found.

In addition, to assess 'within bottle' variability and replication, we took 4, 4L samples from a Niskin bottle deployed at 2019 m at station 48. The average value was 38.4 mBq per kg (2.36 dpm per L) with a standard deviation of 1.3 (0.08) and RSD of 3.4%. A single salinity sample was taken from this bottle, but we recommend taking 4 in the future for comparison, each one after a 4L sample is taken.

Particulate Thorium-228:
For the small particle QMA filters (142 mm), 34 blanks were assessed and for the large particle Ag filters (25 mm), 30 blanks were assessed. Most of these were dipped blanks collected for each particle size using extra filter heads deployed with the McLane pumps, but without a connection to the pumping systems. A few failed pumping effort filters were also counted and assessed (volume < 2 L). Unlike with the shorter-lived Th-234 measurements, which are not generally volume- or activity-limited when counted and which are measured multiple times (subtracting out the influence of the filter itself), Th-228 particulate sample counts per minute (cpm) often drop to cpm similar to those from dipped blanks. We therefore assess the ‘below detection limit’ designation (flag 6) with respect to the cpm of the corrected Rn-220 (Th-228 daughter) RaDeCC measurement for samples and blanks.

Blank QMA counts averaged 0.012 cpm ± 0.007 cpm (1 standard deviation) and blank Ag filter counts averaged 0.009 cpm ± 0.004 cpm. Empty chambers produced similar results (e.g. the 25 mm chambers for Ag filters averaged 0.008 cpm ± 0.005 cpm), suggesting sorption of Th-228 was not a significant issue. We have blank-corrected all of the measurements reported here to account for the influence of the blank filter cpm on the (usually) single RaDeCC measurements. If the blank-corrected sample cpm (initial cpm – blank average) was within the blank standard deviation of zero cpm, the sample was flagged as (6), non-detect. The resulting sample activities were reported here to show that a sample measurement was made, however, the results are negative and should not be used. If the blank-corrected sample cpm was more than the blank standard deviation from zero cpm, the sample activity was reported as (2) or (3). The (3) flag designation is explained above.

Three internal fiber cartridge standards were measured bi-weekly during the time when the large and small particle filters were measured in the counting facility at Woods Hole Oceanographic Institution. The standards were counted on all detectors and used to monitor any potential changes in detector efficiency.

As a part of the calibration process and method development, 8 large particle Th-228 samples were measured at least once on every detector. A few of these samples were measured 3-4 times on a single detector over a few months. These 8 samples were then digested and processed with anion exchange columns to prepare them for traditional alpha spectrometry measurements in the Buesseler lab at Woods Hole Oceanographic. The results from the RaDeCC and traditional counting methods were compared and detector-to-detector consistency (replicability) was assessed.

Since this is a new method, there no certified reference materials for particulate Th-228. However, details can be found in the Charette dataset for HLY1502 radium analyses on how samples of Ra-228, the parent of Th-228, have been intercalibrated using the same instrumentation (RaDeCCs).

Intercalibration and Crossovers: For more on intercalibration and stations, refer to the two attached Intercalibration Reports, under "Supplemental Files".