Sampling
\nSize-fractionated particles were collected using dual-flow McLane Research in-situ pumps (WTS-LV). More details about filter holders and deployments were described in Lam et al. (2017) and Ohnemus and Lam (2015). At most of the stations, two casts of 8 pumps each and two filter holders per pump were deployed to collect samples at 16 depths throughout the water column. At super stations, a 24-depth profile was obtained with three casts. The targeted depths of the wire-out were verified by a self-recording Seabird 19plus CTD at the end of the line and an RBR pressure logger attached to the pump at the middle of the line. The RBR was used for all casts in the entire cruise whereas the CTD was only deployed in the northbound leg due to electronic problems. 142 mm-diameter \"mini-MULVFS\" style filter holders, with multiple stages and baffle systems designed to prevent large particle loss and promote even particle distribution (Bishop et al., 2012), were used. One filter holder/flowpath was loaded with a Sefar polyester mesh prefilter (51 um pore size) and paired Whatman QMA quartz fiber filters (1 um pore size) in series (\"QMA-side\"). The other filter holder/flowpath was also loaded with a 51 um prefilter, but it was followed by paired 0.8 um Pall Supor800 polyethersulfone filters (\"Supor-side\"). A 150 um Sefar polyester mesh was placed underneath all 51 um prefilters and QMA filters as a support to facilitate filter handling. All filters and filter holders were acid leached before use based on the recommended methods in the GEOTRACES sample and sample-handling protocols (Cutter et al., 2010). QMA filters were pre-combusted at 450 \u02daC for 4 hours after acid leaching.
\n
'Dipped blank' filters, including the full filter sets (prefilter on top of paired QMA or paired Supor filters), were also deployed at each cast. Prior to the deployment, these filters were sandwiched in a 1 um polyester mesh filter, placed into acid-leached perforated polypropylene containers, and attached to a pump frame with plastic cable ties. All dipped blank filters were exposed to the seawater for the same amount of time, processed and analyzed as regular samples. As process blanks, dipped blank filters were used for blank subtraction, calculations of uncertainties, and determination of detection limits. A total of 33 dipped blank filter sets were collected and used for blank subtraction and determination of uncertainty and detection limit (Table 3).
\n
In this dataset, data reported from the 51 um prefilter are referred to with a LPT suffix to indicate large particulate total concentrations (>51 um); data reported from the main filters (QMA\u20141-51 um \u2014or Supor\u20140.8-51 um) are from the top filter of the pair only, and are referred to with a SPT suffix to indicate the small particulate total concentrations.
\n
Analytical Procedures
\nParticulate organic carbon (POC) and particulate nitrogen (PN)
\nPOC and PN sample processing was similar to what was described in Lam et al. (2017). Samples were fumed in a desiccator with concentrated HCl and dried in the oven at 60 \u02daC overnight, and then pelletized with tin discs. Tin disc encapsulated samples, from either one 25 mm-diameter punch from the top SPT QMA filters or the entire LPT silver filters, were measured using a CE Instruments NC 2500 model Carbon/Nitrogen Analyzer interfaced to a ThermoFinnigan Delta Plus XP isotope ratio mass spectrometer (IRMS) at the Stable Isotope Laboratory at University of California, Santa Cruz. Isotopic results obtained from the IRMS were calibrated using reference materials Acetanilide (C\u2088H\u2089NO). The effect of dissolved organic carbon sorption is corrected with isotopic values of dipped blanks. The isotopic data are expressed in the standard delta notation (\u03b4) as per mil deviations (\u2030) with respect to international standards of Pee Dee Belemnite (PDB) and atmospheric nitrogen. The precision of the internal standard (Pugel) analyzed along with the samples in the run is 0.07\u2030 for \u03b413C and 0.14\u2030 for \u03b4\u00b9\u2075N.
\n
Particulate inorganic carbon (PIC)
\nA UIC Carbon dioxide coulometer was used for PIC measurement. Briefly, PIC on SPT QMA punches or 1/16 LPT QMA-side prefilter was converted to CO\u2082 by addition of 2 N sulfuric acid. CO\u2082 produced is carried by a gas stream into a coulometer cell where CO\u2082 is quantitatively absorbed by a cathode solution, reacted to form a titratable acid and measured based on the change in current.
\n
Biogenic silica (bSi)
\nAn alkaline leach with 0.2 M NaOH at 85\u02daC was used to leach bSi for both size fractions prior to the measurement on a Lachat QuikChem 8000 Flow Injection Analyzer at UCSC. A 4-h time-series leaching approach was applied to all samples below 500 m to take into account the contribution from lithogenic Si (Lam et al., 2017), where lithogenic Si was of significance in the overall measurement. The intercept of all points in the time series with corrections given volume changes in NaOH at different sampling points was calculated with linear regression and used to best represent the bSi concentrations, assuming bSi is completely dissolved in 1 hour and lithogenic Si is dissolved at a constant rate during the leach (Bar\u00e3o et al., 2015; DeMaster, 1981).
\n
Particulate trace metals (pTM)
\nThe digestion method of pTM is based on a refluxing method (Cullen and Sherrell, 1999; Planquette and Sherrell, 2012) with light modifications similar to the \"Piranha method\" in Ohnemus et al. (2014). In brief, the Supor filter was adhered to the wall by surface tension in a 15 mL flat-bottom screw-cap Savillex vial to avoid immersion. After 4-h refluxing at 110 \u02daC with an ultrapure (ARISTAR\u00ae or Optima\u1d40\u1d39 grade) 50% HNO\u2083/10% HF (v/v) mixture, digestion acids were transferred into secondary vials and heated to near dryness. The residue was heated in 50% HNO\u2083/15% H\u2082O\u2082 (v/v) to dryness at 110 \u02daC. The final residue was re-dissolved with 2 ml 5% HNO\u2083 spiked with 1 ppb In. Two certified reference materials (BCR-414 and PACS-2) were digested routinely alongside the samples to assure the quality of each digestion. Sample solutions were analyzed using an Element XR high-resolution ICP-MS (Thermo Scientific) at the UCSC Plasma Analytical Facility. Elemental concentrations were standardized using multi-element, external standard curves prepared from NIST atomic absorption-standards in 5% HNO\u2083. Instrument drift and matrix effects were corrected using the internal 1ppb In standard and monitored using a mixed element run standard. Concentrations were determined using external standard curves of mixed trace elements standards.
\n
Sampling equipment: Dual-flow McLane Research in-situ pumps (WTS-LV). More details can be found in the patent description (https://patents.google.com/patent/US20130298702) and official website of the manufacturer (https://mclanelabs.com/wts-lv-large-volume-pump/). Other instrumentation described in 'Instruments' section below.
\n
Problem report:
\n(1) Wire angles during pump deployments
\nSevere wire angle for the aft operations were encountered in the operation and wire angles of more than 20\u00b0 during pumping were a consistent issue in the open Arctic water. The actual depth of the pumps is calibrated according to corrections based on the CTD and/or RBR pressure sensors.
\n
(2) Questionable measurements for certain trace metals
\nProblems related with sampling methods: We used Pall Supor800 polyethersulfone filters to collect pTM particles. Those type of filters are known to have high trace metal blanks for Chromium (Cr) (Ohnemus et al., 2014). Particulate chromium (pCr) concentrations are not good to use unless at shelf/slope stations, where particulate Cr concentrations in the water column are high enough.
\n
Problems related to digestion methods: Thorium (Th) data are much lower than the values reported by Morton during intercalibration. Likely, the digestion method we used without submerging the entire filter in the acid mixture seems not to access all Th in the filter.
\n
Problems related to ICP-MS: The median of nickel (Ni) in the LPT dipped blanks is negative. Due to potential contamination from the Ni cone in the ICP, we tend to have a high Ni blank in the acid blank at the beginning of the run and it is hard to wash it off. It is less problematic for SPT Ni, as it has much higher concentrations than LPT. Although acid blanks were carefully according to the shifting Ni baseline, Ni concentrations in both size fractionations are likely to be underestimations, and Ni data should be used with caution.
\n
(3) Instrumental differences (in-situ pump vs. Go-Flo bottles)
\nOverall, our digestion (Sherrell refluxing) of pump samples have consistently lower concentrations than Morton\u2019s digestion of GO-FLO bottle particles. Given the good reproductivity of Go-Flo replicate particles between us and Morton lab applying different digestion protocols in each lab, it is believed that the concentration differences observed are mainly caused by sampling methods (in-situ pump vs. GO-FLO bottle & 0.8 um Supor vs. 0.4 um Supor). Further investigations and intercalibrations between these two sampling systems are needed in other ocean basins.
Blank subtraction:
\nOutliers in dipped blanks (db), as process blanks, for each measurement type were excluded using Chauvenet's criterion (Glover et al., 2011). The median of db filters, except for LSF bSi, was then used in blank subtraction to take into account of the absorption and potential particle loadings from non-targeted depths, especially from the surface ocean during pump recovery. There were anomalously high bSi concentrations of LSF dipped blanks, which were also observed in the GA03 NAZT and GP16 EPZT sections (Lam et al., 2017; Lam et al., 2015). Blank corrections for LSF bSi were made by subtracting the median of LSF failed pump values (pumps with less than 5% of typical water volume filtered). Given a limited number of db measured over the shelf and slope, and similarities between db in the shelf/slope and basin (outliers removed), for major particle composition, such as bSi, POC and PIC, the median of db filters from all stations was used in the calculation. For particulate trace metals (pTM), the median of shelf/slope and basin db are used in the blank subtraction of shelf/slope and basin samples, respectively.
\n
The isotopic effect of dissolved organic carbon and nitrogen sorption is corrected with mean isotopic values of dipped blanks, which is based on an isotope mass balance between db filters and sample filters. The stable isotopic composition of C (d13C) and N (d15N) for the LPT and SPT particles are given as raw values (*_raw), and corrected for the dipped blank. Due to large differences in db d15N values between the shelf/slope and basin, shelf/slope and basin samples were calculated with different process blanks.
\n
Derived parameters: See Supplemental Files, \"Derived Parameter Calculations\" (PDF).
\n
Error propagation:
\nFor most parameters, we could not routinely run replicates, so almost all errors reported are determined from the standard deviation of dipped blank filters (Table 3), converted to concentrations using volume filtered. This assumes that the blank subtraction is the largest source of error. Errors in d13C and d15N are propagated based on an isotope mass balance between db filters and sample filters. Both raw and db corrected isotope data are reported.
\n
For particulate trace metals, there were a subset of SPT samples that were redigested (N=8) and/or rerun, and therefore, errors are calculated by propagating two major errors: uncertainty in the db subtraction, and the combined uncertainty due to heterogeneity in particle distribution and variations in digestion. The former is determined as the standard deviation of the dipped blank filters. The latter is determined from the median relative standard deviation of all repeat samples for that element. Here, since no LPT replicate samples were available, we applied the latter to all SPT and LPT samples. For most elements, the uncertainty from the variability of digestion and particle distribution was the largest source of uncertainty (Figure 1a &, 1b).
\n
Errors in derived parameters are calculated based on rules of error propagation. Although negative concentrations below the detection limit are set to 0 in the calculation of SPM, their errors are still propagated.
\n
Quality Flags:
\nThe detection limit was defined as three times the standard deviation of the dipped blank filters. Values below the detection limit were flagged as QF=6 in the GTSPP convention (also adopted by SeaDataNet and recommended by the GEOTRACES programme). The percentage of all samples that fell below the detection limit is listed in Table 3\u00a0for each parameter and size fraction. Any variations on the methods for blank subtraction or determination of error can be further described for each parameter, as necessary.
\n
Lab quality control (QC) included running standard reference materials for POC&PN (acetanilide), and for pTM (BCR414 and PACS-2\u2014see table 5), as well as participation in intercalibration exercises. Intercalibration of various parameters has been conducted, including comparing bSi & POC profiles with GN04 data at the crossover station (Liguori/ Puigcorbe Lacueva) and historic suspended POC data in the Canada Basin (Griffith). Comparisons of pTM collected using in-situ pumps with Go-Flo & the NIOZ system (TITAN) pTM data (Morton/ Planquette) were also done, helping check for oceanographic consistency.
\n
All data have been assigned quality flags using the GTSPP convention and interpretation:
\n1 = good\u2014passed lab QC and oceanographically consistent;
\n2 = possibly good\u2014oceanographically consistent, but have minor sampling/instrumental issues;
\n3 = possibly bad\u2014not oceanographically consistent, or have major sampling/instrumental issues;
\n4 = bad\u2014failed lab QC (including all failed pumps when only small or no volume was pumped through the filter), or known issue with samples. For a measured parameter from a failed pump, if the filter sample was analyzed and a non-zero volume was recorded so that a concentration could be derived, this concentration is reported with QF=4. These values are generally not reliable because recorded volumes for failed pumps are probably not correct. If a zero volume was recorded, then a concentration could not be derived, and the value is reported as \"NaN\"\u00a0with QF=4. For a derived parameter from a failed pump, this QF=4 applies if any of the parameters on which it is based had a QF=4 but is not \"NaN\".
\n6 = below detection limit;
\n9 = data missing (including all \u201cnd\u201d). For a measured parameter, this QF applies to lost or missing samples that were not measured.\u00a0 For a derived parameter, this QF=9 applies if any of the parameters on which it is based is missing (QF=9) or is \"NaN\".