Dataset: GT10-11 - Particulate composition

Sampling and Analytical Methodology: 

1. Sampling:

Size-fractionated particles were collected using McLane Research in-situ pumps (WTS-LV) that had been modified to accommodate two flowpaths (Lam and Morris Patent pending).  Typically, two casts of 8 pumps each and two filter holders per pump were deployed to collect a 16-depth profile.  The wire-out was used to target nominal depths ('depth_n') during deployment. A self-recording Seabird 19plus CTD was deployed at the end of the line for both cruises. On the second cruise, three RBR data loggers were also attached to pumps #2, #5, and #8 to help correct for actual depths ('depth') during pumping. For the first cruise (KN199-4), the recorded CTD depth was near its target depth and had a small standard deviation over the course of pumping, so we report the target depth ('depth_n') as the final depth ('depth').  For the second cruise, the target depth ('depth_n') is not the same as the final depth ('depth'), since some casts experienced significant wire angles (especially in the western boundary currents), so we corrected for the wire angle based on the recorded depths in the three data loggers and terminal CTD.  

Filter holders used were 142 mm-diameter 'mini-MULVFS' style filter holders with two stages for two size fractions and multiple baffle systems designed to ensure even particle distribution and prevent particle loss (Bishop et al. 2012).  One filter holder/flowpath was loaded with a 51micron Sefar polyester mesh prefilter followed by paired Whatman QMA quartz fiber filters. The other filter holder/flowpath was also loaded with a 51micron prefilter, but followed by paired 0.8micron Pall Supor800 polyethersulfone filters.  These filter combinations were chosen as the best compromise after extensive testing during the intercalibration process (Bishop et al. 2012).  Each cast also had a full set of 'dipped blank' filters deployed.  These were the full filters sets (prefilter followed by paired QMA or paired Supor filters) sandwiched within a 1micron polyester mesh filter, loaded into perforated polypropylene containers, and attached with plastic cable ties to a pump frame, and deployed.  Dipped blank filters were exposed to seawater for the length of the deployment and processed and analyzed as regular samples, and thus functioned as full seawater process blanks.  

All filters and filter holders were acid leached prior to use according to methods recommended in the GEOTRACES sample and sample-handing Protocols (Geotraces 2010).

In this dataset, data reported from the 51micron prefilter are referred to with a 'sink' suffix to indicate the sinking size fraction (>51micron); data reported from the main filters (QMA - 1-51micron - or Supor - 0.8 micron-51micron) are from the top filter of the pair only, and are referred to with a 'susp' suffix to indicate the suspended size fraction.

2. Analytical Methodology:

2.1. Opal (amorphous silica)
A 1/16 subsample of the top 0.8micron Supor filter, equivalent to ~30L, or of the 51micron polyester prefilter above the QMA filter, equivalent to ~60L, was analyzed for amorphous/biogenic Si concentrations using standard spectrophotometric detection of the blue silico-molydate complex. We slightly modified DeMaster’s time-series approach developed for marine sediments to correct for the contribution of lithogenic silica to the leachate (Demaster 1981), using 20mL 0.2N NaOH at 85°C for the leach, and taking a 1.6mL subsample every hour for 3 hours.  The slope of the fit was negligible for shallow samples but generally increased with depth of the sample, a reflection of the increasing importance of lithogenic silica to total silica with depth; we thus proceeded with a 1 hour incubation time for shallow cast samples (<900m), and continued the time-series approach for deep cast samples (>900m).  Dipped blank filters from both shallow and deep casts were used to correct the Supor data.  For >51 micron samples on polyester prefilters, blank corrections were made using the average failed pump values (pumps that never turned on, or that shut off after <5% of programmed water volume was filtered) because of anomalously high prefilter dipped blank values.

The detection limit was three times the standard deviation of dipped blank samples and was 0.26 and 0.19 micronol Si/filter for shallow and deep Supor dipped blank subsamples, respectively, and was 1.05 and 0.35 micronol Si/filter for shallow and deep polyester prefilter failed pump subsamples, respectively.  Values below the detection limit are flagged (QF=4).

The mass of biogenic silica (opal) was calculated assuming a hydrated form of silica: SiO2.(0.4 H2O) (Mortlock and Froelich 1989), or 67.2 g opal/mol bSi.

We use the standard deviation of the dipped blank filters used in the blank subtraction to estimate error in the reported opal value.  The appropriate filter-matched standard deviations were converted to µg opal/L using volume filtered and reported in the opal_susp_sd, opal_sink_sd columns, as appropriate.

2.2 Total Particulate Carbon (TPC)
Total particulate carbon was measured using a Flash EA1112 Carbon/Nitrogen Analyzer using a Dynamic Flash Combustion technique at the WHOI Nutrient Analytical Facility. Suspended particles (1-51micron) were measured for total particulate carbon using one or two 12mm-diameter punches from the top QMA filter, representing the equivalent of 10-20L of material. For the >51micron size fraction, particles from half or a whole 51micron polyester prefilter were rinsed at sea with 1micron-filtered seawater onto a 25mm 0.8micron Sterlitech Ag filter or 25mm pre-combusted Whatman QMA filter before being dried at 60°C.  A quarter of the Ag or QMA filter containing rinsed particles was analyzed for total particulate carbon, typically representing 60-120L of material.
We use the standard deviation of the dipped blank filters used in the blank subtraction to estimate error in the TPC measurement.  For TPC in the suspended (0.8-51 micron) size fraction (TPC_susp), the standard deviation of 8 dipped blank or failed pump QMA filters (6.95 micronol C/filter for QMA).  For TPC in the sinking (>51 micron) size fraction, the standard deviation of 8 dipped blank filters rinsed onto Ag and onto QMA were 0.52 micronol C/filter and 0.59 micronol C/filter, respectively.  The appropriate filter-matched standard deviations were converted to µg C/L using volume filtered and reported in the TPC_susp_sd, TPC_sink_sd columns, as appropriate.

2.3 Particulate Inorganic Carbon (PIC) and CaCO3
PIC was measured using one of four methods noted in data column 'PIC_method':
1.    Directly by coulometry (measurement of CO2 following closed-system conversion of PIC to CO2 upon addition of 1N phosphoric acid to a QMA punch or 1/16 polyester prefilter) (Honjo et al. 1995)
As CaCO3 from the measurement of salt-corrected Ca (using Na for salt correction) (Lam and Bishop 2007) on a 1/16 subsample of Supor or polyester prefilter or 2 QMA punches (2% of filter area) and measured by:
2.    ICP-MS at WHOI following a 2 hr room temperature 25% glacial acetic acid leach, which was dried down and brought back up in 5% HNO3
3.    ICP-MS at WHOI following a 5% (0.6N) HCl leach for 12-16 hrs at 60°C and diluted to 1% HCl
4.    ICP-AES at Boston University following a 5% HCl leach overnight at room temperature
Intercomparability between methods was tested by running select samples in replicate by different methods.  PIC_methods 1,2,3 had good intercomparability.  There was a 20-30% offset in samples analyzed by PIC_method=4 compared to the other methods.  Data from PIC_method=4 were normalized using replicate analyses from a depth profile (GT11-8 for Supor samples; GT11-24 for prefilter samples).  The resulting dataset has improved oceanographic consistency.  When available, the reported error is the standard deviation of replicate analyses (after normalization); if no replicate analyses were made, the reported error is the standard deviation of the dipped blank filters used in the blank subtraction for each method and filtertype, adjusted for volume filtered. The standard deviation of the blank subtraction was 18.3 µg PIC/QMA filter for coulometry and 3.0 µg PIC/prefilter or 11.0 µg PIC/Supor filter for ICP-MS. For ICP-AES, the standard deviation of the blank subtraction was 190 µg PIC/QMA filter, 61 or 12 µg PIC/Supor filter (depending on the run), and 7.1 µg PIC/prefilter.

The mass of CaCO3 is calculated stoichiometrically from the mass of PIC (CaCO3 [µg/L] = 100.08 g CaCO3/12 g C * PIC [µg/L])

2.4 Particulate Organic Carbon (POC)
POC is calculated as the difference between TPC (see 2.2) and PIC (see 2.3). Any negative numbers were set to 0.  Errors were propagated from those from TPC and PIC.

2.5 Particulate Organic Matter (POM)
POM is calculated from POC (see 2.4) using a weight ratio of 1.88 g POM/g POC (Lam et al. 2011).

2.6 Particulate trace metals (pTM)
Methods for particulate trace metal (pTM) digestion and analysis are described in (Ohnemus et al. submitted) and briefly below.  Total pTM concentrations in the suspended fraction (*_susp) were analyzed from 1/16 subsamples of the top Supor (0.8micron) filter.  pTM totals in the sinking size fraction (*_sink) were analyzed from 1/8 subsamples (typically ~150L) of the QMA-side 51micron pre-filter. Pre-filter particles were rinsed at sea onto 25mm Supor (0.8micron) filter discs using 0.2micron-filtered surface seawater collected using clean techniques from an underway Fish system (Bruland et al. 2005).  In Teflon vials (Savillex), samples were first digested using a 3:1 mixture of sulfuric acid and hydrogen peroxide at high heat to remove the Supor filter matrix, then dried. A mixture of HCl/HNO3/HF acids (all acids 4N, heated to 135°C for 4 hrs) was used to digest the material in the remaining pellet, which was then dried, reacted with a small amount of 50% HNO3/15% H2O2 to remove any remaining organics, dried, and resuspended in 5% HNO3¬ for analysis via ICP-MS (Element 2, Thermo-Finnigan).  Elemental concentrations were standardized using multi-element, external standard curves prepared from NIST atomic absorption-standards in 5% HNO3¬.  Standard curves were fitted using weighted least squares fits that consider instrument analytical uncertainties.  Data are reported in units of [nmoles per L of seawater filtered] and have had the median of multiple (typically 12-16) dipped blank filters (analyzed using identical methods) subtracted.  The detection limit of most elements was defined as 3 times the standard deviation of 12-16 dipped blank filters.   We define the detection limit for Ti to be one standard deviation of the dipped blanks filters.  The median and standard deviation of the dipped blank filters and detection limits for the Supor (0.8-51 micron size fraction) and polyester prefilter (>51micron size fraction) are reported in the following table in nmol per whole filter (NB: filter area of 142mm filter is 158.4 cm2):

Table 1: Dipped (filter) blanks and detection limits for 0.8-51um samples for 142mm filters (nmol/filter) (Table 1 PDF file)

In table 1 the errors (*_susp_error, *_sink_error) are reported as the 1-sigma variation of propagated instrumental analytical and standard curve uncertainties, and the variation in subtracted dipped blank filters.

The completeness of our digestion method for trace elements was assessed by digesting and analyzing three certified reference materials: a freshwater plankton CRM from the European Commision Community Bureau of Reference (BCR-414), and two marine sediments from the National Research Council of Canada (PACS-1 and MESS-3).  Certified values with their standard deviation and recoveries from our lab with standard deviations, are presented in the table appended at the end of this document.

2.7 Lithogenic material
Al is usually used as a tracer of lithogenic material since it is the third most abundant element in Earth’s crust after Si and O.  Al has the added advantage that its concentration does not vary much between upper continental crust (UCC Al = 8.04% by weight) and bulk continental crust (BCC Al = 8.41wt%), so the estimate of lithogenic mass is not very sensitive to lithogenic source regions.  However, our data suggests that there is considerable scavenged Al in particles near the coasts, which would lead to overestimates of lithogenic mass in coastal samples.  We thus calculate lithogenic mass two ways: 1) using the UCC Al concentration of 8.04% to calculate lithogenic mass (Litho_AlUCC), and 2) using Ti, a lithogenic tracer that appears to be less affected by scavenging (Litho_TiDust).  Ti has the disadvantage of varying greatly as a function of different source regions (e.g., UCC Ti=0.3wt% and BCC Ti=0.54wt%).   We make the assumption that the source of the lithogenic material is from African dust, and use the concentration of Ti and Al in aerosols collected on four samples between Cape Verde and Mauritania (Shelley and Landing, personal communication) to estimate a Ti composition of 0.6 wt% to estimate lithogenic mass (Litho_TiDust).  We estimate an uncertainty in the lithogenic mass derived from Ti of 7%, which is the propogated uncertainty of the analytical error (1 sigma) of Ti (6%) and the variability in the estimate of the Ti composition of the collected aerosols (4%).  The Ti-based estimate (Litho_TiDust) is the one that we use in subsequent calculations (eg., suspended particulate mass, section 2.9).

2.8 Fe and Mn oxyhydroxides
Fe and Mn in oxyhydroxides were calculated by subtracting Fe and Mn associated with lithogenic material.  Unlike Al, crustal Fe, Mn, and Ti vary as a function of crustal material, but the ratios of Fe and Mn to Ti are less variable. We therefore use Fe/Ti = 8.736 (mole ratio) and Mn/Ti = 0.1268 (mole ratio) derived from aerosols collected on four samples between Cape Verde and Mauritania (Shelley and Landing, personal communication) to subtract the lithogenic contributions of Fe and Mn to derive Fe(OH)3_TiDust and MnO2_TiDust.  For comparison, UCC Fe/Ti and Mn/Ti mole ratios are 10.0 and 0.1745, respectively. Variability in the Fe/Ti and Mn/Ti ratios in the aerosols and analytical errors for Fe, Mn, and Ti were propagated to determine the error for Fe and Mn oxyhydroxides.  The variability in the Fe/Ti and Mn/Ti ratios in the aerosols was 2% and 6%, respectively.  Typical analytical errors for Fe, Mn, and Ti are 3%, 3%, and 6%, respectively. We approximate the formulae for Fe and Mn oxyhydroxides to be Fe(OH)3 (ferrihydrite approximation) and MnO2 (birnessite approximation), with formula weights 106.9 g Fe(OH)3/mol Fe and 86.9 g MnO2/mol Mn, respectively. Negative numbers were set to 0.

2.9 Suspended particulate mass  
Suspended particulate mass in the sinking (>51micron) and suspended (1-51micron) size fractions was estimated as the chemical dry weight of the major particulate phases, which is the sum of POM, opal, CaCO3, lithogenic material from Ti (Litho_TiDust), and Fe and Mn oxyhydroxides, and is calculated as:
SPM = 100.08 g CaCO3/12 g C * PIC [µg/L] + opal [µg/L] + 1.88 g POM/g POC * POC [µg/L] + ...
    Litho_TiDust [µg/L] + Fe(OH)3_TiDust [µg/L] + MnO2_TiDust [µg/L]
Note that the resolution of this data is dictated by the lowest resolution of the component parts.