2. Materials and Methods

2.1 Ocean Station Papa

This study is part of the first EXPORTS field campaign, which took place in late summer 2018 (August 14 to September 9) at Ocean Station Papa (‘Station P’). Station P is a time-series site located in the Northeast Pacific Ocean (nominally 50°N 145°W) that has been regularly monitored since 1949 (e.g., Tabata 1965). It is a high-nutrient and low-chlorophyll system with a ‘muted’ spring bloom owing to persistently low concentrations of iron (Martin and Fitzwater, 1988; Boyd et al., 1996). During our study, Station P was characterized by a shallow mixed layer (average: 29 \(\pm\ \)4 m), and strong vertical and weak horizontal gradients in hydrographic properties (Siegel et al., 2021). Mixed layer macronutrient concentrations were elevated and chlorophylla concentrations were low with a mean of 0.21 μg L-1, which was less than typical August values (2000–2017) for Station P (Siegel et al., 2021). This condition resulted in an isolume depths of the 1% photosynthetically available radiation (PAR) between 70 to 90 m (mean: 78 \(\pm\) 6 m) (Siegel et al., 2021), which was slightly deeper than climatological conditions (Siegel et al., 2021). The site was characterized by a highly recycled food web in the mixed layer (Meyer et al., 2020; McNair et al., 2023), with low POC fluxes (Buesseler et al., 2020; Estapa et al.,2021), which were in line with previous observations performed at Station P (Charette et al., 1999; Wong et al., 2002; Kawakami et al., 2010; Timothy et al., 2013; Mackinson et al., 2015). An in-depth discussion of how POC fluxes measured during EXPORTS compare to previous POC flux measurements is provided in Buesseler et al., (2020).
The observations presented here were conducted from the R/V Roger Revelle , which sampled ecological and biogeochemical properties following an instrumented Lagrangian Float drugged at ~100 m depth. A detailed operational description of the EXPORTS North Pacific field campaign can be found in Siegel et al., (2021).

2.2 Sample collection with the Marine Snow Catchers

Profiles of suspended and sinking particles were collected using three Marine Snow Catchers (MSC; Lampitt et al., 1993, OSIL, UK). The MSC is a large volume (volume: VMSC \(=\ \)89.8 L, height: hMSC\(\ =\) 1.5 m) water sampler with a removable base section (~ 8 L) that enables the partitioning of particles according to a predefined settling time (Riley et al., 2012; Giering et al., 2016). The MSCs were deployed at two to three depths ranging between 20 and 500 m at 13 stations (Table S1 ). Deployment depths were chosen to include (1) just below the mixed layer depth (20–65 m), (2) at ~100 m (Lagrangian float depth), and (3) between 100–500 m. Depths were adjusted to match where particle maxima were detected immediately before the MSC deployments, either via chlorophyll fluorescence profile, or particle counts from the Underwater Vision Profiler-5 (Picheral et al., 2010, Hydroptic, FR). MSC casts were conducted in the afternoons on days 1, 3, and 6 of each of the three 8-day sampling epochs, with additional two-depths casts on days 5 of epochs 2 and 3 (Table S1 ).
A full description of the MSC and its sampling methodology has been published by Riley at al. (2012) and Giering et al. (2016). We modified the protocol slightly as follows: During deployment, the terminal apertures of the top and base sections of the MSC were kept open and were closed at the target depth through a wire-guided messenger trigger mechanism. A tray (height: htray = 4.4 cm, area: Atray \(=\) 0.028 m2, approximate volume: ~ 1 L) was placed at the bottom of the MSC base section prior to each deployment. Immediately following deployment and retrieval, the MSC was secured in an upright position, the initial bulk population of particles was sampled from the MSC’s central tap (time zero, T0 ), and particles were allowed to settle for 2 hours. We operationally defined three particle fractions,top (t ), base (b ) and tray(tr ) according to their sampling location within the MSC after the 2-hour settling period. The top fraction was collected from the central tap in the top section of the MSC and represents the suspended particle pool, i.e., particles that did not sink during the 2 hours settling period. Subsequently, the water contained in the upper part of the MSC was gently drained, and the upper part of the MSC removed. The water in the base section overlying the tray was siphoned off, constituting the base fraction (volume: Vbase; 4.5–5.5 L). Water in the tray was considered as the tray fraction (volume: Vtray 0.33–0.98 L). The base and tray sampled volumes varied across deployments depending on sampling efficiency (i.e., how much volume was lost during sampling). In the laboratory, the tray was visually investigated for the presence of marine-snow-sized particles. In two instances (settling time experiments, see section 2.6) the trayand base fraction were combined before sampling (referred to asbottom fraction).
All three fractions (top , base , tray ) were subsampled for various parameters (Table S1 ). Subsamples for particulate organic carbon and nitrogen (POC and PON) analysis and for counts and identification (particle imaging via the FlowCam) were collected at all stations, except on August 17 when no samples were fixed. Subsamples for the analysis of total particulate carbon (TC), from which we obtained particulate inorganic carbon estimates (PIC) by difference (PIC = TC - POC), biogenic and lithogenic silica (bSi and lSi, respectively) were taken on days 3 and 5 of each epoch. Subsamples for the determination of TEP were collected on days 1 and 5 of each epoch, and the abiotic aggregation potential was determined on seawater collected on day 1 of each epoch.

2.3 Particle imaging and sizing with the FlowCam

Particle size distribution of suspended and sinking particles was assessed using particle imaging. Subsamples (50 mL) were fixed with 37% formalin (hexamethylenetetramine buffered) to a final concentration of 1–2% and stored in dark at 4°C until analysis. Subsamples were analyzed using a FlowCam 8000 Series (Sieracki et al., 1998, Yokogawa Fluid Imaging Technologies, US) in auto-image mode (particles imaged at a pre-defined flow rate) (Table S2 ). To assess relevant particle sizes, we first analyzed subsamples using a x40 objective combined to a 300x1500-μm flow cell and then a x200 objective with a 50x300-μm flow cell. Note that 13 out of 30 suspended particle samples were not analyzed with the x40 objective due to time constraints. The minimum sizes reliably resolved by the objectives were 30 μm and 3 μm, respectively. The maximum size was dictated by the flow cell size and was 300 μm for the x40 objective and 50 μm for the x200 objective.Base and tray fractions were pre-filtered using a 44-μm mesh filter to avoid clogging the 50x300-μm flow cell when using the x200 objective.
Based on the size of the particles and the FlowCam resolution, we chose to restrict our analysis to particles with an Equivalent Spherical Diameter (ESD) between 4 and 128 μm. We calculated the minimum number of particles required to estimate a robust particle size spectrum using the empirical relationship given by Blanco et al. (1994) and assuming that a size bin is well represented if there are at least 10 particles in it (Álvarez et al., 2011). Thus, we imaged around 1,300 particles (i.e., minimum desired count and lowest number of counted particles) to cover a size range of 4 to < 32 μm at x200 and a size range of ≥ 32 to 128 μm at x40.
Imaged particles were sized using FlowCam’s VisualSpreadsheet© software version 4.15.1. and grouped in the five size bins based on their ESD. The complete method settings and example images are presented inTable S2 . Particle concentrations of top , baseand tray were calculated by dividing the counted number of particles in each size bin by the sample volume imaged. Suspended and sinking particle abundances were calculated as explained in section 2.7. Particle size distributions (PSDs) were estimated as differential number size distributions N(D) (# particle mL-1μm-1) for top and tray fractions. The slope of the PSDs was calculated as the slope of a one-degree polynomial fit  to the log10 -transformed PSD andlog10 -transformed arithmetic mean of each size bin (μm) over the entire measured size range (polyfit function, Matlab R2021b).

2.4 Biogeochemical analyses

Typically, 1 L of the T0 , top andbase fractions and ~ 0.1 L of the trayfraction were filtered each for analysis of POC and PON, TC, bSi and lSi, and TEP. For the occasions when we carried out the settling time methodological test, between 0.5–1.5 L of the top andbottom subsamples were filtered instead.
Concentrations of POC and PON were determined by filtering subsamples onto two replicate pre-combusted (450 °C, 30 minutes) GF/F filters (25 mm, Whatmann, UK). The filters were dried at 60°C and stored at room temperature until analysis. Filters were analyzed using a CEC 44OHA elemental analyzer (Control Equipment, US) after treatment with 10% HCl. Replicates were averaged (see average of the relative standard deviations of the filter replicates in Table S3 ). Detection limit of POC and PON ranged between 0.8–16 μg and 0.2–3.9 μg, respectively. All POC values were above the detection limit of the instrument. However, for PON, 24 of 237 values were below detection, seven of which were negative and therefore substituted with zero (μg L-1). Values were corrected to account for non-target carbon on the filter using the average POC and PON mass (12.6 μg for POC; 2.1 μg for PON) of 28 blanks obtained using a multiple volume (0.5, 1 and ~ 2L) regression approach (Moran et al., 1999) of water collected with Niskin Bottles from the CTD-rosette system within the mixed layer (campaign-wide correction). A detailed explanation of the blank correction method and a reconciliation of all the POC and PON measurements obtained using different methodologies during the EXPORTS field campaign is presented by Graff et al., (in review ).
Total particulate carbon (TC) was measured following procedures similar to POC but without prior acidification. Only one filter per fraction was generated due to volume constraints. No measured values fell below the instrument detection limit (1.2 to 4.5 μg). Particulate inorganic carbon (PIC) was calculated by subtracting the uncorrected POC from TC.
Biogenic and lithogenic silica concentrations were determined by filtering samples onto 0.6-μm pore size polycarbonate membrane filters (47 mm diameter, Isopore, Millipore), which were dried at 60°C and stored at room temperature until analysis. Filters were digested in Teflon tubes by adding 4 mL of 0.2 N (normal) NaOH (95°C, 40 minutes) and cooled immediately afterwards. The resulting solutions were neutralized by adding 1 mL of 1 M (molar) HCl and centrifuged (10 minutes, 2500 rpms) to separate lSi from bSi. 4 mL of the solution was diluted with 6 mL of Milli-Q water and assessed via the molybdosilicic acid spectrophotographic method to measure bSi (Strickland and Parsons, 1968). The remaining 1 mL of solution, which was left at the bottom of the Teflon tube together with the filter, was rinsed using Milli-Q water, left to dry, and cooled. To extract lSi, 0.25 mL of 2.5 M (molar) hydrofluoric acid were added. After 48 hours 9.75 mL 1 M (molar) saturated boric acid solution were added, and filters were centrifuged. 8 mL of the resulting solution was added to 2 mL 1 M (molar) saturated boric acid solution and assessed spectrophotometrically as aforementioned.
Concentrations of TEP were determined colorimetrically on triplicate samples. Subsamples were filtered onto 0.4-μm pore size polycarbonate filters (25 mm, Whatmann, UK), stained with Alcian blue, and stored frozen until analysis. Filter blanks, prepared by staining and rinsing wet filters, were processed like the samples. Stained filters were soaked for at least 2 hours in 80% sulfuric acid (Fisher Scientific; 95% w/w), and the absorption at 787 nm was measured spectrophotometrically (Thermo Scientific GENESYS 10S UV- VIS; Passow and Alldredge, 1995). The stained filters were compared to a standard curve developed using Gum Xanthan (Sigma-Aldrich) and hence TEP determinations were expressed as standardized Gum Xanthan equivalents (GXeq) (Bittar et al., 2018). TEP determinations were considered above detection, if the absorbance value of a sample at 787 nm was at least twice the absorbance value of the blank at 787 nm (Passow and Alldredge, 1995). A total of 28 of 112 TEP determinations measured from the MSC were below this detection limit. TEP associated with sinking particles between 300 and 500 m was always at or below detection limit. In addition to the MSC samples, we also determined water column TEP concentration on water collected using Niskin bottles fitted to a CTD rosette (depth range: 5–500 m; n = 153). A total of 22 of 153 TEP measured on the water column sampled with the Niskin bottles were at or below the detection limit. We did not remove TEP values at or below the limit of detection but consider these observations an upper bound estimate of the real in situ concentrations. We expressed TEP in carbon units (TEP-C) using a conversion factor of 0.75 μg C µg L-1/ GXeq L-1 (Engel and Passow, 2001).
Several values of the biogeochemical measurements were considerably higher than the average concentrations at their collection depth. This occurred for ~5% of the total number of observations and these samples are referred to here as outliers . Outliers were defined as observations that were at least twice the average concentration for that depth layer. We refer to background concentrations when these outliers were removed to assess the baseline biogeochemical composition of suspended and sinking particles. The nature of the outliers and the implication of their presence is discussed in section 4.5 below.

2.5 Abiotic aggregation potential experiments

The potential for the abiotic formation of aggregates was evaluated using rolling tanks experiments. Four experiments were conducted using MSC T0 (August 25 at 95 m), top (August 25 at 95 m) or base (August 17 at 55 m and August 19 at 55 and 95 m) sample fractions. Each experiment was conducted for 36 hours in 1.1-L rolling tanks at near in situ temperature (13°C or 4°C) in the dark. Aggregate formation was monitored by visually checking every 6 hours for the appearance of marine snow-sized particles (ESD > 0.5 mm).

2.6 Settling time experiments for MSC

Methodological experiments to determine the effect of settling time on particle fractions partitioning were conducted twice (on August 20 and September 03, Table S1 ) from MSC deployments at 60 and 80 m depth, respectively. Each time, all three MSCs were deployed at the same depth, but settling times after retrieval varied: particles were allowed to settle for 1, 2 and 4 hours after recovery of the MSCs. Subsamples oftop and bottom fractions were analyzed for POC and PON, bSi and TEP.

2.7 Calculations of particle concentrations and fluxes

Following Riley et al. (2012) and Giering et al. (2016), concentrations of suspended (cNS), slow-sinking (cSS) and fast-sinking (cFS) particles were calculated as follows:
cNS = top concentration (2 )
cSS = (basetop concentrations)\(\times\ \)Vbase / VMSC(3 )
cFS = (traybase concentrations)\(\times\ \) Vtray / (Atray \(\times\ \)hMSC) (4 )
Occasionally, after applying the corrections we obtained negative values i.e., the top concentration exceeded the baseconcentration, or the base concentration exceeded the trayconcentration. We assigned a zero concentration to the specific observations. The negative values could reflect a lack of sinking particles, the presence of ascending particles and/or actively moving zooplankton. In the analyses that follow, little differences were found between slow- and fast-sinking particle concentrations (see section 3.2). Hence, these two sinking particle fractions are combined. In four deployments, TEP and TC were measured only in the top andtray fractions. In these cases, sinking TEP and TC concentrations were estimated following equation (4) but subtracting the topfraction from the tray fraction.
The standard particle sinking velocity for the MSC is assumed to be 18 m d-1 as determined geometrically by dividing the sinking distance (height of the MSC; hMSC = VMSC/ AMSC = 1.5 m) by the settling time (t = 2 hours). We assumed that the 18 m d-1 sinking velocity estimate for sinking particles represents a lower bound estimate because the sinking particles could have reached the bottom of the MSC much sooner (Giering et al., 2016).

2.8 Statistical analyses

The relative uncertainties in the calculated concentrations and fluxes were determined by using a Monte Carlo error propagation with mean values obtained by averaging the replicate filters and the precision estimated by calculating the average of the relative standard deviations of the filter replicates (Table S3 ). A 1% uncertainty was assumed for the measured values (VMSC, Atray , hMSC, Vbase , Vtray, hMSC and sinking time). Final estimates of the averages and standard deviations of particle concentrations and fluxes were then calculated from the simulation results. Simple linear regression was used to test (Student’s t-test, 95% confidence) for statistically significant temporal and depth trends displayed by the biogeochemical content of suspended, sinking particles and their partitioning to total. The linear analyses were performed using the function fitlm in Matlab (R2021b). The same approach was performed to test for significant temporal trends displayed by the biogeochemical content of suspended and sinking particles during the settling tests. Paired t-test was used to test for differences among particle fractions. The data presented in this study, and all the data generated during the first EXPORTS field campaign can be found at NASA SeaBASS data repository (https://seabass.gsfc.nasa.gov/cruise/EXPORTSNP).