Samples were collected during August 2021 and November 2021 at or near the BATS site (31°40’ N, 64°10’ W) or at Hydrostation S (32°10’ N, 64°30’ W) on R/V Atlantic Explorer cruises AE2114 and AE2123.

Size-fractionated particle samples were collected using McLane WTS-LV in-situ pumps (4 L min^-1maximum pumping rate; McLane Research Laboratories, Inc.) during all sampling periods. Five to eight depths were sampled between the surface and 200 meters during each cruise. Most pumps were dual-flow, collecting water through two filter holders simultaneously for geochemical and taxonomic analyses, as described by Henderson et al. (2024) and Comstock et al. (2024). Each filter holder was a vertical-intake (McLane) or mini-MULVFS style and contained four 142 mm diameter filter tiers equipped as follows (from top to bottom) for the geochemical analyses reported here: [1] 20 μm Nitex filter, [2] 6 μm Nitex filter (5 μm polyester filter), [3] two stacked 1.2 μm glass fiber filters (GF/C), [4] two stacked 0.3 μm glass fiber filters (GF75). A 150 μm Nitex backing filter was placed beneath the filter(s) of interest on the first three tiers of all filter holders to ensure filter structural integrity. Nitex filters were acid-and methanol-washed before use, and glass fiber filters were pre‑combusted (450°C) for 5 hours. After pump recovery, filter holders were drained with a weak vacuum to remove excess seawater. Filters were photographed, removed and folded with clean forceps, stored in combusted foil, and transported and stored at ‑80°C. Flow meters were placed in-line on each flow path of the pumps, and exact filtered volumes for each flow path were determined; flow rates through filter stacks used for organic analyses averaged <3 Liters per minutes (L/min). We collected dip blanks – filters that did not have any water pumped through them, but were submerged in natural seawater – along with our samples.

Processing of large particle (>20 µm) samples
Samples were stored at ‑80°C until processing. Once in the lab, particles collected on 20 µm Nitex filters were rinsed off the filters onto 47-mm diameter, pre-combusted (450°C, for 5 hr) glass fiber filters with a nominal pore size of 0.7 μm (GF/F) using 0.2 µm-filtered seawater and combusted glass filter towers. Briefly, particles were rinsed from the Nitex filters using an acid-clean squirt bottle to spray across the filter. The Nitex mesh was then sonicated for three minutes in an acid-clean polypropylene Nalgene bottle with more filtered seawater. After sonication, this water was poured into the filter tower. The process was repeated three times, with all filtered seawater being drained from the filter tower with gentle vacuum after each rinse and sonication onto the same GF/F filter. Samples were then freeze-dried and inspected under a dissecting microscope to visually characterize the particles and remove intact zooplankton swimmers or contaminant fibers, which were both rare in the samples.

Analysis of individual carbohydrate monomers
Freeze-dried filters of the 0.3 µm, 1.2 µm, and 20 µm size fractions were also quantitatively split radially by weight for analysis of carbohydrate content for selected samples from August and November 2021.

For carbohydrate monomer analysis, two or three filter splits (replicates/triplicates) were prepared per sample. Filter splits were hydrolyzed (20 h, 100°C) using 0.4 N hydrochloric acid, and hydrolysate was separated from filter material by pushing through combusted glass syringes with combusted glass wool in the tip and 0.2 μm polyethersulfone syringe-tip filters. Samples were neutralized by evaporating acid under nitrogen (N₂), reconstituted in ultrapure water, and filtered again through combusted quartz wool to remove any residual particulates before quantitative aliquots (by volume) were taken for analysis. Frozen samples were transported to the University of California Santa Barbara for analysis via high performance anion exchange chromatography coupled with pulsed amperometric detection (HPAEC-PAD). Here, we quantified neutral, amino, and acidic carbohydrate monomers following Engel and Händel (2011). The mobile phase was as follows: eluent A was 100 mmol L^-1 sodium hydroxide and 200 mmol L^-1 sodium acetate, eluent B was 18 mmol L^-1 sodium hydroxide, eluent C was 1000 mmol L^-1 sodium hydroxide, and eluent D was ultrapure water. Eluent A was filtered through a 0.2 μm nylon membrane filter, and subsequently all eluents were bubbled with N₂ gas for 45 min and degassed before being attached to the system and pressurized with N₂ gas. Individual carbohydrates were separated on a Dionex CarboPac PA10 analytical column (4x250 mm) with a Dionex CarboPac PA10 guard column (4x50 mm). The column, detector, and autosampler were temperature controlled, held constant at 25°C, 30°C, and 4°C, respectively. The flow rate was 1 mL min^-1 and the elution gradient was as follows:

-15 to 22 min: 100% eluent B
23 min: 9:1 eluent D/C
27 to 37 min: 100% eluent A
42 to 57 min: 4:1 eluent D/C
60 min: 100% eluent B

A standard curve was analyzed alongside samples during each run. A stock solution was prepared in ultrapure water to achieve concentrations of 1 mmol L^-1 fucose, rhamnose, arabinose, galactose, glucose, mannose/xylose, ribose, galacturonic acid, and glucuronic acid, and 0.5 mmol L^-1 galactosamine, glucosamine, and muramic acid. The solution was prepared all at once, and then aliquoted and stored at ‑20°C until analysis. Standard aliquots were thawed alongside samples and diluted to concentrations of 10 to 10000 nmol L^-1 per monomer for analysis. Standards of appropriate sizes (i.e., bracketing those of each monomer sample peak) were to determine sample concentrations. To verify consistent instrument performance, a 1000 nmol L^-1 standard was analyzed after every eight samples and compared to the original standard curve. An aliquot of a sample with ample material was also analyzed during every day of analysis to confirm day-to-day consistency. Sample concentrations were calculated from recorded peak areas using the calibration from the standard curve. Blanks of ultrapure water and full process blanks were analyzed alongside samples to check for background carbohydrate content in reagents or contamination. Full process blanks consisted of dip blank filter splits and 0.4 M hydrochloric acid blanks that were processed exactly as samples. Seawater particulate carbohydrate concentrations (nmol L^‑1) were calculated based on the original amount of seawater filtered through the portion of sample analyzed during each run (material extracted from 0.1 to 2.5 L of seawater injected for a single run). We report concentrations here as total carbohydrate carbon as a proportion of POC, where carbon from individual monomers was calculated from the molecular formula for each monomer, summed, and divided by the total POC concentration in that particle size fraction.