Phytoplankton samples for flow cytometry, HPLC and microscopy were collected pre-dawn from Niskin bottles mounted on a 24-place rosette system equipped with a Seabird SBE911 CTD and a Seapoint fluorometer.

Flow cytometry samples (2-mL) for pico-phytoplankton abundance were preserved (0.5% paraformaldehyde) and frozen in LN2, then stored at -80°C until shore-based analyses. On-shore, flow cytometry samples were thawed and stained for 1 h with the DNA stain Hoechst 33342 (1 µg/ml, Monger and Landry, 1993), then analyzed with a Beckman Coulter EPICS Altra flow cytometer (Selph et al., 2011). Listmode data were processed using FlowJo (version 9.7.7, Treestar, Inc.) to delineate Prochlorococcus (PRO), Synechococcus (SYN), and eukaryotic phytoplankton (PEUK). PRO and SYN abundances were converted to carbon using 32 and 101 fg C/cell, respectively (Garrison et al., 2000; Brown et al., 2008). PEUK were mainly ≤2 µm cells, however some were larger and therefore microscopic counts for cells from 2-5 µm were subtracted from the total PEUK abundance, and the remaining cells were converted to carbon and scaled to be that of a cell twice the diameter of SYN (808 fg C/cell).

HPLC pigment samples (2.2-L) were filtered onto GF/F filters, frozen in LN2 and stored at -85°C until shore-based analyses. On shore, samples were sent to Horn Point Analytical Services Laboratory (University of Maryland Center for Environmental Science). There they were extracted, and analyzed using an automated 1100 HPLC system with Agilent temperature-controlled autosampler, Peltier temperature-controlled column oven compartment, PDA detector and ChemStation software. The HPLC method uses a C8 column and a reversed phase, methanol-based solvent system (Van Heukelem and Thomas, 2001; Hooker et al., 2012). MVCHLa and DVCHLa are detected at 665 nm. Carotenoid and xanthophyll accessory pigments are detected at 450 nm.

The pigments used for phytoplankton taxonomic identification were MVCHLa and DVCHLa (sum = TCHLa), monovinyl chlorophyll b (MVCHLb), divinyl chlorophyll b (DVCHLb), chlorophyll c3 (CHLc3), zeaxanthin (ZEAX), fucoxanthin (FUCO), 19’-hex-fucoxanthin (HEX), 19’-but-fucoxanthin (BUT), allophycocyanin (ALLO), peridinin (PER), neoxanthin (NEO), and prasinoxanthin (PRAS). Chlorophyll a contributions for PRO and SYN were subtracted from HPLC pigment data (estimated from flow cytometry and DVCHLa). Similarly, Trichodesmium MVCHLa was assigned from separately collected samples (Selph et al., 2021). The remaining HPLC pigments, except for ZEAX, were entered into the CHEMTAX program (v. 1.95, Wright, 2008) for partitioning into eukaryotic groups.

Initial pigment ratios (accessory pigment:MVCHLa) used in CHEMTAX were those of oceanic species (Higgens et al., 2011) and indicative of the following groups: chlorophytes (CHLOR), diatoms (DIAT), prymnesiophytes - type 6 (PRYM), pelagophytes (PELAG), cryptophytes (CRYPT), prasinophytes - type 3 (PRAS3), and dinoflagellates (A-DINO). Data were divided into 2 groups: shallower and deeper than 60 m, since some of the accessory pigments were only present in deep samples (NEO and ALLO) and the general pattern of pigments showed a different community at depth. The initial ratio matrix was randomized into 60 matrices (0.7 x random number between -0.5 and +0.5), which were then applied to the data sets (Supp. Table I). The resulting partitioning of MVCHLa into these phytoplankton taxa was used to estimate the percent of total carbon biomass in each group.

Microscopy samples from the top two depths (~80%I0 and 40%I0) and the bottom 2 depths (~5%I0 and 1%I0) sampled were used to estimate the biomass of eukaryotic phytoplankton taxa. Phytoplankton taxonomic structure was assessed to the extent possible, separating cells into dinoflagellates, diatoms, and unidentified flagellates. Microscope slides were prepared from 500 mL of preserved sample (260 µL alkaline Lugol’s solution (0.1% final), 10 mL 10% buffered formalin, 500 µL 3% sodium thiosulfate; modified protocol from Sherr and Sherr, 1993), to which 1 mL 0.33% proflavine (w/v) and 1 mL of 4’,6-diamidino-2-phenylindole (DAPI, 0.01 mg/mL) were added. For analysis of cells <10-µm, a slide was prepared from 50 mL subsamples filtered onto a 25-mm, 0.8-µm pore size, black PCTE filter and mounted on a glass slide. For larger (10- to ~50-µm) cells, the remaining sample was filtered onto a 25-mm, 8-µm pore size, black PCTE filter. Slides were frozen (-80°C) until image analysis as detailed in Taylor et al. (2015). Cell biovolumes (BV, µm3) were calculated from length (L) and width (W) according to Taylor et al. (2011). BV was converted to carbon (C, pg/cell) using C = 0.216 x BV0.939 for non-diatoms and C = 0.288 x BV0.811 for diatoms (Menden-Deuer and Lessard, 2000). The FCM-derived PEUK abundance were assumed to represent cells <5 µm, therefore 2-5 µm cells from microscopy were subtracted from total PEUK cells, leaving cells ≤2 µm (not counted with microscopy), and their carbon contents were added to the microscope slide-estimated carbon for a total phytoplankton community carbon estimate. These data were also used to determine carbon:chlorophyll (C:CHL) ratios at the depths where both measurements were taken. Missing intermediate depths (for carbon) were estimated using the 5%I0 C:CHL ratio.