Sample Collection
Samples were collected on two separate research cruises aboard the R/V Kilo Moana in August 2014 and May 2015. Sampling was conducted at the Hawaii Ocean Time Series Station ALOHA (A Long-Term Oligotrophic Habitat Assessment; 22° 45'N, 158° 00'W) and the Bermuda Atlantic Time Series Site (BATS; 31° 40'N, 64° 10'W) in the Central North Atlantic.

Surface water was sampled via the vessel's underway sampling system. The intake pipe is situated on the forward starboard hull section of the vessel approximately 7.5 m below the waterline. The laboratory seawater tap was allowed to flush for 2 hours prior to each sampling. Seawater was pre-filtered through 53 µm Nitex mesh, and pumped through a 0.2 µm polyethersulfone (PES) cartridge filter (Shelco Filters, Micro Vantage, water grade, 9.75" DOE, polycarbonate housing) prior to introduction to the ultrafiltration system. Large volume subsurface water samples were collected using successive casts of a rosette equipped with 24 x 12 L Niskin bottles.

Tangential-Flow Ultrafiltration
The main UF system was constructed using a modified design of the system described in Roland et al. (2009), and expanded on by Walker et al. (2011). Briefly, the system was comprised of four-spiral wound PES UF membranes, having a nominal molecular weight cut off of 2.5 kD (GE Osmonics GH2540F30, 40-inch long, 2.5-inch diameter). The membranes were mounted in stainless steel housings, plumbed in parallel to a 100 L fluorinated HDPE reservoir, with flow driven by a 1.5 HP stainless steel centrifugal pump (Goulds Pumps, Stainless steel centrifugal pump, NPE series 1 x 1-1/4 -6, close coupled to a 1-1/2 horsepower, 3500 RPM, 60 Hz, 3 phase, Open Drip Proof Motor; 5.75 Inch Impeller Diameter, Standard Viton Mechanical Seals). All other system plumbing components contacting seawater were composed of polytetrafluoroethylene (PTFE) or stainless steel.

The system was run continuously at a membrane pressure of 40-50 psi, resulting in permeation flow rates of 1-2 L/min, depending primarily on the temperature of the feed seawater. Sample water was fed into the system using peristaltic pumps and platinum cured silicone tubing at a flow rate matched to the system permeation rates to ensure a constant system volume of approximately 100 L.

Seawater samples of 3000-4000 L were concentrated to a final retentate volume of 15-20 L, drained from the system into acid washed PC carboys and refrigerated (less than 12 hours at 2C) until the next phase of processing. Samples requiring storage for longer than 12 hours were frozen and stored at -20°C. The UF system was then reconfigured to a smaller volume system, consisting of a single membrane having a smaller nominal molecular weight cutoff (GE Osmonics GE2540F30, 40-inch long, 2.5-inch diameter, 1 kD MWCO), and a 2.5 L PES reservoir for further volume reduction and subsequent salt removal (diafiltration). Using this smaller system, samples were reduced to 2-3 L under lower pressure (25 psi, permeation rate = 250 mL/min). Samples were then diafiltered using 40 L of 18.2 MΩ Milli-Q (ultrapure) water, adding water to the sample retentate reservoir at the same rate of membrane permeation. Reduced and diafiltered samples were stored in acid washed PC bottles at -20°C for transport. In the laboratory, samples were further concentrated by rotary evaporation using pre-combusted glassware (450 °C, 5 h). A molecular sieve and a liquid nitrogen trap were placed between the vacuum pump and rotovap chamber to ensure no contamination of isolated material by back streaming of hydrocarbons or other contaminants. After reduction to 50-100 mL, samples were dried to powder via centrifugal evaporation in PTFE centrifuge tubes. Dry material was homogenized with an ethanol cleaned agate mortar and pestle, transferred to pre-combusted glass vials, and stored in a desiccation cabinet until subsequent analyses.

Solid Phase Extraction
Solid phase extraction was conducted using PPL sorbent (Agilent Bondesil PPL, 125 µm particle size, part # 5982-0026) following the general recommendations of Dittmar et al. (2008) and Green et al. (2014), including loading rates, seawater to sorbent ratios, and elution volumes and rates. Between 300 and 500 g of sorbent was used for each extraction, depending on sample volume and DOC concentration, with average loading of 4.2 ± 1.5 L UF permeate per g sorbent representing 1.9 ± 0.6 mg DOC per g sorbent or a DOC to sorbent mass ratio of 1:600 ± 200. This is in line with both the recommendations of Dittmar et al. (2008) (maximum loading = 10 L seawater per g sorbent) and Li et al. (2016) (DOC to sorbent ratio = 1:800). Permeate from the UF system was fed through PTFE tubing to a pair of 200 L HDPE barrels. The permeate water was then acidified in 200 L batches to pH 2 by adding 400 mL of 6 M HCl (Fisher Chemical, ACS Plus grade). Batch samples were mixed continuously during collection, acidification, and loading using a peristaltic pump and platinum cured Si and PTFE tubing positioned at the surface and bottom of each barrel. Acidified batches of seawater permeate were then pumped through the SPE sorbent. SPE flow rates were matched to UF permeation rates (1-2 L/min), such that a pair of 200 L barrels allowed one barrel to be filled while the contents of the other was passed through the sorbent.

Three custom SPE column configurations were used to contain the sorbent material. The column configuration was modified several times for ease of use on subsequent cruises. First, an open, gravity fed, large (49 mm ID x 1000 mm length, 1875 mL volume) glass chromatography column with 40 µm fritted disk and PTFE stopcock (Kimble-Chase, Kontes) was used. Next, we tested a custom built high-pressure SS housing (10 cm ID x 3.5 cm bed height), and finally a parallel combination of 2 medium-pressure glass chromatography columns (Kimble-Chase, Kontes, Chromaflex LC, 4.8 mm ID x 30 cm, 543 mL volume). While all designs proved to be functionally equivalent, the latter parallel combination of 2 medium-pressure glass columns ultimately provided the best configuration in order to maximize flow rates while simultaneously optimizing the ratio of sorbent bed height to loading speed. Further, the commercial availability and ease of use associated with this configuration made it our preferred design.

Following sample loading, the SPE sorbent was desalted with 6 L of pH 2 ultrapure water at a low flow rate (250-300 mL/min). After desalting, the SPE sorbent was transferred to a glass chromatography column (75 mm ID x 300 mm length, 40 µm fritted disk, PTFE stopcock) with ultrapure water rinses to ensure quantitative transfer. Isolated organic material was then eluted from the sorbent with five to six 500 mL additions of methanol. The eluted methanol solution was stored in pre-combusted amber glass bottles at -20°C for transport. Similar to UF samples, the methanol-eluted solutions were first reduced by rotary evaporation to 50-100 mL. Samples were then dried to powder via centrifugal evaporation in PTFE centrifuge tubes. Dry material was homogenized with an ethanol cleaned agate mortar and pestle, transferred to pre-combusted glass vials, and stored in a desiccation cabinet until elemental and isotopic analyses.

Total DOM
Subsamples for dissolved organic carbon (DOC) and total dissolved nitrogen (TDN) concentration measurements were collected into pre-combusted 40 mL borosilicate glass vials following 0.2 µm-filtration. Samples were stored at -20°C until analysis. Subsamples for [DOC] and [TDN] were also taken from the UF system permeate to evaluate mass balance. An "integrated" permeate sample (e.g., Benner et al., 1997) was prepared by sampling and combining equal volumes (100 mL) collected at constant time intervals throughout the ultrafiltration. DOC and TDN concentration measurements were made using the high temperature oxidation method with a Shimadzu TOC-V in the Carlson lab at UCSB (https://labs.eemb.ucsb.edu/carlson/craig/services). DOC concentration measurement errors represent the standard deviation of n=3 replicate measurements. Total DON concentrations were determined by subtracting the sum of dissolved inorganic nitrogen (DIN) species (nitrate, nitrite, ammonia) from TDN. DIN concentrations were determined using a Lachat QuickChem 8000 Flow Injection Analyzer. Ammonia concentrations were below the limit of quantification (0.36 µM) for all samples using QuickChem® Method 31-107-06-1-B. Nitrate and nitrite concentrations were measured as the sum of the two analytes using QuickChem® Method 31-107-04-1-C. The limit of detection for [NO3+NO2] using this method was 0.5 µM and the average precision of replicate standard measurements was ± 1.4 µM. In the case of [DON], measurement errors represent the propagated analytical uncertainty from the subtraction of [DIN] from [TDN]. DOC concentrations measurements were also determined via UV oxidation, cryogenic purification and manometric determination at UC Irvine.