The upper ocean silicon cycle of the subarctic Pacific during the EXPORTS field campaign

Silicon cycling in the high-nutrient low-chlorophyll (HNLC) region of the northeastern subarctic Pacific Ocean was examined as part of the NASA EXport Processes in the Ocean from RemoTe Sensing (EXPORTS) program. The northeast subarctic Pacific was the first ocean region to be designated as HNLC characterized by persistent high macronutrient concentrations and low phytoplankton biomass caused by iron (Fe) limitation of phytoplankton growth (Martin and Fitzwater, 1988; Martin et al., 1994; Boyd et al., 1996). Fe limitation severely restricts the abundance of large diatoms in the northeast subarctic Pacific (Marchetti et al., 2006b) and in all other major HNLC regions including parts of coastal California (Hutchins et al., 1998), the eastern equatorial Pacific (Coale et al., 1996) and the Southern Ocean (Boyd and Law, 2001).

Fe limitation strongly modifies silicon (Si) cycling in HNLC waters through two opposing effects. By reducing diatom growth and biomass, Fe limitation acts to diminish the consumption of silicic acid, Si(OH)₄. In contrast, diatom physiology can respond to low Fe by increasing cellular demand for Si(OH)₄ relative to other macronutrients. In most HNLC regions outside of the northeastern subarctic Pacific, diatom growth is still sufficient for the increased Si demand to overwhelm the potential for growth limitation, causing the preferential depletion of Si(OH)₄ compared to nitrate and orthophosphate (Dugdale et al., 1995; Hutchins and Bruland, 1998; Sarmiento et al., 2004; Bruland et al., 2005; Marchetti et al., 2006a) resulting in secondary Si limitation of diatom Si uptake rates and, in some cases, Si-limited diatom growth (Franck et al., 2000; Nelson et al., 2001; Brzezinski et al., 2008). This phenomenon has led to the suggestion that the description of these systems should be revised to high-nitrate low-silicate low-chlorophyll (HNLSLC; Wilkerson and Dugdale, 1996).

The northeastern subarctic Pacific differs from other HNLC regions in that a low Si(OH)₄ condition is rare, occurring only every few years (Wong and Matear, 1999), with concentrations typically remaining above 15 μM year round. Persistent Si(OH)₄ concentrations ≥15 μM would be considered a Si-replete condition for most diatoms (Martin-Jézéquel et al., 2000), implying that Si limitation is also rare. The lack of frequent Si(OH)₄ depletion in the northeastern subarctic Pacific is likely the result of the lower Fe supply in the northeast subarctic Pacific compared to other HNLC regions. The ferricline in the subarctic Pacific is deeper than the nutricline for macronutrients (Nishioka and Obata, 2017) and corresponds to the depth of winter mixing. Thus, modest vertical mixing injects relatively little Fe into the euphotic zone compared to macronutrients, creating a near chronic low-Fe condition in surface waters of the northeast subarctic Pacific that may exclude large diatoms to a greater extent than in other HNLC regions. The contrasting Si(OH)₄ dynamics between the northeastern subarctic Pacific and other HNLC regions suggest that the effects of Fe on Si cycling in the subarctic Pacific may be unique.

Fe stress affects both diatom physiology and diatom community structure. At the cellular level, low Fe increases diatom Si content relative to that of organic matter by elevating diatom Si content directly through frustule thickening (Hutchins and Bruland, 1998; Takeda, 1998), by reducing organic matter content (Hoffman et al., 2006; Marchetti and Harrison, 2007), or through both of these responses (Takeda, 1998). Low Fe can also shift diatom community structure towards more robust diatom species with inherently higher Si:N and Si:C ratios (Marchetti et al., 2010; Durkin et al., 2012; Assmy et al., 2013). There is also evidence that smaller diatoms may not experience the same degree of Fe limitation as do larger diatom taxa in HNLC waters. In the equatorial Pacific relatively small diatoms of the genus Pseudo-nitzschia were the dominant diatoms under ambient conditions, and experiments showed that they were not significantly stressed by either Fe or Si (Brzezinski et al., 2011a). In the Southern Ocean, even modest-sized diatoms like Fragillariopsis kerguelensis persist under low Fe conditions (Assmy et al., 2013). Thus, diatom responses at the community and physiological levels appear system-specific.

Diatoms in some HNLC regions appear to have two distinct growth strategies that affect their contribution to carbon export. In the Southern Ocean, thin-shelled forms respond rapidly to changes in Fe with a boom-and-bust growth strategy resulting in relatively efficient diatom carbon export during Fe-stimulated blooms (Assmy et al., 2013). In contrast to these “carbon sinkers,” thick-shelled forms persist in relatively low abundance under low Fe relying on their heavy silicification for protection from grazing to facilitate persistence despite slow growth. Taxa employing the latter strategy have been dubbed “silica sinkers” as they are inefficient at exporting organic matter relative to bSi (Assmy et al., 2013). Thus, from both evolutionary and biogeochemical perspectives, low Fe in HNLC regions can impede the export and sequestration of organic carbon by diatoms through species-specific and/or physiological effects on diatom Si:C ratios. How applicable these concepts are outside of the Southern Ocean is unclear. For example, in HNLC regions such as the equatorial Pacific small thin-shelled forms dominate under low Fe (Brzezinski et al., 2011a), but their contribution to export is unknown.

The lack of strong depletion of Si(OH)₄ and other macronutrients in the northeast subarctic Pacific makes this area a true HNLC region with Fe being the predominate limiting nutrient, distinct from all other major HNLC areas. Data indicate that bSi concentrations in the euphotic zone at Ocean Station Papa (OSP; 145°W, 50°N) are <100 nmol Si L^–1 and generally <50 nmol Si L^–1 (Lipsen, 2008). Net silica production rates implied by seasonal net Si(OH)₄ depletion are also low. The annual cycle of Si(OH)₄ concentration shows a maximum concentration of about 23 μM in late winter (Whitney and Freeland, 1999; Harrison, 2002) that begins to decrease in June, ultimately declining by about 7.5 μM through the summer reaching a seasonal minimum of about 15 μM in mid-August (Peña and Varela, 2007). The magnitude of net depletion of Si(OH)₄ implies a relatively low average net silica production rate (production minus dissolution in surface waters) over the 3-month productive period of 7.5 μM ÷ 90 days = 83 nmol Si L^–1 d^–1. This estimate of silica production is conservatively low, as it does not account for the consumption of Si(OH)₄ introduced to surface waters through vertical eddy diffusion across the nutricline. Applying Fick’s (1855) law, assuming a diapycnal diffusion coefficient of 2 x 10^–5 m^–2 s^–1 and a Si(OH)₄ gradient across the nutricline of 3 μmol m^–4 at OSP (Whitney and Freeland, 1999) yields an approximate additional Si supply of 0.5 mmol Si m^–2 d^–1 or 7 nmol Si L^–1 d^–1. Averaging that supply over the upper 70 m leads to a revised average net silica production rate of 90 nmol Si L^–1 d^–1 or 6.3 mmol Si m^–2 d^–1, which is still conservative as it does not account for other forms of vertical nutrient injection. Integrated gross silica production rates measured using silicon-32 tracer in deckboard incubations during June and August–September are lower at 1.2 ± 0.9 mmol Si m^–2 d^–1 (n = 3) and 1.5 ± 1.6 mmol Si m^–2 d^–1 (n = 3; these and all subsequent uncertainty terms are standard deviations), respectively (Lipsen, 2008). Sediment trap studies have shown that the export of bSi is highest in spring (Wong et al., 1999; Timothy et al., 2013) when diatom productivity based on the rate of Si(OH)₄ depletion is also at its annual maximum. Biogenic silica export out of the euphotic zone follows the same seasonal dynamics as Si(OH)₄, reaching an annual minimum in late August to early September (Whitney and Freeland, 1999; Wong et al., 1999).

Another salient feature of OSP is that while cyanobacteria of the genus Synechococcus are abundant, Prochlorococcus is not present (Tortell et al., 1999). Significant Si quotas have been measured for Synechococcus in the equatorial Pacific (Baines et al., 2012), in the Sargasso Sea (Ohnemus et al., 2016), and in laboratory cultures (Brzezinski et al., 2017). Their rates of Si uptake in situ have been inferred from studies of Si use by different particle size fractions in the Sargasso Sea (Krause et al., 2017) and in the South Pacific (Leblanc et al., 2018). In the ocean regions that have been examined, Synechococcus and Prochlorococcus co-occur. As Si use by Prochlorococcus is not known to occur, their absence from OSP reduces uncertainty in interpreting Si uptake by the cyanobacteria in this area.

In the present study, the Si cycle near OSP was examined as part of the EXPORTS program during August and September 2018 near the time of the annual minimum in both the surface Si(OH)₄ concentration and the rate of net Si(OH)₄ depletion. Dissolved and particulate silica concentrations along with rates of silica production and assessments of Si and Fe limitation are compared to rates of bSi and organic carbon export to evaluate the upper ocean Si cycle and the role of diatoms in the region’s biological carbon pump.

The logistics and general findings of the deployment of the EXPORTS program to the subarctic Pacific have been described in detail by Siegel et al. (2021). Briefly, the study was conducted at Ocean Station Papa (nominally 50°N latitude, 145°W longitude) and involved three ships: the R/V Sally Ride and R/V Roger Revelle from the United States that both sampled near OSP from August 16 through September 7, 2018, and the CCGS John P. Tully from Canada that sampled the EXPORTS study site from September 20–22, 2018, beginning 12 days after the US ships departed. The Revelle sampled following a Lagrangian float that was drogued near 100 m for the entire time, while the Ride sampled a wider geographic area in survey mode. Sampling aboard the Revelle and the Ride occurred in three 8-day cycles, referred to as epochs (Siegel et al., 2021), organized around three sets of sediment trap deployments from the Revelle. Particle export was also assessed on the Ride using the ²³⁴Th method tied to elemental ratios obtained from particles collected using in situ pumps (Buesseler et al., 2020; Estapa et al., 2021; Roca-Martí et al., 2021). Particle export was not measured aboard the Tully.

Seawater samples for determining Si(OH)₄ and bSi concentrations were taken on all three ships using mostly common procedures. On the Revelle nutrient samples were taken on casts of a conventional CTD/rosette equipped with 12-L Niskin samplers and on casts of a trace-metal-clean CTD/rosette system using 10-L Go-Flo bottles. Samples for bSi concentration on the Revelle were collected from the trace-metal-clean CTD/rosette system only. Aboard the Ride both nutrient and bSi samples were taken from Niskin samplers using a conventional CTD/rosette system. Both rosette systems were equipped with SeaBird 911 CTDs. Aboard the Tully, all samples were taken from 12-L Niskin samplers on a conventional CTD/rosette system equipped with a SeaBird SBE 911 CTD with samples for Si(OH)₄, bSi, and silica production always collected from the same Niskin samplers at each one of 6 light depths (100, 50, 30, 15, 1 and 0.1% of photosynthetically active radiation, PAR).

On each US ship, seawater nutrient samples were filtered through 0.2-µm polycarbonate filters into acid-cleaned plastic vials and immediately frozen at –20°C. On the Tully, samples for dissolved Si were filtered through 0.6-µm polycarbonate filters into 15-mL plastic centrifuge tubes and kept at 4°C until analysis ashore. On all ships, seawater samples for bSi concentration were processed by filtering 600–1000 mL of seawater through polycarbonate filters. On the Ride, seawater was filtered through 47-mm diameter, 0.6-µm pore-size filters. On the Revelle and on the Tully size-fractionation of seawater samples for bSi analysis was performed by sequential filtration of each sample through 5.0-µm and then 0.6-µm pore-size 47-mm polycarbonate filters. On the Tully an additional sample for total bSi concentration was filtered through a separate 0.6-µm filter. For all bSi samples, filters containing the particulate matter were folded, stored in individual plastic containers and immediately frozen at –20°C to prevent opal dissolution.

At the end of the US expedition, Si(OH)₄ and bSi samples from both the Ride and the Revelle were transported frozen on dry ice to the University of California Santa Barbara. Frozen Si(OH)₄ samples were heated at 50°C for 30–45 minutes to eliminate polymers of Si and then allowed to cool to room temperature before analysis (Becker et al., 2020). The seawater Si(OH)₄ concentrations were determined on a Lachet 8500 series 2 flow injection system with a detection limit of 0.2 μM Si(OH)₄ or 3%, whichever was larger. The concentration of Si(OH)₄ for samples from the Tully was measured with a Beckman DU 530 ultraviolet-visible (UV/Vis) spectrophotometer following Strickland and Parsons (1972) modified to use a reverse-order reagent blank (Brzezinski, 1986) with a detection limit of <0.05 μM, with samples from other casts at the same site analyzed on an Astoria Analyzer using micro-segmented flow analysis. No direct intercalibration was done for Si(OH)₄ analyses between the groups from the US and Canada. However, the automated nutrient analyses from each country were calibrated against certified reference materials (CRM, SCOR-JAMESTEC lot CI in the US, KANSO lot CB for the Canadian cruise). In each case the measured Si(OH)₄ concentrations of the CRMs agreed to within 1% of the certified value. Comparison of the samples from the mixed layer analyzed manually and by auto analyzer during the Canadian cruise showed agreement to better than 1%. Samples for bSi from the US ships were analyzed using the NaOH digestion method described by Krause et al. (2009) using manual colorimetry with a detection limit of 0.05 µmol Si L^–1. Upon completion of the Canadian leg, the frozen bSi samples from the Tully were kept on ice during the short transfer to the University of Victoria and stored at –20°C until analysis following the same NaOH digestion method used for US samples.

Profiles of size-fractionated silica production rates were obtained on the Revelle using the radiotracer silicon-32 (Brzezinski and Phillips, 1997). Nine profiles, each sampling five depths spanning the 55% to the 1% light depths, were obtained across the three epochs using the trace-metal-clean rosette system. All subsampling of Go-Flo samplers was conducted in a trace-metal-clean van and the subsamples were transferred within clear plastic bags to a radioisotope van for tracer addition. Seawater for rate measurements was subsampled into trace-metal-cleaned 300-mL polycarbonate bottles and then spiked with 230 Bq of Chelex-cleaned high-specific activity silicon-32 (15,567 Bq µg^–1 Si). Each sample bottle was capped and the closure sealed with parafilm before being transferred to deck incubators simulating the light levels at each sample collection depth using a combination of neutral density screening and blue plastic film. Incubators were cooled with flowing surface seawater except for those at the two deepest light depths that were held at near in situ temperature using recirculating chillers. Following 24 h of incubation, water samples were size-fractionated through 25-mm diameter 5.0-µm pore size, and then 0.6-µm pore size, polycarbonate filters paralleling the size-fractionation of bSi samples. ³²Si activity on the filters was measured using low-level beta detection as in Krause et al. (2011a).

The degree to which iron and/or Si(OH)₄ limited rates of silica production and rates of primary production was quantified at all eight stations sampled using water collected with the trace-metal-clean rosette at the 40% and 10% light depths. Si-amended (+Si) treatments increased [Si(OH)₄] by 20 μM using a solution of sodium metasilicate, Fe additions (+Fe) targeted an increase of 1 nM Fe using ferric chloride, and a combined treatment (+Si+Fe) used both additions (20 μM silicate, 1 nM Fe). Samples of unamended seawater served as controls. Incubations were conducted for 24 h. The experiment was replicated eight times during the cruise, but treatments were not replicated within single experiments. The silicate stock for the +Si treatment was analyzed for dissolved Fe. The iron concentration in the Si stock caused the 20 μM Si(OH)₄ addition to raise ambient [dFe] by 0.05 nM. Immediately after amendments, Si production samples were spiked with silicon-32, incubated and processed as described for silicon-32 rate profiles. The effect of the same +Si, +Fe and combined +Si+Fe additions on primary productivity was determined in a parallel set of incubations using ¹⁴C-bicarbonate tracer following the light/dark bottle method described by Brzezinski and Washburn (2011). Samples for carbon fixation were size-fractionated through the same diameter and pore-size filters as used for silicon-32 tracer experiments, and sample activity was evaluated at sea using liquid scintillation counting in Ultima Gold XR cocktail.

The cruise tracks and environmental conditions encountered during EXPORTS have been published elsewhere (Siegel et al., 2021). Briefly, salinity in the mixed layer averaged 32.3 with little variation, increasing by about 0.2 at the base of the euphotic zone (1% light depth). Mixed layer temperatures were 14.0–14.2°C decreasing to 6.3–7.0°C at the 1% light depth. Analysis of these physical properties showed the presence of three distinct water types in the mixed layer during EXPORTS (Siegel et al., 2021) and during sampling by the Tully. Water types 1 and 3 were salty and fresh compared to type 2 and were mainly sampled at the beginning and end of the US cruise, respectively. When the Tully sampled near the Lagrangian float after the US cruise, the mixed layer had characteristics of water type 1 that was sampled at the beginning of the US occupation. Sampling by the Revelle occurred predominantly in water mass type 2, with the Ride sampling all three water types during spatial surveys. The euphotic zone (defined at the depth where PAR reached 1% of incident sunlight) at the stations where rate measurements were performed aboard the Revelle was between 63 and 70 m, with shallower depths occurring later in the cruise.

On four occasions the Ride and the Revelle both sampled the upper water column nearly simultaneously when the two ships were within 18 km of each other. In three cases the ships were in the same surface water type based on T/S characteristics (Siegel et al., 2021), while they were in different water types on the fourth (Table 1). To compare Si(OH)₄ concentrations, the values from the upper 20 m (all within the mixed later) were averaged for each ship on each occasion. For bSi concentrations, the values obtained from each ship were integrated to a common depth of 70 m using linear interpolation between sampling depths when necessary. Integrated concentrations were divided by the integration depth to calculate a depth-weighted average bSi concentration in the euphotic zone.

At three of the four stations used for intercalibration the average mixed-layer Si(OH)₄ concentrations on the two ships agreed to within 7–12% (Table 1). On the fourth, where each ship sampled a different water type, dissolved Si values were still within 39% of each other. The level of agreement for bSi concentrations, on the other hand, did not relate to water type. For bSi concentration in the one case when the ships were in different water types, the integrated, depth-normalized bSi concentration measured on samples from the Ride was 6.7 times higher than measured on the Revelle. For the other three comparisons, the ships were in the same water type. In one instance, the normalized concentrations obtained on each ship are within 1% of each other. For the other two comparisons, values are 2.0 and 2.2 times higher for the Ride compared to those from the Revelle (Table 1).

Profiles of Si(OH)₄ concentration measured on the Revelle revealed concentrations consistently >12 μM in euphotic zone, with an integrated depth-normalized average concentration in the upper 70 m of 15.7 ± 1.2 μM (n = 33) and a median concentration of 16.0 μM (Figures 1 and 2). Profiles of Si(OH)₄ concentration from the Ride displayed similar vertical patterns as those from the Revelle (Figure 1A), but depth-normalized integrated concentrations sampled by the Ride in the upper 70 m were lower on average, with a mean of 13.4 ± 2.1 μM (n = 63) and median of 13.6 μM. On both ships, the top of the nutricline, where Si(OH)₄ concentration began to increase relative to surface waters, was between 90 and 100 m (Figure 1A), significantly deeper than the euphotic zone as defined by the 1% light depth (63–70 m). No systematic spatial pattern in average concentrations was apparent in data from either ship (Figure 2). When the Tully sampled the EXPORTS site, the average Si(OH)₄ concentration in the euphotic zone, 16.2 μM, was similar to that observed by the Revelle.

Silicic acid concentrations from the Revelle that were associated with profiles of rate process measurements are listed in Table 2, with the relevant profiles depicted in Figure 3. Euphotic-zone-integrated depth-normalized values averaged 15.4 ± 1.8 μM (n = 9), with a median value of 16.1 μM. Concentrations at the deepest depth sampled shifted from initially relatively low values to higher values on year day 242 and remain high for the rest of cruise. Density, σ_θ, at the deepest depth sampled across all of these profiles averaged 25.42 ± 0.05 (n = 9), indicating that the change in Si(OH)₄ concentration at the 1% light depth was not due to a change in water type.

Concentrations of bSi were generally low during the entire EXPORTS cruise. Concentrations measured on samples from the Revelle were ≤210 nmol Si L^–1 (Figures 1B and 2), with integrated concentrations in the euphotic zone averaging 6.4 ± 2.2 mmol Si m^–2 (n = 9; Table 2). The shape of the profiles was variable, with surface or mid-depth maxima observed on occasion and otherwise relatively uniform values with depth (Figure 4). Significant bSi was present at the 1% light level (Figures 1B and 4), making euphotic zone integrals conservatively low relative to total upper water column inventories.

Although the Ride sampled waters that had lower average Si(OH)₄ concentrations compared to the Revelle, the average depth-normalized integrated bSi concentration sampled on the Ride was higher (Ride: 132 ± 112 nmol Si L^–1, n = 42; Revelle: 83 ± 46 nmol Si L^–1, n = 9; t-test, p = 0.005; Figure 1). The maximum bSi concentration measured on the Ride reached 500 nmol Si L^–1, but was 210 nM on the Revelle (Figure 1B). The average of the normalized bSi concentrations for all samples from both ships was 120 ± 102 nmol Si L^–1 (n = 51). Integrated values for the euphotic zone on the Ride averaged 8.3 ± 4.6 mmol Si m^–2 (n = 42) with a range of 1.8–21.8 mmol Si m^–2. On the Revelle, the mean integrated concentration was 6.4 ± 2.2 mmol Si m^–2 (n = 9), with a range of 3.2–9.4 mmol Si m^–2 (Table 2). Averaged across all samples from both ships, the mean integrated bSi concentration was 8.0 ± 4.3 mmol Si m^–2 (n = 51). The integrated bSi concentration measured by the Tully was 18.4 mmol Si m^–2 (Table 2) which is two to three times higher than the average for either US ship, but not out of the range of values for individual stations sampled by the Ride.

The nine profiles of size-fractionated bSi concentrations from the Revelle were fairly uniform with depth for both size fractions, with occasional surface or mid-depth (25 m) maxima (Figure 4). Integrated depth-normalized concentrations in the 0.6–5.0-µm size fraction averaged 2.1 ± 0.7 mmol Si m^–2 (range of 1.2–3.1 mmol Si m^–2), with the corresponding data for the >5.0-µm fraction being 4.4 ± 2.0 mmol Si m^–2 (range of 0.5–7.2 mmol Si m^–2). The proportion of bSi in the larger fraction was fairly consistent, except at the first station where it was only 17% of total bSi. On average 65 ± 19% of the bSi was >5.0 μm, with the median being 71% (Table 2, Figure 4). Average bSi concentration in the >5.0-µm fraction was twice that in the small fraction, with significant biomass present in both fractions at the deepest depth sampled (1% light; Figure 4). By the time of the Canadian occupation of the study site, the distribution of biomass between the two size fractions reversed compared to that during most of the US occupation. The small fraction comprised 64% of the total integrated bSi stock, with the >5.0-µm fraction dropping to 36%, as similarly observed at the first station occupied by the Revelle (Table 2).

Total silica production rates, calculated as the sum of the rates in the two size fractions across all nine profiles sampled by the Revelle, were low, with a mean of 10.1 ± 9.2 nmol Si L^–1 d^–1. Rates ranged from a low of 0.12 nmol Si L^–1 d^–1 at the base of the euphotic zone (1% light) to a high of 49.9 nmol Si L^–1 d^–1 in shallower waters. Total production rates integrated to the 1% light level averaged 0.62 ± 0.26 mmol Si m^–2 d^–1, with a range of 0.25–1.19 mmol Si m^–2 d^–1 (Table 2).

Profiles of silica production rate (ρ) in both size fractions were relatively uniform, declining slightly with depth with occasional surface or mid-depth maxima (Figure 4). The large size fraction that dominated siliceous biomass also dominated silica production (Table 2). Integrated production rates for the >5.0-µm size fraction averaged 0.50 ± 0.2 mmol Si m^–2 d^–1 compared to 0.12 ± 0.07 mmol Si m^–2 d^–1 for the small fraction. Rates in both size fractions were still significant at the 1% light depth (Figure 4), so that integrals provide conservatively low estimates of upper water column silica production.

Like profiles of ρ, profiles of the specific rate of silica production (ρ ÷ [bSi], V_b) showed little vertical structure (Figure 4). V_b was generally low, implying doubling times of about 2 weeks for both size fractions: 18 and 15 days for the small and large size fractions, respectively. Much higher specific rates were measured at the beginning of the cruise (year day 228, epoch 1, day 2) in the large size fraction, implying a doubling time of 0.9 days. Specific rates at the 1% light depth were often similar to near surface rates, indicating continued silica production in deeper waters.

In the upper euphotic zone, specific production rates were generally higher in the large size fraction than in the small one (Figure 4, Table 2), but specific rates switched to becoming higher in the small fraction at the base of the euphotic zone (Figure 5). The contribution of the small size fraction to silica production averaged across the 9 profiles was fairly constant throughout the upper euphotic zone. There the small fraction accounted for 15–20% of total silica production, with specific production rates that were about half those in the larger size fraction. The contribution of the small size fraction to total silica production rose to 40% near the base of the euphotic zone, where specific rates in the small fraction were over twice those in the larger size class (Figure 5).

Temporal trends observed in Lagrangian mode aboard the Revelle show significant transitions (Figure 6). The depth-normalized silicic acid concentrations in the mixed layer sampled by the conventional CTD/rosette (mixed layer depth defined as the first depth where the potential temperature change from 5 m, or the first recorded depth, exceeded 0.2°C) were low when the Revelle began sampling in water type 1 and remained low during the transition to water type 2 on year day 230. Then, while remaining in water type 2, depth-normalized integrated values increased beginning on year day 231 from 16 μM to 18 μM by year day 233. Concentrations remained at 18 μM until year day 238 when concentrations declined to near 16 μM again and remained there until the end of the cruise.

Changes in siliceous biomass occurred during the same transition periods that were observed for Si(OH)₄ concentration (Figure 6). The bSi concentration in the small size fraction exceeded that of the large size fraction at the beginning of the cruise. The depth-normalized concentration in the >5.0-µm fraction then increased, and that in the small size fraction declined during the transition in water type and from low to higher Si(OH)₄ concentration on year day 230, although the biomass data have limited temporal resolution during this period. Biomass in both size fractions increased during the transition to lower Si(OH)₄ concentrations from year days 231–233 and then declined over the last week of the cruise. Sampling during the Canadian occupation on year day 264 revealed that a bloom had occurred, with bSi tripling in the large size fraction and increasing 6-fold in the small size fraction (Table 2).

Silica production was dominated by the large size fraction throughout the US expedition (Figure 6). Production in the large size fraction declined initially, but then rose between year days 234 and 242 somewhat independently of the changes in Si(OH)₄ and bSi concentrations (Figure 6). Silica production in the large size fraction abruptly peaked on year day 244, 3 days after the observed decline in Si(OH)₄ concentration. Temporal dynamics in silica production were more muted in the small size fraction where rates were essentially constant through time, with a small peak in production that coincided with the increase in production by the large size fraction on year day 244 (Figure 6).

An analysis of variance (ANOVA) was used to investigate the responses of silica production to the +Si, +Fe and combined additions in the nutrient limitation experiments at the 40% and 10% light levels. Light depth, nutrient addition and size fraction were assigned as main effects, with all pairwise and 3-way interactions included. No interaction term was significant at p = 0.05. The analysis showed no significant effect of light depth (F= 1.75, d.f. = 1, p = 0.19), but a significant effect of nutrient addition (F = 5.8, df = 2, p = 0.004) and a significant effect of size fraction (F = 17.5, d.f. = 1, p < 0.001). A similar ANOVA for the response of primary productivity showed a single significant effect related to light level, reflecting a slightly more muted response to treatments at the 10% light level compared to the 40% (F = 9.4, d.f. = 1, p = 0.003).

The results of the ANOVA analyses were further investigated using t-tests (Table 3). V_b in the >5.0-µm size fraction did not show a consistent response to added Si, added Fe or the addition of both nutrients together (Figure 7). Treatment effects were observed in individual experiments, but the mean response of the large size fraction across all experiments was not significant at either the 40% or 10% light depth (Table 3). In contrast, a statistically significant mean response of increased Si uptake by the small size fraction was detected with added Si and with the combined addition of Si and Fe at both the 40% and 10% light depths (the response at 40% light was only marginally significant), with V_b in treatments increasing by an average of 2.5 ± 1.3 times relative to controls (Table 3, Figure 7). Like the large size fraction, the small size fraction did not respond to added Fe, indicating a Si-driven response in the combined Si and Fe treatment. The +Si, +Fe and combined additions did not elicit a significant response of carbon fixation rates for either size fraction (Table 3).

The EXPORTS program sampled the northeast subarctic Pacific during the seasonal minimum in surface Si(OH)₄ concentration. Both bSi and silica production rates in the euphotic zone were found to be low during the sampling by the US ships, but an apparent bloom had occurred by the time the site was sampled by the Canadian ship. Here we interpret these results in the context of data from the larger EXPORTS program together with previously published results to evaluate controls on upper-ocean Si cycling and the role of diatoms in the local biological pump.

Care was taken to compare observations on the Ride and the Revelle by conducting simultaneous CTD casts when the two ships were in proximity to one another. Differences in both Si(OH)₄ and bSi concentrations between ships during the four intercalibration casts were observed. They were not likely the result of analytical differences. The same sampling, preservation and sample shipping methods were used for both sample sets, and the samples from both ships were analyzed in the same laboratory using the same procedures. Data from both the intercalibration casts and regular sampling casts indicate that Si(OH)₄ concentrations were lower, but bSi concentrations higher, on the Ride compared to the Revelle. These differences suggest patchiness in biological activity and nutrient consumption, even within the same water type, implying that different portions of the same physical water type experienced variable biological histories.

Silicic acid concentrations in the euphotic zone were in the range expected from climatology for the late summer. During EXPORTS, bSi concentrations were two orders of magnitude lower compared to Si(OH)₄ concentrations, likely reflecting strong Fe limitation of diatoms. Seasonal profiles of bSi concentrations at OSP in 1999 and 2000 found integrated concentrations of 18.1 ± 0.3 mmol Si m^–2 (n = 2) in February, 41.8 ± 10.6 mmol Si m^–2 (n = 3) in June and 17.7 ± 9.9 mmol Si m^–2 (n = 3) in August–September (Peña and Varela, 2007; Lipsen, 2008), all of which are higher than the overall average from EXPORTS of 8.0 ± 4.3 mmol Si m^–2 (n = 51). The integrated total bSi concentration of 18.4 mmol Si m^–2 observed on the Tully is much higher than the cruise averages for the two US ships, not that different from the historic observations during the same time period and similar to the highest values observed from Ride (Figure 2).

During the US sampling, most of the bSi was >5.0 μm, with about 35% present in the small 0.6–5.0-µm fraction. This finding differs from other size-fractionated biomass measurements during EXPORTS, such as chlorophyll a concentration and primary productivity, which both showed dominance by small phytoplankton cells overall (Meyer et al., 2022). The dramatic increase in bSi concentrations during the Canadian sampling coupled with the shift in dominance to the small fraction suggests a fundamental change in dynamics, but we lack the temporal resolution to diagnose the mechanism. A similar dominance by the small size fraction, but at a lower overall biomass, was observed only once at the very first station sampled by the Revelle, suggesting considerable patchiness.

Size-fractionated measurements of bSi concentration are uncommon, limiting the available data for comparison with the present data set. Krause et al. (2017) observed that only 4% of bSi in the Sargasso Sea was between 0.2 and 3 μm. In the Pacific Ocean, Leblanc et al. (2018) observed that between 11 and 26% of bSi in the South Pacific was between 0.2 and 2 μm, while Wei et al. (2021) found 66% of the bSi in the oligotrophic western tropical North Pacific was 0.2–2.0 µm in size. These values bracket the average of 35% in the 0.6–5.0-µm fraction during EXPORTS, with the value for the western tropical North Pacific being similar to the 64% observed in the small size fraction on the Tully. Rigorous comparison of the data among these studies is difficult given the differences in the filter pore sizes used among studies. The 0.6-µm lower cut-off filter used during EXPORTS may have missed some small Si-containing cyanobacteria (Baines et al., 2012) that would be captured on the 0.2-µm filters used for the Krause et al. (2017), Leblanc et al. (2018) and Wei et al. (2021) studies, while the 5.0-µm upper limit used during EXPORTS would have included more larger particles than the 2–3-µm pore sizes used to define the upper limit of the small size fraction in the other studies. The importance of the small size class at OSP appears to be at the high end of what has been observed, at least during the Canadian sampling and at the first station sampled by the Revelle when the small size class was highly dominant. The identity of the organisms responsible is addressed below.

The measured rates of silica production were low, being only a few tens of nanomoles of bSi produced per liter each day. Specific rates were extremely low, with implied doubling times for bSi in both size fractions of about 2 weeks. Such low rates are consistent with the timing of the cruise during the annual minimum in net Si(OH)₄ drawdown. The main seasonal drawdown of Si(OH)₄ from climatology occurs between May and August (90 days) when Si(OH)₄ concentrations decline by 7.5 μM (Peña and Varela, 2007), corresponding to an integrated net silica production rate of 5.8 mmol Si m^–2 d^–1 assuming a 70-m productive layer. Adding an estimate of silica production supported through vertical eddy diffusion across the nutricline increases the integrated net daily rate to 6.3 mmol Si m^–2 d^–1 (see Introduction). The net silica production rate is the difference between gross silica production and losses due to silica dissolution in surface waters. To compare that rate to the gross silica production rates measured during EXPORTS using silicon-32 tracer, we calculated gross silica production during the spring–summer period as follows: integrated rates of silica production were compared with the export of bSi (see Efficiency of bSi export section), yielding an estimated dissolution:production ratio in the productive surface layer during EXPORTS of 0.44; net silica production, ρ_net, is the difference between gross silica production, ρ_gross, and losses to dissolution. Thus,

and, given the ρ_diss ÷ ρ_gross of 0.44, rearranging Equation (1) yields:

Substituting into Equation (1):

and solving for ρ_gross:

For the spring–summer productive period, the average gross silica production rate was thus estimated to be 6.3 ÷ 0.56 = 11 mmol Si m^–2 d^–1, which is nearly 18 times higher than the average gross silica production rate of 0.62 ± 0.26 mmol Si m^–2 d^–1 (Table 2) measured during EXPORTS.

While the estimate of silica production for spring–summer implies a strong seasonal cycle in silica production, the measured rates of silica production do not. Lipsen (2008) measured integrated silica production rates of 1.2 ± 0.9 mmol Si m^–2 d^–1 (n = 3 profiles) and 1.5 ± 1.6 mmol Si m^–2 d^–1 (n = 3 profiles) at OSP in June and August–September of 1999 and 2000, respectively. Those rates are lower than the seasonal estimate based on Si(OH)₄ drawdown, but higher than observed during EXPORTS. While discerning a clear seasonal pattern in production from the available data is difficult, these comparisons confirm that EXPORTS sampled during a period of low diatom activity, with the potential for significantly different Si cycling dynamics during spring and summer.

The temporal transitions in Si(OH)₄ concentrations and bSi concentrations in the mixed layer (Figure 6) were much faster than can be supported by the measured biological rates. During the transitions, the rate of change in Si(OH)₄ concentration was approximately 1 μmol Si L^–1 d^–1, which is 100 times faster than the average depth-normalized integrated silica production rate of 10.1 ± 9.2 nmol Si L^–1 d^–1 (n = 9). Integrated bSi concentrations in both size fractions changed by 2–4 nmol Si m^–2 d^–1, which exceeds measured integrated production rates by 4–8 times for the large size fraction and 16–33 times for the small. These changes imply a considerable level of patchiness within a given water type or significant shear that partially decoupled the movement of the Langrangian float (100 m) from that of the shallower mixed layer (29 ± 4 m; Siegel et al., 2021), such that the observed changes in the concentration of Si(OH)₄ and bSi over time were influenced by spatial variation among and within surface water parcels established over relatively long intervals of time compared to the temporal resolution of our measurements.

The 2-week doubling times implied by the specific rates of silica production underestimate the doubling times of diatoms. V_b underestimates the specific rate of living cells as it is normalized to total bSi that includes both living siliceous organisms, mainly diatoms, and detrital silica. Values of V_b can be made congruent with the biomass changes if the detrital silica comprised 80% of the bSi present. Such a high detrital fraction is consistent with the findings of Krause et al. (2010) who found that 80–90% of the bSi in the equatorial Pacific HNLC region was detrital. Such high percentages may be characteristic of severely Fe-limited regions where both diatom growth rates and biomass are low.

The lack of evidence for Fe limitation of silica production and primary production in both size fractions in the nutrient limitation experiments may be related to the relatively short 24-h incubations employed. Longer-term “grow-out” experiments lasting several days on the same cruise show strong growth responses and bSi increases in response to the same level of added Fe in the same two size fractions that were used in the present study (see Text S1, Figures S1 and S2). Interestingly, Si addition alone did not increase the growth or biomass of either size fraction in the longer-term experiments (Figures S1 and S2), whereas Si uptake in the small size fraction was clearly stimulated by added Si over 24 h in the experiments reported here. Taken together, these results suggest that Fe limited the growth rate of diatoms in the small size fraction, while their Si uptake was substrate-limited, as has been observed for the diatom community as a whole in the equatorial Pacific (Brzezinski et al., 2008). This hypothesis is consistent with both the short-term and long-term experimental results, as without Fe to permit the growth that would support a persistent demand for Si, cell division and hence frustule deposition would be severely truncated, eliminating any response to added Si on longer time scales.

The lack of experimental evidence for Si limitation of silica production in the large fraction is not surprising. The average half saturation constant for Si(OH)₄ uptake in diatoms is about 2 μM (Martin-Jézéquel et al., 2000). Assuming Michaelis-Menten uptake kinetics, diatoms with a half saturation constant of 2 μM growing in waters with 12 μM to 15 μM Si(OH)₄ would be taking up Si(OH)₄ at 86–88% of their maximum rate (V_m). The response of Si uptake of the small size fraction to added Si is surprising, both because it occurred in the small “non-diatom” fraction and it occurred in a system where Si(OH)₄ was >12 µM at all depths in the euphotic zone, implying inefficient uptake kinetics by the organisms responsible for the response.

A sizeable fraction of the bSi (35%) and Si uptake (19%) was by organisms between 0.6 and 5 µm in size, and this small size fraction dominated siliceous biomass at the very first station sampled by the Revelle (83%) and again during the sampling aboard the Tully (65%) that both sampled water type 1 (Table 1). At roughly half the stations sampled by the Revelle the average V_b in the small size fraction was nearly the same as that in the large size fraction, and the small fraction exhibited much higher specific rates deeper in the euphotic zone compared to the larger size class (Table 1). Notwithstanding the issues with detritus, these specific rates imply active growth and Si use by both small and large cells. The small size fraction also displayed unique Si-uptake physiology not shared by the larger fraction, with a statistically significant increase in Si uptake in response to added Si(OH)₄. Explanations for activity in the small size fraction include the passage of larger cells through the 5.0-µm filters that were then captured on the 0.6-µm filter during serial filtration, Si uptake by cyanobacteria in the picoplankton size class, and the presence of small diatoms or other small silicifiers that were <5.0 μm.

Pseudo-nitzschia spp. were the most abundant large diatom taxa present during the EXPORTS cruise, forming visible chains in micrographs from imaging flow cytometers (Sosik HM, personal communication; https://ifcb-data.whoi.edu/timeline?dataset=EXPORTS). There is some potential for these taxa to contribute to the small size fraction. Marchetti et al. (2008) measured the physical dimensions of the frustules of five Pseudo-nitzschia taxa isolated from OSP and all had transapical widths <5 μm, suggesting that they could pass through a 5.0-µm filter pore if aligned end-on as single cells. However, if the passage of Pseudo-nitzschia was responsible for the observed activity in the small size fraction, then the increase in Si uptake in response to added Si should have been evident in both size fractions as Pseudo-nitzschia dominated the large size fraction. That pattern was not observed making this possibility unlikely.

Small cyanobacteria of the genus Synechococcus are known to contain significant amounts of Si relative to their size (Baines et al., 2012; Ohnemus et al., 2016), and culture studies have shown that Si uptake by Synechococcus increases with added Si up to at least 500 μM (Brzezinski et al., 2017) consistent with observed increase in Si uptake in this size class at the high Si(OH)₄ concentration (30 μM) in experimental treatments with added Si. During EXPORTS integrated abundances of Synechococcus in the euphotic zone (1% light) averaged 1.1 x 10¹² ± 0.4 x 10¹² (1 σ_SD, n = 24) cells m^–2 (Graff JR, personal communication). Estimating their contribution to the standing stock of bSi requires knowledge of their cellular Si content. We used the measured values by Ohnemus et al. (2016) in the Sargasso Sea of 46 amol Si cell^–1 to convert Synechococcus abundance during EXPORTS to bSi concentration. The resulting integrated bSi concentration of 0.05 mmol Si m^–2 is only 2% of the average integrated bSi concentration in the 0.6–5.0-µm fraction, similar in magnitude to estimates from the Sargasso Sea by Krause et al. (2017). This estimated fraction is so low that Synechococcus is not likely to be responsible for the observed Si uptake and response to added Si in the small size fraction.

Also possible is that small relatively under-studied eukaryotes, such as the silicified Bolidophycae, could be contributing to Si uptake in the small size fraction. These sister taxa to diatoms are 10 times more prevalent in the small size fraction (0.8–5 µm) than in larger size fractions in the Tara Oceans global ocean sampling datasets (Ichinomiya et al., 2016) and common throughout the global ocean; however, they typically comprise a minor component of the photosynthetic community. Thus, a novel group of eukaryotes contributing to the Si uptake signature in the small size fraction at OSP is unlikely.

Extremely small diatoms, such as Minidiscus sp., are known from OSP (Clemons and Miller, 1984) and small pennate diatoms in the <5-um size class have been reported at abundances of 10⁴ cells L^–1 (Boyd and Harrison, 1999). Size-fractionated HPLC pigment data from the underway sampling system aboard the Ride (depth of 3–5 m) during EXPORTS show that the fucoxanthin:chlorophyll a ratios and the chlorophyll c:chlorophyll a ratios in the 0.7–5.0-um size fraction were 80–100% of the same ratios in the >5-µm fraction indicating the presence of <5.0-µm diatoms (Roesler CS, personal communication). Therefore, small diatoms most likely account for the silica production in the small size fraction and the uptake response of that fraction to added Si.

Substrate limitation of Si uptake for small diatoms at ambient Si(OH)₄ concentration as high as 15 μM is surprising given the Si uptake kinetics of most diatoms. However, that expectation is based on Si uptake kinetics for diatoms from culture studies, which have generally examined taxa that are relatively large. Also of potential relevance is that the Si additions made during EXPORTS raised ambient Si(OH)₄ concentrations to above 30 μM. In laboratory culture, this amount is near the threshold concentration where diatoms switch from transporter-mediated Si uptake to reliance on diffusive transport (Shrestha and Hildebrand, 2015). The factors that control diffusive flux for large and small cells differ in ways that are consistent with a stronger uptake response to high Si(OH)₄ concentration in smaller cells. The rate of diffusion of a nutrient to the cell surface is a function of cell radius, whereas the Si requirement for the frustule is a function of cell surface area, which varies as the radius squared, diminishing the effectiveness of diffusion as a means to support silica production in larger cells. For example, for a small (4-µm diameter) and large (20-µm diameter) diatom, each with the same frustule thickness, the 2.5-fold increase in cell radius increases the diffusive flux of Si(OH)₄ by a factor of 4, but the Si required to construct the frustule increases by a factor of 25. This discrepancy between small and large cells is a conservative estimate, as larger cells tend to have thicker frustules, which would amplify the contrast. Even with the same thickness, the 2.5-fold increase in the bSi-normalized specific uptake rate, V_b, observed in response to added Si(OH)₄ in the small cells, would truncate to a 2.5 x (5 ÷ 25) = 50% increase in V_b for the larger cell. Given that V_b in the larger size fraction was about twice that in the small fraction on average at ambient Si(OH)₄ concentration (Table 2), the effective increase in uptake in the larger cells would drop to 25%, which is close to the experimental uncertainty for these types of field experiments (Nelson et al., 2001).

Data from the overall EXPORTS program allow euphotic-zone silica cycling to be linked to the export of diatom silica and to diatom organic carbon export out of the surface ocean. One conceptual metric for the efficiency of export is the Ez ratio (Buesseler and Boyd, 2009). For organic carbon, the Ez ratio is defined as the rate of export of organic carbon out of the base of the productive surface layer expressed as a fraction of the rate of net primary production in that layer. There is no fixed convention to define the depth of the productive layer. For EXPORTS, the reference depth for calculating Ez ratios for organic carbon was defined as the depth where in situ chlorophyll fluorescence, as measured by the fluorometers on the CTD/rosette, declined to 10% of the maximum signal measured in overlying waters (Buesseler et al., 2020; Estapa et al., 2021), which is approximately the 0.1% light level and very close to the depth of the nutricline for Si(OH)₄ (Figure 1A). An Ez ratio for silicon would be defined similarly to be the ratio of the rate of opal export at the base of the productive layer expressed as a fraction of net silica production in that layer.

Silicon-32 rate measurements are often referred to as gross silica production rates (e.g., Equations 1–4), as tracer methods measure Si uptake independently of losses due to silica dissolution (Nelson, 1975). This older convention does not take into consideration the efflux of Si from the cell after it is taken up by diatoms, which has since been shown to occur (Milligan et al., 2004), possibly mediated by specific Si efflux transporters (Shrestha et al., 2012). Thus silicon-32 production rate measurements actually reflect the net rate of Si incorporation into particles and are thus conceptually equivalent to net primary productivity (NPP) for organic carbon in that the measurement does not include Si that is lost from cells due to Si efflux, analogous to phytoplankton respiratory C losses for NPP. By contrast, rates of silica production determined using silicon-32 after accounting for losses due to the dissolution of diatom frustules in the euphotic zone would be analogous to net community production (NCP) for organic carbon that accounts for respiratory losses of C within the entire upper ocean food web, and at steady state those rates would be equivalent to the rate of export of bSi out of the productive layer.

The Ez ratio for Si was variable in time and space and differed significantly depending on the method used to measure bSi export. The export of bSi was measured using both neutrally buoyant sediment traps (NBSTs) and surface-tethered sediment traps (SSTs) at 95–105 m once per epoch during EXPORTS (Estapa et al., 2021). Ez ratios, denoted as Ez₁₀₀ ratios, were calculated using those export rates and total integrated rate of silica production in the euphotic zone within each epoch as our best estimate of silica production in the upper 100 m. Those Ez₁₀₀ ratios varied by a factor of two across epochs, averaging 56 ± 15% (n = 3; Table 4). They are biased high, as the sampling to the 1% light depth for silica production rate measurements did not reach the depth limit of silica production (Figure 4). The average implies that the silica dissolution:production ratio in the euphotic zone (D:P = 1 – Ez₁₀₀ ratio) of 0.44 is somewhat lower than the global average of 0.5–0.6 (Tréguer and De La Rocha, 2013).

Biogenic silica export rates during EXPORTS based on ²³⁴Th are available as cruise averages for multiple depths in the upper 100 m (Roca-Martí et al., 2021). Those data were used to estimate the Ez ratio referenced to the base of the euphotic zone, Ez_1% ratio, to align the depth of export with the depth range for silica production measurements and also to estimate an Ez₁₀₀ ratio to parallel the analysis above using data on bSi export from sediment traps. The silica export rate at the average euphotic zone depth of 70 m was estimated to be 0.83 ± 0.24 mmol Si m^–2 d^–1 by interpolating the average silica export rates at 65 and 80 m to 70 m. The flux of bSi at 100 m was lower, 0.73 ± 0.23 mmol Si m^–2 d^–1. The ²³⁴Th-based estimates of the Ez_1% ratio, 1.33 ± 0.64 at 70 m and the Ez₁₀₀ ratio, 1.18 ± 0.62 both exceed unity (Table 4). Considering the Lagrangian sampling strategy employed on the Revelle, those ratios imply that the standing stock of bSi should have diminished substantially during the 3 weeks of the cruise, which was not observed. The discrepancy between the rate of bSi export measured in traps compared to ²³⁴Th is unresolved. Estapa et al. (2021) point out that the traps failed to collect the flux of ²³⁴Th predicted by water column thorium deficits, implying under-sampling of sinking particles by traps, significant zooplankton active migrant fluxes or a spatial-temporal mismatch in the processes sampled by the two approaches.

Given that the EXPORTS cruise occurred during the seasonal low in net Si(OH)₄ drawdown in the upper water column, we estimated the Ez₁₀₀ ratio for bSi during the more productive spring and summer period for comparison. Timothy et al. (2013) measured the flux of bSi from May through August at OSP to be 1.0 mmol Si m^–2 d^–1 at 200 m. To compare these rates to those measured during EXPORTS at 100 m, the 200-m rates must be corrected for the attenuation of the flux of bSi between 100 and 200 m. Estapa et al. (2021) measured the ratio of the flux of bSi at 200 m to that at 100 m to be 0.56 during EXPORTS. Applying that ratio to the 200-m flux measured by Timothy et al. (2013) yields a rate of bSi export during spring and summer at 100 m of 1.8 mmol Si m^–2 d^–1. That value is considerably higher than the average of 0.34 ± 0.26 mmol Si m^–2 d^–1 (n = 3, Table 4) measured using traps during EXPORTS despite potential under-collection of particles by the relatively shallow moored conical traps employed by Timothy et al. (2013).

The Ez₁₀₀ ratio for bSi during the spring through summer period was then estimated using the gross silica production rates determined above (Equation 4) for the spring through summer period of 11 mmol Si m^–2 d^–1, recognizing that Equation 4 uses an older definition of gross silica production which we argued above more appropriately represents net silica production. The resulting Ez₁₀₀ ratio of 1.8 ÷ 11 = 0.16 is three times lower than measured during EXPORTS. While only approximate, the lower value of the Ez₁₀₀ ratio and the much higher rates of both Si production and bSi export in the spring and summer imply a greater potential for both diatom biomass accumulation in surface waters and for higher opal export earlier in the year, as has been documented in part during time-series trap studies (Wong et al., 1999; Timothy et al., 2013). We note that the lower silica production rates obtained by Lipsen (2008) using silicon-32 incubations in June of 1999 and 2000 (average 1.2 ± 0.9 mmol Si m^–2 d^–1, n = 3) yield a correspondingly higher Ez₁₀₀ ratio of 1.5. The estimate from climatological nutrient depletion is likely more representative given the potential for significant variability in daily silica production.

Wong and Matear (1999) estimated that in spring diatoms account for 35–51% of phytoplankton carbon fixation at OSP. That estimate was obtained by converting net Si(OH)₄ depletion to diatom primary production using a Si:C of 0.27, citing Hutchins et al. (1998) as the source of the appropriate ratio in low Fe waters. Applying that same Si:C ratio to the net silica production measured on EXPORTS yields an average diatom carbon fixation across all profiles that is 4.6 ± 2.0% (n = 9) of measured total carbon fixation (Stephens et al., n.d.). A low percent contribution of diatoms to primary productivity is to be expected as the bSi:POC mole ratio of particles in the euphotic zone averaged 0.02 (Roca-Martí et al., 2021).

The measured flux of bSi in traps at 100 m averaged 0.36 mmol Si m^–2 d^–1 (Table 4), implying a maximum carbon export by diatoms of 0.36 ÷ 0.27 = 1.33 mmol C m^–2 d^–1. That value nearly equals the average total organic carbon flux measured in traps, 1.38 ± 0.77 mmol C m^–2 d^–1 at 100 m (n = 9; Estapa et al., 2021). A similar calculation using the silica export at 100 m measured by the ²³⁴Th method, 0.73 ÷ 0.27 = 2.7 mmol C m^–2 d^–1 (Roca-Martí et al., 2021), yields a diatom carbon export that exceeds the measured total carbon flux estimate from the ²³⁴Th method of 2.01 ± 0.56 mmol C m^–2 d^–1 (n = 63). The use of a Si:C ratio of 0.27 to make these estimates assumes that all diatom carbon export occurred as intact diatom cells; however, intact diatoms were only a minor fraction of exported cells (Durkin et al., 2021). A better average may be obtained from an examination of Si:C ratios in exported particles. The bSi:POC mole ratio in material sinking into traps at 100 m, 0.29 ± 0.02 (n = 4; Estapa et al., 2021), was 10-fold higher than the ratio in the euphotic zone, 0.02 (Roca-Martí et al., 2021), reflecting the more rapid remineralization of organic matter compared to opal. Assuming that diatom carbon decreased by 10-fold relative to diatom silica by 100 m lowers the estimated diatom carbon export from 1.33 mmol C m^–2 d^–1 to 0.13 mmol C m^–2 d^–1 for the estimate from traps, and from 2.7 mmol C m^–2 d^–1 to 0.27 mmol C m^–2 d^–1 for the estimate from ²³⁴Th, which would make diatoms responsible for 9–13% of total organic C export. That estimate of the contribution of diatoms to carbon export is 2–3 times greater than their estimated contribution to primary production (3–7% see above) and is consistent with prior observations that diatoms contribute more to C export than to primary productivity in oligotrophic systems (Nelson and Brzezinski, 1997; Brzezinski et al., 2011b).

As this study provides the first detailed examination of the Si cycle in the subarctic Pacific, comparing the observations obtained to those from other HNLC and open-ocean systems is valuable. Table 5 compares the major elements of the silicon cycle at OSP to the HNLC regions of the Southern Ocean and equatorial Pacific, and to Si cycling at other open ocean sites off Hawaii at station ALOHA and off Bermuda at the Bermuda Atlantic Time Series (BATS) site. A caveat of this exercise is that the parameter estimates for the Southern Ocean, BATS and ALOHA are based on seasonal to annual data, while much of the data for the equatorial Pacific and the subarctic Pacific are from fewer sampling events.

The Southern Ocean dominates all other systems in terms of siliceous biomass, silica production and silica export (Table 5) and is thus largely excluded from further discussion. For the remaining regions, integrated bSi concentrations at OSP exceed those in the subtropical gyres (ALOHA, BATS), but are less than those in the upwelling zone of the eastern equatorial Pacific. Integrated silica production rates are intermediate, with rates at OSP being most similar to those at BATS, but less than those measured in the equatorial Pacific and greater than at ALOHA.

Si limitation at OSP occurred only in the small size fraction where it affected the rate of Si uptake, but not likely the growth rate. In contrast, substrate limitation of Si uptake is pervasive in the other three systems (Table 5) and occurs in both the Antarctic and Pacific sectors of the Southern Ocean (Franck et al., 2000; Nelson et al., 2001). Limitation of Si uptake is sufficiently severe at BATS that Si limitation of growth rate may occur (Brzezinski and Nelson, 1996). Size-fractionated Si uptake rates are not available for the other systems, but experiments in the equatorial Pacific showed that the small Pseudo-nitzchia species that dominated the ambient diatom community were Si replete (Brzezinski et al., 2008; Brzezinski et al., 2011a) rather than being uptake limited as observed at OSP. Co-limitation of growth rate by Fe and of Si uptake by Si occurs both at OSP (this study) and in the equatorial Pacific (Brzezinski et al., 2008; Brzezinski et al., 2011a), but the effect is observed for large diatoms in the equatorial Pacific while it was confined to small diatoms at OSP, with the larger diatoms being solely limited by iron (see Figures S1 and S2). Given that severe Si(OH)₄ depletion is rare at OSP (Wong and Matear, 1999), the unmet demand for Si by small diatoms appears held in check by iron limitation of their growth rates, which combined with the direct evidence for strong Fe, but not Si, limitation of larger diatoms (Martin and Fitzwater, 1988; Martin et al., 1994; Boyd et al., 1996) indicates that the northeastern subarctic Pacific is unique in that it is a true Fe-limited HNLC region while dynamics in all other low Fe regions are better described as high-nitrate low-silicate low-chlorophyll (HNLSLC) (Wilkerson and Dugdale, 1996).

Of the four systems outside the Southern Ocean, the export of bSi at OSP measured by both sediment traps and ²³⁴Th is the highest, with a high Ez ratio during EXPORTS that is similar to that at ALOHA (Table 5). Our indirect estimate of the Ez ratio during spring and summer, 0.16, falls to levels observed at BATS. Lower water temperatures may be a significant factor controlling the Ez ratio for opal at OSP. Water temperatures at OSP are more than 10°C colder than at ALOHA or BATS. Silica dissolution is highly temperature-dependent, with a Q₁₀ of between 2.2 and 2.6 (Kamatani and Riley, 1979; Kamatani, 1982; Natori et al., 2006), which would lead to dissolution rates in OSP surface waters that are less than half of those in the two warmer systems. No direct measure of silica dissolution rates are available from OSP, but the average dissolution:production ratio estimated from traps as (1 – Ez₁₀₀) of 0.44 implies that this ratio at OSP is relatively low and similar to ALOHA, which has been described as a low productivity, high export system (Brzezinski et al., 2011b). While slower rates of silica dissolution at lower temperatures may enhance opal preservation at OSP, clearly this explanation cannot explain the similarity in Ez between ALOHA and OSP. Overall, in comparison to the equatorial Pacific and the North Pacific and North Atlantic subtropical gyres, OSP is intermediate in bSi concentration, intermediate in rates of silica production, but high relative to the magnitude and efficiency of bSi export.

The upper ocean silica cycle in the subarctic Pacific near OSP was examined as part of the EXPORTS program. The conditions encountered were consistent with climatology and indicated that sampling occurred during the annual minimum in net Si(OH)₄ depletion in surface waters. Biogenic silica concentration was low, being in the tens of nanomolar range. Rates of silica production were also low and more similar to those found in the subtropical gyres of the North Pacific and North Atlantic than to other open-ocean HNLC regions. However, estimates suggest that silica production rates may be ten times higher during the more productive spring and summer period.

A fair fraction of bSi and Si uptake occurred in the 0.6–5.0 µm fraction, though on average the >5.0 µm fraction dominated at 65% of biomass and 81% of Si uptake during EXPORTS. A bloom event appeared to occur after the EXPORTS cruise as revealed by sampling 12 days later aboard the Tully that showed strong increases in bSi concentrations in both size fractions (Table 2). The bloom occurred in water type 1 similar to the first station sampled by the Revelle early in the cruise. Both samplings of water type 1 saw dominance by the small size fraction irrespective of the variation in biomass.

Iron limitation of silica production and primary productivity was not detected in short-term experiments lasting 24 h, in contrast to longer-term grow-out experiments during EXPORTS that revealed strong Fe limitation of both the large and small size fractions. Si limitation of Si uptake was detected in the small, but not the large size fraction, potentially due to diatom reliance on diffusive transport under the experimental conditions. Indirect evidence points to the response in the small size fraction as being due to <5.0-µm diatoms rather than to cyanobacteria or to larger diatoms that passed through the 5.0-µm filter. Those small diatom taxa appear to be co-limited, with growth rate limited by Fe and Si uptake rate restricted by ambient Si(OH)₄ concentration, as observed for large diatoms in the HNLC region of the equatorial Pacific. Exceptionally strong Fe limitation of all diatoms in the subarctic Pacific may explain the lack of preferential drawdown of Si(OH)₄ relative to nitrate at OSP (Wong and Matear, 1999) compared to other Fe-limited regions where such preferential drawdown is common. The northeastern subarctic Pacific is thus unique, in that it is a true HNLC region with high concentrations of all macronutrients, including Si(OH)₄.

About a third of the bSi produced in the euphotic zone was exported through 100-m depth as estimated from sediment traps. Conditions at OSP most resembled those at station ALOHA off Hawaii in that OSP appeared as a low productivity, high export system relative to bSi. However, the drivers of this pattern differ between the two regions, with strong Fe limitation reducing production and colder temperatures lowering silica dissolution at OSP compared to low Si(OH)₄ concentration and near permanent stratification lowering silica production at ALOHA, with greater relative losses to dissolution in the warmer subtropical waters. The estimated contribution of diatoms to organic carbon export during EXPORTS (9–13%) was disproportionately high compared to their estimated contribution to primary productivity (3–7%) as has been observed in other oligotrophic systems (Nelson and Brzezinski, 1997).

Data sets used in this article can be found at: EXPORTS, 2018, SeaBASS, DOI: http://dx.doi.org/10.5067/SeaBASS/EXPORTS/DATA001. EXPORTS, 2018., BCO-DMO, https://www.bco-dmo.org/program/757397. DataVerse, https://dataverse.scholarsportal.info/privateurl.xhtml?token=f93ba246-9955-48cc-8bf9-21e7312db9a8.

The supplemental files for this article can be found as follows:

Text S1. Results of long-term investigations of nutrient limitation.

Figure S1. Response to iron or silicic acid addition.

Figure S2. Response to silicon or iron stress.

Thanks go to the captains and crews of the R/Vs Revelle and Ride and the CCGS Tully. The authors wish to thank Heidi Sosik for access to data from the Imaging FlowCytobot and Collin Roesler for interpretation of size-fractionated HPLC pigment analysis. They thank Dr. Salvatore Caprara and PhD student Travis Mellett for their assistance in water collection and incubation setup.

The authors would like to acknowledge support from the U.S. National Science Foundation awards NSF-OCE 1756442 (MAB), NSF-OCE 1756433 (KNB), NSF-OCE 1756816 (BDJ), and Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant (DEV).

The authors declare that they have no conflict of interest.

Contributed to conception and design: MAB, BDJ, KNB, DEV.

Contributed to acquisition of data: SMK, JLJ, MAB, BDJ, KNB, DEV.

Contributed to analysis and interpretation of data: MAB, BDJ, KNB, DEV.

Drafted and/or revised the article: MAB, BDJ, KNB, DEV, SMK, JLJ.

Approved the submitted version for publication: MAB, BDJ, KNB, DEV, SMK, JLJ.

How to cite this article: Brzezinski, MA, Varela, DE, Jenkins, BD, Buck, KN, Kafrissen, SM, Jones, JL. 2022. The upper ocean silicon cycle of the subarctic Pacific during the EXPORTS field campaign. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2021.00087

Domain Editor-in-Chief: Jody W. Deming, School of Oceanography, University of Washington, Seattle, WA, USA

Guest Editor: Ivona Cetinic, NASA GSFC/USRA, Greenbelt, MD, USA

Knowledge Domain: Ocean Science

Part of an Elementa Special Feature: Accomplishments from the EXport Processes in the Ocean from RemoTe Sensing (EXPORTS) Field Campaign to the Northeast Pacific Ocean