Dissolution Dominates Silica Cycling in a Shelf Sea Autumn Bloom

Autumn phytoplankton blooms represent key periods of production in temperate and high‐latitude seas. Biogenic silica (bSiO2) production, dissolution, and standing stocks were determined in the Celtic Sea (United Kingdom) during November 2014. Dissolution rates were in excess of bSiO2 production, indicating a net loss of bSiO2. Estimated diatom bSiO2 contributed ≤10% to total bSiO2, with detrital bSiO2 supporting rapid Si cycling. Based on the average biomass‐specific dissolution rate (0.2 day−1), 3 weeks would be needed to dissolve 99% of the bSiO2 present. Negative net bSiO2 production was associated with low‐light conditions (<4 E·m−2·day−1). Our observations imply that dissolution dominates Si cycling during autumn, with low‐light conditions also likely to influence Si cycling during winter and early spring.


Introduction
Despite representing <10% of the ocean, shelf seas represent~30% of primary production (~15 Pg C/year, Field et al., 1998). Annually, spring blooms, triggered by stratification and optimal growth conditions (Lindeman & St. John, 2014;Taylor & Ferrari, 2011), dominate production, with weaker rates during the stratified summer terminated by increased autumn mixing (Findlay et al., 2006;Painter et al., 2016;Perry et al., 2008;Wihsgott et al., 2018). During summer, phytoplankton adapt to low nutrient conditions in surface waters and low light at the base of the surface mixed layer (SML), forming a subsurface chlorophyll maximum (SCM; Hickman et al., 2009). Weakening stratification is associated with a secondary biomass peak, considered an "autumn bloom," which may reflect elevated production and/or redistribution of SCM material (Findlay et al., 2006;Painter et al., 2016;Perry et al., 2008).
Ratios of D:P represent the fraction of bSiO 2 production supported by upper-ocean Si-recycling, or expressed as 1-D:P, the fraction of bSiO 2 production supported by "new" Si from the deep sea . Most studies measuring rates of production and dissolution have occurred during spring and summer, with coastal studies almost exclusively from upwelling sites (Krause et al., 2010;Tréguer & De La Rocha, 2013). Despite the importance of Si dissolution in controlling upper-ocean Si dynamics, and the fate of diatom carbon, uncertainty remains of what controls the variability in D:P ratios, or how this impacts the Si cycle Tréguer & De La Rocha, 2013).
With this in mind, this study examined bSiO 2 production and dissolution, alongside measurements of bSiO 2 pool composition and size, during an autumn shelf sea bloom. The specific objectives were to (1) measure the magnitude of bSiO 2 production and examine the composition of the diatom community; (2) examine rates of bSiO 2 dissolution when the water column is "resetting" before winter; and (3) compare ratios of dissolution to production (D:P) and the factors driving them in order to provide information on Si recycling.
Depth of the SML was determined from density profiles, from the ships CTD and mooring time-series, by identifying where potential density increased by 0.02 kg m −3 above the 10-m value Wihsgott et al., 2019). A 2π PAR irradiance sensor (Skye Instruments, SKE 510) on the RRS Discovery and a quantum PAR meter (LiCor Inc., USA) on the mooring-measured incident irradiance (E 0 ). Average SML irradiance (Ē SML ) was determined following Poulton et al. (2018). Mooring fluorescence data were determined by a Seapoint fluorometer, standardized using fluorescent sulfate microspheres (FluoSpheres, ThermoFisher Scientific) with nighttime data only to avoid daytime quenching (Wihsgott et al., , 2019.

Silica Cycling
Micromolar (μmol/kg) concentrations of nitrate + nitrite (NOx) and silicic acid (Si (OH) 4 ) were measured with a Bran and Luebbe segmented flow colorimetric autoanalyzer following Woodward and Rees (2001). Samples were taken in acid-cleaned, aged, HDPE bottles, and clean sample handing procedures followed. All samples were analyzed within 1-2 hr of collection. Certified reference materials were used daily (KANSO, Japan), and analytical procedures followed International GO-SHIP recommendations (Hydes et al., 2010). The typical uncertainty of the analytical results were between 2% and 3%, and the limits of detection for nitrate and phosphate were 0.02 μmol L −1 , while Si(OH) 4 was always higher than the detection limits. Measurements of particulate silica concentrations (bSiO 2 ; μmol Si kg −1 ) were made on 0.5-L samples filtered onto 0.8-μm polycarbonate filters, digested in 0.2 M·NaOH at 85°C for 1 hr and neutralized with 0.2 M·HCl (Ragueaneau & Tréguer, 1994). Sample analysis was with a SEAL QuAAtro autoanalyzer (Poulton et al., 2006) using standard techniques. Daily measurements of gross bSiO 2 production (ρ P ) and net bSiO 2 production (ρ N ) were made at the six depths, and followed Krause et al. (2010) (see also Poulton et al., 2006). Duplicate water samples were spiked with 0.03-μCi high specific activity (2.06 μCi/mL) 32 Si-labeled Si (OH) 4 and incubated for 24 hr in 70-mL polycarbonate bottles (Corning TM ). Incubations were filtered onto 25-mm 0.8-μm polycarbonate (Whatman Nucleopore TM ) filters under gentle vacuum. Filters were digested with 0.2-M NaOH in 20-mL plastic scintillation vials (PerkinElmer UK), at 85°C for 1 hr. After cooling, 0.2-M HCl was added and 15 mL of UltimaGold TM liquid scintillation cocktail. Activity on the filters was determined on a 1220 Quantulus TM ultralow level liquid scintillation counter (Wallac, Finland) after >100 days, allowing secular equilibrium.
Daily ρ P and biomass-normalized specific uptake (V P ) were calculated after Krause et al. (2010). Triplicate unspiked samples, collected alongside samples for ρ P determination were incubated in parallel for 24 hr, with differences between initial and end bSiO 2 concentrations considered as net bSiO 2 production (ρ N ), with dissolution defined by the difference between ρ P and ρ N (Krause et al., 2010). Biomass-normalized specific dissolution (V D ) was calculated after Krause et al. (2010).

Chlorophyll-a, Primary Production and Diatom Biomass
Chlorophyll-a (Chl) concentrations (mg m −3 ) were measured on 0.2-L water samples filtered onto 25-mm Whatman GF/F filters, and extracted in 6-mL 90% acetone (Sigma-Aldrich, UK) at 4°C for 18-24 hr. Chl fluorescence was measured on a Turner Designs Trilogy TM fluorometer using a nonacidification module calibrated with solid and pure Chl standards (Sigma-Aldrich, UK). Daily net primary production (NPP) rates were determined following Poulton et al. (2018). Four samples in 70-mL polycarbonate bottles from each depth were spiked with 1,258-1,628 kBq of 14 C-labeled sodium bicarbonate (PerkinElmer, UK), with 1-mL of borate-buffered formaldehyde (~2% final solution) added to one bottle as a formalin-killed blank. Samples were incubated in parallel to ρ P and terminated by filtering onto 25-mm 0.45-μm Whatman Nucleopore TM polycarbonate filters, with extensive rinsing to remove unfixed isotope. Organic (NPP) carbon fixation was determined using the microdiffusion technique (as in Poulton et al., 2018) with 1-mL of 1% orthophosphoric acid added to remove 14 C-particulate inorganic carbon and 10-mL of UltimaGold TM (PerkinElmer, UK) liquid scintillation cocktail. Filter activity was determined on-board a Tri-Carb 3100TR (Perkin-Elmer) liquid scintillation counter.
Diatom cell counts were analyzed from each sampling depth by light microscopy of samples in 250-mL brown glass bottles, preserved in acidic Lugol's solution (2% final solution), and analyzed with a Leica DMI4000B inverted microscope (Widdicombe et al., 2010). Diatom bSiO 2 was estimated by applying the average cell Si content (8.14 pmol Si cell −1 ) from Brzezinski (1985) to the diatom cell counts.

Celtic Sea Hydrography
At CCS, ocean cooling commenced in early October, steadily deepening the SML until the end of December, primarily by wind-induced mixing (Figure 1b; Wihsgott et al., 2019). Integrated Chl (Figure 1b) increased parallel to SML deepening, up until 12 November, when there was an increase to 93 mg m −2 , followed by a decline and little variation until the end of December. Average SML irradiance (Ē SML ) calculated from the CCS mooring data declined rapidly from 6 E·m −2 ·day −1 to <4 E·m −-2 ·dat −1 from 24 October ( Figure 1c). Ē SML then declined slowly until mid-December, with values <3 E·m −2 ·day −1 from 24 October, and <2 E·m −2 ·day −1 from 25 November.
Ship measurements showed similar trends to the mooring data (Table 1; Figures 1b and 1c). Deeper SML occurred at CS2 than CCS, with SML temperatures warmer (~0.5°C) at CS2 than CCS. Chl integrals from the cruise (Table 1) were similar in magnitude to the mooring data, due to similarity in SML and euphotic zone depths . Chl ranged from 37-71 mg m −2 , with lower integrals at CS2 than CCS, and an increase of 10-30 mg m −2 between early and late November (Table 1). Incidental irradiance varied from 7.7-12.1 E·m −2 ·day −1 over November (average 9.6 E m −2 day −1 ; Table 1). Little difference between CS2 and CCS was seen in terms of Ē SML , with an average of 1.9 E·m −2 ·day −1 (20% of E 0 ; Table 1), broadly similar to the mooring data ( Figure 1b).
Vertical ρ P profiles were uniform across the SML and for most stations were <20 nmol Si·L −1 ·day −1 (Figure 2d). Rates of ρ P were higher at CS2 on 18 November (range 25-50 nmol Si·L −1 ·day −1 ) in association with high bSiO 2 concentrations (Figures 2b and 2d). Euphotic zone integrated ρ P ranged from 0.28-2.14 mmol Si·m −2 ·day −1 , with an average of 0.78 mmol Si·m −2 ·day −1 and no differences between sites or time ( Table 1). The ratio of NPP to ρ P (mol C: mol Si) for integrals, ranged from 11-153 with an average of 69, 9 times higher than the average diatom cellular ratio of C:Si (7.6, Brzezinski, 1985), and implies a low diatom contribution to NPP Rates of ρ N were either slightly positive or strongly negative across the water column (Figure 2e), with ρ N ranging from 68 to −215 nmol·L −1 ·day −1 and an average of −29 nmol·L −1 ·day −1 . Negative ρ N implies net dissolution rather than net production (positive ρ N ), with inferred ρ D rates well above ρ P . Rates of ρ D normalized to total bSiO 2 ranged from −0.3 to 0.5 day -1 (Figure 2f), with an average of 0.2 day -1 (Table 1). Integrated ρ D rates were generally in excess of ρ P for most sampling sites (Table 1), apart from CS2 on 18 November. High V D reflects the strongly negative ρ N and high bSiO 2 concentrations observed at all sampling sites. Ratios of D:P was strongly negative for all sampling sites (apart from CS2 on 18 November; Table 1), indicating that Si production could be fully supported by recycled Si (OH) 4 .

Discussion
High NPP and Chl (Table 1 and   ), and (f) biomass-specific dissolution rate (V D , day −1 ). Diatom bSiO 2 was estimated by applying the average cell Si content from Brzezinski (1985) to the diatom cell counts. Different sampling dates and locations are given as symbols (see Table 1 for dates and locations). Error bars are from triplicate measurements for each sampling depth.
(where growth equals losses), Wihsgott et al. (2019) determined that the autumn bloom started on 6 October (see Figure 1b), terminating 50 days later (20 November) when the SML deepened below the critical depth.
Our observations characterize elevated carbon fixation and biomass during the late stages of this period.
Though the shelf edge (CS2) showed similar Chl and NPP to CCS, on 18 November it exhibited higher nutrient levels, bSiO 2 and Si-uptake, positive net bSiO 2 production, with a D:P ratio (0.8) indicative of "new" Si supporting 20% of bSiO 2 production. The Celtic Sea has an energetic internal wave field (Green et al., 2008;Vlasenko et al., 2014), with breaking waves at the shelf edge enhancing nutrient supplies . During November 2014, stratification supported an internal tide (Wihsgott, 2018), with warm water at the shelf break from internal tidal mixing redistributing heat vertically (see Figure 2 in Ruiz-Castillo et al., 2018). Our observations indicate that this autumn diapycnal nutrient transfer supports bSiO 2 production; however, the limited nature of our observations means that the remainder of this paper will only consider the Si dynamics at CCS.
However, consideration of only ρ P (gross production) fails to recognize the significant rates of bSiO 2 dissolution. Rates of bSiO 2 dissolution (Table 1) are similar to other studies, for example, in the Sargasso Sea (0.1-1.3 mmol Si m −2 day −1 , Krause et al., 2010) and Monterey Bay (0.6-6.5 mmol Si m −2 day −1 , Brzezinski et al., 2003). In these cases, however, bSiO 2 production was higher than bSiO 2 dissolution so that net production was positive Krause et al., 2010). At CCS in autumn, dissolution rates were higher than ρ P , so D:P ratios were >1 (Table 1), and ρ N (net production) was negative (Figure 2e). In effect, new bSiO 2 material was produced as fast as it was dissolved, although the two pools were not necessarily the same. The level of detrital bSiO 2 not associated with diatoms cells was estimated to be high at CCS (>90%), and rapid dissolution of dead diatoms has also been observed in the Bay of Brest and Southern Ocean (Beucher, Treguer, Corvaisier, Hapette, & Elskens, 2004;Beucher, Treguer, Corvaisier, Hapette, Pichon, & Metzl, 2004).
Ratios of D:P > 1, indicative of dissolution rates exceeding production rates, are thought to be possible only over short timescales or in limited localities (Tréguer & De La Rocha, 2013). High ratios have been observed in the Bay of Brest and Southern Ocean, and like the Celtic Sea, these are linked to high proportions of detrital bSiO 2 from grazing and cell death following blooms (Beucher, Treguer, Corvaisier, Hapette, & Elskens, 2004;Beucher, Treguer, Corvaisier, Hapette, Pichon, & Metzl, 2004). Applying the average biomass-specific dissolution rate (V D~0 .2 day −1 ; Table 1) implies an exponential decline in bSiO 2 that would take 22 days to reduce bSiO 2 levels to <1% of initial levels. If these biomass-specific dissolution rates continued during winter, this would support the upper-ocean dissolution of the autumn bSiO 2 pool prior to the spring bloom.
The source of the autumn bSiO 2 detrital pool also needs to be considered. Due to limited coastal influence at CCS (see Ruiz-Castillo et al., 2018) and an absence of estuarine diatoms, a terrestrial source for the material is unlikely. The depth (145 m) and persistent thermocline would also limit any potential sedimentary bSiO 2 .
In fact, the bSiO 2 pool represents only 3% of the estimated total water column inventory of Si (406 mmol Si m −2 , based on a 145-m water column and winter Si (OH) 4~2 .8 μM, see Hydes et al., 2001), or 20% of the SML inventory of Si (55 mmol Si m −2 ; Table 1). The total autumn bSiO 2 pool is also equivalent to only 10% of estimated Celtic Sea spring bloom bSiO 2 production (115 mmol Si m −2 ), based on the Si (OH) 4 decline from winter (2.8 μM) to summer (0.5 μM; see Hydes et al., 2001). Taking the average diatom cell Si content from Brzezinski (1985) (8.14 pmol Si cell −1 ), the total autumn bSiO 2 pool is roughly equivalent to only 30 cells ml −1 . Though the detrital bSiO 2 is important for autumn Si cycling, contextually it is a small proportion of the total Si pool (re)cycled annually.
Diatoms are common features of shelf sea SCM (Barnett et al., 2019;Hickman et al., 2009). The summertime SCM at CCS contained 2.1 mg Chl m −3 , whilE the SML had 0.3 mg Chl m −3 (Wihsgott et al., 2019). When stratification broke down in early October (Figure 1b), nutrients and Chl were redistributed across the SML, with SML Chl increasing to 2.2 mg Chl m −3 , indicating that resupply of replenished nutrients by SML deepening was the driver of the autumn bloom (Wihsgott et al., 2019). Diatoms and bSiO 2 from the SCM may have been redistributed into the SML in early October, with (net) bSiO 2 production of only 0.3 mmol Si·m −2 ·day −1 (30% of measured production rates) needed to acquire the November bSiO 2 pool. Hence, November bSiO 2 is actually not high and low contribution to the total Si inventories, high detrital composition, and rapid dissolution rates imply that during the late autumn bloom diatom Si dynamics are limited by factors other than nutrient availability.
Light-limitation of the Celtic Sea autumn bloom has been described in terms of low growth rates , and interactions between the SML and critical depth (Wihsgott et al., 2019). In the context of Si dynamics, vertical ρ N profiles show little pattern, with both positive and negative ρ N relative to depth (Figure 2e). If ρ N are averaged for each incubation light-dose (Figure 3), then a vertical pattern is observed, with less negative ρ N at higher irradiance levels (>4.8 E·m −2 ·day −1 ) and negative ones at lower levels (<2.3 E·m −2 ·day −1 ). This implies positive (net) bSiO 2 production at irradiance levels >4 E·m −2 ·day −1 , higher than the compensation irradiances of Wihsgott et al. (2019) for net growth (1.2 and 3.0 E·m −2 ·day −1 ), or the Ē SML for late October to mid-December ( Figure 1c and Table 1). This indicates that positive net bSiO 2 production was light-limited during the latter stages of the autumn bloom and dissolution-dominated water column Si dynamics; while our incubations encouraged positive (net) production by alleviating light limitation.
Fundamental different mechanisms between blooms in spring (increasing stratification and irradiance) and autumn (declining stratification and irradiance; Figure 1b) may lead to differences in Si cycling. Spring diatom growth is encouraged by alleviation from light limitation and terminated by Si limitation as Si (OH) 4 is converted to bSiO 2 and exported, so (net) bSiO 2 production and export dominate Si-cycling. In contrast, our autumn observations imply that a large fraction of detrital bSiO 2 , combined with suboptimal light conditions, limit net bSiO 2 production, with bSiO 2 dissolving, so that (net) bSiO 2 dissolution dominates.
In an ecosystem with a large bSiO 2 pool, the fraction of this pool that is living or detrital will determine whether continued production or dissolution dominates Si cycling. This balance between production or loss may also occur during the early stages of the spring bloom, as weak stratification establishes itself, but is eroded via mixing events until it overcomes these conditions (e.g., Daniels et al., 2015;Taylor & Ferrari, 2011). Attention should focus on determining the bSiO 2 pool composition and its growth dynamics in terms of living and detrital components to better understand surface ocean Si cycling.