Sampling expeditions were executed in the NSCS onboard the R/V “Shi Yan 3” during March and April 2014. In total, 40 surface sediment samples (0 – 2 cm) were collected using a stainless-steel box sampler. The sampling locations are indicated in Figure 1A . The collected sediments were sealed in polyethylene Ziploc bags, and the air removed at the same time. Then, they were immediately frozen at - 20 °C until further processing in the laboratory. Note that our sampling sites were mainly located in the northern continental shelf area of the South China Sea, and a few stations (i.e., 7, 8, 21, 34, and 88) were located in the NSCS shelf-break area ( Figure 1A ). A conductivity temperature depth (CTD) system was used to measure the bottom water temperature, salinity, and depth.

Apparent solubility and dissolution kinetics of bSiO₂ of eight surface sediments with different sediment components ( Figure 1A ; Table 1 ) in the NSCS were measured using stirred flow-through reactors ( Figure 2 ) (Van Cappellen and Qiu, 1997a; Gallinari et al., 2002; Rickert et al., 2002; Gallinari et al., 2008). Each reactor consisted of a cell with a suspended Teflon-coated magnetic stirring bar at the bottom, with a volume of 50 cm³. Both cell ends were closed by nylon filters with a pore size of 0.45 μm, and the filters were supported by PTFE grids. The filters and grids were held in place by screw-on caps. The flowing solution could be pumped in or out of the headspaces without a barrier through narrow channels of the caps. The well-closed flow-through reactors sat on top of a multipoint magnetic stirrer, which were put into an accurately temperature-controlled incubator. All parts of the system were interconnected with PTFE pipes. In the experiments, up to 2 g of homogenized sediment sample with no pretreatment were added to the reactor (25°C) and an input solution of a well-characterized composition (0.7 M NaCl, buffered at pH 8.0 ± 0.1 using bicarbonate, with known different dSi concentrations) was pumped through the reactor at a constant flow rate until the dSi concentration in the reactor outflow stabilized. Because of the relatively low contents of bSiO₂ in the NSCS, the flow rates were maintained between 0.8 - 3.1 ml/h to ensure a measurable difference of dSi concentration between the inflow and outflow due to dissolution or precipitation. Lower flow rates were under higher degrees of undersaturation. The dSi concentration in the input solution was progressively decreased to produce a transition from a precipitation to a dissolution regime. In general, a group of experiments lasted 10 - 18 days according to flow rate and time to reach the steady-state. The dSi concentration and pH in the input solution were monitored throughout the whole course to ensure no influential fluctuations. All of the experiments were carried out under ambient atmospheric pressure and sustained for about 80 days. It is worth noting that the homogeneous process perhaps enhances the reactivity of bSiO₂ by crushing the silicate minerals in the flow-through experiment like in wet alkaline digestion (DeMaster et al., 1983; Ward et al., 2021). Van Cappellen and Qiu (1997a) were the first to introduce flow-through experiments into the study of bSiO₂ dissolution in the Southern Ocean, however, the numerously subsequent studies followed the homogeneous process and not well quantified this influencing factor yet (Gallinari et al., 2002; Loucaides et al., 2008; Wu et al., 2017). Regardless, the sediment homogeneous process remains a commonly used method of sample preparation and we followed this procedure.

Sediment samples were freeze-dried in the laboratory. It is noteworthy that the grinding effect on the measurement of bSiO₂ by wet alkaline digestion could enhance the calculated bSiO₂ concentrations up to ~50% (DeMaster et al., 1983; Ward et al., 2021). Therefore, to improve the measurement reproducibility (< 2%) in low bSiO₂ content samples from marginal seas, the sediments were as gently as possible ground using an agate pestle and mortar (trying to avoid crushing the silicious structures and silicate minerals) for homogenization. The sedimentary bSiO₂ contents were analyzed using a modified alkaline leaching method (Liu et al., 2002) which combined the methods described in DeMaster (1981) and Mortlock and Froelich (1989). In detail, ~ 100 mg sediment samples were placed in 50 ml polypropylene centrifuge tubes after removing organics and carbonates with H₂O₂ and HCl. Exactly 40 ml of 2% Na₂CO₃ solution was added to the samples. The tubes were covered tightly and mixed homogenously using a vortex rotating machine, then incubated in a water bath preheated to 85 °C. After 1 h, the tubes were removed and centrifuged. Then, 125 μl of clear centrifugation supernatant was pipetted from each extraction solution for dSi analysis. This procedure was repeated for 1 – 8 h. Note that after each sampling, the tubes were stirred vigorously to resuspend the sediment then placed again in the water bath. All sampling steps were completed quickly to minimize dSi loss on solid surfaces. The concentrations of dSi in the leaching solution were determined using the molybdate blue spectrophotometric method (Mortlock and Froelich, 1989). The bSiO₂ content was calculated by a first order kinetic model, that is, the dSi we measured was a function of time t (Koning et al., 1997; Rickert et al., 2002):

where [bSiO₂%]₀ represents the content of bSiO₂ (wt.%) in the sediment sample; [dSi%] _t donates extracted content of SiO₂ at time t; k Na 2 CO 3 . d b Na 2 CO 3 . e apparent rate constants in Na₂CO₃ solution that summarize the influence of biogenic and detrital matrix on the dSi concentration, respectively (Koning et al., 1997; Rickert et al., 2002). The coefficient of variation (i.e., relative stand deviation) for five parallel extractions was< 2.00%, indicating good reproducibility. The result of inter-laboratory comparison through this method demonstrates good accuracy of bSiO₂ measurement (Liu et al., 2002; Wu et al., 2015).

The OC and TN contents were measured using a CHNOS Elemental Analyzer (Vario EL-III; Elementar Analysensysteme GmbH, Germany) after inorganic carbon was removed via acidification with 1 M HCl. The analytical precision of OC and TN was< 6% based on duplicate measurements (Liu et al., 2010). The data of OC contents have been published in Yang et al. (2018). The content of CaCO₃ was measured using acid-base titration (Müller, 1966), and the coefficient of variation was 1.0% for five parallel extractions. To calculate the detrital material contents, the OC was converted to organic matter contents assuming a multiplier of 2.8 g of organic matter per g of carbon (Redfield composition) (Sayles et al., 2001). Therefore, detritus % = 100 – (bSiO₂% + CaCO₃% + OC% × 2.8).

The specific surface area (SSA, m² g^-1) of the sediments were determined by the conventional N₂-BET method using an ASAP2460 full-automatic surface area analyzer (ASAP2460, Micromeritics, USA). Samples were heated overnight at 90 °C and for one hour at 150 °C prior to analysis. The multi-point BET surface area was calculated from the N₂ absorption isotherm. The sediment grain size was analyzed using a laser particle size analyzer (Mastersizer 2000; Malvern Instruments Ltd., Malvern, Worcestershire, UK) capable of analyzing grain sizes from 0.02 to 2000 μm. Three sediment size categories were distinguished: (1) clay: D< 4 μm; (2) silt: 4< D< 63 μm; and (3) sand: D > 63 μm. The samples were analyzed in duplicate, and the analytical precision was< 2%. The mean grain size was reported as Φ, where Φ = -log₂ D (D unit: mm) (Krumbein, 1934). The data of grain size have been published in Yang et al. (2018).

The contents of bSiO₂ in surface sediments of the NSCS varied from 0.64% to 2.06%, with an average of 1.04% ± 0.35% ( Figure 3A ). As a whole, the bSiO₂ contents presented a strong zonal distribution tendency and varied with the water depth isobath, which decreased from the inner shelf (< 50 m) to the outer shelf (50 - 200 m), and then increased to the upper slope (> 200 m). Furthermore, higher levels were predominantly observed in the region with fine-grained (D< 63 μm) sediments and lower bSiO₂ concentrated in the sandy (D > 63 μm) sediments. The horizontal distribution of the OC contents showed similar patterns with that of bSiO₂, which were 0.1% to 1.03%, with an average of 0.49% ± 0.27% ( Figure 3B ). The bSiO₂/OC mole ratios varied from 0.24 to 1.06, with lower values at the inner shelf (0.45 ± 0.28) and upper slope (0.42 ± 0.11) and higher values at the outer shelf (0.51 ± 0.14). The CaCO₃ and detrital contents were 16.6% ± 12.4% (3.4% - 54.8%) and 80.9% ± 12.4% (42.7% - 92.8%), respectively. The distribution trends of these two components were opposite, increasing gradually from the inner shelf to the upper slope for CaCO₃ and decreasing gradually for detritus ( Figures 3C, D ).

The dissolution rates of sedimentary bSiO₂ were plotted as a function of the outflow steady-state dSi concentrations ( Figure 4 ). The uncertainties of the solubilities were calculated by the narrow fluctuation of [Si_out] after steady state conditions were reached. The apparent solubility of bSiO₂ in surface sediments of the NSCS ranged from 227 to 519 μM with a mean value of 298 ± 94 µM, which is much lower than the solubility of the fresh diatom frustules (~ 1100, 0 – 4 °C) (Hurd, 1973). In general, the solubility of the inner shelf and upper slope are higher than that of the outer shelf (except station 10). The highest solubility value was observed at station seven on the upper slope, where the bSiO₂ content was highest. The lowest value was recorded at station two on the outer shelf. The dissolution rate constant of the inner shelf and upper slope was higher than that of the outer shelf, which was from 0.67 to 1.53 yr^-1, with an average of 1.12 ± 0.3 yr^-1 ( Table 2 ). The east side (station 13) of the Pearl River estuary is evidently higher than the west side (station 79). Station one and 79 recorded the highest and lowest values, respectively. In addition, the reaction order (m >1) ranged from 1.1 to 3.42. A relative higher value was observed at station seven.

The surface sediments in the NSCS were mainly composed of silt (4 - 63 μm) and sand (> 63 μm). Fine-grained sediments were found primarily in the Pearl River estuary and adjacent coastal region, and secondarily in the upper-slope area (station 7, 8, 21, and 34), whereas coarser sediments occurred mainly in the outer-shelf area (50 – 200 m), especially in the northeast ( Figure 5 ). The highest values of specific surface area (SSA) were recorded at station one and seven, and the lowest values at station two, four, and five ( Table 1 ).

The distribution of bSiO₂ ( Figure 3A ) and bSiO₂/OC ratio ( Figure 6 ) revealed a mismatch between the (diatom biomass)/(total phytoplankton biomass) ratio and the sedimentary bSiO₂/OC ratio at the inner shelf and outer shelf, and also a mismatch between the bSiO₂ primary productivity and the bSiO₂ sediment records at the upper slope of the NSCS. The sedimentary bSiO₂ record is believed to a good indicator for the surface bSiO₂ primary production (Schrader and Sorknes, 1991). The wide shelf of the NSCS receives a large discharge of terrigenous nutrients and results in overall enhanced biological productivity (Han et al., 2012). But the primary production presented distinct regional characteristics at the inner shelf (< 50 m), outer shelf (50 – 200 m), and upper slope (> 200 m). The satellite remote-sensed spatial distribution of Chl-a showed a gradient decrease from nearshore to offshore, and changes roughly along the 50 m isobath with concentrations of ~ 1 mg m^-3 in all seasons. In addition, the total biomass of phytoplankton and the proportion of diatoms in the mixed layer also showed a gradient decrease from the nearshore to the offshore throughout the year (Xiao et al., 2018), which is consistent with the variation of surface primary productivity (< 40 m: 1.73 ± 0.12 g-C m^-2 d^-1; 40 – 120 m: 0.67 ± 0.35 g-C m^-2 d^-1; > 120 m: 0.43 ± 0.02 g-C m^-2 d^-1) (Pan et al., 2015). This evidence indicated that even though there is no report of primary production of bSiO₂ in the NSCS at present, it can be speculated that the distribution of bSiO₂ primary production should be in a decreasing trend from nearshore to offshore. Therefore, the high sedimentary contents of bSiO₂ accumulated in the inner-shelf region (1.27% ± 0.39%) ( Figure 3A ) is mainly attributed to the enhanced bSiO₂ primary production by the large discharge of terrigenous nutrients, especially in the area affected by the perennial westward plume of the Pearl River (station 13). However, a mismatch between the bSiO₂ primary productivity and the bSiO₂ sediment records at the upper slope of the NSCS was observed. High contents of bSiO₂ (1.43% ± 0.42%) were recorded in the upper-slope region (0.71% ± 0.24%) ( Figure 3A ). It is obvious this was not related to the euphotic bSiO₂ primary production. Our flow-through experimental results showed a dissolution kinetics-controlled preservation of bSiO₂ in this area (discussed in section 4.4).

In addition to primary production, sediment grain size plays a role in the distribution of bSiO₂. As Figure 5 shows, the distribution of mean grain size (Φ) was similar with the distribution of bSiO₂ (%), and fine-grained sediments were mainly distributed in the inner-shelf and the upper-slope regions. The strong correlation between “Φ” values of mean grain size and bSiO₂ contents (r = 0.71, p< 0.01) indicated fine-textured particles (silt and clay) have greater physical protection for bSiO₂. Compared with different biogenic material (OC and TN) in different sediment types, we found a different preservation patterns of bSiO₂ in the sediments of the NSCS. As a part of organic matter, a great linear regression of TN versus OC in sediments with an intercept of zero demonstrates the two components are preserved in a well coupled manner, whether in finer or coarser sediments ( Figure 7A ). However, the positive linear relationship between OC (%) and bSiO₂ (%) has a higher slope (3.0) in fine-grained (Φ > ~ 5.5) sediments than in coarser (Φ< ~ 5.5) sediments (0.78) ( Figure 7B ), which suggests not only that fine-grained particles have better preservation performance for biogenic material, but also with each equal additional amount of OC stored, the fine-grained sediments will retain 2.8 times more bSiO₂ than coarser sediments in the NSCS. Furthermore, the average molar ratio of bSiO₂/OC in finer sediments (0.34 ± 0.07) (mainly at the inner shelf and upper slope) is lower than in coarser sediments (0.57 ± 0.17) (mainly at the outer shelf) ( Figure 6 ). Therefore, there is a mismatch between the (diatom biomass)/(total phytoplankton biomass) ratio and the sedimentary bSiO₂/OC ratio in the inner shelf and outer shelf. The outer-shelf region with coarser sediments has been described as a sandy swath, and the sandy swath is the product of glacial sea level low stand and exposition to winnowing by strong bottom current recently (Wang and Li, 2009). The higher SiO₂/OC ratio, relative to the inner shelf and upper slope, could potentially be due to the higher energy level of this area. High-energy conditions could lead to more resuspension of sediments than low-energy environments and prompt longer oxygen exposure time, which in turn leads to lower OC contents through selective degradation of labile young fractions and reduced TOC content (Van der Voort et al., 2018). Compared with the silicic acid degree of unsaturation, the demand of oxygen is much smaller for the dissolution of bSiO₂. The mismatch at the inner shelf is discussed in section 4.3.

The solubilities of bSiO₂ in the NSCS sediments (298 ± 94 µM) is seriously inhibited, which exhibits a low level compared with some marginal seas and open oceans ( Figure 8 ). It has been proven that aluminum is probably the most significant factor affecting spatial variations in measured bSiO₂ solubility (Dixit et al., 2001). Field data and laboratory experiments reveal that the structural incorporation of Al into the silica framework by substituting Si atoms in the silica lattice reduces the solubility whether during the living diatom biomineralization process, settling after death, or deposited on the seafloor (Van Bennekom et al., 1991; Van Beueskom et al., 1997; Gehlen et al., 2002; Van Cappellen et al., 2002; Koning et al., 2007). The Pearl River delivers a high concentration of Al (360 – 690 nmol L^-1) to the NSCS and observes a shelf–to-slope transportation pattern (Zhang et al., 2020). Due to this severe terrigenous effect, the dissolved Al concentrations of seawater in the NSCS was reported to be at a high level (8 – 59 nmol L^-1, with an average of 30 ± 11 nmol L^-1) with a seaward decline (Guo, 2016), which is higher than some marginal seas, such as the Arabian Sea (3 – 15 nmol L^-1) and the Weddell-Scotia Seas (1 – 3 nmol L^-1) (Van Bennekom et al., 1991; Schüßler et al., 2005), and is much higher than some open oceans, such as the Pacific Ocean (0.3 – 5 nmol L^-1), the North Atlantic (6 – 25 nmol L^-1), and the Southern Ocean (0.33 – 0.8 nmol L^-1) (Orians and Bruland, 1986; Kramer et al., 2004; Middag et al., 2011). The scavenging proportion of the dissolved Al by phytoplankton was reported as up to 9.4 – 25.7% in summer (Zhang et al., 2020). Culture diatoms in Al-enriched solution exhibit a 20% lower solubility than in Al-depleted solution at 3 °C (Gallinari et al., 2002). The “primary uptake” of Al in the water column probably explains part of the reason why the solubility of the NSCS sediment is lower than some other sea areas. More important is the “Al – detrital – bSiO₂” interactions during early diagenesis of bSiO₂ in sediments. On one hand, the concentration of Al in pore water generally is several or tens of times (100 – 500 nmol L^-1) higher than seawater (Koning et al., 2007), which drives the “secondary uptake” of bSiO₂ debris. The well correlation between the dissolved Al concentration in pore water with the detrital/opal ratio in sediments confirms the detrital materials as an important source of Al (Dixit et al., 2001). The increase in Al levels often happens at the sediment-water interface (Van Cappellen and Qiu, 1997b). The high detrital matter content (80.9% ± 12.4%) in the NSCS sediments, which is higher than the Arabian Sea (30% – 50%), the Peru-Basin (66% ± 1%), and the Scotia Sea (36% - 66%) (Rickert, 2000), provides an abundance of Al to restrain the bSiO₂ dissolution and solubility. On the other hand, as the bSiO₂ deposited at the sediment-water interface, the continuously releasing Al (also including other ions, i.e., Mg²⁺, K⁺, and F^-) from detrital-rich sediments can induce reprecipitation with dSi and bSiO₂ to form authigenic silicates or adsorb onto the surfaces of bSiO₂ to further reduce the solubility (Van Cappellen and Qiu, 1997b; Dixit et al., 2001; Michalopoulos and Aller, 2004; Loucaides et al., 2010), which is called reverse weathering. It has been proven a crucial pathway that can remove large amounts of silica from ocean within a short time (months to years) (Michalopoulos and Aller, 2004; Rahman et al., 2017; Tréguer et al., 2021). Rickert et al. (2002) used sediment samples pretreated with 10% HCl and H₂O₂ in the Scotia Sea and found that the solubility increased by 104% and the dissolution rate constant increased by 590% compared to unpretreated samples. The specific surface areas also increased by a factor of two to three times due to the pretreatment procedure (removing the organic coating and partially authigenic minerals). These results were attributed to the detrital minerals or authigenic aluminosilicates forming on the bSiO₂ surface, some of which could readily dissolve in mild HCl solution. Loucaides et al. (2010) reinforced the view by long-term incubation experiments (fresh diatom frustules were separated from the terrigenous sediments by a dialysis membrane, only allowing the exchange of dissolved species), and demonstrated the interaction between bSiO₂, seawater, and lithogenic minerals dramatically inhibits the solubility of bSiO₂ by forming new mineral precipitates on bSiO₂ surface. Our results showed that a strong correlation between solubility and detrital or detrital to bSiO₂ ratio ( Figure 9 ). Gallinari et al. (2002) used different diatom species mixed with variable amounts of silicate minerals in a batch experiment, which presented a similar correlation between solubility and detrital to bSiO₂ ratios. This evidence supported our conclusion that the “Al – detrital – bSiO₂” interactions play important retarding roles in solubility space variation of bSiO₂ in the sediments of the NSCS.

Variations of specific surface area (SSA) must be considered when comparing the dissolution properties of sedimentary bSiO₂ and siliceous plankton in marine surface waters (Van Cappellen et al., 2002). It is generally believed that the SSA can provide a reliable proxy to represent the availability of solid surface area for solution reaction (Hurd, 1973). The dissolution rate constant of freshly exposed diatom frustules measured in labs can be as high as ~16 yr^-1, and the SSA of diatom frustules varied from 16 to 260 m² g^-1 (Van Cappellen et al., 2002), which are much higher than that in the NSCS sediments (~ 1.12 yr^-1 and ~ 7.1 m² g^-1) ( Table 1 ). In this regard, the lower rate constants and SSA values in sediments reflects that the preferential dissolution of the part with higher SSA in heterogeneous siliceous components, or the highly reactive part of bSiO₂ was regenerated during the settling process in the water column.

The reaction rate constants measured under the same conditions should reflect the relative differences of reactive surface areas (or reactive surface sites) (Rickert et al., 2002). However, it seems that sediment bulk SSA can be used as a proxy for the reactivity of bSiO₂ in the NSCS at some stations, such as station one and seven, which exhibit the highest dissolution rate constants corresponding to the highest SSA values. An abnormal phenomenon occurred at station four, which presented a relatively higher dissolution rate constant but very low SSA value. Our data overall presented a weak correlation between the dissolution rate constants and the sediment SSA, particle sizes, or bSiO₂ loadings ( Tables 1 , 2 ). Rickert et al. (2002) reported that the correlation of reaction rate constants and sediment SSA were only found in almost pure bSiO₂ samples (cultured phytoplankton or acid-cleaned samples from the water column), excluding pure siliceous oozes (bSiO₂ contents > 49%), whereas the adsorption capacity of the bSiO₂ for cobalt ions well corresponded to the reactivity changes (Van Cappellen and Qiu, 1997b). The BET surface area of siliceous ooze cores in the Southern Ocean basically remains unchanged with increasing depth, whereas the reaction rate constants still decrease by three times when the depth reaches 30 cm (Van Cappellen, 1996). However, elemental and microscopic analyses showed no changes in mineral composition or any diagenetic alterations throughout the sediment cores. This evidence implies that the reactivity of the bSiO₂ surface has altered but not the geometric surface area determined by nitrogen BET, which has been called a progressive “aging” process of the bSiO₂ surface (loss of reactive surface sites) (Van Cappellen, 1996). This “aging” process may be related to surface charge density (Dixit and Van Cappellen, 2002; Loucaides et al., 2010). It is worth noting that all of the FTIR (Fourier-transform infrared) spectroscopic evidence (Schmidt et al., 2001), the apparent reactivity rate constants obtained from leaching alkaline solution (Koning et al., 1997), and from flow-through experiment (Rickert et al., 2002) manifest that the loss of reactive surface sites has already begun as the bSiO₂ settles down in the water column. The sediment flux of the Pearl River is transported to the NSCS in the form of a plume, which presents distinct seasonal variations influenced by the East Asian monsoon: during summer, under the effect of weak southwest winds (~ 6 m s^-1), the water mass at the inner-shelf moves eastward, but the suspended particles of the Pearl River plume are still predominantly transported to the west due to the prevalent coastal current ( Figure 1B ); during winter, the coastal current is strengthened by the strong northeasterly winds (~ 9 m s^-1), which can resuspend sediments and keep them transporting westward ( Figure 1C ) (Liu et al., 2014). The dissolution rate constants of bSiO₂ on the east side of the Pearl River estuary (0.67 yr^-1, station 79) are distinctly lower than those on the west side (1.22 yr^-1, station 13). Furthermore, the sedimentation rate on the west (~ 1cm yr^-1) side of the Pearl River estuary is much higher than on the east side (~ 0.5 cm yr^-1). We suggest that resuspension-deposition and higher sedimentation rate maintain the fresh bSiO₂ in the surface sediment on the west side of the Pearl River estuary and result in the high reactive of bSiO₂ in the surface sediments. It is completely different for the degradation of OC. Wei et al. (2020) reported older terrigenous OC on the west (1950 yr BP) side of the Pearl River estuary than on the east (1510 yr BP) due to resuspension-deposition. Repeated resuspension-deposition cycles could promote oxygen exposure time and accelerate the remineralization of fresh OC, and leave the refractory older OC, whereas redox oscillations have little impact on the dissolution of bSiO₂. This is probably also the reason why the highest (diatom biomass)/(total phytoplankton biomass) ratio in surface water but the lowest bSiO₂/OC ratio was recorded in the sediments of the west inner-shelf region ( Figure 6 ). Besides the “aging” process, the vertical transport of bSiO₂ through the fast settling of aggregates or fecal pellets also play a role (Shi et al., 2019), trying to maintain the initial reactivity of bSiO₂ of the surface ocean, such as at station one with lower water depth and higher reactivity. In this case, the most important parameter could be the degree of coupling between surface waters and the seabed. For example, the formation of blooms and the rapid sedimentation of diatoms, either in the form of large aggregates or embedded within fecal pellets of large euphausiids (Nelson et al., 1995), could lead to the deposition of fresh diatoms with dissolution properties similar to those encountered in surface waters. Our sediment samples are located in a wide range of water depths (40 – 773 m), and subjected to intense turbulence by active water dynamics or sediment resuspension due to the buoyancy of river plumes, wind force, or terrigenous impact (Cao et al., 2020). The partly unsatisfactory agent of SSA for the reactivity surface of our samples presumably reflected the different “aging” or other water column processes (i.e., forming diatom aggregation or fecal pellets) of bSiO₂ particles when sinking in the water column.

The dissolution of bSiO₂ is very sensitive to temperature changes, and preservation status largely depends on the difference of in situ temperature (Kamatani and Riley, 1979; Van Cappellen and Qiu, 1997b). To further discuss the in situ preservation mechanism of bSiO₂ in the NSCS sediments, the measured reaction rates were reconstructed under in situ conditions. We applied an average activation energy (E_app) of 60 kJ/mol to obtain the rate constants at in situ temperature using the equation (Rickert et al., 2002):

where R = 8.314 J K^-1 mol^-1 and T (K) donates the temperature, k refers to the rate constant (yr^-1). The pressure was not considered because it has little influence within water depths of 800 m (Loucaides et al., 2012). The dissolution rate constants of bSiO₂ in the NSCS sediments at in situ temperature were 0.94 ± 0.13 yr^-1, 0.61 ± 0.13 yr^-1, and 0.30 yr^-1 at the inner shelf, outer shelf, and upper slope, respectively ( Table 2 ). The upper depth (2 cm) interval corresponds to the zone of net silica dissolution. Then, the net dissolution rate of bSiO₂ at the eight stations of the 2 cm surface sediments were reconstruct by combining the experimental derived non-linear dissolution law (Eq. (2), Figure 10 ) with the in situ pore water dSi concentration (averaged of 0 - 2 cm), which were 0.20 ± 0.01 yr^-1, 0.18 ± 0.11 yr^-1, and 0.001 yr^-1 at the inner shelf, outer shelf, and upper slope, respectively ( Table 2 ). The reconstructed in situ dissolution rate in the upper slope (station seven) is orders lower than that at the shelf. This suggests that the differences of bSiO₂ preservation between the inner shelf, outer shelf, and upper slope were the results of a wide range of in situ temperature-induced differences of dissolution rates constants and the departure from equilibrium-induced differences of in situ dissolution rates. These results showed dissolution kinetics-controlled evidence that explains the mismatch between the (diatom biomass)/(total phytoplankton biomass) ratio and the sedimentary bSiO₂/OC ratio in the inner shelf and outer shelf, and also a mismatch between the bSiO₂ primary productivity and the bSiO₂ sediment records in the upper slope of the NSCS. The mismatch of bSiO₂/OC ratio at the outer shelf is due to high-energy conditions and the mismatch at the inner shelf is attributed to the resuspension-deposition, which have been discussed in section 4.1 and section 4.3. The high reconstructed in situ rate constants (0.94 ± 0.13 yr^-1) and dissolution rate (0.20 ± 0.01 yr^-1) strengthened the fact of the lower bSiO₂/OC ratio observed at the inner shelf. Wei et al. (2020) suggested that surface sediments at the slope displayed a relatively higher terrigenous OC content and older terrigenous OC than that at the outer shelf, likely due to cross-shelf transport from the Pearl River and sediment winnowing on the outer shelf. Our results suggest that the lower in situ reactivity (0.30 yr^-1) and dissolution rate (0.001 yr^-1) allowed sedimentary bSiO₂ to be well preserved in this area.

The degree of undersaturation of the pore fluid represents the thermodynamic driving force for the dissolution process and is directly dependent on the solubility of the bSiO₂ (Van Cappellen and Qiu, 1997a). The dissolution rates were plotted against the departure from saturation state ( Figure 10 ), which exhibits a non-linear dissolution kinetic, which was also observed in different deposition settings (Van Cappellen and Qiu, 1997a; Van Cappellen and Qiu, 1997b; Rickert et al., 2002; Ragueneau et al., 2009). The non-linear dissolution kinetics implied a higher dissolution rate in a highly unsaturated silicic solution, that is, more efficient recycling of bSiO₂ in dSi-depleted upper sediments compared with deep sediments, which is meaningful for benthic Si regeneration. However, the non-linear dissolution kinetics must be severely inhibited by the high contents of detritus in the NSCS sediments. The overall dissolution rate constants of bSiO₂ in the NSCS sediments was 1.12 ± 0.3 yr^-1 at 25°C, which is lower than some marginal seas, such as the West Antarctic Peninsula Shelf [2.5 yr^-1, Gallinari et al. (2008)], the Arabian Sea [1.48 ± 1.4 yr^-1, Rickert (2000)], the Scotia Sea [1.81 ± 1.4 yr^-1, Rickert (2000)], and the east China seas (containing the Yellow Sea and the East China Sea, 1.53 ± 1.39 yr^-1 [Wu et al. (2017); Wu and Liu (2020)]; it is also lower than some open oceans, such as the Southern Ocean [2.39 ± 0.2 yr^-1, Van Cappellen and Qiu (1997b)], the equatorial Pacific [1.59 yr^-1, Gallinari et al. (2002)], and especially the North Atlantic [7.39 yr^-1, Ragueneau et al. (2001)] ( Figure 8 ). The extremely high reaction rate constants of the bSiO₂ in sediments or the sediment trap of the North Atlantic Ocean are believed to be an important reason for the very low bSiO₂ contents (~1%) and poor preservation efficiency (12%) in the sediments compared with the Southern Ocean (Rickert, 2000; Ragueneau et al., 2001). Inversely, the relatively low reaction rate constants in the NSCS may be responsible for the well preservation efficiency of the bSiO₂ (27%) (Ma et al., 2022). Therefore, our results support the view that continental margin sediments play significant roles as silica sinks in the marine silica budget (DeMaster, 2002).