Benthic diatoms modify riverine silicon export to a marine zone in a hypertidal estuarine environment

Riverine dissolved silicon (DSi) and biogenic silica (BSi) are modulated along the estuarine gradient by several biotic and abiotic processes governed by physical forcings. An important area controlling silicon transport in alluvial estuaries with large intertidal mudflats is the benthic diatom-dominated biofilm system. Here, the hypertidal Severn Estuary, UK, has been used as a case study to improve our understanding of silicon transport in these benthic-dominated systems. We present the first time-series dataset of Si concentrations in the Severn. River and tidal hydrodynamics drove spatio-temporal changes in DSi. The longitudinal profile of DSi followed the classical view of dilution with downstream transport. Despite low riverine supply of BSi and low siliceous-phytoplankton production, relatively high BSi concentrations were measured in the Severn Estuary (maximum of 14.9 mg/L), which accounted for over 70% of the total bioavailable silicon present and were characterised by isotopically heavy waters (δ30Si of + 0.9 to + 1.1‰). Benthic biofilms (microphytobenthos) on the intertidal mudflats contained significant biomass (measured as chlorophyll a concentration with a maximum of 116.8 ± 16.2 µg/g dw. sed) with high productivity, driven by their photoprotective adaptions to these harsh intertidal environments, contributing to isotopically heavy mudflat water (δ30Si of + 1.19 to + 2.03‰), and resulting in high benthic BSi content in the spring (0.74 ± 0.03%) and summer (0.76 ± 0.05%). The fast-flowing tidal currents resulted in high bottom shear stress which likely exceeded the erosion thresholds of the biofilms, transporting the sediment-BSi matrix into the water column. Suspended particulate matter (SPM) and BSi remained tightly coupled in the estuarine water column (bioflocculation), and experienced the series of erosion–deposition events, burial/dissolution and export out of the estuary. Our novel observations improve understanding of the complex processes governing Si transport in hypertidal, benthic-dominated estuaries, and highlights the importance of tightly coupled benthic-pelagic systems in influencing the terrestrial silicon export to a marine zone.


Introduction
Diatoms have an obligate growth requirement for dissolved silicon (DSi). Diatoms uptake DSi and through biomineralization form their particulate silica frustules, hereafter referred to as biogenic silica (BSi). These siliceous organisms contribute up to 40% of global marine productivity, and through high BSi export rates to the deep ocean, are a critical component of the biological carbon pump (Tréguer et al. 2018(Tréguer et al. , 2021. DSi enters the hydrosphere and is transported to the ocean via rivers and groundwater (Frings et al. 2016), accounting for 70-85% of all DSi inputs into the ocean (Tréguer & De La Rocha 2013;Tréguer et al. 2021). Rivers are a key network in transporting BSi from the terrestrial environment to the marine zone (Tréguer et al. 2021). Studies into estuarine Si cycling have gained momentum with new insight into the role of the coastal ecosystem on biogeochemically modifying the riverine Si inputs to a marine zone (e.g., Pastuszak et al. 2008;Carbonnel et al. 2009Carbonnel et al. , 2013Leynaert et al. 2011;Li et al. 2019;Zhang et al. 2020;Sarath et al. 2022), including the importance of estuarine benthic-pelagic coupling (e.g., Raimonet et al. 2013;Welsby et al. 2016).
Estuarine benthic ecosystems are potential areas of high biogeochemical cycling activity and are among some of the most productive ecosystems globally. Estuarine biogeochemical processes interact on different spatio-temporal scales, induced by a wide range of physical forcings (Regnier et al. 2013). As a result, pelagic DSi and BSi fractions, hereafter PDSi and PBSi, are biogeochemically modified along the estuarine gradient due to the interaction of hydrological, geochemical, and biological processes over different spatio-temporal scales. However, these coastal modifications to the terrestrial Si inputs are likely to have consequences for the coastal and marine ecosystems, for global biogeochemical cycles, and ultimately, for the estimates of both global terrestrial and marine Si budgets (e.g., Tréguer et al. 2021). Current global Si budget estimates have considered estuarine Si cycling; however, they have primarily focused on two processes: the pelagic Si biological production and reverse weathering, neglecting the potential Si supply from the estuarine benthic ecosystem. By disregarding the potential estuarine benthic Si cycle, estimates of the Si budgets are accompanied by large uncertainties. Further, to date, estuaries and the coastal environment remain under-sampled (Brewin et al. 2016), leaving changes in the Si budgets to go unnoticed through the filter of land-ocean exchange.
Despite previous estuarine Si cycling studies, which are often centred on the Scheldt Estuary (e.g., Struyf et al. 2005aStruyf et al. , 2005bCarbonnel et al. 2009Carbonnel et al. , 2013 Bay of Brest (e.g., Ragueneau et al. 2005), Chesapeake Bay (e.g., Conley and Malone 1992) and Yangtze (Li et al. 2019;Zhang et al. 2020), knowledge of the potential Si modifications in an estuary which is dominated by a biologically productive benthic system, are less well-known. Substantial research has been carried out on the role that saltmarshes and wetlands play in cycling Si in the coastal environment (e.g., Struyf et al. 2005aStruyf et al. ,b, 2010Jacobs et al. 2008;Carey and Fulweiler 2014), but few have studied the complex benthic diatom-dominated biofilm ecosystem on intertidal mudflats. The role of the biofilms in biostabilizing the sediment has been extensively researched (e.g., Paterson 1989;Underwood and Paterson 1993;Sutherland et al. 1998;Stal and De Brouwer 2003;Perkins et al. 2004) but the contribution of Si to the coastal budget, is often overlooked. Here, we aim to draw attention to the importance of this neglected benthic system in modifying riverine DSi and BSi exports to a marine zone. In doing so, we also use Si isotope measurements as an exciting new approach to understand Si transport in these complex and tightly coupled benthic-pelagic systems. Only a few studies have applied Si isotopes to estuaries: the Scheldt Estuary (Delvaux et al. 2013), the Changjiang Estuary (Zhang et al. 2015), and those of the world's largest rivers: the Amazon, Yangtze and Pearl rivers (Zhang et al. 2020). Knowledge of the extent estuaries alter the isotopic Si signature prior to reaching the oceans, is therefore poorly constrained.
The Severn Estuary (UK) is a suitable case study to investigate the effect of benthic-pelagic coupling on the transport of DSi and BSi in an estuary. The Severn is heterogeneous because of complex hydrogeomorphological dynamics and has the second largest hyper-tidal range on Earth. This large tidal range results in the development of large intertidal mudflats, inhabited by diatom-dominated microphytobenthic (MPB) biofilms at the sediment-water interface (SWI) (Underwood 2010). The biofilms are subjected to high resuspension rates (Manning et al. 2010) associated with the extreme tidal dynamics (Neil and Couch 2011), creating a potential export of Si from the benthic system into the water column. Due to high turbidity in the water column (Manning et al. 2010), which limits phytoplankton growth (Underwood 2010), the photosynthetically productive diatom biofilms (Welsby et al. 2016) have a high potential to biologically mediate Si through the uptake of benthic DSi (BDSi) and through biosilicification, the production of benthic BSi (BBSi). With the potential strong benthic-pelagic coupling between the benthic ecosystem on the intertidal mudflats and the Severn's water column, the biogeochemical forcings, alongside the estuarine hydrodynamics, are expected to significantly influence the transport of Si.
We hypothesized that a proportion of the pelagic BSi may originate from the resuspended MPB biofilms inhabiting the intertidal mudflats of the Severn Estuary. Our first aim was to investigate the effects of multiple forcings that influenced the transport and transformations of PDSi and PBSi along the Severn Estuary. These forcings were investigated by analysing the: river discharge rates; salt and sediment transport (Suspended Particulate Matter, SPM); pelagic biomass (chl a); Total Organic Carbon (TOC) concentrations; and macronutrient (nitrate, nitrite, ammonium, and phosphate) concentrations; and Si input from external sources including the estuarine tributaries and DSi inputs from groundwater and rainwater. Our second aim was to determine the relative importance of the benthic system of the intertidal mudflats of the Severn Estuary as a source of Si to the water column. This was addressed by determining the BDSi and BBSi budgets of the MPB biofilms inhabiting the intertidal mudflats of the Severn Estuary, using stable silicon isotopes to investigate uptake mechanisms, and by exploring the multiple factors that could influence the uptake of BDSi and the subsequent build-up of BBSi in benthic regions over spatio-temporal scales.

Sampling methodology
Ecological surveys were carried out each season in 2016 in the River Severn, Severn Estuary and upper Bristol Channel (Fig. 1). The sampled periods coincided with the seasons, and for clarity, are referred to hereafter as; winter (January-March), spring (April-June), summer (July-September) and autumn (October-December). However, due to constraints, these sampled days, although span the entire season, do not necessarily reflect the temporal change associated with each season, which would require daily sampling. Distances along the river-estuary system were calculated using UK Ordnance Survey basemaps in ArcGIS. A transect was taken, aboard a research vessel, from the upper estuary at high tide, along the longitudinal axis, to the outer estuary. Water was collected from the surface (< 2 m deep) and at depths (> 15 m deep) using a 5 L Niskin bottle. A transect of four stations was taken along each of the main tributaries: River Wye, River Usk, and River Avon, along with a sampling station in Cardiff Bay (Rivers Taff and Ely). These tributaries discharge into the estuary between ca. 15 and 45 km from the lower estuarine boundary. Sampled intertidal mudflat sites, which coincided with spring cycles, ranged from the upper Severn Estuary at Severn Beach, Site 1 (51.57° N, 2.67° W), to the mid-estuary on the English coast at Portishead, Site 2 (51.49° N, 2.77° W), and on the Welsh coast at Newport Wetlands, Site 3 (51.53° N, 2.92° W), to the outer estuary at Sand Bay, Site 4 (51.38° N, 2.97° W).
Three replicates (3 × 1 L) of river and estuary water were sampled to measure PBSi and chlorophyll (chl) a concentrations from the filtered matter. Estuarine pelagic chlorophyll fluorescence was measured using the Water Pulse Amplitude Modulation (PAM) fluorometer (Walz, Effeltrich, Germany) on sub-samples of freshly collected water, using the cuvette system of the WATER-ED Emitter-Detector Unit. However, water column productivity was too low for any measurements (F < 150 rel. units). Acid cleaned (10% HCl) HDPE bottles were filled with 50 mL filtered water through PALL Acrodisc PF 32 mm syringe filters with 0.8/0.2 µm pore space supor membranes for analysis of PDSi, Total Organic Carbon (TOC), and macronutrient (nitrate, nitrite, ammonium, and phosphate) concentrations, along with Si isotopes. Three water sample replicates (3 × 1 L) were collected for suspended particulate matter (SPM) concentrations (g/L). River water pH was logged using pH-Pal-Pocket testing (ETI Instruments, U.K.) calibrated to a pH buffer of 4, 7 and 10. Riverine conductivity was measured using a conductivity probe calibrated using potassium chloride, and converted to salinity (PSU). Air and water temperatures were recorded using a hand-held thermometer.
Microphytobenthic biofilms at the SWI were sampled during daytime low tide periods, between ~ 11:00 h and 16:00 h. At each mudflat, 15 replicates were taken, along a ~ 100 m transect, at the mid-shore tide height, parallel to the surf zone, using a semi-random sampling scheme targeting areas of extensive biofilms, with ~ 5 m between each sample of biofilm. Three field replicates were taken at each station to capture the natural variability of the system on the metre scale. Sediment mini-cores of a diameter of 2.5 cm were extracted for the surficial 5 mm biofilm for chl a analysis and BBSi content. Variable chlorophyll fluorescence of undisturbed diatom-dominated biofilms was logged at each station as a proxy for photosynthesis (see Online Resource 1). A sample (50 mL) of the mixed mudflat water (a mixture of surface and pore fluids) was placed into acid (10% HCl) cleaned 50 mL centrifuge tubes, centrifuged at 4000 rpm for 10 min, with the supernatant decanted and filtered through PALL Acrodisc PF 32 mm syringe filters with 0.8/0.2 µm pore space supor membranes for the concentrations of BDSi, phosphate, nitrate, nitrite, ammonium and total organic carbon (TOC).
Sediment porosity (ϕ) (Eq. 1) was calculated at each station, as a function of the wet and dry sediment mass, and the wet bulk density ρ s (assumed to be 1.5 for unconsolidated sediment), and the density of water ρ w (assumed to be that of seawater, 1.02).
Rainwater was collected in the vicinity of the Severn Estuary, once a month (max. 12 h), for 4 months, in 2016 using a pre-cleaned 1L plastic bottle. Groundwater data from boreholes were obtained from the Environment Agency (EA), UK, a national organisation that collects environmental monitoring data. Sunshine hours were from the UK Met Office (Met Office, 2016), and, at each station, ambient irradiance (mmol photons/m 2 /s) was measured using the Mini Quantum Sensor attached to the Walz PAM fluorometer.
Tidal velocities were compared to the bottom shear stress to assess erosion capacity. Bottom shear stress (τ b in N/m 2 ) was determined by the quadratic friction law (Eq. 2): (1) = Mass of wet sediment (g) − mass of dry sediment (g) ∕ mass of wet sediment (g) × s ∕ w where, ρ is the density of pure water (1000 kg/m 3 ), g is the acceleration due to gravity (m/s 2 ), V is the tidal current velocity in the estuary (m/s), and C is the Chézy coefficient, with typical values of 40 m 1/2 /s for estuaries (Volta et al. 2014).

Laboratory analysis
Approximately 25% of the surficial biofilms were removed from the mudflat sediment mini-cores and dried in an oven at 85 °C for 24 h to determine the percentage loss of weight upon drying. Benthic and pelagic samples (~ 0.05 g) were measured for BSi using the weak alkaline extraction method adapted from DeMaster (1981). BSi concentrations were measured using the standard Heteropoly Blue method, modified from the original molybosilicate acid spectrophotographic method on a Hach Lange DR3900 spectrophotometer (Strickland and Parsons 1972). Blanks were processed alongside each sample. The internal error of the spectrophotometer was 2%.
Chl a content was used as a proxy for biofilm and water column photosynthetic biomass and was determined following Mackereth et al. (1978). Briefly, the chl a from the filters or the sediments were extracted for 24 h using 4 mL acetone buffered with MgCO 3 to prevent degradation of chl a to phaeophytin, with samples kept in the dark at 4 °C. The samples were then centrifuged at 2000 rpm for 15 min, and then analysed in a JENWAY 6305 spectrophotometer (absorbances at 630 nm, 647 nm, 664 nm, and 750 nm). Chl a was then calculated using the optical densities at the different wavelengths corrected for scattering at 750 nm (Eq. 3).
Aliquots of all filtered mudflat and river waters (200 µL) were diluted with Milli-Q H 2 O (9.8 mL) for DSi concentration analysis (mg/L) following the Heteropoly Blue method (Strickland and Parsons 1972). Sub-samples (2.0 mL) were measured for phosphate concentrations (PO 4 -P, mg/L), following the LCK 349 method. All phosphate concentrations were below 50 mg/L and did not interfere with the BSi measurements. Nitrate (NO 3 -N) and nitrite (NO 2 -N) concentrations (mg/L) were measured from sub-samples (2.0 mL) using the LCK method 339 and 341, respectively. Ammonium concentrations (NH 4 -N, mg/L) were measured from subsamples (5 mL) using the 304 LCK method. Due to interferences of chloride ions on these macronutrients, HDPE bottled estuarine water samples were immediately frozen and were analysed for PDSi and other macronutrient concentrations (mg/L) on a SEAL analytical AAIII segmented flow colorimetric auto-analyser at Plymouth Marine Laboratory following the methodology recommended by GO-SHIP (Hydes et al. 2010). TOC concentrations were measured using the Non-Purgeable Organic Carbon (NPOC) technique (Visco et al. 2005) at the School of Geographical Sciences, University of Bristol.
Filtered pelagic and benthic mudflat water from the summer and autumn sampled periods of 2016 were analysed for their Si isotope value (presented here as δ 30 Si, see Eq. 4). Saline water samples were prepared following the MAGnesium Induced Co-precipitation (MAGIC) method using sodium hydroxide (Karl and Tien 1992) with ) modifications, prior to column chromatography, using the cation exchange resin, adapted from George et al. (2006). The BioRad resin (DOWEX 50 W-X12, 200-400 mesh) was pre-cleaned, and rinsed with acid (3 M, 6 M and concentrated HCl) and MQ, before the sample and standards were passed through the columns. The determination of Si isotopes ( 28 Si, 29 Si, 30 Si) was performed on a Thermo Scientific Neptune Plus High Resolution multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS) at the Bristol Isotope Group laboratories. Raw measurements were blank corrected and corrected for instrument mass bias which affects fractionation, using a standard-sample bracketing protocol and Mg isotope correction (Cardinal et al. 2003). The standards and samples have 29 Si and 30 Si values that fall along the expected mass-dependent fractionation line, with 30 Si: 29 Si = 0.512 30 Si ± 0.04, r 2 = 0.997, n = 42, which fall between the mass-dependent relationship for kinetic and equilibrium fractionation (Georg et al. 2006). The long-term external reproducibility of the measurements was assessed using the sponge standard LMG08 (Hendry et al. 2011), with values measured at: δ 29 Si = − 1.71 ± 0.07‰ and δ 30 Si = − 3.47 ± 0.06‰, n = 17. These values compared well to reported mean values (Hendry et al. 2011) of: δ 29 Si = − 1.72 ± 0.08‰, and δ 30 Si = − 3.37 ± 0.17‰. Full replicate δ 30 Si measurements of river samples agreed within ± 0.08‰ or better.
where, x corresponds to 29 Si or 30 Si, SMP is the sample, and STD is the standard Quartz RM8546 (NBS-28).

Statistical analysis
Data was tested for normal distribution (p > 0.05) and equal variance (p > 0.05). Data that met these assumptions were tested using Two-Way ANOVA (spatial and temporal difference). However, when the Two-Way ANOVA showed significant interaction, data was tested using One-Way ANOVA, followed by a post hoc test. Data which failed to meet the assumptions of normal distribution were log transformed. Data that failed to meet the assumptions of both normality and equal variance were tested using the nonparametric Kruskal-Wallis test, followed by a post hoc Mann-Whitney pairwise test. To test the linear associations between Si and biological, nutrient, and environmental factors in the benthic and pelagic systems, Pearson's product momentum correlations were applied. A Principal Component Analysis (PCA) was applied to normalized data to assess the interaction between Si and biological, nutrient and environmental factors in the pelagic and benthic systems. The Environment Agency's (EA 2016) monitoring data has been used to validate the data presented here.
Relatively higher PBSi concentrations were measured in the estuary each season compared to the river and marine zone. Peak PBSi concentrations were observed in the upper estuary in the summer and autumn sampled periods (max. 14.9 mg/L and 10.0 mg/L, respectively). PBSi concentrations displayed significant variation between sampling periods (H = 20.35, d.f = 3, p < 0.001). A significant difference was found between the winter and the other sampled periods (post hoc analysis p < 0.05), but no difference was found between samples analysed in the spring, summer and autumn periods. No linear correlation was observed between PDSi and PBSi in the Severn. PDSi dominated 76% of the proportion of Si (PDSi + PBSi) in the River Severn, whereas in the estuary BSi dominated, especially in more typically biologically productive periods: spring (72%), summer (77%) and autumn (77%), compared to the winter (38%) ( Table 1).

Estuarine tributaries (River Wye, Usk and Avon)
PDSi concentrations in the tributaries (min. 0.2 mg/L, max. 5.7 mg/L) were comparable to the PDSi concentrations measured in the River Severn (min. 0.2 mg/L, max. 3.8 mg/L) (see Online Resource 2). Environment Agency (EA 2016) monitoring PDSi data for the Avon at Keynsham in 2016 were in a similar range between 1.2 and 5.6 mg/L. PBSi concentrations in the tributaries were higher with a wider range (min. 0.0 mg/L, max. 19.4 mg/L, avg. 1.9 ± 0.9 mg/L) compared to PBSi concentrations measured in the River Severn (min. 0.1 mg/L, max. 2.5 mg/L). However, the highest PBSi concentrations of 19.4 mg/L were recorded in the River Usk ~ 12 km upstream of the river mouth, with a visible tidal influence: high salinity levels of 10 PSU and high SPM concentrations of up to 2.6 g/L.

Rainwater and groundwater supply of DSi
DSi concentrations in the rainwater remained below 0.3 mg/L. Environment Agency groundwater DSi concentrations data from boreholes in the southwest UK in the catchment area for the Severn Estuary ( Fig. 1), ranged between 6.7 and 9.9 mg/L (Environment Agency 2017).

Biological mediation of silicon: pelagic productivity
Water column biomass concentrations (chl a in mg/L) were negligible in all sampled periods in the turbid Severn Estuary and Bristol Channel (Fig. 3) and remained below 0.001 mg/L in the river. Similarly, Environment Agency monitoring chl a data was also low in the Severn (avg. 0.003 mg/L), with relatively higher concentrations in the river compared to the estuary. Total Organic Carbon (TOC) concentrations in the River Severn were between the limit of detection and 4.5 mg/L but lacked significant correlation with PBSi in all seasons. Turbidity was high with low secchi disk depths in winter (0.18 m), spring (0.15 m), summer (0.2 m), and autumn (0.3 m). Significantly higher (H = 17.286, d.f = 3, p < 0.001, post hoc analysis p < 0.05) riverine SPM concentrations were measured during the winter (max. 2.1 g/L), compared to the spring (where concentration were always below the limit of detection), summer (max. 0.01 g/L) and autumn (max. 0.08 g/L). In the tidally dominated Severn Estuary, seasonal changes in river flow rates had less of an influence on the estuarine turbidity, with high SPM concentrations, predominantly a composition of mud/clay in the upper Severn Estuary (Fig. 3). Environment Agency (EA 2016) monitoring SPM data also displayed peak SPM concentrations in the upper Severn Estuary, and followed a similar seasonal pattern to our data: 1.1 g/L (winter), 3.2 g/L (spring), 1.5 g/L (summer), and 1.7 g/L (autumn) (Fig. 3). No clear correlation was observed between SPM and PBSi during the winter, but a positive linear correlation was observed in the more typical biologically productive periods: spring (n = 36, r 2 = 0.52, p < 0.01), summer (n = 36, r 2 = 0.84, p < 0.001) and autumn (n = 36, r 2 = 0.86, p < 0.001) (Fig. 3). Bottom water samples in the Severn Estuary had higher concentrations of SPM and PBSi compared to surface water samples each sampled period in 2016 (Fig. 3). Macronutrient concentrations were high in the River Severn compared to the Severn Estuary and Bristol Channel, similar to a decrease in DSi at higher salinity levels (Table 2). In each sampling period, nitrate was the most abundant macronutrient in the River Severn (annual avg. of 4.4 ± 0.7 mg/L) and the Severn Estuary (annual avg. of 1.5 ± 0.3 mg/L) and displayed greater temporal variation in the river compared to the more saline end-member. Environment Agency's (EA 2016) monitoring nitrate data were also similar with a range between 0.6 and 8.2 mg/L. Following Redfield-Brzezinski nutrient ratios of C:Si:N:P of 106:15:16:1, PDSi was potentially limiting relative to nitrate during typically biologically productive periods in the river and estuary (see Online Resource 3). Water column temperatures in the Severn were relatively lower during the winter (5.4 ± 0.3 °C) and autumn (9.3 ± 0.03 °C) compared to the spring (13.2 ± 0.1 °C) and summer (17.9 ± 0.03 °C) ( Table 2). In the Severn Estuary, pH fluctuated within the range of 6.9 and 7.9. In the River Severn, pH fluctuated between 6.7 and 8.3 (Table 2).
To assess the interconnections between Si and environmental variables in the water column of the Severn, PCAs were undertaken for each sampling period in 2016 (Fig. 4). PC1 and PC2 accounted for most of the variance in the winter (71.8%), spring (71.6%), summer (87.2%), and autumn (87.7%). In the spring, summer, and autumn, PBSi clustered with SPM with high loadings.

Biological mediation of silicon: benthic productivity
To determine whether the benthic Si export may have contributed to the high estuarine BSi concentrations, the benthic DSi (BDSi) and benthic BSi (BBSi) budgets on the intertidal mudflats were assessed.
To interpret the spatio-temporal variability in these BDSi and BBSi budgets (Fig. 5), the combined impact of multiple consumption/production processes were considered, with an emphasis on the benthic biological mediation of Si alongside environmental forcings. Relatively lower chl a content was measured during the winter (30.6 ± 5.0 µg/g dw. sed.) and increased during the spring (35.7 ± 3.9 µg/g dw. sed.) and summer (55.1 ± 6.9 µg/g dw. sed.) reaching significantly higher levels (H = 55.82, d.f = 3, p < 0.05; post hoc analysis p < 0.01) during the autumn (116.8 ± 16.2 µg/g dw. sed.) (Table 3). No significant linear correlations were calculated between BBSi and chl a. TOC concentrations on the mudflats ranged between 0.4 and 28.1 mg/L in the sampling periods of 2016, with a significantly difference between sampled sites (H = 25.08, d.f = 3, p < 0.001, post hoc analysis p < 0.001) ( Table 3). No significant linear correlations were calculated between BBSi and TOC.
Benthic biofilms during the spring and summer sampling periods were exposed to higher air and water temperatures (Table 3). Despite metrological seasonal changes, water content in the mudflats lacked significant variability between sampling periods: winter (41 ± 0.02%), spring (53 ± 0.02%), summer (43 ± 0.02%), and autumn (45 ± 0.02%). Sediment porosity lacked seasonal variation, with fine sediment fractions (< 63 µm) dominating the intertidal mudflats. The total number of sunshine hours were on average higher in the spring (177 h) and summer (154 h) compared to the winter (65 h) and autumn (84 h). Irradiance levels at the time of sampling were relatively high (> 1253 µmol photons/m 2 /s) during the summer, and relatively lower (< 150 µmol photons/m 2 /s) during the autumn and winter. The MPB biofilms were dominated by photosynthetically productive pennate diatoms, adapted to the mudflat environment (see Online Resource 1). Rapid Light Curves (used to explore the relationship between irradiance and photosynthesis) from the winter and autumn sampled periods saturated, with maximum relative Electron Transport Rates averaging 126 ± 7 rel. units in the winter and 142 ± 28 rel. units in the autumn. In contrast, the majority of RLCs from the spring and summer failed to saturate. In addition to light stress, the

Silicon isotopes in the Severn
Pelagic δ 30 Si DSi in the summer ranged between + 0.64 and + 1.04‰ and between + 0.63 and + 1.08‰ in the autumn (Fig. 6; Online Resource 4). Estuarine waters were isotopically heavier, with δ 30 Si DSi deviating from the theoretical conservative dilution line, indicating non-conservative behaviour between the freshwater and saline end-members. Between 40 and 80 km, considered here as the river-estuary transition zone, δ 30 Si DSi increased, corresponding to a rise in PBSi and SPM concentrations (Fig. 3). δ 30 Si DSi displayed an exponential relationship with both PBSi and SPM (see Online Resource 5). The external input of DSi with an isotopically heavy signature may have originated from the intertidal mudflats of the Severn Estuary. Studies on the Si isotopic signatures on mudflats has only recently begun, with few methodologies Fig. 6 Silicon isotope evolution in the River Severn-estuary-marine zone. A Summer δ 30 Si DSi . B Autumn δ 30 Si DSi . δ 30 Si DSi presented with a LMG 2SD of 0.19. PDSi concentrations (mg/L) (grey squares) and PBSi concentrations (mg/L) (black triangles) presented with 1SE (n = 6 per station). Pelagic δ 30 Si DSi (grey circle) presented with 2SD of long-term reproducibility. C SPM concentrations (mg/L) presented with 1SE (n = 6 per station). D Salinity (psu), presented with 1SE (n = 6 per station). Environment Agency monitoring data (2016) presented to fill in the transition zone where no data was collected in the present study established to handle the high background matrix (e.g., Hughes et al., 2011), which result in significant challenges during analysis by MC-ICP-MS. The mudflat water samples analysed were characterised by high δ 30 Si DSi values (up to + 2.03‰ in summer). To our knowledge, these are the first δ 30 Si DSi values from the River Severn, estuary and intertidal mudflats.

Pelagic DSi dynamics
The high river discharge and strong tidal hydrodynamics were likely key drivers influencing the spatial variation in the transport and biological consumption of DSi and other dissolved macronutrients, with their longitudinal profiles following the classical view of downstream transport and dilution in estuaries (Chou et al. 2007) (Fig. 2). Changes in external forcing drove temporal variations in PDSi, with high rainfall and consequently high discharge rates supplying DSi to the system. This enhanced the residual downstream transport, shifting the salt intrusion zone downstream (Online Resource 6) and resulted in significantly high winter PDSi concentrations, similar to those recorded by Morris et al. (1984) Table 4). However, the δ 30 Si DSi of the River Severn were within the range reported for 557 rivers (− 0.14 to + 4.66‰) (see Table 4). The non-conservative behaviour and heavy isotopic compositions of the river waters are consistent with biological uptake by pelagic diatoms, as has been observed in other estuary systems (e.g., Zhang et al. 2020). However, the Si input from the tributaries (Usk, Wye, Avon and Cardiff Bay) and groundwater would also have influenced the spatio-temporal variation in PDSi concentrations and Si isotopic signatures. To what degree this groundwater DSi supply fuelled and controlled the Severn's DSi and isotopic budget remains unknown, advocating the need for further research into this nutrient pathway, and the influence it may have on enriching nutrient levels in coastal environments.

Pelagic BSi dynamics
Compared to PDSi, the concentration of PBSi were relatively high in the Severn Estuary (Fig. 3). Unlike other systems where PBSi is controlled by pelagic primary production, the high turbidity (Fig. 3) in the Severn Estuary reduced light availability, and likely restricted phytoplankton growth, resulting in a lack of chlorophyll fluorescence readings and low water column primary productivity. Low phytoplankton production has been observed previously in the Severn Estuary (Joint and Pomroy 1981;Underwood 2010) and other tidal-dominated estuarine systems, e.g., marine-brackish parts of the Scheldt (< 0.05 chl a mg/L, Carbonnel et al. 2009). Despite tidal systems having long residence times (see Gao et al. 2020), the lack of pelagic production (regardless of seasonal changes) likely allowed the transport of nutrients out of the estuary preventing any significant DSi retention (calculated following Vollenweider 1968, see Online Resource 7). The lack of DSi retention contrasts with the high average values used in estimates of global Si budgets: 20% (DeMaster 1981; Tréguer and De La Rocha 2013); 25% for temperate rivers (Tréguer et al. 1995); 15% for 19 worldwide estuaries (Roubeix 2007).

High benthic productivity
Consideration of the relatively low riverine/tributary supply of PBSi, the relatively low marine PBSi concentrations and the low phytoplankton productivity, would indicate that the relatively high estuarine PBSi originated from an alternative source. Compared to the pelagic system, the benthic diatom-dominated biofilms on the intertidal mudflats had high productivity (rETR max of up to 223 rel. units, Table 3) driven by their adaption to the harsh mudflat environment (see Online Resource 1), including behavioural down-regulation through cell motility away from high irradiance levels and down-regulation via the induction of non-photochemical quenching (Perkins et al. 2010). These adaptations likely allowed the diatoms to maximise photosynthetic efficiency which aided the growth of biomass-rich biofilms (116.8 µg/g dw. sed. Table 3), and the production of BBSi (Fig. 5).
Biologically productive, rich biofilms are also supported by the isotopically heavy mudflat waters (Fig. 6). Benthic biological activity and the   (2015) preferential uptake of 28 Si (i.e., isotopic fractionation) led to isotopically heavy residual mudflat porewater, with values reaching highs of > 2.0‰. However, alongside the benthic biological mediation of Si, several abiotic processes could have also influenced Si concentrations and led to complex isotopic signatures. For example: transformation of silica into clay minerals (reverse weathering); dissolution of BSi and higher plants and regeneration of DSi following deposition under specific conditions (factor of saturation state, pH, temperature); dissolution of clays (resulting in isotopically light waters); and  Fig. 3), which would have been subjected to the estuarine hydro-geodynamic processes of deposition, burial, and resuspension. The calculated bottom shear stress was comparable to published critical erosion threshold values of the Severn Estuary biofilms, τ 0cr : 3.33 N/m 2 (Yallop et al. 2000). Uncles (2010) report maximum mean spring tidal (MST) current velocities of 3 m/s in the Severn Estuary, with mean neap tides (MNT) of approximately half the MST velocities. MST current velocities resulted in a bottoms shear stress of 55.2 N/ m 2 > > 3.33-10.8 N/m 2 , suggesting the MST currents would have easily exceeded the erosion threshold of the biofilms. Even the MNT currents, with a bottom shear stress of 13.8 N/m 2 > 3.33-10.8 N/m 2 , would have led to some erosion. It is plausible that the erosion of the biofilms led to the transport of BSi to the water column on spring-neap tidal cycles, resulting in high pelagic BSi and SPM concentrations (Fig. 3).

Retention vs export of BSi in the Severn
The resuspended benthic diatoms (BSi) are typically bound by a polymer-rich biofilm formed from extracellular polymeric substances (EPS), (Yallop et al. 2000). The fast-flowing tidal currents associated with high bottom shear stress would have exceeded the erosion threshold of these biofilms despite their biostabilising role. In the turbulent, saline water column of the estuary the biofilms rich in organic matter and the SPM may consolidate to form one entity, known as 'bioflocculation' (Manning et al. 2013). These bioflocs explain the significant linear correlation (Fig. 3) and the strong coupling between PBSi (benthic signature) and SPM (predominately mud) (Fig. 4) as BBSi and sediment originate from the intertidal mudflats and remained as one entity in the water column. Furthermore, once current velocities decrease and with bioflocculation increasing floc size and enhancing sinking velocities, SPM and PBSi concentrations in the estuarine bottom water increased (Fig. 3), contributing to BSi retention in the estuary: winter (50%), spring (2%), summer (2%) and autumn (24%) (see Online Resource 7). These estimates are in line with BSi retention values from Chesapeake Bay (65%, D'Elia et al. 1983) and the Bay of Brest (48%, Ragueneau et al. 2005). This erosion-deposition-recycling cycle makes calculating the transport of BBSi in the Severn complex, and therefore further research is required to quantify the export of BBSi to the water column in the Severn Estuary. Compared to the estimated global riverine amorphous silica (including BSi) supply of 1.9 ± 1.0 Tmol/yr in Frings et al. (2016), surprisingly the Severn Estuary accounted for approximately 16% of this budget. Therefore, the consumption and production of Si by the benthic ecosystem are significant in hypertidal estuarine systems with large intertidal mudflats like the Severn. It is acknowledged that the biogenic silica content was generally measured on thicker biofilms, and that the BBSi flux per unit area is potentially overestimated given the 'patchy' nature of these biofilms (Orvain et al. 2003). Regardless, these estuarine systems and their tight benthic-pelagic coupling (driven by BBSi) must be included in Si budget mass balance estimates.
Si isotopes: strong benthic-pelagic cycling of Si The Si isotopic data further supports the idea of strong benthic-pelagic cycling of Si in the Severn Estuary (Fig. 7). Coupled to the decrease in PDSi, there was a progressive and systematic downstream enrichment in heavy isotopes (Fig. 6), similar to the observations of Cockerton et al. (2013) in the Nile, Delvaux et al. (2013) in the Scheldt, and Weiss et al. (2015) in the Elbe, Germany. The resuspension of BBSi along with the mixing of isotopically enriched mudflat porewater, likely led to the isotopically heavy estuarine waters (Fig. 6). Future research on Si isotopes from mudflat settings should explore the effects of high organic carbon content, and other background matrix issues, on mass spectrometry analysis. Further, one of the key issues in estimating global δ 30 Si DSi is that they are extrapolated from riverine values, which do not include important variations induced by the complex estuarine dynamics.

Conclusion
For the first time, a time-series dataset of Si concentrations in the Severn has been produced, which has improved our knowledge of the pelagic-benthic Si cycle in hypertidal estuaries. These findings highlight the importance of documenting the way in which an estuary modifies Si along its path to the ocean, prior to incorporating the riverine Si flux into global Si budgets. This is particularly important in benthicdominated systems where multiple erosion-deposition cycles interfere with the riverine input signals.
Biotic and abiotic processes on the vast intertidal mudflats controlled the Severn's Si cycle. They have been shown to modify the benthic Si budgets, Fig. 7 Summary of dissolved silicon (DSi), biogenic silica (BSi) transport and silicon isotope signatures in the riverestuary-marine continuum of the Severn. A Flux of PDSi and PBSi from the River Severn and tributaries into the Severn Estuary. Benthic algal biofilm productivity-export of BBSi and δ 30 Si DSi from the intertidal mudflats into the water column of the Severn Estuary. B Epipsammic diatoms attached to sand grains in the intertidal sand flats, e.g., Site 4 (Sand Bay). C Epipelic diatoms move freely between sediment grains in the intertidal mudflats. D Water column mixing. E Organic matter, sourced from the rivers, land run-off and re-suspended biofilms aid in the bioflocculation of suspended matter. ETM zone: Estuarine Turbidity Maximum zone in the upper estuary together with the isotopic signatures of the mudflat water. Upon immersion, these benthic fractions and isotopic signatures are transported into the water column, preventing the build-up of BBSi despite its continuous supply by the biologically productive benthic diatoms. Once in the pelagic realm, the BSi likely undergoes a series of erosion-deposition events and can be removed from the system through burial/ dissolution (retention) or can be transport out of the estuary on the tide. Here we show that the Severn Estuary has tight benthic-pelagic coupling. However, compared to other estuarine systems where this coupling is driven by pelagic production, in the Severn, the coupling is driven by benthic production (BBSi). Considering a large fraction of BSi is exported from the mudflats, and unlike DSi, could be transport further afield (DSi would likely be used up in more productive, less turbid areas of the Severn), the BSi could dissolve and replenish the Bristol Channel and even the Celtic Sea, with 'new' DSi ready to be consumed by phytoplankton, the latter supporting several important fisheries. Through continued field campaigns and modelling efforts, future research should investigate the transport of this benthic intertidal Si to the estuarine water column, and the transport of this estuarine-sourced BSi out to the Bristol Channel and Celtic Sea.