Interactions between sediment microbial ecology and physical dynamics drive heterogeneity in contextually similar depositional systems

This study focuses on the interactions between sediment stability and biological and physical variables that influence the erodibility across different habitats. Sampling at short‐term temporal scales illustrated the persistence of the microphytobenthos (MPB) biomass even during periods of frequent, high physical disturbance. The role of MPB in biological stabilization along the changing sedimentary habitat was also assessed. Key biological and physical properties, such as the MPB biomass, composition, and extracellular polymeric substances, were used to predict the sediment stability (erosion threshold) of muddy and sandy habitats within close proximity to one another over multiple days, and within emersion periods. The effects of dewatering, MPB growth, and productivity were examined as well as the resilience and recovery of the MPB community after disturbance from tidal currents and waves. Canonical analysis of principal components (CAP) ordinations were used to visualize and assess the trends observed in biophysical properties between the sites, and marginal and sequential distance‐based linear models were used to identify the key properties influencing erodibility. While the particle size of the bed was important for differences between sites in the CAP analysis, it contributed less to the variability in sediment erodibility than key biological parameters. Among the biological predictors, MPB diversity explained very little variation in marginal tests but was a significant predictor in sequential tests when MPB biomass was also considered. MPB diversity and biomass were both key predictors of sediment stability, contributing 9% and 10%, respectively, to the final model compared to 2% explained by grain size.

Variability in response to physical forcing is an inherent property of natural systems (Black et al. 2002) and represents a significant challenge for modeling and predicting the behavior of natural sediment beds. It is also commonly suggested that an important source of this variability, biogenic stabilization (Tolhurst et al. 2009;Parsons et al. 2016), is largely confined to fine cohesive sediments (mud flats) rather than more sandy substrata. However, this approach neglects the heterogeneous composition of natural beds that vary both spatially and temporally (Rainey et al. 2003;Chapman et al. 2010). Understanding the biogenic processes that generate variation and heterogeneity in natural systems will support our ability to model system behavior more accurately in the future.
Biological stabilization of cohesive (muddy) sediment is often attributed to microbially produced extracellular polymeric substances (EPS) that increase the cohesion between sediment particles, and forming biofilms (Hubas et al. 2018). The production of EPS is typically attributed to bacteria and microphytobenthos biomass (MPB) (Lubarsky et al. 2010;Chen et al. 2017). While MPB are more abundant on cohesive sediments, recent studies show that microbially produced EPS can also hinder bedform development and inhibit erosion in noncohesive (sandy) (Malarkey et al. 2015;Chen et al. 2017) and mixed sediment beds (Parsons et al. 2016). The balance between physical disturbance (hydrodynamic stress) and bed erodibility is complex (Beninger et al. 2018). Regular physical forcing can restrict the accumulation of fine sediment and MPB on the bed (Mariotti and Fagherazzi 2012) preventing the MPB standing stock from developing fully (Blanchard et al. 2001). However, once developed, after a period of calm conditions, increasing erosive stress may be resisted. The biomass and nature of the MPB, epipelic (e.g., Paterson and Hagerthey 2001) or epipsammic (e.g., Harper and Harper 1967;Hickman and Round 2007), will contribute to the variability of the response to stress.
If a system is largely abiotic, then the introduction of biota can create greater heterogeneity or homogenize the system. For example, greater microphytobenthic diversity has been linked to higher grazer diversity (Balvanera et al. 2006) which through differences in bioturbation can increase habitat heterogeneity (Hale et al. 2015). Furthermore, patchy biofilm distribution and growth have been associated with positive biophysical feedbacks as the system becomes inherently more patchy leading to spatial self-organization and more fine sediment accretion and eventually influencing large geomorphological features (Weerman et al. 2010). However, the introduction of biota can also cause different sediments to become more similar to one another. For instance, the presence of large infauna can "smooth out" the effects of flow on sediment resuspension across different sediment types (Li et al. 2017). The former processes of increasing heterogeneity suggest that microbial growth and EPS accumulation can not only have a localized effect, but if growth becomes extensive, biostabilization is capable of influencing ecosystem functionality at various spatial and temporal scales (Orvain et al. 2012;Ubertini et al. 2015) and although variability may increase (Chapman et al. 2010) this can have system-wide implications. Many studies have focused on seasonal and interannual variability (Montani et al. 2003;van der Wal et al. 2010). However, the mechanisms that drive changes to the structure of the system (van de Koppel et al. 2001) and resilience and recovery from disturbance can occur on relatively short spatial (Spilmont et al. 2011) and temporal scales (Orvain et al. 2012). Furthermore, the variability observed from small spatial scales over short timeframes in intertidal environments can be of the same order of magnitude as both seasonal and annual variability (Seuront and Leterme 2006). The importance of considering temporal scales has been highlighted in previous soft sediment studies (Tolhurst et al. 2005a;Hewitt et al. 2007). However, short-term temporal dynamics that may influence EPS accumulation, biofilm development, and biostabilization have not been well characterized across different sediment types. The development of biofilms depends on the balance between growth and detachment, with hydrodynamic stress being a primary driver of benthic biofilm detachment (Telgmann et al. 2004). We therefore require further information on the interactions between biofilm properties, biostabilization, hydrodynamic stress, and subsequent resistance to erosion over multiple emersion periods and within different habitats. This information is essential to assess the role of biostabilization, both from ecological and dynamic perspectives (de Brouwer et al. 2000;Underwood and Paterson 2003;Mariotti and Fagherazzi 2012). MPB influence on sediment stability and this key ecosystem function augments their important roles in; the transfer of energy to higher organisms (MacIntyre et al. 1996); the bentho-pelagic exchange of sediment and nutrient cycling (McGlathery et al. 2004). The importance of MPB in these ecosystems highlights the need to understand the dynamics governing their presence across different habitats. Frequent resuspension of MPB cells and related EPS may prevent the formation of substantial biofilms, and therefore limit their biostabilization potential (Aspden et al. 2004); however, an "inoculum" often remains in place  leading to rapid recolonization under suitable conditions. We hypothesize that biofilm properties such as MPB biomass, colloidal carbohydrate concentrations, and the MPB diversity will influence biostabilization of various sediment types. Furthermore, we hypothesize that the biogenic influence will persist over various temporal scales (emersion on consecutive days), as the microphytobenthic community and biotic characteristics tolerate regular, high intensity tidal inundation. As laboratory experiments cannot generally capture the natural variability in large, complex, and dynamic marine systems (Tolhurst et al. 2009), we examined these short-term dynamics in closely associated muddy and sandy habitats in the Dee Estuary, England. This estuary is subject to strong current velocities and frequent wave action, resulting in turbid waters with a high suspended load (Amoudry et al. 2014), but has various sedimentary habitats in close proximity to one another, making it an excellent model system. Suspended sediment often affects water quality, which limits light availability for sediment dwelling photosynthetic organisms during tidal inundation (Pratt et al. 2014), and the physical disturbance from flow itself may prevent the accumulation of EPS and biofilm development on the bed (Blanchard et al. 1997;Ubertini et al. 2015).

Study sites and sample collection
The Dee is a hypertidal estuary located on the border between England and Wales in the Eastern Irish Sea. The estuary is tidally dominated with a mean spring tidal range of 7-8 m (Moore et al. 2009). The geomorphology of the flats causes a tidal asymmetry that is flood dominated, resulting in significant accretion of fine sediments in the upper estuary (Halcrow 2013). Three sites were selected between Hilbre Island and West Kirby (Fig. 1), based on the geomorphology of the bed surface at the sampling time. The first site (sandy 1) was dominated by noncohesive sediment (sand) with waveinfluenced 2D current ripples. Site 2 (sandy 2) was similar but had active 2D and 3D ripples. The third site (muddy) was composed of muddy sand, with either a flat bed or relict current ripples (Lichtman et al. 2018).
Surface sediment samples were collected at four time points during tidal exposure over 3 d at each site from 23 rd to 31 st May 2013 (sandy 1: 23 rd -25 th , sandy 2: 26 th , 28 th -29 th ; and muddy: 28 th -29 th and 31 st May). A full description of the physical conditions during the campaign at the adjacent sites can be found in Lichtman et al. (2018) and the Supporting Information. In brief, sampling dates at sandy 1 coincided with the tides transitioning from neaps into peak springs and there was also increased wave action due to high winds on 23 rd -24 th May. Despite the strong wave action at sandy 1, which caused the maximum wave-current bed shear stresses during wave cycles to be larger, the peak current bed shear stresses during inundation were greater at sandy 2 than sandy 1. Slightly weaker currents were observed at the muddy site on day 3, as the tides moved from peak springs toward neaps. Nonetheless, comparable maximum depth-averaged flood/ebb currents were measured across all sites (0.4-0.7 m s −1 ) and the maximum water depth at each site ranged between 2 and 3.3 m (Supporting Information Table S1).
At each site, 20 surface sediment samples were collected each day (n = 5 × 4 time points). The first sampling occurred 30-60 min after sediment exposure each day with sampling repeated quarterly during low tide until 60 min before inundation. Samples were collected within 5 m of rigs deployed by NOC Liverpool (Lichtman et al. 2018) and University of Plymouth across an area of approximately 3 m 2 . Sediment cores (2 mm depth, surface area = 250 mm 2 ) were frozen and stored in liquid nitrogen using the contact core method described in Ford and Honeywill (2002) and Brockmann et al. (2004). Cores were subsequently stored frozen (-80 C) in the dark until processed. To capture both epipelic and epipsammic microalgal cells, replicate surface scrapes (n = 5, 10 × 10 × 2 mm depth) were collected and stored in 2.5% w/w glutaraldehyde/filtered seawater solution from time point 2 (T2) only.

Sample processing
Water content (%) was calculated from wet and freeze-dried core weights before sediment organic matter (%) was determined by loss-on-ignition at 450 C for 4 h. Chlorophyll a (Chl a) pigments were extracted with 90% acetone following the trichromatic method of Jeffrey and Humphrey (1975). The colloidal and total carbohydrate fractions of the EPS were determined using the phenol-sulfuric acid assay (DuBois et al. 1956) following Underwood and Paterson (2003). Due to differences in contents vs. concentrations caused by the varied water content of sediment samples, both chlorophyll and carbohydrate measurements are expressed as concentrations per unit area (mg m −2 ; Tolhurst et al. 2005b). The effective particle size distribution (PSD; Grabowski et al. 2012) was determined using a Malvern Mastersizer 2000 laser diffraction analyzer (Malvern Instruments Ltd, 2013) and summarized using GRADISTAT software (Blott and Pye 2001) prior to statistical analysis with D 50 and mud content (% < 63 μm) used for further analysis. The relative difference in erosion threshold required to suspend a user-defined sediment concentration of 0.01 kg m −3 was measured using the portable in situ cohesive strength meter (CSM; Paterson 1989;Tolhurst et al. 1999). In addition to the surface erosion threshold, the undrained shear strength was measured using a 33 mm Pilcon shear vane (5 cm depth).

Low temperature scanning electron microscopy
Fragments of contact core samples, frozen in liquid nitrogen (−196.8 C, 1 atm), were mounted on mechanical stubs and examined using a JOEL 35CF SEM fitted with a low temperature scanning electron microscopy (LTSEM) (Oxford Instruments CT 1500B) following the procedure given in Paterson (1995).

Statistical analysis
Analyses were performed using "R" statistical software, version 3.1.1 (R Development Core Team 2013) through the R studio graphical interface (v. 0.98.1083) and in PRIMER software (V.6, PRIMER-E, Ivybridge, UK). Differences in biophysical variables were determined between sites, emersion times, and days, after assumptions were tested (Zuur et al. 2007;Pinheiro et al. 2012). Necessary transformations were applied to conform to assumptions for parametric statistical testing (ANOVA), where possible or nonparametric Kruskal-Wallis (H) tests were employed. Corresponding post hoc Tukey's or Dunn-sidak tests were applied to detect differences between specific groups. The relationships between the different sediment properties and stability measurements were also assessed using Spearman rank correlations. No significant differences were observed across the different timepoints during emersion for stability or biochemical properties, therefore timepoints were pooled resulting in 20 replicate samples from each day and site. Multivariate analysis of the data using canonical analysis of principal components (CAP) was employed on square-root transformed data to assess the response of multiple biophysical variables across the sites using constrained ordination taking account of the correlation structure of data (Andersen and Willis 2003).
The MPB community composition between the sites and days was also examined using CAP, based on Bray-Curtis dissimilarity matrices (Somerfield 2008). Differences in species between sites and days were tested using permutational multivariate analysis of variance in addition to exploring Shannon's diversity index (H 0 ) and Pielou's evenness index (Magurran 2004).
To determine whether the variation in sediment erosion thresholds could be explained by differences in the measured biophysical properties of the sediment across all sites, data were pooled and distance-based linear models (DistLM) were employed (Anderson et al. 2008). Temporal factors (time since emersion and sampling day) were included as explanatory variables along with the various biophysical properties. Marginal and sequential tests were examined using Akaike's information criterion (AICc) and a backward elimination process to identify the best combination of predictors, that maximized the explained variation with the most parsimonious model (Clarke and Gorley 2006).

Sediment bed properties and stability
The percentage of mud (< 63 μm) was significantly higher at the third site (Muddy site, Table 1; 27%, H (2) = 93.93, p < 0.001) compared to the sandy sites but there was no significant difference in mud content between the two sandy sites (0.8% and 1%). For the sandy sites, clean particles were visually observed with very little associated organic matter ( Fig. 2A,B) and the grain size distributions were similar (Supporting Information Fig. S1). In contrast, the muddy site exhibited more varied and organic-rich sediments (Fig. 2C,D), but total organic content was relatively low across all sites (< 2%) and no significant site differences were detected. The water content, colloidal carbohydrate, and Chl a concentrations were all significantly higher at the muddy site compared to both sandy sites (all p < 0.001, Table 1). Differences were between all three sites, with higher contents and concentrations at the muddy site, followed by sandy site 2 and then sandy site 1. The shear strength of the bed was also significantly different between sites, yet this was due to lower strength at sandy site 2, as strengths were similar between sandy site 1 and the muddy site (p > 0.05).
At sandy 1, both the colloidal carbohydrates (EPS) and erosion thresholds varied significantly over the sampling days (H (2) = 27.12, p < 0.001 and H (2) = 13.76, p < 0.001) but with opposing trends (Table 1). Colloidal carbohydrates were lowest on day two when the erosion threshold was highest, with a decrease in threshold coinciding with an increase in mud and organic content. Overall, the erosion measurements at sandy 1 correlated very poorly with all measured biological and physical variables, but there was a negative relationship between Chl a concentration (MPB biomass) and the erosion threshold at this site over the sampling days. At the second site (sandy 2), the erosion threshold decreased over the sampling days to its lowest on day 3 (1.8 kPa). This was alongside significant reductions in colloidal carbohydrate concentrations from 320 AE 115 mg m −2 (day 1) to 229 AE 54 mg m −2 (day 3, Fig. 3; F 2,59 = 6.57, p < 0.001). At the muddy site, the erosion threshold was at its highest on day 1 (14.5 kPa) and the lowest on day 2 (2.7 kPa) also coinciding with the lowest colloidal carbohydrate concentration (428 AE 110 mg m −2 , p < 0.001) and the strongest wave action (Lichtman et al. 2018).
The D 50 of the bed varied very little between days at sandy 1 (Table 1), although there was a small but statistically significant difference at sandy 2, D 50 increasing from 202 AE 3 μm on day 1 to 213 AE 10 μm on day 3 (H (2) = 25.88, p < 0.001). No significant differences in D 50 were detected at the sites during different emersion points measured (data not presented).

Microalgae community analysis
Significant spatial and temporal differences were detected in the microalgae community across sites, as well as differences between individual days within the sites (p < 0.01). Diversity (H 0 index) and evenness (Pielou's index) were significantly different between sites (F 2,24 = 4.91, p < 0.05) but differences were relatively small (H 0 at sandy 1 = 3.14, sandy 2 = 3.32 and muddy site = 3.62, and Pielou's index sandy 1 = 0.92, sandy 2 = 0.87 compared to muddy site = 0.94). A few cosmopolitan species such as Nitzschia frustulum var. inconspicua were present in almost all samples, across all sites, with Achnanthes punctulata present in greater numbers at the sandy sites. There was a greater abundance of low profile than motile species at the sandy sites and there were no high-profile species noted. Interestingly, the muddy site had a similar average abundance of low-profile epipelic species to that of the sandy sites, had fewer small epipsammic cells, but had greater numbers of motile and high-profile species that dominated this site. The high proportions of Navicula gregaria (4%), Amphora coffeaeformis var. perpusilla (4%), and Pleurosigma aestuarii (4%) at the muddy site appeared to have the greatest effect on site differences in community composition (Fig. 4), whereas other smaller species such as Opephora mutabilis (4%) and Cocconeis sp 1 (4%) were more abundant at the sandy sites. Diversity and evenness did not change over time at the sites, but the abundance and turnover of key species did vary. At sandy 2, a decline in Opephora mutabilis contributed 9.3% to the overall dissimilarity (42%) between days while the decline in the large motile Navicula digitoradiata and Pleurosigma aestuarii (6% and 5%, respectively) were the greatest contributors to the overall dissimilarity (46%) between days at the muddy site.

Biophysical influences on erodibility
Inspection of the CAP plots (Figs. 5, 6) and the resulting trace statistic (p < 0.001) confirmed a strong overall difference between the sites based on biophysical properties. The first axis of the plot (CAP1) was partitioned between several biological variables relatively evenly, including Chl a, both carbohydrate fractions, the diversity and number of MPB species present, and water content. Together these variables and the D 50 of the bed, which exhibited a strong anticorrelation to the biological properties, dominate this axis. On the second axis (CAP2), the D 50 of the bed and the undrained shear strength were important factors. A clear spatial separation was observed  from the superimposed scatter plot (Fig. 5) as well as a temporal component based on draining throughout the emersion period. The close relationships between several variables in the CAP plots were in agreement with correlation analysis ( Table 2). As there were no clear dominant biological or physical factors, the majority of properties within both axes were retained for further exploration. Water content was highly correlated to several variables (Table 2) and there were no significant effects of dewatering detected from within each tidal exposure period. Therefore, water content was removed from further models. Various single and sequential predictor variables significantly explained the variation in the sediment erodibility across all sites in DistLM (Table 3). When properties were considered individually, both Chl a and organic content significantly explained the greatest variation (at 9% and 8%, respectively). While Chl a exhibited a negative effect across all sites, this was primarily driven by the negative relationship at sandy 1, and although MPB diversity and abundance were not good single predictors, they were valuable in sequential tests after consideration of the MPB biomass estimates (Chl a concentration). While the D 50 was marginally insignificant in both marginal and sequential tests, it was important to retain in the latter yet surprisingly the mud content of the sediment was not selected as a good predictor of erosion threshold across the sites.

Discussion
Our results illustrate that the MPB maintain a key role in sediment dynamics, by surviving harsh environmental conditions, and quickly re-establishing biostabilization. MPB continued to exert an influence on a key ecosystem function; sediment stability across different sediments. Mariotti and Fagherazzi (2012) proposed that, given equal intensities of disturbance, the biomass of a biofilm determines whether or not it will be eroded and our results support this. Importantly, our results suggest that this biostabilization can exert influence on different sediment types. In energetic environments, the frequent turnover and reworking of the sediment may be expected to remove biofilms, hence these habitats are often depicted as abiotic (Fig. 7). In very energetic systems, like our sandy site 1, the formation of a fluffy biofilm or layer of cells and EPS may not have created a stable matrix and therefore leads to a "low biostabilization" scenario. The lack of incorporation into the bed explains the negative relationships between key biofilm properties and sediment erosion measurements observed at sandy 1. However, as grain size was reduced and mud content increased, even slightly, this positively influenced sediment stability, promoting stronger relationships between the biochemical properties of the biofilm and sediment stability at sandy 2 and the muddy site.
While frequent resuspension of MPB cells and related EPS may prevent the formation of substantial biofilms, and therefore limit their biostabilization potential (Aspden et al. 2004), an inoculum often remains in place (Chen et al. 2019) and this can still exert biostabilizing effects on the sediment as we have illustrated. This persistence of the biofilm and its stabilizing properties means that a biofilm can develop rapidly, if conditions become favorable (see Fig. 7; Chen et al. 2019). Previous studies of relationships between EPS carbohydrates and  sediment stability have estimated that 2-3 d are required (Lundkvist et al. 2007). However, as we have illustrated in situ that growth does not begin anew at the start of each tidal cycle and biofilms present across different sediment habitats, although invisible, maintain their biostabilization potential. The stabilizing effects of MPB may therefore take less time to develop and become more significant in the natural environment .
At the sandier sites, the fine sediment and organic matter which was captured in the suspended sediment traps (data not presented) may have settled onto the sediment surface during slack water, but did not accumulate uniformly on the bed. At sandy 2, a fine organic coating was observed on larger sand grains on ripple crests ( Fig. 2A) whereas grains from ripple troughs were almost free of OM (Fig. 2B) suggesting that OM was being "caught" from the suspended sediment in the flow. In contrast, the surface sediment at the muddy site was characterized by a much thicker blanket of fine cohesive particles, rich in organic matter and MPB cells, confirmed by LSTEM  images (Fig. 2C,D) and biochemical analysis ( Table 1). The presence of this organic-rich material can result in positive feedbacks to the system, whereby the adhesive organic EPS and MPB cells trap and bind more fine material, maintaining a muddy bed and reducing suspended sediment concentrations. This stabilizing effect can often be limited to warmer seasons when MPB growth is higher (Borsje et al. 2008) and periods of lower physical disturbance (Widdows and Brinsley 2002) in temperate regions, but appears to prevail under higher shear stress in this instance. An increase in the D 50 of the bed and a decrease in the organic content at the muddy site 28 th -31 st was accompanied by a sharp decrease in bed stability (see Table 1), suggesting the removal of organic material and resulting increase in particle size can destabilize these beds. Organic material creates cohesion between sediment particles, stabilizing the sediment when it is bound to particles (Black et al. 2002;Manning et al. 2010;Zhang et al. 2018). However, MPB and OM transported to a particular area during the tide can also form a "fluff" layer on the bed surface that is easily resuspended if it is not incorporated into a biofilm (Orvain et al. 2003). This is likely the case at sandy site 1 as high MPB biomass (Chl a concentration) and EPS were observed on days when the erosion threshold was reduced. The cells and EPS detected in the sediment surface were therefore unlikely to have formed a protective film. Indeed, in sandy sediments (like sandy site 1), small diatoms tend to attach themselves to the grains and coat individual sand grains in EPS rather than forming a substantial biofilm per se. The different mechanisms by which MPB and EPS develop in dynamic sandy sites may explain the negative relationship between Chl a and erosion threshold at sandy site 1. Substantially more EPS is produced and excreted by epipelic (motile) diatoms, like the taxa dominating the muddy site. While previously it has been thought that > 50% of the MPB community must be epipelic species (Underwood et al. 1995;Underwood and Paterson 2003), in the Dee Estuary it appears that the proportion may be much lower. These differences in the relationships often hinder attempts to generalize MPB biomass effects on erodibility, and lead to significant differences between studies. Despite the importance of accurately forecasting erosion threshold parameters for sediment transport predictions (Sanford 2008), the influence of biological cohesion across different habitats is rarely considered in these models (Le Hir et al. 2007). This is primarily due to the complexity of intertidal systems and differences in biological and physical processes across sediment gradients. There is likely a threshold of development under which very coarse sandy sites may not be positively influenced by MPB biostabilization, or under extreme conditions like the significant wave action at sandy site 1 prior to sampling (Lichtman et al. 2018). Nonetheless, other sandy sites (such as our sandy site 2) can be positively influenced by biostabilization (Larson et al. 2009). Cells and EPS that are not incorporated as a biofilm can be easily suspended and recorded as erosion by the CSM system and these results suggest that characterizing the MPB community can help to explain these differences (Fig. 7A).
In this study, EPS (as colloidal carbohydrates) were positively related to sediment erosion thresholds, even at sandy site 2. At this site, the community was composed primarily of low profile pioneer species, and had a limited number of motile forms (discussed further in microalgal communities section). This relationship suggests that even when the EPS matrix does not form a substantial biofilm on the surface, it still offers some form of protection to the underlying sediment (Fig. 7B). Laboratory (Malarkey et al. 2015;Parsons et al. 2016) and field investigations ( Table 3. The percentage variation in the erosion threshold of the sediments across all sites, explained by various biophysical properties. Both marginal (single predictor) and step-wise sequential results for DistLM are presented.
have recently illustrated the influence of low EPS contents distributed deeper into the sediment bed. In these investigations, microbially produced EPS hampered sediment transport, bedform development, and bedform migration without the presence of a visible biofilm on the surface (see also Chen et al. 2017). These vectors of change are undoubtedly part of the short-term variation due to changes in the spring/neap cycle and daily weather fluctuations, but these changes can also be the first steps toward a transition toward an alternative state (van de Koppel et al. 2001).

The microalgal communities
Over the relatively short sampling period described here, the microalgae community at the sandy sites appeared to remain in early successional stages, whereas the community at the muddy site had already developed into a more vertically structured community composed of stalked, filamentous, and motile microalgae (Winsborough and Golubic 1987). The latter forms can withstand stronger flow velocities and abrasion from moving sediments by migrating into the muddy sediment or creating filaments. However, adnate forms such as Achnanthidium are well equipped to resist flow (Passy 2007). Certain diatom species are indicators of the flow regime, with particular species exhibiting preference for high flow such as Achnanthidium spp. (Passy 2007). While flow differences can result in different communities, the effects can also be dampened by differences in turbidity (Soininen 2004). Nonetheless, information on community composition together with information on the biophysical properties of the sediment can be useful for determining the differences in erosional resistance and potential biostabilization across different habitats. Our results indicate that the number of species was altered over the Increasing the local resistance to erosion in some patches. (C) A fully colonized substratum where biofilm development has created a more uniform sediment surface, once again reduces the variability of the system but further increases the erosional resistance. Top left: Spatial variation in erosion resistance across the bed. Condition A = homogenous abiotic grains producing constant and predictable erosion thresholds. Condition B = a highly heterogeneous system with an erosion threshold influenced by the complexity of local conditions, and the patchy distribution of MPB and bacterial biofilms. Condition C = coherent biofilm increases sediment stability and reduces erodibility in a consistent manner across the bed. This creates a more homogenous response to erosional stress until bed failure. Erosive loss from areas of biofilm growth (C) can lead to the depositional gain of MPB at other sites (B). This may lead to the development of a substantial biofilm (C) or the subsequent resuspension of the MPB again. At more energetic sites (A), fine sediment and MPB are deposited during slack tide, but these are resuspended on the next tide, maintaining a relatively homogenous system where MPB may be present but there is no stabilizing effect of the biofilm due to frequent resuspension. These states may alter as local conditions change including seasonal, light, nutrient and temperature differences (which would stimulate the MPB and biofilm growth), and hydrodynamic conditions which increase erosional stress on the surface sediment.
days at the different sites and species composition changed over short temporal scales. However, in each site the different guilds (low profile, high profile, motile, etc.) remained dominant due to their adaptations to the flow environment. While MPB community succession can be revealed through the microscopic identification of cells, this is time-consuming and could be complimented by next generation sequencing of the prokaryote and eukaryote communities for longer-term studies (Hicks et al. 2018). This would provide a more comprehensive microbial community analysis in relation to biostabilization  as the diversity of prokaryotes has been linked to hydrodynamic regimes (Besemer et al. 2009). Such an approach would be incredibly useful for capturing the transformation of sites that are frequently disturbed and dynamic in nature, into more stable muddy habitats over longer timescales. MPB composition and structure can reflect differences in flow regimes (Krajenbrink et al. 2019) and as the community changes, primary productivity, and the production of EPS exudates will vary, with knock on effects on various ecosystem structure and functions (Hope et al. 2019). As the hydrodynamic effects on MPB communities can modulate the effects of others stressors (Villeneuve et al. 2011;Polst et al. 2018), understanding the interaction between the community and hydrodynamics across different sediment habitats is essential. For instance, Achnanthes spp. and Nitzschia inconspicua were observed in all Dee Estuary sites. These cells are often one of the first species to inhabit recently disturbed sediment (Cardinale 2011), and are cosmopolitan (Sabater 2000). They can grow prostrate to the surface or adnately (Berthon et al. 2011;Cardinale 2011), which helps them withstand high flow velocities (Passy 2007). We have however illustrated that these are not displaced by the development of the biofilm as they were still present in our muddier site. These could be key species to examine for the effects of other stressors in these environments. These pioneers promote the rapid recolonization of the sediment bed after physical disturbance, instigating a biological succession, which promotes greater biodiversity and heterogeneity in the biofilm and among higher organisms (Balvanera et al. 2006). This can pave the way for a more heterogeneous community of microorganisms and a complex habitat that can increase biostabilization .
The survival of algal cells during tidal inundation or their deposition from the water column establishes a potential for microbial growth and eventual biofilm formation if conditions allow (Fig. 7B,C). This was evident from the differences between sites in the Dee Estuary. At sandy site 1, the MPB community was composed of pioneer species that turned over with the prevailing hydrodynamic conditions. At the opposite end of the spectrum, the MPB community at the muddy site was more stable and composed of larger epipelic species. However, at sandy site 2, the community was distinct and appeared to be intermediate between the two other sites. In these transitioning sites, the maximum variation in the sediment surface erosion is expected (Fig. 7B) as the surface is patchy. As MPB communities develop and grow on the sediment surface, this drives the capture of more cohesive material and this positive feedback enhances the development of a more homogenous and stable surface dominated by biofilms (Fig. 7C).
Disturbance from tidal flow can exert the same effects as large bioturbating fauna oxygenating the sediment surface layers Precht and Huettel 2003). These processes are important for soft sediment ecosystem functions such as sediment oxygenation, biogeochemical cycling and, depending on the organic enrichment of the sediment, degradation processes (Widdicombe and Austen 2001). The contribution of large infauna has recently be discussed elsewhere (Hillman et al. 2019) and are of course important to consider in many habitats. Low numbers of large fauna were observed at these sites, therefore we focused on the physical processes and the interaction with microbial organisms that are known to stabilize and disturb the bed. The close spatial association of visibly different sedimentological properties suggests that biophysical factors may contribute to the variation over short distances despite the similarity of dynamic context.
Understanding the biophysical factors influencing sediment stability across different habitats allows us to begin to discern how and why mixed beds occur and the mechanisms by which they alternate between muddy, mixed, and sandy habitats. Muddy sites can capture fine sediment, due to the cohesive nature of fine particles, MPB, and EPS in the surface layers (Table 1). This cohesiveness can prevent fine particles from winnowing during inundation and result in higher erosion thresholds. These biophysical properties can lead to the formation of a cohesive matrix that can effectively trap additional material from the water column and improve the clarity of the overlying water. Positive correlations between mud content, organic content, EPS carbohydrates, and MPB biomass/ community indices were apparent at the muddy site (Fig. 5) indicative of biofilm development where higher numbers of motile diatoms were present. It has previously been suggested that relatively high proportions of motile diatoms, and hence high EPS concentrations, are required to trap new deposits of sediment (Underwood 1997) and counteract the physical forces that resuspend sediment, and this can lead to positive feedbacks. Van de Koppel et al. (2001) highlighted these feedbacks, and proposed that ecosystem engineering (Jones et al. 1994), principally by MPB, can mediate changes in bed sedimentology.
At the sandy habitats, the regular physical disturbance from waves and currents over the tidal cycle prevented the accumulation of larger MPB, which limits biofilm development (Blanchard et al. 2001). Over the course of this relatively highresolution investigation, the D 50 of the sandy site 1 increased despite the fine nature of the material frequently collected in suspension traps at the sites (data not presented). This was primarily due to the prevailing wind wave action at this site during this period (Lichtman et al. 2018; Supporting Information Table S1). Fine cohesive sediment has to be removed frequently in sandier habitats, through resuspension or winnowing to impede the development of biofilm growth but MPB are still present and still exert influence over the sediment dynamics. When conditions are altered this can allow the MPB to proliferate, significantly increasing the erosion threshold and instigating a transition to finer sediment. These differences in the erosive nature of the bed and the fate of settling material is key to maintaining the differentiation between patches and increases overall habitat heterogeneity (Weerman et al. 2011) and functioning of soft sediment ecosystems (Thrush et al. 2008).
By investigating the short term, temporal dynamics influencing the MPB community, and the feedbacks between the biomass, community composition, exudates of MPB, and biostabilization potential, we can begin to understand the conditions required to instigate the changes that lead to transitions and postulate how the microbial organisms in these habitats can persist and continue to exert an influence on sediment stability.

Conclusion
The relative influence of MPB and EPS on sediment stability and transport remains poorly understood across different sediment habitats. The results of the study suggest various biological properties of the bed associated with the MPB significantly influence the short-term variability in the erodibility of different surface sediments. Importantly, we illustrate that while MPB diversity explained very little variation in marginal distance based linear tests, primary producer diversity was a significant predictor when MPB biomass was also considered in sequential tests. We emphasize the importance of considering the microbial diversity when assessing their influence on ecosystem functions such as sediment stability. Further evidence of biological cohesion across natural habitats of increasing complexity and at multiple spatial and temporal scales is required in order to understand the biological influence on sediment dynamics. Further data with natural gradients of sand and mud should be examined and the influence of larger benthic organisms included to document the influence of biological properties across different habitats, under differing physical conditions and with increasingly complex communities. This will facilitate the use of these variables in future sediment transport models.