How strong are salt marshes? Geotechnical properties of coastal wetland soils

Salt marshes are globally distributed, vegetated intertidal wetlands and marsh edge erosion is common on many shores. To understand how and why marsh edge erosion occurs, the response of salt marsh substrates to applied shear and vertical stress must first be quantified. This response is likely influenced by marsh substrate biological, geochemical and sedimentological composition. However, currently there is little systematic research into the between‐marsh variability in these properties and how they affect both marsh edge erosion processes and the ability of a marsh to maintain its position vertically within the tidal frame.


| INTRODUCTION
Salt marshes (vegetated intertidal platforms) provide important regulating, provisioning and cultural ecosystem services (Foster et al., 2013;Spalding et al., 2014). One particular service is that of hazard regulation, whereby marsh surfaces are efficient storm surge, current and wave dissipaters during infrequent high-energy events . It has therefore been argued that the presence of such wetlands often lowers the risk of flooding and reduces erosion and maintenance costs of landward-lying flood protection structures (Beaumont et al., 2008).
However, salt marsh area loss has been recorded on many coastlines worldwide (Blankespoor et al., 2014; e.g. Northwest Europe: Baily & Pearson, 2007;van der Wal & Pye, 2004). Future marsh areal loss is also of concern, both from ongoing and near-future accelerated sea-level rise where landward marsh migration is not possible (Schuerch et al., 2018), and changes in temperatures and precipitation patterns which can adversely affect the marsh vegetation (Strain et al., 2017). This widespread area loss and concern over external pressures has raised important questions regarding the future persistence of salt marshes and the ecosystem services they support.
In European salt marshes, marsh platform elevation, tidal inundation and sedimentation are strongly linked, through a form-process feedback (French, 2006). Understanding changes in the vertical stability of the marsh platform is therefore critical. One component of this is the ability of the marsh substrate to withstand vertical stress. For marsh platforms exposed to loading from exceptional water depths during storm events on the Gulf and East Coasts of the United States, Cahoon (2003) observed a range of field responses to storm surge inundation, from surface compression and near-complete rebound, to an overall increase or decrease in elevation. However, how this range of responses can be generated remains poorly understood.
Whilst much recent saltmarsh research has looked at marsh response to sea-level rise and, as a consequence, has had a strong focus on vertical performance and stability, the loss of saltmarsh area has drawn attention to the need to know more about lateral stability.
Empirical studies suggest that the rate of marsh edge erosion may correlate with wave power (e.g. Leonardi et al., 2016), but models used to simulate marsh edge erosion assume a spatially homogeneous substrate (vertically and horizontally) and thus far do not incorporate different substrate properties. Understanding marsh edge erosion processes and marsh stability first requires an understanding of how the substrates behave when subjected to an applied force. Marsh edge erosion processes include (a) cliff undercutting followed by cantilever failure or gravitational slumping once the overlying section weight exceeds the substrate strength (e.g. Allen, 1989Allen, , 2000Bendoni et al., 2014;Francalanci et al., 2013) and/or (b) tension crack formation during substrate drying and subsequent toppling failure (e.g. Bendoni et al., 2014;Cola et al., 2008;Francalanci et al., 2013).
While there is a general consensus among existing studies that such processes contribute to marsh edge erosion, observations of such erosion are rare and difficult to capture (e.g. Godfrey et al., 2020). As such, the erosion processes causing retreat are generally inferred from marsh morphology change after erosion has taken place and are therefore poorly understood.
What is less well understood is precisely how marsh edge erosion occurs: firstly how biophysical processes achieve marsh substrate removal and secondly how the geotechnical and sedimentary properties that may relate to spatial variations in substrate resistance to erosion vary within and between marshes of different substrate composition. This paper deals with the latter.
We use a laboratory study because the specific substrate properties controlling the mechanics of vertical and horizontal substrate deformation or failure are best tested on substrates in the laboratory, free from the confounding influences that accompany the observation of storm surges and tidal inundations in the field.
Here, 'marsh stability' describes the ability of marsh surface and margin substrates to resist the shear and vertical stress applied by gravity, waves, tides, storm surges, overburden and water table changes over timescales of hours or more. In a vertical direction, we distinguish between total stress (applied stress) and effective stress (total stress minus pore water pressure). In this paper, substrate shear strength relates particularly to the bulk resistance of the marsh substrate to shear stress. Shear strength can be divided into a cohesional component (resulting from, amongst others, electrostatic forces between particles, root cohesion and apparent cohesion from suction; Barnes, 2010) and a frictional component (resulting from interparticle contacts; Knappett & Craig, 2012). Frictional strength is only important when the substrate is subjected to vertical or, in some cases, horizontal stress, and when that stress causes particle interlocking. Shear strength is distinct from erodibility, although the difference between the two terms is poorly defined in the literature (see Brooks, 2020).
Erodibility relates to the critical erosion shear stress (the substrate resistance to grain-by-grain erosion), and this is not measured in this study. Given that substrate composition (e.g. particle size, root content and organic content) will affect cohesion (Zhang & Hartge, 1990) and the degree of particle interlocking (friction; Alias et al., 2014;Kara et al., 2013), this paper aims to understand the influence of substrate composition on bulk substrate shear strength in a laboratory setting.
There comes a point where substrates are subject to shear stress which exceeds the frictional and cohesional strength of that substrate and erosion follows. It is important to understand how substrate composition affects frictional, cohesional and ultimately shear strength to explain how marshes erode and therefore how salt marsh stability may change in the future.
This paper applies a range of geotechnical tests to two sedimentologically contrasting salt marsh substrates, with the aim of better constraining the range of salt marsh substrate properties that control marsh substrate strength and deformation and showing how, within this range, variations in geotechnical properties link to substrate composition. Fundamentally, the application of such methods offers an opportunity to better understand how a marsh substrate behaves under stress application from waves, currents or water level changes (e.g. Callaghan et al., 2010;Tonelli et al., 2010) in a closely controlled environment which cannot be achieved using a shear vane or cohesive strength meter (e.g. see Watts et al., 2003).
Shear box and ring shear tests are able to determine peak and residual substrate shear strengths, respectively, in response to longerterm changes in the sediment overburden or water table (two influential factors on the stress environment which have hitherto been rarely considered in salt marsh studies). These tests also quantify the cohesional strength and the effective angle of shearing resistance (Table 1), which can be linked to salt marsh substrate composition.
While ring shear tests generally consider the response of a previously failed substrate to large shear stresses, the roots were removed for these tests in this study, allowing us to assess the influence of roots on substrate strength. Oedometer tests determine the magnitude of vertical deformation caused by a vertical effective stress and whether such deformation is either permanent or recoverable (Table 1) Finer, cohesive sediments (with a lower D 50 ) generally exhibit greater strength in shear (e.g. Ford et al., 2016;Lo et al., 2017).
Uniformity coefficient (U c , defining the range of grain sizes) A high U c (>4) indicates that there is a wide range of particle sizes within the sample. This usually allows a greater packing density, which is then reflected in a higher bulk density, lower void ratio and therefore more interparticle contacts.
Shear box Peak substrate shear strength for a given vertical effective stress (τ f ) A higher peak shear strength indicates that the substrate can resist a higher applied shear stress.
Effective cohesion (c 0 ) This reflects the soil shear strength when exposed to neither vertical stress nor confining pressure (Head, 1981) and reflects the electrostatic forces between particles.
Effective angle of shearing resistance (φ 0 ) The component of soil shear strength provided by interparticle friction (Head, 1981). This increases as interparticle contacts increase.

Ring shear
Residual shear strength for a given vertical effective stress (τ f ). This reflects the strength retained after failure.
Substrates can often withstand a lower stress (indicated by residual strength) if they have failed previously. Usually, strength reduction occurs as clay platelets reorientate to border the failure plane. As roots were removed for these tests (in line with BS1377, and unlike for the shear box tests), this reduction in strength compared to the peak shear strength was hypothesized to also reflect the loss of roots.
Residual effective angle of shearing resistance (φ r 0 ) As above, this usually reflects the component of soil shear strength provided by interparticle friction. As roots were removed for ring shear tests in this study (see above), any reduction in the residual effective angle of shearing resistance may also reflect the loss of roots.
Oedometer Axial (vertical) displacement in response to a vertically applied effective stress.
Indicates the vertical displacement (and therefore change in surface elevation) in response to a vertically applied effective stress. If this is irrecoverable (see OCR below), this has implications for marsh platform inundation frequency and sediment supply.
Change in void ratio (a proxy for bulk density) in response to this vertically applied effective stress Increased bulk density is known to increase substrate shear strength (Watts et al., 2003).
Pre-consolidation pressure (maximum previous vertical effective stress to which the soil has been subjected) In the case of salt marshes this most likely reflects the influence of desiccation on pore water pressures.
Overconsolidation ratio (OCR) Tells us about the magnitude of vertical displacement, and the substrate vertical recovery potential, following unloading. When OCR < 1, the sample is normally consolidated so has not previously been exposed to a vertical effective stress greater than that it was exposed to immediately prior to sampling. A sample is overconsolidated when OCR > 1, so the sample has previously been exposed to a higher vertical effective stress than that immediately prior to sampling. If OCR < 1, large, irrecoverable vertical displacement is expected. If OCR > 1, small, recoverable vertical displacement can be expected, until a stress is applied which exceeds the pre-consolidation pressure.
not been on understanding marsh substrate response to applied stress. Rather, previous studies have been applied to 'decompact' vertical marsh sequences for the purpose of improving sea-level reconstructions from marsh cores (Brain et al., 2011(Brain et al., , 2015(Brain et al., , 2017. By improving understanding of the links between marsh substrate composition and behaviour, this paper does not intend to mimic real-world processes, but does provide a first step towards improving the understanding of salt marsh stability. This paper characterizes the intrinsic soil geotechnical properties, then considers how these properties might affect the response of the substrate to applied stress in natural settings.

| FIELD SITES
This study focuses on two UK open coast marshes with minerogenic content >85%, as is characteristic of Northwest European marshes (Allen, 2000). However, the two marshes have contrasting clay, silt and sand contents.

| Tillingham marsh
Tillingham marsh is located on the central section of the Dengie Peninsula, Essex, UK ( Figure 1) and reaches 800 m wide. The marsh-mudflat transition is characterized by a well-developed, shorenormal ridge-runnel morphology (Schuerch et al., 2019), with mudflats stretching approximately 4 km offshore. Over the past 100-150 years, the Dengie marshes have demonstrated dynamic behaviour, with evidence of both advancing and retreating phases (Harmsworth & Long, 1986;Pye & French, 1993;van der Wal & Pye, 2004).
Tillingham marsh sediments are clay/silt-dominated (Ford et al., 2016). The near-horizontal vegetated platform lies at an eleva-  ODN). Such water levels would have exposed the marsh surface to hydrostatic forces (the force induced by water loading on the submerged marsh surface) of up to 14 kPa.
However, the relatively rapid (minutes to hours) application of this hydrostatic force means that no vertical deformation would be expected, unless full hydraulic disconnectivity were to occur at the marsh surface.

| Warton marsh
Warton marsh is located in Morecambe Bay (Figure 1 'Undisturbed' samples were taken at 0-30 and 30-60 cm depth at each marsh edge site and also at 0-30 cm depth at each tidal flat site. The sampling procedure followed the block sampling method described in BS5930:2015(BSI, 2015 and is depicted in Figure 3. While it is recognized that no sample can be fully undisturbed, this method is known to minimize disturbance as much as possible compared to other methods (see Carr et al., 2020) and samples were therefore suitable for geotechnical testing. All samples were stored at 5 C prior to analysis.

| Characterizing substrate composition
Particle size analysis was undertaken on all bulk and undisturbed samples. Organic matter was removed by treatment with 30% w/v hydrogen peroxide overnight and subsequent heating to 90 C in a water bath for minimum 2 h. Sodium hexametaphosphate (4.4% concentration) was used as a deflocculating agent. Samples were analysed using a Malvern Mastersizer 2000. The D 10 , D 50 (median) and D 60 grain sizes were used to quantify the uniformity coefficient (U c , defining the range of particle sizes) following Knappett and Craig (2012): where D 60 = 60th percentile grain size (μm), D 10 = 10th percentile grain size (μm).
To determine the root mass content, 500 g samples were wet sieved through a 425 μm sieve to separate out the roots. The residue (root mass) was then left in an oven at 40 C overnight and the sediment dried at 105 C until the masses remained constant between successive measurements. Root mass content was calculated using where m roots = root mass content as a percentage of dry mass (%), m residue = sieve residue (roots) mass after drying at 40 C (g), m totaldry = total dry root mass and dry sediment mass (the latter after drying at 105 C) (g). Some studies have heated the roots at 60 C (Howes & Teal, 1994), while others have air dried the material (Ouyang et al., 2017). This study found a balance between the drying speed, while minimizing the loss of volatile root components, which can occur at higher temperatures (Hackney & de la Cruz, 1980).

| Strength and compressibility testing
Geotechnical tests of direct shear strength, residual shear strength and compressibility were undertaken in the geotechnical soils laboratory at the BGS using a shear box, ring shear and oedometer.

| Shear box
Shear box and ring shear tests can record the drained strength of soils (i.e. where the rate of changes in total applied stress is slower than the rate of dissipation of excess pore water pressure). Therefore, they provide vital information on the response of the marsh substrate to cumulative long-term changes in the total applied stress (longer than approximately a week, or the time over which it takes excess pore water pressures to dissipate). Theoretically, changes in the applied stress could relate to stress changes induced by erosion of the overburden (although such surface erosion is known to be negligible on salt marsh surfaces; Spencer et al., 2015), or changes in the elevation of the water table. The effect of these longer-term processes is rarely considered in salt marsh studies. Therefore, while this paper does not intend to mimic real-world erosion processes, it aims to understand the behavioural properties of marsh substrates, while incorporating other, rarely studied methods of stress application. Shear box and ring shear tests were therefore performed on the samples to understand factors which may be contributing to substrate strength and to understand the influence of the longer-term stress environment (e.g. from changes in the water table), rather than to indicate when and why particular erosion processes might occur.
Undisturbed field samples were subsampled in the laboratory to fit flush to a 100 mm Â 100 mm Â 20 mm shear box, following the undisturbed sampling technique in Figure 3. Each full shear box test required three duplicate specimens, which were restrained laterally by the apparatus, subjected to a vertical load and then sheared along a horizontal plane. Specimens were consolidated overnight and sheared the following day, then replaced with the next specimen for the next test stage. All tests were undertaken using a Wykeham Farrance Model No. 25402 shear box.
The shear box test method followed the consolidated/drained method in BS1377-7 (BSI, 1990a) and Head (1994). Samples were saturated using distilled water and the shear rate for each test stage was dictated by T 100 (the time at which the sample would reach 100% consolidation, assuming that the consolidation rate during primary consolidation continued throughout the test; Bishop & Henkel, 1962). The precise shear rate was determined by first using T 100 to calculate the maximum time to failure (t f ), based on Gibson and Henkel (1954): where t f = time to failure (min), T 100 = theoretical time to 100% consolidation (min).
The likely horizontal shear deformation (mm) of the sample at the point of failure was then found based on values in Head (1994) for the sediment type in question. Likely horizontal shear deformation was therefore taken to be 10 mm for samples from Tillingham (a plastic clay) and 2 mm for samples from Warton (a dense sand). This likely horizontal shear deformation was divided by t f to give the maximum displacement rate (mm/min) for use in the test. This displacement rate was between 0.001 and 0.1 mm/min, in accordance with Bolton's (1991) recommendations for clay and silt-dominated substrates under fully drained conditions. The displacement rate was therefore sufficiently slow to allow drainage and prevent the build-up of excess pore water pressure, allowing the assumption of zero pore water pressure and constant vertical effective stress, σ n ', throughout each test. The test was therefore a drained test where σ n ' is equal to the total stress, σ n . Although the samples were taken from <60 cm depth, the applied vertical stress was restricted by the minimum stress which could be applied in the shear box. Vertical total stresses (σ n ) therefore started at 10 kPa for the first specimen, increasing to 20 and 40 kPa for the second and third specimens. This typified effective stresses equivalent to 1, 2 and 4 m water overburden (assuming full hydraulic disconnectivity and rapid dissipation of pore water pressures), or approximately 0.5, 1 and 2 m sediment overburden, respectively.
The shear stress (τ, kPa), applied to the shear surface at all times throughout the shear stage, was calculated using where P = horizontal shear force (load ring calibration multiplied by load dial reading, N) applied to the specimen and A = initial plan area of the specimen (mm 2 ).
Stress-strain curves were produced from which the peak shear strength (τ f , kPa) was derived. Where the material did not reach τ f , the final shear stress reached was taken as τ f . The derived τ f values for each specimen were then plotted against vertical effective stress (σ n , kPa) to construct a failure envelope determined by the Mohr-Coulomb equation: where τ f = shear strength (kPa), c' = effective cohesion (kPa),

| Ring shear
The ring shear apparatus comprised an annular cell (height = 5 mm, inner radius = 35 mm, outer radius = 50 mm). Ring shear samples were prepared from the trimmings of the second shear box specimen for each location/depth. Trimmings at their natural moisture content were sieved at 1.18 mm to remove most roots and any larger particles, in accordance with BS1377-7 (BSI, 1990a). The sample was carefully smeared into the annulus using a thin-tipped palette knife, saturated with distilled water and consolidated at the lowest vertical total stress (14.4 kPa).
The test procedure followed BS1377-7 (BSI, 1990aHead, 1994) and was undertaken using the lowest vertical total stresses allowed by the machine (14.4, 26.6 and 51.1 kPa for stages one, two and three, respectively). The same sample remained in the apparatus for all three consolidation and shear stages, so only one sample was needed to obtain all three peak shear strength values.
After each stage, the average shear stress (τ, kPa) applied over the shear plane was calculated using where L = distance between the two points of force application on the tension beam (mm), F 1 = final force on dial 1 (N), F 2 = final force on dial 2 (N), r 1 = inner radius of test specimen (mm), r 2 = outer radius of test specimen (mm).
The residual angle of effective shearing resistance was calculated as in the shear box test.

| Oedometer
Samples were tested on a GDS Automatic Oedometer System (GDS Instruments Ltd, 2014). Tests were undertaken in accordance with BS1377-5 (BSI, 1990b and Head (1994), and duplicate tests were run on two 'replicate' samples from each undisturbed sampling location, where possible. Samples were saturated with distilled water.
The one-dimensional consolidation test was run as an eight-stage axial (vertical) load-unload test comprising six consolidation stages, with a maximum load of 400 kPa, followed by two unloading stages ( where H s = equivalent height of solid particles (mm).
From this, the new void ratio at the end of a loading stage (e), defined as the initial void ratio (e 0 ) minus the change in void ratio (Δe), could be calculated: The overall results of each test were displayed as graphs of void ratio (e) against vertical effective stress (σ'). The e-log σ' relationship was used to derive the maximum previous stress to which the sediment had been subjected. This was then used to determine the overconsolidation ratio (OCR; see Knappett & Craig, 2012) using Equation (10). This highlighted whether structural changes due to applied effective stress would be reversible or irreversible: where P c = maximum previous vertical effective stress to which the soil has been subjected (pre-consolidation pressure, kPa), P 0 = in-situ effective stress (kPa) to which the soil was exposed upon excavation. whereas U c was higher (p = 0.028) at Tillingham (mean = 13.00) than at Warton (mean = 6.00), based on the Kruskal-Wallis test (Kruskal & Wallis, 1952).
Root mass content was higher and more variable at Tillingham (mean = 0.78%; 1 SD = 0.61%) than at Warton (mean = 0.23%; 1 SD = 0.30%). While root mass at Warton decreased clearly with depth, the change in root mass was less consistent with depth at Tillingham ( Figure 4).

| Shear behaviour
Stress-strain curves indicated that the substrate at both Tillingham and Warton required constant application of shear stress to produce deformation ( Figure 5), although the substrate begins to deform within seconds of the stress being applied. Mohr-Coulomb failure envelopes from the shear box tests denoted a low c' (<7.0 kPa) at both sites (Table 3). The φ' was higher at Warton than Tillingham for all equivalent locations, ranging between 29.9 and 36.1 at Tillingham, but between 33.4 and 43.5 at Warton (Table 3).
At both sites, the difference between φ' and φ r ' was greatest at 0-30 cm depth at the marsh edge site ( F I G U R E 4 Root mass variation with depth at both Tillingham and Warton. Samples are plotted as the depth of the upper limit of the sample (e.g. 0-10 cm sample is plotted at 0 cm) F I G U R E 5 Stress-strain curves for Tillingham (a-c) and Warton (d-f) shear box samples expected, the shear stress a sample could withstand for a given vertical effective stress was lower for the residual strength than for the peak strength by 0.1-25 kPa depending on the sampling location and vertical effective stress ( Figure 6). This is denoted by the lower φ r ' compared to φ' at all sites (Table 3).

| Consolidation behaviour
Tillingham oedometer samples underwent little initial axial (vertical) displacement in each stage (<5 mm), sometimes for up to 3 min after initial effective stress application. The consolidation curve then began T A B L E 3 Angle of internal friction (φ', ), effective cohesion (c 0 , kPa) and residual angle of internal friction (φ r 0 , ), as determined by shear box and ring shear tests. The difference between the angle of internal friction and the residual angle of internal friction (φ' -φ r ') is also shown. All values are given to one decimal place  as expected (with slow initial compression, rapid 'primary' compression, followed by slow 'secondary' compression), albeit delayed ( Figure 7). At Warton, consolidation in each loading stage occurred rapidly (generally under 5 s) after the initial load application, and samples consolidated less than at Tillingham (Figure 8). For example, after application of the maximum pressure (400 kPa), no Warton samples consolidated by more than 3 mm, while up to just over 8 mm consolidation occurred at Tillingham for the same load (Figures 7 and 8).
The initial void ratio at Tillingham (1.82-2.77) was higher than at Warton (0.98-1.15; Figure 9). During unloading, the Tillingham samples dilated, but the initial void ratio was not reached with any sample and only 51-66% of the initial void ratio was recovered. The Warton samples dilated during unloading, with all samples except those from 30 to 60 cm depth at the marsh edge returning to their initial void ratio (Figure 9).
At Warton, the pre-consolidation pressure was higher, particularly at 30-60 cm depth at the marsh edge, and on the tidal flat. This resulted in a larger OCR at Warton (mean of 10.75, compared to mean of 7.03 at Tillingham; Table 4).
Few other studies report c' and φ' for salt marsh sediments. Only one other marsh/tidal flat has been tested using shear box and ring shear tests: Bothkennar, Scotland. The results presented here are comparable to estuarine silts at Bothkennar, where c' = 1.5 kPa, φ' = 34 and φ r ' = 30 (Hight et al., 2003;Lehane & Jardine, 1992). The angle of shearing resistance for both peak and residual values was therefore high at the Venice Lagoon, Bothkennar, Warton and Tillingham, in the context of known φ' and φ r ' values in the geotechnical literature. This may indicate a high angular grain content and low platy clay content (Giang et al., 2017). However, these studies did not discuss the relevance of these results for marsh stability further.
The higher φ' at Warton reflects the lower void ratio (Figure 9) and results in a consistently higher shear strength at Warton than at equivalent locations at Tillingham. A lower void ratio increases φ' because it indicates a higher packing density, more interparticle contacts and greater interlocking between particles (Barnes, 2010). While both larger particle sizes and a higher U c (the latter not a characteristic of Warton sediments) have been shown to increase the φ' and thus the peak substrate strength (Alias et al., 2014;Kara et al., 2013), not all studies reach the same conclusion (e.g. Latha & Sitharam, 2008). Rather, many studies infer that the void ratio may be the most important influence on φ', with larger particle sizes and a higher U c both reducing the void ratio (Latha & Sitharam, 2008;Selig & Roner, 1987).
For example, Vangla and Latha (2015) note that particle size does not affect peak shear strength when the initial void ratio remains constant. This paper therefore supports existing research (Crooks & Pye, 2000;Crooks et al., 2002), which uses separate density and undrained shear strength methods, in finding that a lower void ratio (or higher bulk density) increases the shear strength of marsh substrates.
At Warton, the higher φ' on the marsh surface compared to φ' at depth likely reflects compaction of the upper marsh stratigraphy, possibly due to grazing (e.g. Lambert, 2000). Compaction increases grain interlocking and friction between particle contacts, leading to higher values of φ' (Barnes, 2010). Grazing-induced compaction may therefore locally increase the substrate shear strength in the uppermost layers. While some autocompaction may occur in deeper sediment layers (Allen, 1999), the results of this study are consistent with studies elsewhere that have shown that grazing-induced compaction at depth (below 20 cm) can be negligible or absent (Elschot et al., 2013).
The higher φ' on the marsh surface compared to φ' at depth could also be related to changes in moisture content and therefore pore water pressure, caused by root suction, desiccation or reductions in the level of the water table. However, further research would be needed to assess the relative influence of compaction and changes in moisture content on φ'.

| Peak and residual shear strength
Substrates can demonstrate a reduction in strength following previous failure, usually reflecting clay softening/reorientation to the failure plane (Knappett & Craig, 2012). This was not the case in the substrates tested here, as there was no reduction in τ past the point at which the maximum strain was reached ( Figure 5) roots. Comparing φ r ' and φ' thus begins to separate the influence of clay softening/reorientation from that of root presence. The fact that the largest decrease in φ' from the peak to the residual scenario at F I G U R E 9 Pressure applied (kPa) vs. void ratio (e), where a higher void ratio indicates a lower bulk density at the end of each stage for samples from Tillingham (a) and Warton (b) both Tillingham and Warton occurred at the marsh surface (Table 3), where root mass was greatest (Figure 4), corroborates the hypothesis that roots are contributing to this loss of frictional strength.
The ability to isolate the strength provided by roots, as demonstrated through the techniques deployed in this study, is key to identifying the response of marsh substrates to forcing. While many studies have identified the role of roots in increasing the strength of the uppermost marsh substrate layer (e.g. Bendoni et al., 2014Bendoni et al., , 2016Van Eerdt, 1985a), no studies have previously isolated the strength derived from roots in a three-dimensional salt marsh substrate sample.
This research gap has limited our ability to understand the link between marsh edge erosion processes and retreat rates.
Previous studies have explored the tensile strength of the roots in vegetated substrates, and found that this may reflect vegetation density and type (Van Eerdt, 1985a) and that it affects the morphology of tension cracks which are narrower and shallower than in unvegetated substrates (Francalanci et al., 2013). Van Eerdt (  Future marsh evolution models should incorporate such vertical variations in shear strength and recognize the relative contributions made by root systems. Pollen and Simon (2005) and Simon and Pollen (2006) developed a model for the prediction of critical conditions for riverbank stability on the basis of vertically varying controls (e.g. sediment and vegetation composition) that determine the likely mechanism of failure (e.g. planar or cantilever failure) following undercutting. What is needed now is a similar approach for saltmarsh margins. We have considerable data on the critical erosion shear stress of marsh substrates (e.g. Watts et al., 2003) and therefore the likelihood and potential rate of undercutting of the marsh platform.
However, more geotechnical studies such as that presented in this paper are needed to understand controls on substrate failure type, timing and rate following undercutting, across a range of substrate types, marsh margin morphologies and hydrodynamic settings. Our results can constrain the geotechnical behaviour of individual marsh sediment layers and, in combination with further empirical evidence for margin failure processes, lead to an improved understanding of the processes involved in marsh margin failure (e.g. tension crack formation and cantilever or toppling failure) and the influence of marsh composition on why certain processes occur. This will ultimately allow the salt marsh community to produce similar failure models to those provided for riverbank failure whereby the Mohr-Coulomb failure envelope is modified to include the root reinforcement of soil strength (e.g. Pollen & Simon, 2005;Simon & Pollen, 2006). Using such a model will allow the vertical extent and rate of undercutting to be varied, and the resulting influence on the failure processes and rates to be compared.

| Consolidation behaviour
One mechanism of wetland elevation change (in particular in response to storms) is the process referred to by Cahoon (2006) as 'soil shrinkage', which refers to the expulsion of water from voids within the soil. In the geotechnical engineering literature, this is known as 'consolidation'. It is important to note that this is distinct from the process of compaction (expulsion of air from voids) or the process of 'autocompaction' noted in salt marshes, with the latter denoting the compression of the substrate under its own weight and accompanying increase in dry density (see Allen, 2000) due to a combination of consolidation (expulsion of liquid) as well as compaction (expulsion of air). This distinction is important as it has implications for the hydraulic conductivity (and therefore drainage) of the soil and must be taken on board by the salt marsh community to achieve accurate interpretation of observed compression of marsh substrates.

| Timing/magnitude of expected displacement
The results of this study show that the rate of marsh substrate response to applied vertical effective stress varies between locations.
While axial displacement at Tillingham was delayed (by up to 3 min; Figure 7), it occurred rapidly at Warton following initial load application (generally in under 5 s; Figure 8). This likely reflects the higher permeability of the Warton substrate, caused by the higher sand and lower clay content, allowing faster drainage and consolidation (Head, 1994). The slower consolidation response at Tillingham may reflect resistance provided by roots, which, once crushed, allow consolidation to continue at a faster rate. The composition (organic and minerogenic) of a substrate can therefore influence the rate of the substrate response to applied vertical effective stress.
Samples at Warton consolidated less (up to 15% of initial height) than at Tillingham (up to 45% of initial height). This likely reflects the higher initial void ratio at Tillingham (1.82-2.77) than at Warton (0.98-1.15), and the higher root content at Tillingham. In marsh substrates on the Mississippi Deltaic Plain with organic matter contents of 2.36-5.37% by volume (Nyman et al., 1990), roots retain moisture and often form a macropore structure within the marsh substrate (DeLaune et al., 1994), especially those roots which are hollow, such as S. anglica roots. Roots can thus provide a space for water and air, increasing the void ratio. Finally, roots have some elasticity and can be compressed more than solid particles when loaded (Head, 1994). Furthermore, in Cowpen Marsh, Tees Estuary, UK, Brain et al. (2011) found Although the weight of water applies a total vertical stress on the marsh surface, the applied stress will be taken up by additional pore water pressure, and not changes to the substrate structure. The only scenarios which would allow deformation to occur would be if there were full hydraulic disconnectivity at the marsh surface, or if the grain size were large enough to provide free drainage (neither of which is likely to be achieved in salt marsh settings). Therefore, at least at Tillingham and Warton marshes, deformation due to the consolidation process (and therefore the 'soil shrinkage' process referred to in Cahoon, 2006) seems unlikely.
While Cahoon (2006) proposes various mechanisms (sediment erosion/deposition, compaction, root decomposition/growth, soil swelling and folding of the root mat) for why marsh platform elevation changes may occur, the precise mechanisms are never able to be fully constrained, as the surface elevation table apparatus groups processes into surface and below-ground surfaces. Specific mechanisms should, however, be tested on substrates in the laboratory, and this is often easier than directly observing the impact of storm surges in the field.
These should be tested on both minerogenic and organic substrates.
In so doing, we will be able to better understand the response of salt marsh substrates to loading during storm surges.

| Vertical recovery potential following unloading
The pre-consolidation pressure at Warton was higher than at Tillingham, implying that Warton sediments had previously been exposed to higher vertical effective stresses. The sediments were therefore lightly to heavily overconsolidated at Tillingham and heavily overconsolidated at Warton (OCR > 3; see Barnes, 2010), similar to the overconsolidation found by Brain et al. (2011) at Cowpen Marsh, UK. Such overconsolidation resulting from the high pre-consolidation pressure could occur due to vegetation removing moisture from the soil, alongside tidal exposure and a reduction in groundwater level, with the latter allowing desiccation and producing capillary suction stresses (Brain et al., 2011). The higher pre-consolidation pressures at Warton may reflect the coarser grain size and the higher marsh platform elevation within the tidal frame compared to Tillingham, which result in faster drainage and a longer duration and frequency of subaerial exposure, enhancing desiccation. During summer, neap tides rarely reach the marsh surface at Warton, allowing desiccation to occur, alongside a reduction in water table elevation (Barras & Paul, 2000). The combination of coarser grain size and a higher elevation within the tidal frame at Warton has thus enhanced desiccation and increased the pre-consolidation pressure and the OCR.
The pre-consolidation pressures at both Tillingham and Warton are unlikely to be reached by storm surges recorded near these sites.
While substrate deformation in response to a storm surge is unlikely (as it would require full hydraulic disconnectivity at the marsh surface), any deformation would be fully recoverable if the applied vertical load did not exceed the pre-consolidation pressure. The high preconsolidation stresses in the substrates tested here therefore imply that vertical total stresses applied by storm surges would not be able to produce irrecoverable compression. As such, the desiccation pro- Incomplete recovery will lower the marsh within the tidal frame and expose it to higher, more frequent, hydrostatic pressures from increased water depths. The increased flooding frequency may also increase sediment supply and near-surface accretion. There is therefore a need to understand the pre-consolidation pressure and OCR of salt marsh substrates, in order to understand the degree to which present marsh substrates are consolidated, how they may consolidate under future vertical loads, and whether they would remain consolidated after load removal. Understanding these three aspects of substrate characteristics and behaviour will allow for better near-future forecasts of marsh vertical stability.
It is clear from the results of this study that desiccation of Tillingham and Warton marsh substrates resulted in lower magnitudes of vertical displacement during loading than if desiccation had not taken place, and that this vertical displacement was also recoverable. Desiccation processes may thus increase marsh resistance to compression.
The desiccation processes that generated the high pre-consolidation stresses will also have applied suction to the substrate, reducing the void ratio. As a reduced void ratio is linked to an increased substrate strength, desiccation processes may increase resistance to both compression and shear failure processes. Desiccation processes and the resultant suction stresses should therefore be included in marsh evolution models, due to both their influence on marsh substrate shear strength and also the resistance of the marsh to compression. By measuring suction and understanding the OCR and pre-consolidation pressure of a marsh substrate then feeding this data into models, autocompaction models could more accurately indicate both the magnitude of autocompaction and also whether the compression is recov-

| CONCLUSIONS
Shear box and ring shear tests have allowed the analysis of salt marsh and tidal flat substrate response to the application of shear stress under laboratory-controlled drained conditions. The substrates tested here required continued shear stress application to produce deformation. These tests allowed separation of the cohesional and frictional components of shear strength and therefore provided an insight into possible controls on the shear strength of these substrates (particle size, the shape of the particle size distribution curve and void ratio).
By comparing peak and residual shear strength, and the peak and residual friction angle, these tests allowed the separation of the influence of sediment and roots on marsh substrate shear strength.
The application of oedometer tests to salt marsh substrates has provided insight into both the magnitude of axial displacement and the recovery potential of marsh substrates following application of a vertical effective stress. Both the expected magnitude of displacement and recovery potential of a marsh substrate were affected by past stress conditions on the marsh (particularly due to desiccation). The high pre-consolidation stresses measured here highlight that, even under the worst possible conditions (full hydraulic disconnectivity and rapid dissipation of pore water pressures), vertical total stresses applied by storm surges would not be able to produce irrecoverable displacement, as has been observed in some US marshes.
Future work is now needed to assess the heterogeneity in marsh substrate behaviour in response to applied shear stress across space.
Furthermore, linking the shear box and ring shear data to tensile strength measurements of roots could provide evidence as to why the greatest reduction in strength between the peak and the residual scenarios is identified at the marsh surface. While this study assessed the impact of monotonic stresses on substrate shear strength, future work could look at cyclic application of shear stresses. This would more closely resemble the forces experienced in the field, mimicking the cyclic oscillations of pressure induced by changes in incident waves and tidal levels. Combining the drained shear box and ring shear tests with undrained strength measurements would also improve understanding of marsh substrate behaviour in response to applied stress, and thus marsh change over longer timescales, than the nearinstantaneous timescale associated with in-situ undrained measurements.
While this research did not intend to mimic real-world processes, the understanding of soil geotechnical properties will allow improvement of models of marsh edge erosion processes in the future (e.g. Bendoni et al., 2014), ultimately improving our understanding of these processes and, therefore, marsh stability. Finally, the measurements reported in this paper will allow for future models of marsh geomorphological evolution to better represent these dynamic responses, along with their implications for the hydrological, sedimentological and ecological function of marsh systems.

DATA AVAILABILITY STATEMENT
Datasets related to this paper will be accessible from PANGAEA data repository as soon as a the DOI is available.