In-situ loading experiments reveal how the subsurface affects coastal marsh survival

Coastal marshes are key vegetated landforms within tidal environments that provide multiple ecosystem services, yet they have experienced substantial global losses in recent decades. Their long-term survival depends on maintaining elevation within the intertidal zone by keeping pace with relative sea-level (RSL) rise through vertical accretion driven by physical sediment deposition, biological productivity, and chemical processes. Autocompaction of shallow, recently deposited, unconsolidated sediments is a critical component that reduces net elevation gains and contributes to RSL rise. Although SET-based monitoring and geotechnical modeling have revealed that subsurface processes strongly influence marsh surface elevation, major uncertainties remain regarding the magnitude of natural loads that drive autocompaction, the impact of subsurface heterogeneity on compaction variability, and nonlinear feedbacks between marsh processes. To address these gaps, the study introduces a novel in-situ, meter-scale loading experiment in two Venice Lagoon marshes to directly measure controlled-stress subsurface deformation and pore-pressure responses, enabling characterization of autocompaction dynamics under near-field, undisturbed conditions.

Study design and sites: A meter-scale in-situ loading experiment was developed to emulate a one-dimensional (oedometric) compression test in the field, targeting the upper ~1 m where most marsh autocompaction occurs. Experiments were executed at two sites in the Venice Lagoon, Italy: (1) Lazzaretto Nuovo (LN; northern basin; platform elevation ~0.55 m above msl) in July 2019, and (2) La Grisa (LG; southern basin; ~0.5 m above msl) in October–November 2020. LN displacements were logged manually; LG used automatic 1-min logging. Loading apparatus and protocol: Eight 500-L polyethylene tanks (0.78 × 0.69 × 1.04 m) were arranged in two rows on wooden pallets over a reinforced geotextile, forming a loaded area of ~2.25 × 1.75 m (~4.0 m²). Tanks were interconnected at the base with plastic tubes to ensure equal water levels and uniform loading using seawater pumped from the nearest creek. The maximum cumulative load was ~40 kN, producing vertical effective stresses from ~1–2 kPa up to ~10–11.3 kPa at full load. Two to four loading–unloading cycles were performed with increasing tank filling, maintaining each load for >20–24 h to capture elastic, plastic, and viscous (creep) responses. A temporary high tide at LN partially submerged the platform during full load. Displacement monitoring: Five vertical displacement transducers were deployed within the central area between tank columns: CO at the surface and C10, M10, E10 at 0.1 m depth (center, mid, edge positions respectively), plus a deeper sensor below the load center (C50 at 0.5 m at LN; C40 at 0.4 m at LG due to a stiff silty-sand layer). Sensor placement captured the expected deformation bowl with maximum settlement under the load center. A rigid H-shaped steel reference frame for transducers was anchored to over-consolidated Pleistocene sediments using two steel piles with tips at ~6 m depth. High-precision spirit leveling monitored the reference frame and pallet corners; a distant benchmark (~50 m away, ~6 m depth) provided stability control. Pore-pressure monitoring: Five water-pressure loggers were installed along a line orthogonal to the displacement array at depths 0.2 m (P20A, P20B), 0.5 m (P50A, P50B), and 1.0 m (P100). An additional logger was placed several meters away to capture ambient tidal effects. Pressure logging recorded tidal influences and small perturbations caused by (un)loading operations. Site stratigraphy and coring: Multiple cores (1–5 m) along transects (LN ~200 m; LG ~300 m) characterized depositional environments and subsurface heterogeneity. LN’s upper 1 m comprises ~0.8 m homogeneous silty clay overlain by ~0.2 m silty clay to clay with sparse plant remains (minerogenic-dominated accretion). LG’s upper 1 m includes a basal 0.45 m laminated very fine sand to silty clay (fining upward), overlain by 0.25 m silty sand and 0.10 m silt with plant remains, capped by ~0.2 m organic clay rich in roots (organogenic-dominated accretion). Laboratory testing: Undisturbed samples from the upper 1 m were subjected to standard geotechnical characterization: grain size by sieving and hydrometer (ASTM D6913; ASTM D7928), organic matter by loss-on-ignition (ASTM D7348), one-dimensional oedometer compression with wet saturation and double drainage, and permeability via falling head during oedometer stages. LN oedometer samples at 0.2, 0.3, 0.45 m; LG at 0.16, 0.65, 0.8 m. Compression (Cc), recompression (Cr), virgin compression slope (Co), tangent modulus M, and hydraulic conductivity k were derived across stresses ~3–100 kPa. Environmental controls: Tidal stage was measured in an adjacent channel and referenced to the stable benchmark; a boardwalk minimized anthropogenic disturbance. At LN, thunderstorms interrupted measurements and required partial reinstallation and sensor resets; LG completed continuous logging over Oct 27–Nov 2, 2020. Data handling: Field settlements at specified depths were converted to deformations (percent strain over intervals 0–0.1 m and 0.1–0.5/0.4 m) for comparison with lab oedometer-derived strains at similar stress increments.

• Autocompaction magnitude and variability: Under similar applied loads, minerogenic LN experienced maximum settlements of ~6–7 mm at the surface/0.1 m (CO, C10, M10) under ~11.3 kPa sustained ~22 h, whereas the organogenic, root-rich LG marsh reached ~32 mm surface settlement after ~72 h at ~11.3 kPa. This demonstrates strong spatial variability controlled by subsurface composition and heterogeneity.
• Depth-dependent deformation: At LN (full load), settlements were ~6–7 mm (CO/C10/M10), 2.5 mm (C50), and 0.7 mm (E10). At LG: cumulative settlements relative to onset were 32 mm (CO), 18 mm (C10), 15 mm (M10), 1 mm (C40), 2 mm (E10). Rebounds after unloading were ~50–60% of peak at LN; at LG they were ~50% (CO), 40% (C10), 60% (M10), −7% (C40), 67% (E10), leaving permanent settlements of ~22, 13, 10, 2, and 2 mm respectively.
• Pore-pressure dynamics: Tidal submergence produced major pressure peaks at all depths; (un)loading operations induced smaller perturbations: ~0.05–0.06 m at 0.2 m depth and ~0.02 m at 1.0 m at LN; up to ~0.04 m at LG. Dissipation times increased with depth (LN: ~1 h at 0.2 m vs 3–4 h at 1.0 m; LG: ~2–3 h). Sensors at similar depths showed differing responses at LG, reflecting strong lithologic heterogeneity.
• Creep behavior: With constant load after pore-pressure dissipation, clear secondary (viscous) compression was recorded for the first time in marsh soils. At LG, creep coefficients were Ca ≈ 2.5×10⁻2 (0–0.1 m) and Ca ≈ 1.3×10⁻2 (0.1–0.4 m), about an order of magnitude higher than values inferred for nearby engineered embankment trials (10⁻3–4×10⁻3).
• Elastic vs plastic components and recoverability: The ratio of compressibility during loading to unloading (s) was ~2 in situ, indicating unexpectedly high recoverability compared with laboratory Cr/Cc inferences (s ~10) and with shallow lagoon soils elsewhere (~15). This is attributed to reinforcement by intact root and rhizome networks in the upper ~0.20–0.25 m.
• Field–lab consistency and heterogeneity effects: Field strains over 0–0.1 m and 0.1–0.5/0.4 m intervals align with lab oedometer strains at similar stresses, validating the oedometric (1-D) assumption for the meter-scale setup. However, displacement ratios with depth deviated strongly from homogeneous elastic predictions (e.g., CO to C40 ratio ~15 at LG vs theoretical ~1.2), underscoring vertical heterogeneity.
• Geotechnical properties: LN soils were silty with LOI ~10–15% in top 0.2 m, k ≈ 7×10⁻9–2×10⁻8 m s⁻1 at 20 kPa, and Co ≈ 0.4 (0.2 m) decreasing to ~0.1 (0.3–0.45 m). LG had high sand fractions (60–85%) in mid-depths, shallow LOI averaging ~20% (peaks 25%), and k from 5×10⁻10 m s⁻1 (0.16 m) to ~10⁻8 m s⁻1 (0.65–0.8 m).

The experiments directly address how subsurface properties govern marsh surface elevation change under additional loads simulating sediment accretion. Results show that autocompaction is a first-order control on marsh elevation trajectories and varies substantially across short spatial scales due to lithological and biological heterogeneity. Organogenic, root-rich upper layers exhibit large primary and secondary compression but also surprising recoverability during unloading, likely due to in situ vegetation reinforcement. The minerogenic LN site, with silty clays and low organic matter, compacted much less under the same load. Pore-pressure responses and dissipation times reflect depth-dependent hydraulic conductivity and compressibility, confirming saturation and rapid equilibration under small experimental load perturbations relative to tides. The validated field–lab consistency supports using the meter-scale in situ oedometric framework to derive depth-averaged compressibility for use in biomorpho-geomechanical models. These insights challenge models that neglect autocompaction or assume homogeneity, as elevation gains from accretion can be largely offset by compaction depending on subsurface composition. Incorporating subsurface heterogeneity, vegetation effects on stiffness, and creep is essential to more realistically forecast marsh resilience to RSL rise and to evaluate restoration strategies involving sediment additions.

This study introduces a novel, cost-effective in-situ loading experiment that captures the hydro-geomechanical behavior of tidal marsh subsurface under controlled stresses at meter scale. Applied at two contrasting Venice Lagoon marshes, the method reveals pronounced differences in autocompaction (≈6–7 mm vs ≈32 mm under ~11.3 kPa), strong depth and lateral heterogeneity, significant creep, and unexpectedly high recoverability in root-reinforced shallow soils. Field measurements align with laboratory oedometric behavior, validating the 1-D assumption and enabling depth-averaged parameterization. The findings highlight that predictions of marsh elevation change and resilience to RSL rise must explicitly include autocompaction and subsurface heterogeneity. Future work should (i) expand testing across diverse geomorphic settings and seasons, (ii) couple in situ tests with high-resolution subsurface imaging and root biomass mapping, (iii) integrate creep and vegetation effects into physics-based marsh evolution models, and (iv) evaluate the effectiveness of sediment-based restoration by quantifying subsurface responses prior to interventions.

• Environmental interruptions: Severe thunderstorms disrupted LN measurements, causing gaps, sensor resets, and partial reconstruction of the setup.
• Temporal scope: Experiments spanned days, capturing primary consolidation and early secondary compression but not longer-term multi-seasonal or interannual behavior.
• Depth coverage: Displacement sensors were limited to 0.5 m depth at LN and 0.4 m at LG due to a stiff layer; deeper contributions to compaction (>0.5–1 m) were not directly measured.
• Load range and area: Applied stresses were modest (~1–11.3 kPa) over ~4 m²; responses to larger or more spatially extensive loads were not tested.
• Instrument adjustments: Pore-pressure sensors required initial equilibration to field saturation; some sensors at similar depths (LG) recorded divergent behaviors due to small-scale heterogeneity.
• Site specificity and heterogeneity: Results reflect two sites with distinct depositional histories; lateral variability within and among marshes may limit generalizability without broader sampling.
• Tidal influence: High tides submerged the LN platform during full load, slightly modifying effective stress on tanks and background pressures.