Evaluating the Long-Term Effectiveness of Marsh Terracing for Conservation with Integrated Geospatial and Wetland Simulation Modeling

Abstract

Coastal marshes provide essential ecosystem services, yet they are vulnerable to anthropogenic stressors and climate change, particularly sea level rise (SLR). Restoration approaches like marsh terracing have emerged as nature-based strategies to enhance resilience and reduce habitat loss. This study applies the Sea Level Affecting Marshes Model (SLAMM) to assess the potential of marsh terraces to mitigate future losses, while also examining the model’s limitations, including its assumptions and capacity to reflect complex marsh processes. A geospatial approach was used to generate 3D representations of terraces through morphostatic modeling within digital elevation models (DEMs). Under a no-restoration scenario, SLAMM projections show that all marshes analyzed are at risk of total loss by 2100. In contrast, scenarios including terracing demonstrate a delay in net marsh loss, extending the persistence of key marsh habitats by approximately a decade. Although marsh degradation remains likely under high SLR conditions, the results underscore the utility of marsh terraces in prolonging habitat stability. Additionally, the study demonstrates the feasibility of integrating restoration features like terraces into DEMs and wetland models. Despite SLAMM’s simplified erosion and accretion assumptions, the model yields important insights into restoration effectiveness and long-term marsh dynamics, informing more adaptive, forward-looking coastal management strategies.

1. Introduction

Estuarine and coastal ecosystems (ECEs), including tidal wetlands such as salt, brackish, and freshwater marshes, are dynamic systems located at the land–sea interface that provide essential ecosystem services [1]. As such, these natural buffers perform a wide array of critical regulatory, habitat, and productivity functions, such as protecting shorelines from erosion, mitigating floods, and improving water quality by filtering pollutants and nutrients [2,3]. Globally, functional coastal wetlands are estimated to provide ecosystem services—such as flood control, climate regulation, natural hazard mitigation, water purification, and soil formation—worth approximately $194,000 (USD) per hectare annually [4]. In the United States alone, these services are particularly vital to the 40% of the 2020 U.S. population living in coastal areas [5], and their contribution to flood and storm protection delivers an estimated annual benefit of $23 billion to coastal communities [4].

Despite their importance, ECEs are among the most vulnerable and heavily utilized natural systems worldwide [6,7,8]. Globally, approximately 50% of salt marshes, 35% of mangroves, 30% of coral reefs, and 29% of seagrasses have been lost or degraded [9,10,11,12,13], resulting in the decline of key ecosystem services such as fisheries productivity, nursery habitats, and water filtration and detoxification functions [7]. This degradation has also led to biodiversity loss, increased biological invasions, diminished water quality, and weakened coastal defenses against flooding and storms [14,15,16]. Climate change, particularly accelerated sea level rise (SLR), poses an additional threat to tidal marshes, with sea levels projected to increase by 30–100 cm by 2100 [17]. In certain areas, rising sea levels may cause tidal marsh submergence and habitat migration as salt marshes shift landward and upward with tidal range, encroaching upon and replacing tidal freshwater and brackish marshes [18,19]. This decline in tidal marsh areas and alteration of habitat types may result in significant changes to the ecosystem services these wetlands provide, with potential ramifications for coastal resilience and community well-being. This has led to an urgent need for restoration efforts to enhance coastal resilience and recover critical ecosystem services.

Restoring marsh habitats can deliver significant benefits, including improved water quality, reduced flood impacts, and the preservation of biodiversity, yet these benefits are sensitive to life history, genetic and phenotypic variation, adaptive capacity, and rates of environmental change [20]. Such restoration efforts not only enhance habitat connectivity for fish and wildlife but also bolster recreational opportunities by creating vibrant natural areas for birding, fishing, and other nature-based activities. Among various restoration techniques, marsh terracing has emerged as a promising nature-based solution (NbS) for combating marsh loss [21]. Marsh terracing involves constructing narrow ridges of soil in open water areas to reduce fetch and lower significant wave height, thereby dissipating wind-driven wave energy, encouraging sediment deposition, and facilitating the re-establishment of emergent marsh vegetation [22,23]. Despite its demonstrated success in the northern Gulf of Mexico, its application in other regions remains relatively unexplored.

Despite the growing interest in marsh terracing as a restoration technique, research on its hydrodynamic performance, longevity, and capacity to restore marsh ecosystems remains limited [22]. Existing studies have employed a range of approaches to evaluate terrace performance. For instance, remote sensing techniques have been used to assess terrace dynamics [24], while wave energy in marsh terraces was modeled using the Simulating Waves Nearshore (SWAAM) tool [25]. Keller et al. [26] used Delft3D Flexible Mesh numerical modeling to identify optimal terrace configurations for restoration sites, quantifying wave attenuation and sediment deposition, and developing performance metrics for future projects.

The Sea Level Affecting Marshes Model (SLAMM) is a widely used computational tool designed to simulate the long-term impacts of sea level rise (SLR) on coastal wetlands. It integrates geospatial and environmental variables to represent dominant coastal processes and predict wetland transitions over time [27,28]. In a study of the Albemarle-Pamlico Estuarine System in North Carolina, SLAMM has undergone hindcasting and validation to assess the accuracy of the model output quality [29]. In combination with the InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) model, SLAMM has also been effectively used to simulate sea level rise scenarios and identify impacts on ecosystem services [30]. Despite its broad applicability, SLAMM and similar models have seen limited use in evaluating the hydrodynamic effects of marsh terraces and other constructed nature-based solutions (NbS), particularly in terms of their capacity to attenuate wave energy.

A key gap in the current literature is the lack of spatially explicit, quantitative modeling of designed conservation techniques—such as marsh terraces—within simulation environments that support scenario analysis, comparative evaluation, and performance assessment. While marsh terraces and living shorelines are occasionally represented in 2D GIS environments or CAD-based design tools, they are seldom incorporated into integrated modeling frameworks that simulate marsh evolution under SLR scenarios for impact analyses. Marsh terraces have been extensively implemented in the northern Gulf of Mexico, where their performance has been evaluated using aerial imagery across multiple design types and post-installation periods [24]. However, these assessments often lack predictive capacity. Although no SLAMM simulations have been found published directly on marsh terracing, SLAMM has found utility for predicting future distribution potential of marshes in Great Bay National Estuarine Research Reserve (NERR), New Hampshire, where it helped identify sites for living shorelines [31]. Another analogous case involved simulation of fringing marshes in North Carolina [28], where SLAMM was used to evaluate no-action versus intervention by dredged sediment fill and marsh planting. Although neither study analyzed SLAMM for marsh terraces, the demonstrated use and value for similar nature-based designs supports an investigation of marsh terraces.

Understanding the short- and long-term benefits of marsh terracing projects, particularly across various construction and life cycle phases, could enhance their permit review and adoption as a restoration tool. This study addresses critical knowledge gaps by developing and demonstrating a novel methodological approach that integrates 3D geospatial representations of marsh terraces into the SLAMM wetland simulation framework. The research makes three key methodological innovations: (1) the development of a comprehensive geospatial workflow for incorporating engineered restoration features into digital elevation models with appropriate ecological zone assignments; (2) the first application of SLAMM to evaluate marsh terracing effectiveness in the Mid-Atlantic region; and (3) the novel use of morphostatic terrain modeling to represent mid-channel restoration features disconnected from existing shorelines.

The study’s novelty lies in its pioneering integration of restoration engineering design with landscape-scale ecological modeling, enabling prospective evaluation of restoration performance under climate change scenarios. Unlike previous SLAMM applications that focus on natural landscape evolution, this research demonstrates how detailed engineering specifications can be translated into model-ready geospatial datasets, providing a replicable methodology for evaluating constructed nature-based solutions. This approach fills a critical gap between restoration design and long-term performance assessment, offering a screening tool for restoration planning that considers both immediate habitat creation and long-term persistence under sea level rise.

Referring to a pilot project in Back Bay National Wildlife Refuge, this research aims to inform the ecological effectiveness of marsh terracing and guide its future implementation in coastal settings. Specifically, the study pursues three main objectives: (i) Quantify ecological and conservation benefits of the proposed restoration under future SLR scenarios, including improvements in habitat longevity, shoreline erosion mitigation, and reduction in wetland to open water conversion; (ii) Evaluate the performance of a marsh simulation model in capturing the impacts of terracing, focusing on key indicators such as sediment deposition and projected morphological changes; (iii) Assess the limitations and applicability of the modeling approach, including its assumptions, scalability, and ability to accurately represent the dynamic processes of marsh ecosystems.

2. Materials and Methods

2.1. Study Site

The Back Bay National Wildlife Refuge (NWR) (latitude 36.67212, longitude −75.91564) is situated in the southeastern corner of the City of Virginia Beach, VA, USA (Figure 1a). It is bordered to the north by Little Island Park, a city park, and to the south by False Cape State Park. Established in 1983, the refuge serves to protect and provide critical habitat for migrating and wintering waterfowl. The refuge and its surrounding areas support a high diversity of plant species, which, in turn, provide essential breeding and foraging habitats for a variety of terrestrial wildlife, birds, and waterfowl. In addition to its ecological significance, the refuge is a focal point for nature-based recreation, attracting substantial visitor engagement through activities such as hiking, biking, wildlife observation, kayaking, and surf fishing. The bay has attracted a variety of interdisciplinary ecological research even as human uses and coastal development have increased [32].

The Back Bay ecosystem is particularly fragile, facing threats from both natural processes and anthropogenic disturbances. Over the past century, area coastal wetlands have been lost to development and stabilization [33], and the refuge has experienced significant ecological degradation, including the loss of more than 800 ha of marshland and approximately 70% of its submerged aquatic vegetation [34] (Figure 1b). Studies have also documented the loss of marshes and replacement by invasive Phragmites australis [35]. The primary drivers of these losses are wind tide flooding and sea level rise. Wind tide flooding occurs when strong southerly winds force water from the Currituck Sound into the refuge, inundating its marshes. Sea level rise is particularly significant to the region owing to subsidence, with vertical land motion resulting from a glacio-isostatic forebulge resulting in subsidence in excess of 2 mm/year [36]. The compounding effects of wind tide flooding and sea level rise result in frequent and severe flooding events. These environmental challenges not only threaten the integrity of marsh habitats but also exacerbate flooding in adjacent low-lying residential areas and disrupt critical transportation infrastructure. Consequently, these issues amplify the vulnerability of both ecological systems and human communities within and around the refuge.

The Virginia Coastal Resilience Master Plan [37] projected that up to 89% of the state’s existing tidal wetlands may be lost to open water by 2100 due to sea level rise, coastal development, and other stressors. To mitigate wetland loss, the City of Virginia Beach has initiated a marsh terrace project within the Southern Rivers Watershed as part of the city’s Flood Protection Program. The proposed restoration design aims to restore over 300 hectares of marsh and seagrass habitats by creating a network of approximately 40 individual marsh terraces and barrier islands within the historically degraded Bonney Cove project site within the Back Bay NWR (Figure 1).

Figure 1. (a) Study area location within Back Bay National Wildlife Refuge and City of Virginia Beach portion of the Currituck Sound Watershed. Arrows denote the broad wind tide flooding process induced by sustained southerly winds from the Currituck Sound. (b) SLAMM domain zoom view of Bonney Cove project site (~72 km² total area.) Blue polygons approximate historical wetland extent in 1868 georeferenced maps compared to yellow lines of present-day shoreline from the NOAA NGS Continually Updated Shoreline Product (CUSP) [38]. The pop-out map shows the Bonney Cove Project site. (c) The generalized wind tide flooding pathways through which floodwaters spread throughout the project site. A weather station point within the Back Bay NWR indicates the location of the data collected for (d) a wind rose diagram of wind speeds (knots) and directions in frequency (percentage of time). Observations were measured from 2015 to 2025.

Figure 1. (a) Study area location within Back Bay National Wildlife Refuge and City of Virginia Beach portion of the Currituck Sound Watershed. Arrows denote the broad wind tide flooding process induced by sustained southerly winds from the Currituck Sound. (b) SLAMM domain zoom view of Bonney Cove project site (~72 km² total area.) Blue polygons approximate historical wetland extent in 1868 georeferenced maps compared to yellow lines of present-day shoreline from the NOAA NGS Continually Updated Shoreline Product (CUSP) [38]. The pop-out map shows the Bonney Cove Project site. (c) The generalized wind tide flooding pathways through which floodwaters spread throughout the project site. A weather station point within the Back Bay NWR indicates the location of the data collected for (d) a wind rose diagram of wind speeds (knots) and directions in frequency (percentage of time). Observations were measured from 2015 to 2025.

2.2. Modeling Approach

This study adopted the Sea Level Affecting Marshes Model (SLAMM), version 6.7, developed by Warren Pinnacle Consulting, Inc. [39], to estimate the changes in the marsh landscape for every decade until 2100 under two scenarios: (1) a baseline scenario in which no modifications to the existing landscape are implemented, and (2) a terrace restoration scenario, which incorporates the addition of marsh terraces according to the proposed restoration plan. Key spatial data inputs for SLAMM include elevation, slope, initial habitat distribution, geomorphic parameters, tidal datums, water levels, and SLR projections. Additional technical details on data acquisition and processing are provided in

Table 1. Summary of geospatial data and parameter inputs used in Sea Level Affecting Marshes Model (SLAMM) simulations for the Back Bay project area.

Data Input	Source(s)	Description/Processing Summary
Digital Elevation Model (DEM) File	National Geodetic Survey [40] NOAA OCM Partners [41]	Chesapeake Bay USGS topobathy 1 m DEM in deep water and upland areas overlaid with 1 m 2019 topobathy LiDAR survey raster for low elevation coastal dry land, wetlands, and shallow waters (<~4 m depth) with newer, high-quality data. Both DEMs were downloaded in NAVD88.
SLAMM Categories File	Chesapeake Bay Program [42] Dewitz [43] U.S. Fish and Wildlife Service [44,45] Warren Pinnacle Consulting, Inc. et al. [46]	The lookup table provided in the SLAMM 6.7 program files was used to assign SLAMM codes based on National Wetland Index (NWI) categories for wetland polygons in the study area. The remaining unjoined NWI polygons were manually updated in the lookup table using the naming scheme provided in the SLAMM Technical Documentation. Dry upland (developed and undeveloped) classes were delineated using the Chesapeake Bay Foundation and National Land Cover Database LULC data products.
SLOPE File	DEM derivative grid	Derived a slope angle raster from the input DEM file, indicating each cell’s slope in degrees.
VDATUM File	DEM File NOAA NGS et al. [47]	To adjust the original DEM from geodetic vertical datum NAVD88 to the Mean Tide Level (MTL) for SLAMM modeling, the NOAA NGS VDATUM tool was used. VDATUM generated a raster of correction values by subtracting the MTL VDATUM output from the original NAVD88 DEM. Areas outside of interpolated tidal correction coverage used parameter MTL-NAVD88 (−0.083 m) as a constant correction.
Site-Specific Parameters	Dewberry et al. [27] U.S. FWS [48]	Previous SLAMM studies in the Back Bay National Wildlife Refuge (NWR) determined the parameter values for erosion, accretion, and historical SLR rates.
Tidal Datum and Water Level Parameters	NOAA NGS et al. [47] NOAA CO-OPS [49]	SLAMM parameters great diurnal tide range and MTL datum elevations relative to NAVD88 were derived from a reference NOAA tide gauge. The Sewells Point, VA, tide gauge (ID 8638610) station was selected for its proximity to the study area, its long-term establishment, and availability of tidal datum adjustments to NAVD88. Salt elevation and water level changes relative to MTL are also derived from inundation frequency data analysis measured at this gauge.
Sea Level Rise Projections	NOAA CO-OPS Sweet et al. [50] U.S. Army Corps of Engineers [51]	Sea level rise projections customized for the study site were based on Sewell’s Point and NAVD88 datum [49].

.

2.2.1. Elevation Data

The primary elevation dataset represents baseline conditions in the study area as a composite 1 m resolution topo-bathymetric digital elevation model (DEM) referenced to the North American Vertical Datum of 1988 (NAVD88) (Figure 2). This DEM was compiled from the USGS Coastal National Elevation Database (CoNED) topo-bathymetric DEM (2016) covering the Chesapeake Bay, and the 2019–2020 NOAA National Geodetic Survey (NGS) Post-Florence topo-bathymetric LiDAR Survey conducted over coastal Virginia and the Carolinas (Table 1).

To facilitate compatibility with SLAMM’s use of Mean Tidal Level (MTL) vertical datum, the DEM was transformed from NAVD88 to MTL using NOAA’s VDatum software [46]. Details on the conversion process are described in Appendix A. The baseline topo-bathymetric DEM at the Bonney Cove project site overlaid with contour lines of MTL and other relevant tidal elevations above NAVD88 specific to Sewells Point is shown in Figure 2b.

2.2.2. Wetland Data and Land Cover Classification

Wetland data representing baseline conditions (Figure 3a,b) were obtained from the National Wetlands Inventory (NWI), developed by the U.S. Fish and Wildlife Service (US FWS) [45]. Refer to

Table A1. SLAMM v. 6 wetland code and categories with corresponding NWI descriptions after Dewberry et al. [27] and Warren Pinnacle Consulting, Inc. et al. [39].

SLAMM Class	SLAMM Category	NWI Wetland Class Description
1	Developed Land	Dry upland and developed/impervious areas. Requires manual delineation with a separate dataset from NWI. By default, SLAMM assumes that developed land will be defended against sea level rise unless otherwise specified.
2	Undeveloped Land	Dry upland and undeveloped areas. Requires manual delineation with a separate dataset from NWI.
3	Non-Tidal Swamp	Palustrine non-tidal water regimes with forest and scrub-shrub (living or dead) cover.
4	Cypress Swamp	Palustrine non-tidal water regimes with needle-leaved deciduous forest and scrub-shrub (living or dead) cover.
5	Non-Tidal Inland Freshwater Marsh	Non-tidal water regimes; palustrine with emergent cover, and lacustrine and riverine systems with non-persistent emergent vegetation cover.
6	Tidally Influenced Freshwater Marsh	Palustrine and riverine freshwater tidal water regimes with emergent vegetation cover.
7	Transitional Saltmarsh/Scrub Shrub	Marsh border; Estuarine intertidal water regimes with scrub-shrub and forest cover.
8	Regularly-Flooded Saltmarsh	Low marsh; Estuarine regularly flooded intertidal water regimes with emergent marsh vegetation cover.
10	Estuarine Beach	Estuarine intertidal water regimes with unconsolidated shores.
11	Tidal Flat	Estuarine intertidal water regimes with unconsolidated shores (mud or organic) and aquatic beds, and marine intertidal water regimes with aquatic beds.
12	Ocean Beach	Marine intertidal water regimes with unconsolidated shores (cobble-gravel, sand).
13	Ocean Flat	Marine intertidal water regimes with unconsolidated shores (mud or organic), low energy coastline.
15	Inland Open Water	Palustrine, lacustrine, and riverine systems with unconsolidated bottoms and aquatic beds.
16	Riverine Tidal	Open water riverine systems.
17	Estuarine Open Water	Estuarine subtidal open water regimes.
18	Tidal Creek	Estuarine intertidal water regimes with streambeds.
19	Open Ocean	Marine subtidal and intertidal water regimes with aquatic beds and reefs.
20	Irregularly-Flooded Saltmarsh	High marsh; Estuarine irregularly flooded intertidal water regimes with emergent marsh vegetation cover.
22	Inland Shore	Shoreline not pre-processed using tidal range elevations; Palustrine, lacustrine, and riverine systems with unconsolidated bottoms, rocky shores, and streambeds.
23	Tidally Influenced Swamp	Palustrine tidally influenced swamp water regimes with scrub-shrub and forest cover.

and the accompanying text in Appendix B for additional information on the NWI dataset, the NWI to SLAMM class conversion process, and descriptions of the land cover categories. Dry upland areas, absent from the NWI dataset, were classified as either developed or undeveloped land primarily with the high-resolution (1 m) Chesapeake Bay Land Use and Land Cover (LULC) 2022 dataset [42]. The National Land Cover Database (NLCD) 2021 Land Cover dataset [43] validated these upland classifications and addressed any remaining gaps.

2.2.3. Sediment Change Rates

Sediment change rates were derived from previous studies conducted at Back Bay NWR, which estimated these parameters using regional research at sites with similar vegetation. Erosion and accretion rate parameters were primarily sourced from Dewberry et al. [27], supplemented by additional values from the U.S. Fish and Wildlife Service et al. [48]. Where multiple sources reported values for the same parameter, precedence was given to the most recent study. A summary of the sediment change rate parameters is provided in

Table 2. SLAMM site parameter inputs and source literature.

Parameter		Value	Source Summary
NWI Photo Date (YYYY)		2009	True Color, 1 m, 2009 U.S. National Agriculture Imagery Program (NAIP) Source Imagery for U.S. FWS NWI dataset within the study area [44].
DEM Date (YYYY)		2019	2019–2020 NOAA NGS Topo-bathy Lidar DEM: Coastal VA, NC, SC [40]
Direction Offshore [n,s,e,w]		East	Direction of open ocean water from shoreline
Erosion Rate (Horizontal m/yr)	Marsh	0.06	The value used by Dewberry et al. [27] for the Back Bay NWR based on an averaged long-term marsh erosion value obtained from the Virginia Institute of Marine Science (VIMS) Shoreline Evolution study of the City of Virginia Beach [51].
	Swamp
	Tidal Flat
Accretion rate (mm/yr)	Reg. Flooded Marsh	3.7	The value from Dewberry et al. [27] and the U.S. FWS [48] for the Back Bay NWR based on Sedimentation Erosion Table (SET) data collected in Cedar Island, NC.
	Irreg. Flooded Marsh
	Tidal-Fresh Marsh
	Inland Fresh Marsh
	Tidal Swamp	1.1	The value used by the U.S. FWS [48] for the Back Bay NWR based on an average of fresh wetland accretion rates within the region.
	Swamp	0.3
Beach Sedimentation Rate (mm/yr)		0.5	The value used by Dewberry et al. [27] and the U.S. FWS [48] for the Back Bay NWR. Also a commonly used average beach sedimentation rate in SLAMM applications.
Historic SLR Trend (mm/yr)		3.29	The value used by the U.S. FWS [48] for the Back Bay NWR based on average values recorded at the Portsmouth, VA, and Oregon Inlet, NC Tide gauges.
MTL to NAVD88 (m)		−0.083	Sewells Point, VA NOAA station tidal elevation and datum values relative to NAVD88.
Great Diurnal Tide Range (m)		0.841	Estimated from Corolla Sound, NC, and Beggars Creek Bridge and Sewells Point, VA.
Salt Elevation (m above MTL)		0.928	Inundation frequency analysis using ~20 years (2003 to 2023) of monthly highest recorded water levels.

.

2.2.4. Sea Level Rise and Tidal Parameters

Historical SLR rates were initially based on previous SLAMM applications at Back Bay NWR, which calculated the mean SLR trend between Portsmouth, Virginia (ID 8638660; 3.76 mm/year) and Oregon Inlet Marina, North Carolina (ID 8652587; 2.82 mm/year), resulting in a region SLR rate of 3.29 mm/year [48] (Table 2). Primary tidal parameters, including MTL, the Great Diurnal Tide Range (GT), and salt elevation, were derived from the NOAA Sewells Point, VA tide gauge (ID 8638610), Beggars Bridge Creek USGS gauge (ID ID0204300267), and Corolla, NC gauge (ID 02043433). Data for the present tidal epoch (1983–2001), historical SLR trends, and monthly water level records at this station were sourced from the NOAA NOS/CO-OPS database. Salt elevation and additional inundation frequency thresholds were approximated from 20 years (2003–2023) of modern tidal records [27,39], using the methods specified in Figure A1 and Figure A2, and associated text in Appendix C.

Projected relative sea level rise (RSLR) values sought a scenario accepted by local jurisdictions and permitting authorities and adopted used the Sweet et al. [50] intermediate-high scenario, incorporating global sea-level rise trends and regional vertical land movement rates (2.469 mm/year) measured at the Sewells Point, VA NOAA tide gauge (Figure 4). RSLR values were integrated into SLAMM at decadal intervals from 2020 to 2100 (

Table A2. Values for the intermediate-high sea level rise scenario in the SLAMM simulations retrieved from 2022 U.S. Army Corps of Engineers (USACE) Sea Level Change Calculator [52]. Values are relative to NAVD88 and based on the NOAA tide gauge at Sewells Point, Norfolk, estimated subsidence rate of 0.002469 mm/yr.

Year	NOAA RSLR (m, NAVD88)
2020	0.19
2030	0.33
2040	0.49
2050	0.68
2060	0.90
2070	1.14
2080	1.42
2090	1.71
2100	2.04

). Ohenen et al. [36] also estimated approximately 2 mm/year subsidence for the nearby areas, but with spatially variable uncertainties.

2.2.5. Simulation Setup and Protection Scenario

All input rasters, including the DEM, land cover, slope, and tidal corrections, were clipped to the study area boundary and, where necessary, resampled to a 1-m cell resolution for consistency. Each raster was standardized to maintain consistency in spatial dimensions (i.e., the number of rows and columns) across all datasets before being converted to ASCII format, ensuring compliance with the SLAMM computational framework. The protection scenario implemented for the simulations assumed that all land potentially vulnerable to SLR was subject to inundation or erosion processes [39]. Under this scenario, wetlands were permitted to migrate inland, and all dry upland areas—both developed and undeveloped—could transition to other habitat categories as conditions changed. Additionally, simulations did not include an input dike raster, following the assumption that the presence of dikes, levees, or other forms of coastal armoring would not protect wetlands and drylands from RSLR.

2.3. Geospatial Processing and Terrace Design Integration

A geospatial workflow developed in ArcGIS Pro ModelBuilder was used to integrate the marsh terrace designs into the site’s topo-bathymetric DEM and generate zones of appropriate tidal marsh classes surrounding the newly restored structures. This approach ensured that terrace elevations, wetland classifications, and spatial transitions were accurately represented for SLAMM simulations referencing a 95% design set [53]. Detailed geospatial pre-processing steps for input SLAMM datasets for the restoration simulation are summarized in Figure A3 in accompanying Appendix D.

The Bonney Cove project site covers about 2.3 km² of shallow open water, historically more widely covered in marsh habitat [54]. The proposed restoration design features a network of marsh terraces– elongated islands formed from dredged sediment– strategically positioned in Bonney Cove to optimize the available marsh edge habitat, which is highly valuable to aquatic and avian species (Figure 5a,b). Native marsh vegetation species (e.g., cordgrasses, needlebrush, groundsel, wax myrtle, and bald cypress), identified from local surveys, have been designated for planting plans on top of the upper dredged sediment layer of the terraces above the waterline. Terrace patterns and densities at the project site reflect design priorities for maximizing edge-to area ratios while also modifying hydrodynamics (e.g., reducing wave energy, wave heights, flow velocity) by shortening wind fetches in the prevailing southernly wind directions. Rock armor will be installed on the side slopes to protect certain terraces exposed to open water in the north (T100, T101, and T105 in Figure 5a) and the south (T135, T137–T140 in Figure 5b). Prior modeling (i.e., XBeach) of the project’s individual element designs incorporated empirical site-specific observations, show how the marsh terraces are anticipated to increase friction against the movement of flood pathways, lowering water levels north of Bonney Cove by approximately 30%, reducing wave heights within Bonney Cove by approximately 45%, and have a slight influence on decreased flow velocities expected to promote submerged aquatic vegetation (SAV) growth in between the terraces [54,55].

Terrace configurations were classified into two primary types based on flat-top width: 4.57 m and 9.14 m. A subset of terraces, termed hybrid terraces, combine both configurations with a 4.57 m top on one side and a 9.14 m top on the other (Figure 5c). Elevation and habitat dataset modifications were based on cross-section blueprints enumerating dimensions, slope gradients, and wetland classifications, as shown in Figure A4. The inset maps of Figure 5c show examples of contour lines designating extruded elevations following these design plans.

2.3.1. Wetland Zone Classification

Terrace slopes were divided into tidal wetland zones matching SLAMM classifications, including upland, high marsh, low marsh, and tidal flat zones. Elevation thresholds for these zones were determined using cross-sectional measurements and approximate tidal datum elevations (details in Figure A4). A buffering method was applied to delineate wetland boundaries based on distance from the terrace tops (see Figure A3 and Figure A4, and associated text in Appendix D). These buffers were subsequently used to generate SLAMM input files for habitat classes under restoration conditions, with polygons representing upland, high marsh, low marsh, and tidal flat zones (Figure 6a,b). Figure 6c compares the SLAMM class distribution over the different terrace configurations. After completing DEM modifications, these data were finalized for compatibility with the SLAMM modeling framework.

2.3.2. Interpolation and DEM Modification

To facilitate the accurate integration of terrace features, the original topo-bathymetric DEM was resampled from a 1 m cell size to a higher 0.30 m (1 ft) cell size, aligning with the high resolution and unit increments specified in the design set. To incorporate terrace features into the DEM, linear interpolation methods were applied to create new elevation data, as detailed in Figure A3, Figure A4 and Figure A5, and accompanying text in Appendix D. The modified topo-bathymetric DEM featured terrace elevations smoothly integrated into the underlying baseline natural bathymetry (Figure 7a,b). A magnified view of the extruded elevations displaying the different terrace configurations is shown in Figure 7c. The final DEM, representing conditions with marsh terraces, was used to generate the slope and VDATUM correction data inputs for the SLAMM simulation under the restoration scenario. For this second model run, values used in the site parameters and sea level rise projections (see Section 2.2.3 and Section 2.2.4) remained consistent with the baseline simulation.

3. Results

Results suggest that marsh migration follows a dynamic and nonlinear trajectory in the study area. Model results show that irregularly flooded marshes, which experience infrequent wind-driven inundation, will transition into regularly flooded marshes as rising water levels increase tidal exchange and saturate the soil more frequently (Figure 8a). Over time, the prolonged submersion will lead to the conversion of regularly flooded marshes into transitional marshes. As inundations continue owing to sea level rise, these transition zones gradually erode and diminish in areal extent, ultimately transforming to mudflats, which are largely devoid of vegetation and subject to continual inundation (Figure 8b). Irregularly and regularly flooded marsh area will decrease at high rate, as a part of the marsh evolution, while the low tidal zone and mud flats would initially increase 563% by 2040 and then drop by 76.5% by 2070 and will virtually disappear by 2100 (

Table 3. Anticipated shifts in land use and land cover types between 2020 and 2100, assuming no restoration efforts. Percentage changes are calculated relative to the baseline year of 2020. These percentage values are visually represented using color coding, with losses indicated in red and gains in green.

Cover Class	km2 by Decade						% Change by Decade
	2020	2040	2050	2070	2090	2100	2040	2050	2070	2090	2100
Low Tidal Zone and Mud Flats	0.294	1.952	0.536	0.069	0.008	<0.001	563.9	82.3	76.5	97.3	99.9
Irregularly Flooded Marsh	1.074	0.007	0.004	0	0	0	99.3	99.6	100	100	100
Regularly Flooded Marsh	1.094	0.113	0.074	0.044	0.002	<0.001	89.7	93.2	95.9	99.8	99.9
Transition Salt Marsh	0.197	0.082	0.051	<0.001	0	0	58.4	74.1	99.9	100	100

). Indeed, the only marked gains of marsh extent are evidenced in the initial construction of marsh terraces and the accommodation space they provide to colonization through 2050.

Modeling results for the baseline scenario where no terrace restoration has been implemented reveals that all the preexisting marsh classes evaluated are vulnerable to the SLR scenario used in the study and are predicted to lose 100% of their area by 2100 (Figure 9).

SLAMM results indicate that the marsh terraces will delay the net loss of habitat, thereby extending the duration of various habitat classes until the end of the century. Although the loss of the marsh is ultimately unavoidable owing to accelerating relative SLR, the terraces can be expected to prolong the marsh presence by approximately 10 years in these structures. This extension is particularly noticeable in the increase of tidal flat areas, along with some prolonged effects on different habitat classes (Figure 10). The terrace restoration project will also increase the cover of regularly flooded marsh, tidal flat and transitional salt marsh through the end of the century (

Table 4. Projected changes in land use/land cover type from initial time (2020) to 2100 with marsh terracing. Percent change calculations are based on change relative to 2020. The % change values are color-coded based on direction of change (i.e., losses in red and gains in green).

Cover Class	km2 by Decade						% Change by Decade
	2020	2040	2050	2070	2090	2100	2040	2050	2070	2090	2100
Low Tidal Zone and Mud Flats	0.329	1.980	0.562	0.094	0.025	0.017	501.8	70.8	97.1	99.2	99.4
Irregularly Flooded Marsh	1.110	0.042	0.039	0.017	0.003	0.003	96.2	96.5	98.5	99.7	99.7
Regularly Flooded Marsh	1.139	0.153	0.103	0.061	0.021	0.055	86.5	90.9	64.6	98.2	95.2
Transition Salt Marsh	0.265	0.151	0.120	0.069	0.050	0.003	43	54.7	63.8	81.1	98.8

).

Among the various land categories, irregularly flooded marshes and transitional salt marshes show the highest expansion under the restoration scenario (

Table 5. Projected percentage increase in land use/land cover type between the restoration and no-restoration conditions for the study site. The values are in hectares. The % of change values (the difference between no-restoration and restoration conditions) are color coded based on direction of change from no-restoration to restoration (i.e., gains are shown in green).

Cover Type	Scenario	2020	2050	2080	2100
Irregularly Flooded Marsh	No Restoration	107.4	0.44	0	0
	Restoration	111	3.9	0.3	0
	% Change	3.4	786.4
Regularly Flooded Marsh	No Restoration	109.4	7.4	0.2	0
	Restoration	113.9	10.1	1.8	5.2
	% Change	4.12	36.5	800
Tidal Flat	No Restoration	26.8	53.6	6.5	0
	Restoration	30.3	56.1	8.7	1.4
	% Change	13.1	4.7	33.8
Transition Salt Marsh	No Restoration	19.7	5.1	0	0
	Restoration	26.5	11.9	6.3	0.003
	% Change	34.5	133.3

). Model simulations suggest that the terrace restoration project is more effective in preserving transitory salt marsh habitat under high SLR projections compared to no restoration. Specifically, the restoration scenario results in an average increase of about 133.3 % in salt marsh area by 2050, relative to the no-restoration scenario.

4. Discussion

This study presents the first modeling evaluation of a marsh terracing restoration effort in the Mid-Atlantic region, using the Sea Level Affecting Marshes Model (SLAMM) as a screening tool to explore the long-term evolution of wetland habitats under sea level rise and subsidence. While SLAMM does not directly simulate hydrodynamic processes, it provides a useful framework for assessing potential landscape-scale changes in marsh distribution and persistence under alternative restoration scenarios. Limitations also include the representation of bay bathymetry and shorelines as morphostatic landforms, which restricts the potential of dynamic coastal erosion and sedimentation to affect marsh accretion rates on complex, spatially heterogeneous patterns. In addition, wave action and shoreline change rates are approximated, tidal ranges are estimated from nearby areas with variable tide versus wind tide predominates, and historic rates of shoreline change may be inadequate to extrapolate to future conditions with changing climate and storminess. A number of other caveats should be mentioned with SLAMM modeling. First, the wetland data used in this study may have inherited error or inaccuracy for wetland type classification or spatial positional error. Although no gross errors were obvious in our study area, some inaccuracy present in the NWI mapping could propagate into the future simulations. Second, SLAMM’s limited spatial representation of salinity gradients and shoreline erosion rates are generally static over time, whereas future SLR could affect non-linear changes in the salinity regime or erosion rates. Further, the characterization of wave action is quite limited in SLAMM, with rates being dependent on limited observations in the region and the study area not very precisely compartmentalized by fetch exposures and wave shear stresses that would be represented in a numerical wave process model. Each of these limitations should be considered in future data collection for improvement and quality assurance or control. Nonetheless, the modeling techniques and ample abundance of high resolution DEMs and bathymetry do provide a foundation for future study and improved, integrated modeling. By including marsh terracing features in the restoration scenario, the model allowed us to examine how this intervention may influence trajectories of marsh resilience, conversion, and loss.

Comparisons between de facto baseline conditions and the simulated alternative terracing scenario suggest that strategic implementation of terraces can enhance habitat stability and facilitate wetland migration, especially in areas vulnerable to submergence. These findings highlight the potential of marsh terracing to serve as a viable adaptation strategy for sustaining critical coastal ecosystems under future climate stressors. Importantly, this application of SLAMM supports early-stage restoration “spatial screening” or feasibility planning by identifying locations where marsh terracing may yield the greatest ecological conservation benefits. While more detailed hydrodynamic modeling is needed to fully capture the interaction between terraces, sediment transport, and wave energy, our approach demonstrates a novel use of SLAMM’s landscape evolution capabilities to assess the potential of strategically placed nature-based infrastructure to restore both geomorphological and ecological functions.

While previous applications of SLAMM have primarily focused on simulating marsh advancement through shoreline-adjacent interventions such as sediment augmentation (e.g., [28]) or assessing resilience of existing marsh extents, our approach simulated the establishment of marsh terraces in open-water areas disconnected from the current shoreline. Specifically, we modeled terrace construction in Bonney Cove, a site historically protected by natural barriers that have since eroded [54]. This degradation transformed the area into open water and contributed to the development of a secondary channel, intensifying hydrological connectivity between the upper and lower bays and exacerbating flood risks.

Unlike traditional shoreline restoration, the proposed offshore terrace system in Bonney Cove is designed to interrupt floodwater pathways using engineered nature-based features informed by hydraulic modeling. By incorporating these mid-channel terrace features into the SLAMM domain, our study expands the model’s application from retrospective assessments to prospective landscape planning. Although SLAMM does not simulate morphodynamic feedbacks, incorporating terrace features into the model domain, guided by engineering design specifics, allows for an anticipatory evaluation of habitat conversion potential under sea level rise scenarios. This mid-channel restoration concept thus combines ecological restoration with proactive flood mitigation, demonstrating how spatially targeted nature-based infrastructure can inform adaptive management in complex coastal environments. As previously noted, the limited characterization of waves and erosion in SLAMM could underestimate the positive effects of the terraces for their limitation of fetch in the north and south (particularly with prevailing SSW and NNE winds) which would result in lower wave heights, reduced shoreline change, and increased accretion rate. Incorporating feedback mechanisms with an external or coupled wave model, for instance, may improve this process characterization in future studies.

Prolonging the functional lifespan of marsh habitats through restoration techniques such as marsh terracing can yield significant ecological benefits [56]. Our modeling results suggest that the implementation of marsh terraces in Bonney Cove has the potential to extend marsh persistence by approximately 10 years under current sea level rise projections. These findings are based on morphostatic assumptions, which do not account for feedback between vegetation, sediment dynamics, and hydrology.

From an ecological standpoint, even modest extensions in marsh longevity can play a critical role in maintaining the integrity of coastal ecosystems. Marshes serve as essential buffers against shoreline erosion, enhance water quality by filtering pollutants and trapping sediments, and provide habitat for a wide range of species, including commercially and ecologically important fish, birds, and invertebrates [57,58]. They also support primary productivity and nutrient cycling, contributing to the resilience of adjacent aquatic and terrestrial systems. The spatial configurations and vegetation of the terraces are additionally anticipated to provide wave attenuation, supporting the establishment of SAV by decreasing turbidity, reducing bottom shear stress, and increasing light penetration, as shown in previous modeling of the project site layout using XBeach [54].

However, to maximize the ecological benefits of terracing interventions, a more comprehensive, morphodynamic approach should be considered. This would involve integrating vegetation dynamics, sediment transport, and elevation feedback into future modeling and management efforts. For example, actively planting marsh vegetation on terrace platforms and implementing routine maintenance (e.g., sediment augmentation, invasive species management) could accelerate habitat development and increase structural complexity, improving the suitability of restored areas for wildlife use and enhancing overall ecosystem function over time. By extending marsh life and functionality, restoration efforts such as terracing not only delay habitat loss but also create opportunities for marsh migration and adaptation in response to sea level rise. This underscores the importance of designing restoration projects with long-term ecological performance and adaptability in mind.

The pilot initiative in Virginia Beach aims to restore critical marsh habitats while also enhancing recreational opportunities. By expanding habitats for terrestrial wildlife and waterfowl, the project supports biodiversity conservation, offers vital breeding and foraging grounds, and promotes nature-based recreation. These co-benefits are expected to positively influence the local economy through increased ecotourism and support for small businesses. This study developed a robust modeling framework that is essential to quantify ecological effectiveness, evaluate trade-offs, and generate transferable insights. Given the novelty of marsh terracing as a restoration strategy in the Mid-Atlantic region, such an approach is particularly valuable for informing adaptive management. It provides the scientific foundation needed to evaluate performance under changing environmental conditions and to support evidence-based decision-making.

While SLAMM provides valuable insights into potential wetland evolution under sea level rise scenarios, the model operates under several key assumptions and limitations that must be acknowledged when interpreting results. SLAMM employs a morphostatic approach, treating bay bathymetry and shorelines as fixed landforms, which restricts the model’s ability to capture dynamic coastal erosion, sedimentation processes, and complex morphological feedbacks that can significantly affect marsh accretion rates in spatially heterogeneous patterns [59]. The model’s empirical framework simplifies complex biogeochemical and ecological interactions, assuming that specific vegetation types thrive within fixed elevation ranges without accounting for adaptation or species migration over time [59,60]. Additionally, SLAMM does not simulate vegetation changes that are known to occur when sedimentation exceeds rates of sea level rise, potentially leading to shoreline progradation that the model cannot capture.

Previous validation efforts have revealed important insights about SLAMM’s performance and limitations. The most comprehensive evaluation was conducted by Wu et al. [60], who used neutral models to assess SLAMM’s prediction accuracy over a 10-year period in the lower Pascagoula River basin, Mississippi. This study found that SLAMM could simulate wetland change more accurately compared to random constraint match (RCM) and growing cluster (GrC) neutral models, with higher correct predictions and lower false alarms [60]. However, the evaluation also revealed that model performance decreased as the time-scale of retrospection increased due to compounding errors, suggesting limitations in long-term projections [59]. Earlier validation work by Park et al. [61] at Pelican Pass, Louisiana, found SLAMM predictions within 1% of observed landscapes over a 13-year period, though this represents one of the few explicit accuracy assessments in the literature.

SLAMM’s treatment of critical processes also presents limitations. The model approximates wave action and shoreline change rates using simplified relationships, and tidal ranges are often estimated from nearby areas where wind-driven versus astronomical tide dominance may differ significantly from the study site. Historical rates of shoreline change used in parameterization may be inadequate for extrapolating to future conditions with changing climate and storminess patterns [59]. Furthermore, SLAMM generally lacks the ability to model SAV distributions, an important limitation given the ecological significance of these habitats (although SLAMM has been customized for at least one instance, the Yaquina Bay Estuary of Washington state [62]). The model’s morphostatic assumptions also preclude representation of barrier island migration, overwash processes, and sediment transport dynamics that can be critical in some coastal settings. In addition, modeling results for Back Bay resemble similar projected losses in the Venice Lagoon’s subtidal and intertidal zone leads to reduce geodiversity, ecosystem services, and the overall health of lagoonal systems [63]. Despite these limitations and the consequences of no-action, SLAMM remains valuable for landscape-scale screening and comparative scenario analysis, particularly when used in conjunction with local field data and expert knowledge to inform restoration planning and adaptive management strategies.

5. Conclusions

This study evaluates a marsh terrace project’s performance over future decades using wetland simulation and geospatial modeling of restoration features. Results show the value of marsh terraces in extending the longevity of wetlands in Back Bay, while the simulation approach provides insights into future geomorphic representation and modeling. While acknowledging modeling limitations to represent coastal morphodynamic behavior among storms, sedimentation, and vegetative processes, it is nonetheless notable that the methods herein show value as a screening approach to coastal landscape conservation for mitigating wetland loss and improving coastal resilience. The techniques developed for DEM terrain analysis and modification (detailed in Appendix A, Appendix B, Appendix C and Appendix D) are also extensible in other regions, especially where LiDAR and moderate to high-resolution vegetation mapping of wetlands are available.

The SLAMM allowed us to estimate the annual evolution of different marsh habitats. From these fundings future studies should consider the importance of modeling the expected outcome of restoration efforts. For instance, ecological and hydrodynamic benefits of a marsh terrace restoration project can be quantified with a variety of parameters, including potential to habitat, reduced wave energy, and mitigation of shoreline erosion. Model performance improvements may also be investigated, such as using a higher fidelity wave model to simulate fetch limitation, wave energy attenuation, and sediment deposition. In addition, the creation of relatively sheltered still waters between marsh terraces could provide accommodation space for SAV (a habitat not characterized directly in SLAMM.)