Sediment Exchange Across Coastal Barrier Landscapes Alters Ecosystem Extents

Barrier coastlines and their associated ecosystems are rapidly changing. Barrier islands/spits, marshes, bays, and coastal forests are all thought to be intricately coupled, yet an understanding of how morphologic change in one part of the system affects the system altogether remains limited. Here we explore how sediment exchange controls the migration of different ecosystem boundaries and ecosystem extent over time using a new coupled model framework that connects components of the entire barrier landscape, from the ocean shoreface to mainland forest. In our experiments, landward barrier migration is the primary cause of back‐barrier marsh loss, while periods of barrier stability can allow for recovery of back‐barrier marsh extent. Although sea‐level rise exerts a dominant control on the extent of most ecosystems, we unexpectedly find that, for undeveloped barriers, bay extent is largely insensitive to sea‐level rise because increased landward barrier migration (bay narrowing) offsets increased marsh edge erosion (bay widening).

• Cross-landscape interactions are explored in new model framework connecting entire barrier system from shoreface to mainland forest • Simulations suggest barrier migration leads to rapid back-barrier marsh loss, while barrier stability enables back-barrier marsh recovery • Bay extent is insensitive to sea-level rise rates because of offsetting responses from barrier migration and marsh-edge erosion processes

Supporting Information:
Supporting Information may be found in the online version of this article. Sediment connectivity between ecosystems is critical in determining the evolution of ecosystem boundaries and their influence on migration rate and ecosystem extent. The width of the back-barrier environment depends on the trajectory of barrier transgression relative to the slope of the underlying substrate , and the size of fine-grained deposits associated with back-barrier environments can affect the rate of barrier migration (Brenner et al., 2015). Barrier overwash may provide an important source of sediment for adjacent back-barrier marshes to maintain elevation relative to sea level (Lorenzo-Trueba & Mariotti, 2017;Walters et al., 2014). However, barrier rollover via overwash can also result in lateral back-barrier marsh loss (Deaton et al., 2017). Additionally, because wider (narrower) bays produce larger (smaller) waves, a positive feedback arises that tends to lead to back-barrier bays either being marsh-filled or devoid of marsh . Further, change in mainland marsh extent depends on the slope of the adjacent forest, such that shallow-sloping forest uplands favor net expansion of mainland marsh even under accelerated RSLR (Kirwan et al., 2016).
Each of the ecosystems found in barrier systems are typically modeled in isolation or in ways that consider connections between two adjacent ecosystems via sediment fluxes (e.g., Lauzon et al., 2018;Mariotti & Carr, 2014;Mariotti & Fagherazzi, 2010;Walters et al., 2014). To investigate the role of coupled interactions across the barrier landscape as a whole, we developed a new model framework, BarrierBMFT, that connects the entire barrier system from the shoreface to mainland forest. Using this model, we ran experiments across a range of RSLR rates and external suspended sediment concentrations to explore how competition between the migration of different ecosystem boundaries influences the extent of each ecosystem over timescales of decades to centuries.

Model Development
BarrierBMFT (Figure 1) is an exploratory model framework that couples Barrier3D (I. R. B. Reeves et al., 2021), a spatially-explicit model of barrier evolution, with the Coastal Landscape Transect model (CoLT; Valentine et al., 2023), which integrates carbon cycling dynamics into the Bay-Marsh-Forest Transect Model (BMFT) of Kirwan et al. (2016). See Text S1 in Supporting Information S1 for a detailed description of the model framework, boundary conditions, and its components.

Barrier3D and CoLT
Barrier3D simulates the cross-shore and alongshore topographic evolution of a barrier segment ( Figure 1a) in response to RSLR and individual storms, including dune, overwash, and shoreline dynamics. Barrier3D assumes invariant beach width and slope. Dunes in the model grow logistically (Houser et al., 2015), with the shape of the logistic curve controlled by , the characteristic dune growth rate for the barrier segment. Probabilistically simulated storm events lower dune heights and result in overwash deposition landward of the dune crest, which can extend the back-barrier shoreline landward. RSLR and the cumulative volume of sediment removed from the upper shoreface by overwash contribute to ocean shoreline change. As such, barriers in the model tend to migrate landward when dunes are low but are stationary when dunes are tall.
CoLT simulates topographic evolution and carbon accumulation across a cross-shore transect connecting bay, marsh, and forest ecosystems ( Figure 1b). The position of the marsh-bay boundary is modeled as the balance between progradation (via deposition of sediment in the bay at the marsh margin) and erosion (via wind wave action) of the marsh edge (Mariotti & Fagherazzi, 2010). Bay depth evolves according to the mass balance between bay fetch, internal marsh-bay sediment exchange, and the suspended sediment concentration of an external source (SSC x ) that represents tidal inlet or riverine exchange with the back-barrier system (Mariotti & Carr, 2014). Marsh platform accretion in the model depends on the sum of mineral and organic contributions of mud-sized particles, which varies according to the distance from the marsh edge, local suspended sediment concentration at the marsh edge, and autochthonous organic matter production. The position of the marsh-forest boundary migrates passively as a function of upland slope and RSLR.

BarrierBMFT
In the BarrierBMFT coupled model framework, two CoLT simulations drive evolution of back-barrier marsh, bay, mainland marsh, and forest ecosystems, and a Barrier3D simulation drives evolution of barrier and back-barrier  (Figure 1c). As these model components simultaneously advance, they dynamically evolve together by sharing information annually to capture the effects of key cross-landscape couplings. In the CoLT simulations, the change in location of one marsh edge modifies the fetch and bay depth felt by the opposite marsh edge such that lateral erosion or progradation of one marsh ecosystem influences the other nonadjacent marsh ecosystem. Storm overwash deposition in Barrier3D influences organic content of marsh soils and alters the morphology of back-barrier marsh and bay ecosystems in CoLT. Dune dynamics in Barrier3D regulate the volume of overwash transported to the back-barrier marsh and bay ecosystems, and back-barrier marsh morphology influences the volume and transport of overwash deposition in ways that affect barrier geometry and ocean shoreline change rates. Overwash deposition tends to shift the barrier-marsh and marsh-bay boundaries landward. On the other hand, marsh expansion into the barrier uplands in CoLT from passive flooding of the barrier interior via RSLR tends to shift the barrier-marsh boundary seaward. While not the focus of this work, carbon dynamics of the entire coupled coastal system are ripe for future exploration with BarrierBMFT.

Model Simulations
By simulating across wide ranges of key input parameter values and given the simplified nature of model parameterizations, our experiments investigate coupled dynamics relevant to most coastal barrier systems. To ground the model in reality and provide a common starting point for our simulations, we parameterize the initial morphology ( Figure S1 in Supporting Information S1) and storm climatology for Smith Island and Smith Island Bay within the Virginia Coast Reserve (VCR), USA ( Figure 1d). Smith Island is a narrow, low-lying barrier representative of typical undeveloped transgressive barriers on the US East Coast, with a long-term retreat rate of 6 m/yr. Initial barrier morphology is derived from a digital elevation model (NOAA, 2017) for a portion of Smith Island discretized into 10-by-10 m cells. Initial marsh widths of 500 m, for both the back-barrier and mainland marsh, are based on estimates from aerial imagery. We use an empirical record of storm activity for the VCR (I. Reeves et al., 2022) processed through a copula-based multivariate sea storm model (Wahl et al., 2016) to produce the suite of 10,000 synthetic storms used in Barrier3D. See Table S1 in Supporting Information S1 for all input values used in our experiments.
To explore how the evolution of each coastal ecosystem influences the extent of the others over time, we ran simulations across a range of RSLR (3-15 mm/yr) and SSC x (40-80 mg/L) values, parameters critical to barrier and marsh migration and horizontal and vertical dynamics of the marsh platform. Although the model can be used to study regressive barrier systems, we focus on transgressive systems for consistency with current and projected behavior of most undeveloped barrier coastlines under accelerated global RSLR. Each simulation lasts 400 years and uses a mainland forest slope of 0.005, typical of barrier coastlines (Kirwan et al., 2016). To account for storm stochasticity within BarrierBMFT, we ran each parameter combination 50 times and report the mean of these 50 simulations as the result. Additionally, we ran this suite of simulations with a relatively fast characteristic dune growth rate ( = 0.75 yr −1 ), for which dunes tend to be at or near their maximum height for a majority of the simulation, and a relatively slow characteristic dune growth rate ( = 0.45 yr −1 ), for which dunes tend to be near their minimum height. In Figures S2-S5 in Supporting Information S1, we present the sensitivity of our results to several input parameters, demonstrating that varying their values produces minimal quantitative differences in our results and does not change the fundamental conclusions we draw from them.

Results
Average change in planimetric width over the simulation duration varies across each of the ecosystems represented in BarrierBMFT and across RSLR-SSC x parameter space ( Figure 2). Increases in barrier width ( Figure 2a) tend to be greater with lower RSLR rates and higher SSC x . Lower RSLR rates facilitate wider barriers by driving less ocean shoreline erosion and passive flooding of the back-barrier shoreline (I. R. B. , while barrier width tends to be greater at higher SSC x because back-barrier deposition is enhanced, which reduces accommodation needed for overwash to extend the back-barrier shoreline landward. Following Kirwan et al. (2016), mainland marsh tends to expand landward at higher rates of RSLR (Figure 2d), as rates of upland migration are more sensitive to RSLR than rates of edge erosion in shallow-sloped environments like barrier systems. Higher SSC x on the other hand leads to increased deposition at the seaward marsh boundary and therefore also favors wider marshes. However, low suspended sediment supply across a marsh that has widened extensively by high RSLR rates (e.g., SSC x = 40 mg/L, RSLR = 15 mm/yr, Figure 2d) can lead to drowning of the mainland marsh interior and limited net gain or even net loss in mainland marsh extent.
Forest width declines as a function of RSLR (Figure 2e) as the intertidal-subaerial margin passively migrates up the slope of the mainland. With a mainland slope representative of barrier coastlines (0.005), the width of the coastal zone (distance from the ocean shoreline to mainland marsh-forest boundary) tends to widen as marsh expansion into the shallow forest uplands outpaces coastal squeeze produced by landward barrier migration ( Figure 2f); widening of the coastal zone is enhanced at higher RSLR rates. With a steeper mainland slope more representative of glaciated coastlines (m = 0.01; Kirwan et al., 2016), the coastal zone tends to narrow because of limited mainland marsh expansion into adjacent forest ( Figure S2 in Supporting Information S1).
Back-barrier marsh extent decreases relative to the initial conditions across the parameter space (Figure 2b). The universal loss in the back-barrier marsh extent in our simulations is attributed primarily to barrier migration: back-barrier marsh is progressively buried by overwash deposition as a transgressive barrier moves landward, and is unable to compensate for this loss at the barrier-marsh margin by prograding into the bay. The loss of back-barrier marsh extent, however, tends to be less severe at lower RSLR rates and higher SSC x (Figure 2b). Back-barrier marsh extent after 400 years, on average, is greatest at lower rates of RSLR and higher SSC x because the ratio of accommodation space being filled to accommodation being created is greater, allowing the marsh to prograde. As explained further below, the loss of back-barrier marsh extent is also less severe for faster dune growth rates ( Figure S3 in Supporting Information S1). For all other ecosystems, the slower dune growth scenario yields negligible differences in ecosystem extent compared to the faster scenario.
The characteristic dune growth rate affects back-barrier marsh extent by regulating the rate and timing of marsh burial via overwash (Figure 3). With relatively fast dune growth rates that enable dune heights to recover rapidly following storms (Figure 3a), the tendency for long periods of barrier stability provides windows for marsh recovery following periods of loss via barrier migration (Figure 3b). Marsh recovery can occur both by progradation of the marsh edge into the bay, which tends to occur with high suspended sediment concentrations and low RSLR rates, and by passive flooding of the barrier interior from RSLR, which is dependent on the slope of the barrier interior. Alternating periods of barrier stability and transgression lead to alternating periods of back-barrier marsh contraction and expansion (Figures 3a and 3b). This is demonstrated by histograms of average annual dune height and annual marsh extent of simulations with high SSC x (80 mg/L) and low RSLR (3 mm/yr; Figures 3c and 3d) for the faster and slower dune growth scenarios. Relatively fast dune growth rates result in a bimodal distribution of average annual dune heights (Figure 3c), as well as a roughly uniform distribution of annual marsh extent ranging from 0 to approximately 600 m (Figure 3d). This indicates that the width of the back-barrier marsh oscillates over time in response to discontinuous barrier migration (I. R. B. Reeves et al., 2021). With relatively slow dune growth rates, consistently low dunes ( Figure 3c) and continuous barrier migration result in a distribution of annual marsh extent skewed toward 0 m (Figure 3d). Thus with slower growing dunes, the persistence or recovery of back-barrier marsh is significantly limited in our simulations even under conditions of high SSC x and low RSLR.
Our simulations also suggest that bay extent is primarily dependent on SSC x rather than RSLR (Figure 2c).
Higher SSC x leads to bay narrowing as depositional marsh processes dominate, while lower SSC x leads to bay widening as erosional marsh processes dominate. Bay extent is less sensitive to RSLR because of competition between the migration of the mainland and back-barrier marsh edges. While higher rates of RSLR tend to increase mainland marsh edge erosion and drowning, which widens the bay (Figure 4a), higher rates of RSLR also increase barrier migration rates, which tend to narrow the bay (Figure 4b). In our model simulations, therefore, the tendency for higher rates of RSLR to widen the bay via enhanced erosion of the mainland marsh edge is offset by the tendency for higher rates of RSLR to narrow the bay via faster barrier migration.

Discussion and Conclusions
The goal of this work is to explore and explain the complex behavior of large-scale barrier systems induced by couplings between ecosystems across the landscape, rather than to predict the evolution of a particular location or the effects of specific conditions. In support of this goal, our coupled model uses broad and simplified parameterizations to represent the compound effects of real-world processes. As a result, this modeling approach inevitably and intentionally neglects some noteworthy processes. For example, BarrierBMFT does not take into account couplings between back-barrier stratigraphy and barrier migration, and instead assumes that all sediment exposed on the barrier shoreface is 100% sand. In reality, sources of finer back-barrier sediment that outcrop on the shoreface can accelerate barrier migration patterns (Brenner et al., 2015). Similarly, BarrierBMFT assumes the coastal landscape contains no lithified geologic structures or immobile clasts that could impede migration of ecosystem boundaries. While the model resolves the impacts of individual storms on the barrier, it does not consider impacts from discrete storm events on the back-barrier ecosystems, which can produce nonlinear shifts in back-barrier ecosystem boundaries and sediment transport (e.g., Carr et al., 2020;Turner et al., 2006). Further, the model does not incorporate the dynamic exchange of organic matter from the marsh to the bay, and instead assumes that any organic matter eroded from the marsh in excess of 5% of the total eroded mass is lost via decomposition or dispersal (e.g., Spivak et al., 2019), despite the fact that allochthonous organic matter in some cases can be recalcitrant and thousands of years old (Hopkinson et al., 2018;Van de Broek et al., 2018). The effects of seagrass in the bay on associated ecosystems (e.g., I. R. B. Reeves et al., 2020) are also not considered in the model. While explicitly including the processes mentioned above would add layers of complexity to our results, we believe our model parameterizations capture the fundamental dynamics of barrier systems.
Our model results suggest that barrier migration is a primary cause of back-barrier marsh loss in coastal barrier systems. Over the broad ranges of parameter values used, our simulations suggest back-barrier marsh extent is more sensitive to rates of barrier transgression than rates of marsh edge erosion or progradation. These results are supported by observations from Deaton et al. (2017), who find that a majority of all marsh loss in the VCR from 1860 to 2010 can be attributed to barrier migration. However, this coupling can be expected to be minimal along barriers where sediment connectivity is limited by anthropogenic (e.g., Rogers et al., 2015) or ecologic (e.g., I. R. B.  factors. Tall dunes also limit sediment connectivity between the barrier and back-barrier marsh. In our model experiments, intermittent periods of tall dunes and resultant barrier immobility can allow for back-barrier marsh recovery, through both passive flooding of the barrier interior and marsh progradation into the bay (Lorenzo-Trueba & Mariotti, 2017). This suggests that back-barrier marshes are dynamic features of barrier systems that depend on periods of relative barrier stability. While overwash deposition is known to provide intertidal platforms for back-barrier marsh colonization and development (e.g., Osgood et al., 1995;Tyler & Zieman, 1999), our model suggests that back-barrier marshes created in this way will be short-lived because the same conditions that favor their creation also favor their destruction. Walters et al. (2014) found statistically significant peaks of negligible (0-100 m) and narrow (150-700 m) back-barrier marsh widths in the VCR, as exemplified in aerial imagery (Figures 3e and 3f). These observations align with the range of back-barrier marsh widths simulated in the relatively fast dune growth scenario (0-750 m; Figure 3d). Previous modeling studies (LorenzoTrueba & Mariotti, 2017;Walters et al., 2014) have proposed that barrier overwash maintains back-barrier marshes within this observed narrow width range by providing additional sediment to the marsh platform. Our model results offer an alternate, though not exclusive, explanation: that observations of narrow marshes in the VCR could be attributed to discontinuous barrier retreat (Lorenzo-Trueba & Ashton, 2014; I. R. B. Reeves et al., 2021), whereby back-barrier marsh expansion that occurs during periods of barrier stability is counterbalanced by loss during periods of transgression.
Although RSLR has a primary influence on barrier, marsh, and forest extent, we find that bay width is relatively insensitive to RSLR. Back-barrier suspended sediment supply is the dominant control of bay extent over time in our model, because barrier response to RSLR tends to offset the mainland marsh-edge response. Although this finding is counterintuitive, it is supported by a near zero net change in tidal prism among the barriers of the VCR from 1860 to 2010 despite an average landward barrier migration rate of 5.1 m/yr (Deaton et al., 2017). In environments with little to no potential for upland marsh migration, such as those with relatively steep mainland slopes like the VCR (Deaton et al., 2017) or anthropogenic barriers to upland flooding (Kirwan et al., 2016), this offset effect will come to an end when the mainland marsh eventually disappears. Whenever this occurs, bay extent will tend to decline with increasing rates of RSLR as barrier transgression rates increase (Figure 4a). Conversely, in systems where the barrier remains relatively immobile for centuries (e.g., because of significant development or "geomorphic capital;" Mariotti & Hein, 2022), bays will tend to widen in response to RSLR without barrier migration to offset mainland marsh erosion. For shallow-sloping coastal plains typical of barrier systems, our results suggest that marsh migration into mainland forests tends to outpace landward barrier migration rates, leading to overall widening of the coastal zone ( Figure 2f). While higher rates of RSLR might be expected to enhance coastal squeeze, our results indicate that higher RSLR rates may instead enlarge the coastal zone of most barrier systems.
Our model findings underscore the importance of cross-landscape sediment transport couplings in coastal barrier systems. While their importance has been previously suggested (e.g., Deaton et al., 2017;Fitzgerald et al., 2018), cross-landscape couplings remain largely conceptual because observations are typically made at the scale of individual ecosystems and are difficult to quantify experimentally at the scale of entire landscapes. Creative empirical measurements (e.g., overwash sediment flux into subaerial environments) are needed to resolve couplings across real-world barrier landscapes and test model hypotheses. As such, our exploratory model serves an important role in capturing the most basic cross-landscape couplings of barrier coastlines, and indicates that barrier processes could account for a significant proportion of change in future extent of adjacent and non-adjacent associated ecosystems.