Invasive Phragmites provides superior wave and surge damage protection relative to native plants during storms

Phragmites marshes, which are found in every continent except in Antarctica, are being removed by resource managers in the US because it is considered an invasive species with little ecosystem service value. Here we present a comprehensive study on the ecosystem service value of an invasive Phragmites marsh vs a native Typha marsh for flood protection during tropical cyclones. Using a vegetation-resolving three-dimensional surge-wave model and observed vegetation and building data, we assessed the value of the Piermont Marsh in buffering Piermont Village, New York, USA from wave, flood, and structural damage during Superstorm Sandy in October 2012. Observed and simulated wind and water level data along the Hudson River were used as boundary conditions. Model results showed that the Marsh, with predominantly invasive Phragmites australis, dissipated more than half of the wave energy, but negligible flood, at the Village during Sandy. River-borne debris could not be transported across the Marsh to the Village. If Phragmites were replaced with the shorter, native cattail, Typha angustifolia, simulations of Sandy suggested that Piermont Marsh’s wave and debris buffering capacity would be preserved. However, had Sandy occurred in non-growth season when Typha is much shorter and sparser, the Marsh would be unable to buffer the wave and debris. Simulated residential structure damage during Sandy (>$10 M) agreed well with reported losses. If the Marsh were absent, the total loss would have increased by 26%. Since damage is dependent on the storm characteristics, we estimated the protective value of the Phragmites marsh for a 1% annual chance flood and wave event to be more than $2 M. This confirms the significant value of Piermont Marsh in protecting Piermont Village from flood and wave damage. To develop a balanced restoration plan, marsh managers should consider biodiversity as well as the significant ecosystem service value of Phragmites-dominated marsh for flood protection.


Introduction
Coastal wetlands are natural and nature-based features which can protect coastal communities from damage due to coastal inundation during storms, high tides, and sea-level rise (SLR). Coastal communities and government resource managers are eager to understand and take advantage of the ecosystem service value of coastal wetlands for flood protection. Recent studies (see e.g. Sheng Blake et al (2013). Image stated to be in the public domain).
property damage along the tri-state coasts. Their predictions, however, were not verified against the National Flood Insurance Flood Program (NFIP) loss payout data ($3.9 B for NJ alone) and their model did not incorporate detailed wetland data (plant species, height, and density). Using actual loss and marsh data, Lathrop et al (2019) found that the presence of Spartina marshes in south NJ had little role in buffering exposed residential structures from storm surge, due to the low (0.3-1 m) marsh height, consistent with the high NFIP payouts there. Sheng et al (2021) conducted a comprehensive study on the role of tidal marshes in affecting coastal inundation along NJ/NY/CT coasts during Sandy, a hypothetical extreme Black Swan storm, and the 1% annual chance flood and wave event.
During Sandy, residents in the Piermont Village (figure 1), approximately 15 miles upstream of New York City (NYC) along Hudson River, experienced minor flood damage (with $3.56 M NFIP payout) in comparison to other coastal communities in the tri-state region due to sheltering of the Piermont Marsh to the south. The tall (>2 m) and dense vegetation (>200 stems m −2 ) in the Marsh is predominantly (92%) the non-native invasive common reed, Phragmites australis. Phragmites grows on all continents except Antarctica and is among the most prolific invasive plants in North American wetlands, including fresh, brackish, and salt marshes located along the Hudson River (Kettenring et al 2011). Before the widespread invasion of Phragmites, cattail, Typha angustifolia, a native species that experience more significant seasonal die-back than Phragmites was regionally prevalent in brackish, tidal marshes like Piermont. Due to Phragmites' ability to alter biodiversity and ecosystem functions, marsh managers have focused on marsh restoration in Phragmites-dominated ecosystems via invasive plant removal which can be very challenging and costly (Kettenring and Adams 2011). On the other hand, numerous studies have found that Phragmites marshes do offer valuable ecosystem services by providing habitats to many fish and bird species (Benoit and Askins 1999, Weinstein and Balletto 1999, Weinstein et al 2000, Hanson et al 2002, Weis and Weis 2003, Weis 2005, Kiviat 2013. Phragmites marshes are effective in removal of nutrient (Toyama et al 2016) and heavy metal (Windham et al 2001), carbon sequestration (Lal 2004), and combating SLR (Rooth andStevenson 2000, Theuerkauf et al 2017). The role and value of Phragmites marsh in flood protection, however, is not well understood. While residents of the Piermont Village would like to keep the Piermont Marsh as a barrier to reduce storm-induced flood damage, marsh managers have been considering a potential restoration plan which may include partial Phragmites removal. Residents and managers agreed that, prior to developing any restoration plan, a study is needed to understand the role of Piermont Marsh in protecting residents of Piermont Village from storm-induced flood damage.
The physiological and phenological differences between the two species in this study, P. australis and T. angustifolia, as well as the positioning of this marsh adjacent to a built community affected by an intense storm event, make Piermont Marsh an ideal study site to investigate the role of marsh vegetation in protecting coastal communities from storminduced flooding and waves. With abundant vegetation data (plant species, distribution, height, and stem density) as well as other data (water level, wave, and damage) during Sandy, this study used a threedimensional vegetation-resolving surge-wave model CH3D-SWAN Zou 2017) to quantify the role Piermont Marsh played in buffering residents of the Village of Piermont from massive flood-related damage during Sandy and other potential storms. Value of the Piermont Marsh was assessed by comparing the effects of two vegetation types (Phragmites, Typha, and a no vegetation control) on storm surge, wave energy, flood extent, and structural damage during storm events. Following the schematics shown in figure SI 1 (available online at stacks.iop.org/ERL/16/054008/mmedia), this study focuses on the simulation of flood and wave in the high-resolution Piermont region while using results from the large-scale model simulation for the tri-state region (see SI and Sheng et al 2021 for description). The surge-wave model CH3D-SWAN and simulation results are described in sections 2-4, while the economic analysis is described in section 5, followed by conclusions in section 6.

A 3D vegetation-resolving surge-wave model
CH3D-SSMS is an integrated storm surge modeling system, incorporating a coastal surge model CH3D (Sheng 1986(Sheng , 1989) and a wave model SWAN (Booij et al 1999). CH3D and SWAN are dynamically coupled to simulate the time-varying surge and wave fields. Details of the coupling mechanism can be found in Sheng et al (2010). CH3D uses the Reynoldsaveraged Navier-Stokes (RANS) equations to compute water elevation and current velocities. To represent the vegetation-induced drag forces to the mean flow, extra profile drag D P and skin-friction drag D S are included in the RANS equations. The drag forces were formulated as (Lewellen and Sheng 1980): where i = 1, 2 refer to x and y coordinate, respectively, u i is the mean horizontal flow velocity component in the ith coordinate, C p is the profile drag coefficient, C f is the skin friction coefficient, A w is the wetted area per unit volume, A f is the frontal area per unit volume, q is the square root of twice the turbulence kinetic energy. All variables in equations (1) and (2) can vary in the vertical direction, which usually has 4-16 layers. The profile drag coefficient C p is predetermined with typical values between 0.1 and 1.0, and the skin friction coefficient C f is computed from: where ν is the kinematic viscosity, which equals 1.004 × 10 −6 m 2 s −1 , and c 1 is an empirical constant set to 0.125 (Schlichting 1968), consistent with previous studies of Sheng and Zou (2017) and Lapetina and Sheng (2015). According to the experimental study of Tanino and Nepf (2008), the profile drag coefficient C p depends on the solid volume fraction occupied by the vegetation elements ϕ, stem diameter b v , and stem Reynolds number Re c =ūb v /ν (Re c is the Reynolds number for current, andū is the mean flow), and it has an inverse relation with Re c , up to Re c = O ( 10 3 ) , then converges to a constant value. Meanwhile, Sheng and Zou (2017) found that the model results did not show high sensitivity to C p in a range from 0.1 to 0.3 which are comparable to those used by den Hartog and Shaw (1975), Uchijima and Wright (1964), and those determined by Nepf (1999) for dense vegetation in a laboratory flume. Doubling the C p values resulted in negligible changes in the simulated inundation, hence no attempt was made to implement more complex forms of C p which vary with stem density and Reynolds number based on vegetation length scale (see e.g. Mazda et al 1997, Nepf and Vivoni 2000, Wamsley et al 2010, Nepf 2012, Hu et al 2015. Marsh breakage was not considered for simplicity. In the vegetation-resolving SWAN model of Suzuki et al (2012), vegetation is treated as cylinder units, and the plant-induced forces (drag and inertia forces) are modeled using the formula of Mendez and Losada (2004) to account for irregular waves and depth-varying bottom, i.e. a vertical layer schematization for representing vegetation structure and calculating the dissipation term in the spectral action balance equation as: where ρ is the water density, C D is the bulk drag coefficient, N v is the stem density, k is the wave number, σ is the angular frequency, α v is the submergence ratio (vegetation height over water depth), h is the water depth, and H rms refers to the root-mean-square wave height. For irregular waves, σ and k are associated with the peak wave period. Ozeren et al (2014) found that the inertial force is much less significant compared with the drag force for the Keulegan-Carpenter number (KC) greater than 10, thus the ignorance of inertial force is a reasonable assumption for this study in which the KC's of all simulated cases are much greater than 10. A theoretical expression for the wave dissipation length L D when the root-mean-square wave height at the marsh edge H rms,0 is reduced by a ratio α was derived. Assuming rigid vegetation and linear wave theory in shallow waters: Just like the turbulent eddy coefficient, drag coefficient C D has no exact solution and has to be determined empirically. We considered eight empirical formulas for C D based on mostly laboratory studies with artificial and live vegetation ( Although the experiments used to fit the ensemble of empirical formulas included vegetation with stiffness from flexible to rigid and spatial scale from laboratory flume to the field, each formula is limited to certain vegetation types and hydrodynamic regimes. Before applying CH3D-SWAN to the Piermont, we conducted simulations of two laboratory experiments, one involving flow in and around vegetation (Zong and Nepf 2010), and another involving breaking wave over a vegetated sloping bed (Wu et al 2011), to validate the accuracy of the model and to determine the sensitivity of model results to drag coefficients. Moreover, the model was applied to simulate wave dissipation data measured by Garzon et al (2019) in the Chesapeake Bay region. Results of these simulations agreed very well with measured data and are presented in the supplementary information.

Study site
Piermont Marsh, one of four main tidal wetlands under the management of Hudson River National Estuarine Research Reserve (HRNERR), is located 40 km north of NYC, immediately south of the Village of Piermont. It is a relatively narrow marsh (approximately 1.15 km 2 ) bounded by the Hudson River with 3.2 km shoreline to the east and the uplands Tallman Mountain State Park to the west. The width of the Marsh starts from 100 m at the southern tip and increases to 600 m at the northern end. Phragmites dominates expansive swaths of the Marsh, including the entire margin of the Marsh abutting the Hudson River, as well as a complex network of tidal creeks.
The Marsh is a semidiurnal microtidal system with 1.21 m tidal range, and the average elevation of marshland is around 0.8 m, higher than the mean higher high water (0.63 m) (United States Geological Survey [USGSa]). Therefore, the peak high tides are usually below the marsh platform, even for spring tide conditions. With meteorological forcing, water from the Hudson River can flood the marsh platform irregularly. However, due to climate change, the coastal area around NYC is experiencing accelerated SLR with approximately twice the global rate (Horton et al 2015). The effect of SLR on the Marsh is not addressed here for simplicity but will be addressed in a future article.

Vegetation mapping and measurement
As shown in the vegetation map (figure 2(a)), Piermont Marsh is a typical example where Phragmites monocultures have overtaken previously biologically diverse brackish marsh plant communities (Mills et al 1996). In 2014, the area of Phragmites marsh in Piermont increased to 0.94 km 2 , about 92% of the total marsh area.
Natural resource managers in the eastern US, including those at HRNERR and New York State Department of Environmental Conservation (NYS-DEC), are concerned about the reduction of local biodiversity and productivity by the prolific Phragmites (Chambers et al 1999). Restoration efforts that replace Phragmites marshes with native species have often been suggested by marsh managers; for example, the restoration of native Typha has been proposed and is underway in three other HRNERR tidal wetlands, i.e. Tivoli Bays, Stockport Flats, and Iona Island Marsh.
Besides biodiversity-conservation considerations, the coastal protection functions of marshes, including those in the HRNERR, are also receiving more attention, especially with accelerating SLR and intensification of storms in the future. Vegetation characteristics, e.g. diameter, density, height, morphological structure, and mechanical properties, play significant roles in the vegetation-flow interactions and affect energy dissipation as well as flow pattern (Tempest et al 2015), which may change landscape morphology over the long term. Phragmites stems are considerably taller and stiffer and possess a larger diameter than previously dominant native plants, such as Typha, in the region (Bellavance and Brisson 2010). The sediment accretion rate in Phragmites marshes also tends to be higher than in those with native species, such as Spartina alterniflora (New Jersey Department of Environmental Protection (NJDEP)  (2020). Image stated to be in the public domain. 2020). Therefore, it is necessary to develop a better understanding of Marsh's unique capability to attenuate storm-induced flood and wave damage on residential structures before carrying out any restoration project.
Two field surveys were conducted in May (late spring) and September (early fall) 2017 to measure the vegetation structure in both seasons in the study area. The height, diameter, stem density, and leaves were measured along transects (red solid lines in figure 2(a)) for Phragmites and native species (Scirpus spp. and Spartina patens) in the Piermont Marsh. The same data were collected for Typha in the nearby Iona Island Marsh (18 miles north of the Piermont Marsh). The vertically varying frontal area per volume A f and the wetted area per volume A w were then calculated (figures 2(b) and (c)) from measured stem density and diameter and leaf area following Lapetina and Sheng (2014). A Typha plant is characterized by long strap-like leaf blades branching out from shoot base; whereas the leaf blades of a Phragmites plant are relatively shorter and grow between the middle and upper levels of the stem. Denser leaf blades of Typha greatly increase the wetted area compared with Phragmites. In May, canopies are mixed with short young, broken, and unbroken dead stems, resulting in multiple layered morphological structures-and this vertical variation of stem density is more pronounced for Typha-whereas the morphology is relatively uniform in September as live plants are fully grown. The frontal areas of Phragmites and Typha at the bottom layer in late spring are greater than those in early fall due to the existence of broken plant stems (left from last growing season) and the seasonal emergence of short new shoots. Phragmites produce more litter on the surface, and this litter decays more slowly, and therefore traps more sediment for a longer time.

Model setup 3.3.1. Hydrodynamic model setup
A high-resolution curvilinear grid, nested within a regional-scale domain ( figure 1(a)), was developed for the Piermont area, nearby uplands, as well as the surrounding tidal river (Hudson River) and creek system ( figure 1(c)). The grid had an average cell size of 6.45 m, the minimum cell size reached 0.62 m in the tidal creeks, and the resolution of land cells was 1-3 m, sufficient to resolve the flow in the main tidal creeks and overland flooding. Hudson River bathymetry collected by NYSDEC with 1 m resolution in 2014-2015 was used. For shallow waters and land area that the dataset did not cover, a seamless bathymetry and topography dataset (1/9 arc-sec resolution) from Hurricane Sandy Digital Elevation Models (National Center for Environmental Information) was used. All the vertical positions were referenced to the NAVD88.
This study examined the impacts of the Piermont Marsh on flooding, waves, currents, and structural losses during Sandy. The open boundary conditions along the southern and northern boundaries were extracted from a large-scale domain Sandy simulation. As shown in figure 3(a), the peak water levels at the boundary cells lied in between the surveyed high-water marks at NYROC07562 (USGSb) and NYWES07662 (USGSc), indicating high confidence of the boundary conditions for the fine-scale simulations. Winds ( figure 3(b)) recorded at the Hudson River Environmental Conditions Observing System weather station at Piermont Pier were used to provide the meteorological forcing. Before Sandy's landfall, the study area experienced north-easterly wind. As the eye of Sandy approached NJ shore, wind in Piermont gradually increased to 19.65 m s −1 and switched clockwise to easterly-south-easterly. In the meantime, as the peak storm tide reached Piermont from NYC, it was 9.5 ft (2.9 m).

Input for vegetation module
Due to high Re (O ( 10 3 ) ) during storm conditions, the profile drag coefficient C p is in the convergence region. As in Sheng and Zou (2017), a series of C p values (0.1-0.3) were tested, with negligible difference for the flooding results. Therefore, C p was set to 0.2 to account for the lower modulus of elasticity of marsh vegetation compared with rigid cylinders.
For simulating wave dissipation by vegetation, we calculated the values of C D using the empirical formulas discussed earlier and the vegetation and wave orbital velocity data of Piermont Marsh. We chose to use a value of 0.86, about half of the value given by Ozeren et al (2014), for Piermont Marsh due to the good performance of Ozeren et al (2014) for the laboratory experiments and the more flexible nature of Phragmites stems compared to the Birch dowel used by Chapman et al (2015). Unless indicated explicitly, the vegetation characteristics surveyed in September 2017 and C D of 0.86 were used in the model. Sensitivity analysis showed that the increased drag coefficient would slightly reduce the dissipation distance without changing the results at Piermont Village.
Multiple-layered A f and A w shown in figure 2 were input to the model to represent the Phragmites and Typha marshes, and a uniform morphological structure with a density of 600 stems m −2 , stem diameter of 3.5 mm, and height of 0.9 m, was used to compute A f and A w for the small native marsh patches where Scirpus and S. patens were dominant. In comparison, Piermont Marsh had a stem density of ∼100 stems m −2 , stem diameter of ∼0.5 cm, and a height of 3.38 m.

Simulation results
To understand the role of Piermont Marsh in reducing damage to the Village of Piermont during Sandy, we simulated the flooding, currents, and wave in the Piermont Marsh and Village area. The results are summarized below.

Simulated inundation during Sandy
As shown in figures 4(a) and (b), inundation in the Village was barely reduced (<2 cm) by the Marsh, due to its relatively narrow width (∼500 m) along the mostly Easterly peak wind direction. The results were similar if the Marsh were replaced by Typha, as shown in figure 4(c). During Sandy, the simulated maximum water level was 2.88 m at a house in the southern Village and 2.86 m at the Village Pier ( figure 4(d)), within 3% of the observed values. If the Marsh were removed, the corresponding values would become 2.89 m and 2.86 m. The residential properties of the Village are shown in figure 4(f).

Simulated currents during Sandy
As shown in figure 5, surface currents in the southern Village would be very strong if the Marsh were removed, while the existing Marsh as well as a 0917 (September 2017) Typha Marsh both significantly reduced the surface currents from 60 cm s −1 (current speed before the water interacted with the marsh) to less than 5 cm s −1 , which would minimize the transport of debris from outside the southern edge of the Marsh into the Village. With the 0517 (May 2017) Marsh structure, the existing Phragmites Marsh would significantly reduce the currents, while the Typha Marsh with considerably shorter and sparser plant canopy would not.

Simulated wave height
As shown in figure 6, without the Marsh, the significant wave height at the southern Village would have been more than 50 cm, barely dissipated from the wave height at the edge of the Marsh of about 60 cm. With the existing Marsh or a 0917 Typha Marsh, wave height was significantly dissipated to about 15-20 cm. Within about 150 m, the wave height was dissipated by more than 50%, consistent with the dissipation   length shown in equation (5). With the 0517 Marsh structure, the existing Phragmites Marsh would still significantly reduce the wave height while the seasonally lower Typha Marsh would not.

Economic analysis of the role of Piermont Marsh during Sandy
As a first step to assess the ecosystem service value of the Piermont Marsh in reducing property damage due to flood and wave during storms, we conducted a parcel-based economic analysis using the simulated inundation and wave along with the best available building footprint data (NYS GIS Clearinghouse) and damage functions from the U.S. Army Corps of Engineers (U.S. Army Corps of Engineers (USACE) 2015). The height of the wave crest (WC) curve was used in regions where the depth limited controlling wave height H C was greater than 1.5 ft, while flood depth curve was used elsewhere. Here, WC is defined as: where d is the still water elevation. Our estimated property damage of $3.72 M for the 41 properties that received $3.56 M NFIP payouts had an error of 4.4%. The actual claims might have been greater than the payouts because many claimers were either uninsured or under-paid. Table 1 shows that flood and wave accounted for ∼70% and ∼30% of the structural loss. If all the 500+ buildings in the Village were included in the damage analysis, the estimated loss of $11.9 M (with 71% being flood damage) is less than the total estimated loss of $20 M (including losses that were not included in our analysis) by the Village of Piermont (2014), with an avoided loss of $902 K (7.6% of the actual loss). The relative wave loss was 26.2%. Sheng et al (2012) found that surge dissipation varies significantly with local wetland type, crossshore wetland width, and storm characteristics. While the Marsh was not effective in reducing flood damage during Sandy, it could significantly buffer flood damage during a different storm with different intensity and wind direction. Therefore, to calculate the value of Piermont Marsh in reducing structural loss, it is necessary to consider an ensemble of storms with lesser intensity but higher frequencies. To this end, we used the joint probability method with optimal sampling JPM-OS (Condon andSheng 2012, Yang et al 2019) to generate a set of optimal storms to represent all possible storms predicted by the North Atlantic Stochastic Hurricane Model NASHM (Hall andJewson 2007, Hall andYonekura 2013). Following Yang et al (2019), we calculated the 1% annual chance flood and wave maps for Piermont and the associated structural loss due to flood and wave. Table 1 shows that the estimated structural loss for the 1% annual chance event was $18.82 M where wave accounted for ∼60% of the loss. The relative avoided loss due to flood and wave was 11.34% and 12.1%, respectively.

Conclusions
A comprehensive study was conducted to examine the role of the Phragmites-dominated Piermont Marsh in buffering wave and flood-induced structural loss in the Village of Piermont during Sandy. The study used a vegetation-resolving three-dimensional surge-wave modeling system CH3D-SSMS with 0.62-6.45 m resolution, with detailed marsh data and forcing conditions provided by simulated and observed wind and water level data at the air-sea interface and the model open boundary, respectively. Before application to Piermont, the model was used successfully to simulate two laboratory experiments, one on flow in vegetation and one on wave dissipation over a sloping beach, plus one field measurement in the Chesapeake Bay. Appropriate drag coefficients were determined.
Our results showed that during Sandy, the high wave energy in the Hudson River was significantly (56%) dissipated by the Phragmites-dominated Piermont Marsh before reaching the Piermont Village. On the other hand, only 1% of the flood was dissipated by the relatively narrow marsh width subject to peak easterly wind. The Marsh was able to significantly reduce the current and prevent the riverborne debris from transporting across the Marsh and reaching the Village. If the tall and rigid Phragmites were replaced by the slightly shorter Typha found in nearby Iona Marsh, the wave and debris buffering capacity of the Marsh would have been preserved during Sandy when both plants were tall and dense in the fall. However, if the Marsh were replaced by Typha and experienced a Sandy-like storm during the nongrowth season (e.g. May), the Marsh's buffering capacity would be greatly reduced relative to the existing Marsh. The shorter and sparser Typha marsh would not have provided much buffering during the frequent N'easters in winter and spring.
Using the simulated flood and wave over the Piermont Village with residential property data and damage functions, our estimate of structural loss due to flood and wave during Sandy was $11.93 M, compared to the more than $20 M losses including rebuilding and income losses reported by Village of Piermont. If the Piermont Marsh were absent, the total loss would have increased by 8% to $12.83 M due to 26% increase in wave-induced damage. The estimated losses for the 1% annual chance event at Piermont Village are $18.82 M and without the Marsh, they could reach $20.95 M. Thus, Piermont Marsh is preventing $21 300 in annualized loss. While the avoided losses during Sandy were mainly associated with wave-induced damage, the avoided losses of the 1% annual chance event were a combination of flood (11%) and wave (12%) induced losses. Since the 1% annual chance flood is generated by storm events that are more frequent but less intense than Sandy, our results are consistent with the claims by Rezaie et al (2020) that wetlands are more efficient in buffering more frequent storms with lower intensities relative to less frequent, but more energetic events like Superstorm Sandy.
Our comprehensive study confirmed that the value of marsh for flood protection depends significantly on vegetation type, storm characteristics, and local community. The 200+ acre tall (>2 m) and dense Piermont marsh is found to provide significant protection for Piermont Village during storms. If the marsh were made of short (<1 m) and sparse Spartina along New Jersey coast, they would not have been able to protect the Village from Sandy, consistent with the finding of Lathrop et al (2019) for South Jersey. Importantly, extensive model verification with available data (vegetation, water level, wave, and structural loss) during Sandy confirmed the robustness of our study. However, although detailed vegetation data was used, we did not include the damage of vegetation due to excessive wind or hydrodynamic loading. The damage assessment used actual building data but relied on empirical damage functions. To reduce uncertainties in the future, species-dependent storminduced vegetation damage can be investigated, and more robust building-specific damage functions can be developed. Nevertheless, our study provided realistic assessment on the value of the Piermont Marsh for flood protection, which can be used as guideline for resilience and restoration planning by the Village and marsh managers. This study demonstrated the significant value of the invasive common reed, Phragmites, in protecting coastal communities from storm-induced structural loss. Moreover, Rooth and Stevenson (2000) and New Jersey Department of Environmental Protection (NJDEP) (2020) found that Phragmites enables marshes to elevate faster than native species. Therefore, natural resource managers considering eradicating Phragmites to enhance biodiversity and other ecosystem functions should weigh the tremendous value of Phragmites in protecting coastal communities from structural damage due to storms and SLR to develop site-specific restoration plans that balance biodiversity and coastal resilience. The role of Piermont Marsh in protecting Piermont Village from future flood and wave damage will be examined in an upcoming paper.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.